Sunderland Andrew 2009

Characterising and improving a magnetic gradiometer for
geophysical exploration.
by
Andrew Sunderland, B.Sc.(Hons)
Submitted in partial fulfilment of

the requirements for the degree of
Doctor of Philosophy
School of Physics
The University of Western Australia
July, 2009
ABSTRACT
Characterising and improving a magnetic gradiometer for geophysical

exploration.
by Andrew Sunderland, B.Sc.(Hons)
Supervisors: Prof. D.G. Blair, Dr L. Ju, and Dr A.V Veryaskin
Magnetic gradiometers are powerful tools for mineral exploration. The magnetic
field contains valuable information about the mineral content of the surveyed terrain.
The magnetic gradient specifies the amount of spatial variation in the direction and
magnitude of the magnetic field. Surveys that measure the magnetic gradient pro-
vide vastly more information about geological targets than the magnetic field alone.
This technology could have enormous benefits in terms of new discoveries and lower
exploration costs.
The magnetic gradient is normally calculated by subtracting the outputs of two
total field magnetometers which are separated by a baseline. In 1997, a direct string
magnetic gradiometer (DSMG) was developed that directly measures magnetic gra-
dients using only a single string as its sensing element. This thesis describes research
conducted to improve the sensitivity and performance of the DSMG.
The main advantage of the DSMG is that only gradients can induce second har-
monic vibrations in the string. Thus, the DSMG is insensitive to uniform magnetic
fields that we are not interested in, such as the global magnetic field of the Earth.
By using inductive electronics to measure second harmonic string vibrations, we can
select to measure the magnetic gradient of nearby targets.
Recent work has shown that a magnetic gradiometer with a noise floor of
√
0.01 nT/m/ Hz should be sufficiently sensitive for geophysical exploration. In order
to reach this goal, this thesis presents an investigation of all noise sources affecting
the DSMG.
The dominant noise in the DSMG is thermal noise. Random thermal motions of
the molecules inside the string will cause the entire string to vibrate by a microscopic
amount. This random vibration is indistinguishable from weak magnetic gradients.
The signal to noise ratio of the DSMG can be increased by using more power or by
reducing the coupling of the string with the thermal reservoir.
In this work I show that the DSMG’s performance depends on its ability to dissi-
pate heat whilst minimising air damping. By combining a high current, an optimised
temperature and low pressure, the thermal noise level of the DSMG has been de-
√ √
creased from 0.65 nT/m/ Hz down to 0.18 nT/m/ Hz. I show that the thermal
noise floor can be decreased further by optimising the string dimensions, power con-
sumption, and materials. I then present the design parameters for the next generation
√
of sensor, which will reach the target sensitivity of 0.01 nT/m/ Hz for airborne geo-
physical applications.
In addition to its terrestrial applications, the DSMG is also suitable for deployment
in space. Gas damping is negligible in high vacuum and no vibration isolation is
required. This means that longer strings with low resonant frequencies can be used.
Using theoretical modelling, I show that a space borne DSMG should be able to
match the white noise level of SQuID based magnetic gradiometers and have a lower
1/f noise corner. Deployment in space could be the most viable application of the
DSMG because of the ease of operation and enhancement of sensitivity.
If the thermal noise level is reduced then other sources of noise will start to become
more important. When rotated in the Earth’s magnetic field, the DSMG detects a
pseudo magnetic gradient despite the field being almost uniform. A possible cause is
magnetically susceptible parts which are magnetically aligning with the Earth’s field.
I have conducted a thorough investigation of magnetic susceptible parts in the DSMG
and reported the results in this thesis.
In the DSMG, a pair of inductive pickup coils are used to measure the string’s
√
displacement with a root mean square accuracy of 10−11 m/ Hz. This is adequate at
present but the inductive electronics may not be sensitive enough after other improve-
ments in the DSMG are implemented. Here, I present a new capacitive displacement
√
readout with a high sensitivity of 10−13 m/ Hz.
The thesis also presents some magnetic gradient measurements in the lab and
the results of a ground survey in the field. These trial measurements are used to
characterise the DSMG and demonstrate its effectiveness for airborne surveying.
Acknowledgments
I would like to thank my supervisors Prof. David Blair, Dr Li Ju, and Dr Alexey
Veryaskin. Without their assistance and support, writing this thesis would not have
been possible. Prof. Blair’s knowledge of all areas of physics proved invaluable and
he was a leading force behind all my publications. I must address special thanks to
Dr Ju for her guidance, her door was always open if I needed help.
I would like to thank Dr. Alexey Veryaskin, Dr Wayne McRae, and Mr Howard
Golden for their invaluable contributions to the research. Dr Veryaskin’s expertise
and great knowledge of magnetic gradiometry was vital. His encouragement since the
first day of my PhD project inspired my research. It was a great pleasure for me to
work with him. I must thank Dr. McRae again for proof-reading many versions of
my thesis and constantly reviewing my work. I would also like to thank Mr Golden
for his expertise in geophysics and many useful discussions and suggestions.
I would like to thank the workshop staff, especially Peter, David and Steve, for
their workmanship and invaluable ideas.
To those friends who have given me support and warmth that have made my time
at UWA so much richer, I give my thanks. Thank you Sascha, Yan, Fan, Kazumi,
Jerome (thank you for Wednesday nights at the pub), Jean-Charles, Lucienne, Eric,
Andrew Wooley, Haixing, Zhongyang, Sunil, Viet, and Michelle (thanks for being the
best housemate). Especially, thanks to Mr. Slawomir Gras for his great friendship and
to my lovely girlfriend Sundae for being the best girl in the world and her continued
love and support.
Finally, I would like to express gratitude to my parents and my brothers, especially
Ian for his help in the last year.
iii
Contents
1 Introduction 1
1.1 Tools for Geophysicists . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Magnetic field of the Earth . . . . . . . . . . . . . . . . . . . . . . . . 2
1.3 Magnetic minerals . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.4 What is a magnetic gradient? . . . . . . . . . . . . . . . . . . . . . . 6
1.5 Example magnetic gradients . . . . . . . . . . . . . . . . . . . . . . . 7
1.6 Advantages of measuring the magnetic gradient . . . . . . . . . . . . 10
1.6.1 Gradient measurements in general . . . . . . . . . . . . . . . . 10
1.6.2 Magnetic gradients in particular . . . . . . . . . . . . . . . . . 11
1.6.3 Magnetic gradient tensor versus Total field gradient . . . . . . 11
1.6.4 Space applications . . . . . . . . . . . . . . . . . . . . . . . . 12
1.6.5 Borehole applications . . . . . . . . . . . . . . . . . . . . . . . 13
1.6.6 Archeological applications . . . . . . . . . . . . . . . . . . . . 13
1.6.7 Other applications . . . . . . . . . . . . . . . . . . . . . . . . 13
1.7 Limitations of magnetic gradient measurements . . . . . . . . . . . . 13
1.8 Common mode rejection . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.9 Existing magnetic gradiometers . . . . . . . . . . . . . . . . . . . . . 15
1.9.1 Fluxgate gradiometers . . . . . . . . . . . . . . . . . . . . . . 16
1.9.2 SQuID gradiometers . . . . . . . . . . . . . . . . . . . . . . . 17
1.9.3 Total field gradiometers . . . . . . . . . . . . . . . . . . . . . 18
1.9.4 Other magnetic gradiometers . . . . . . . . . . . . . . . . . . 19
1.10 History of the Direct String Magnetic Gradiometer . . . . . . . . . . 20
1.10.1 The beginning 1997-2001 . . . . . . . . . . . . . . . . . . . . . 20
1.10.2 Principal of Operation . . . . . . . . . . . . . . . . . . . . . . 21
1.10.3 Vibration noise . . . . . . . . . . . . . . . . . . . . . . . . . . 24
v
1.10.4 Airborne trials 2003-2004 . . . . . . . . . . . . . . . . . . . . . 25
1.11 Target sensitivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
1.12 Noise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
1.13 Thesis outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
2 Characteristics of the DSMG 37

2.1 Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
2.2 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
2.3 Iron Sphere measurements . . . . . . . . . . . . . . . . . . . . . . . . 42
2.4 Laboratory noise performance . . . . . . . . . . . . . . . . . . . . . . 45
2.5 Field testing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
2.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
3 Displacement readout 53
3.1 Preface: Inductive readout . . . . . . . . . . . . . . . . . . . . . . . . 53
3.1.1 Principal of operation . . . . . . . . . . . . . . . . . . . . . . 53
3.1.2 Mutual inductance . . . . . . . . . . . . . . . . . . . . . . . . 54
3.1.3 Dynamic range . . . . . . . . . . . . . . . . . . . . . . . . . . 56
3.1.4 Sensitivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
3.2 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
3.3 Amplitude measurements . . . . . . . . . . . . . . . . . . . . . . . . . 60
3.4 Capacitance versus distance . . . . . . . . . . . . . . . . . . . . . . . 63
3.5 Phase measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
3.6 Noise measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
3.7 Applying the capacitive readout to a DSMG . . . . . . . . . . . . . . 75
3.7.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
3.7.2 Thermal noise . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
3.7.3 Rationale for the new readout . . . . . . . . . . . . . . . . . . 77
3.7.4 Common mode rejection ratio . . . . . . . . . . . . . . . . . . 77
3.8 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
3.9 Postscript: Space applications . . . . . . . . . . . . . . . . . . . . . . 80
4 Heading error 83
4.1 Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
4.2 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
4.3 Direct String Magnetic Gradiometer . . . . . . . . . . . . . . . . . . . 87
4.4 Choice of low susceptibility materials . . . . . . . . . . . . . . . . . . 88
4.5 Comparison of materials and contamination reduction . . . . . . . . . 90
4.5.1 Surface contamination . . . . . . . . . . . . . . . . . . . . . . 90
4.5.2 Intrinsic susceptibility of materials . . . . . . . . . . . . . . . 93
4.5.3 Torlon . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
4.6 Error analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
4.7 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
4.8 Postscript: Heading error calculation . . . . . . . . . . . . . . . . . . 101
4.8.1 The frame . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
4.8.2 Parts near the one quarter and three quarter points . . . . . . 103
4.8.3 Parts near the end points of the string . . . . . . . . . . . . . 104
4.8.4 Heading error calculation . . . . . . . . . . . . . . . . . . . . . 105
4.8.5 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
4.8.6 Out of plane vibration . . . . . . . . . . . . . . . . . . . . . . 107
5 Sensitivity and optimisation 115

5.1 Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
5.2 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
5.3 Dynamics of a bounded current carrying string . . . . . . . . . . . . . 118
5.4 Noise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
5.4.1 Electronic noise . . . . . . . . . . . . . . . . . . . . . . . . . . 120
5.4.2 Vibration noise . . . . . . . . . . . . . . . . . . . . . . . . . . 121
5.4.3 Thermal noise . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
5.5 The loss angle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
5.5.1 Air damping . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
5.5.2 Clamp losses . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
5.5.3 Violin mode losses . . . . . . . . . . . . . . . . . . . . . . . . 127
5.6 Noise measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
5.7 Optimising the magnetic gradiometer . . . . . . . . . . . . . . . . . . 133
5.7.1 Optimal temperature . . . . . . . . . . . . . . . . . . . . . . . 134
5.7.2 Maximum temperature . . . . . . . . . . . . . . . . . . . . . . 137
5.7.3 Optimum pressure . . . . . . . . . . . . . . . . . . . . . . . . 138
5.7.4 Optimum wire diameter . . . . . . . . . . . . . . . . . . . . . 143
5.8 Sensitivity as a function of power . . . . . . . . . . . . . . . . . . . . 144
5.9 Comparison of materials . . . . . . . . . . . . . . . . . . . . . . . . . 145
5.10 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
5.11 Appendix A: Elastic energy and Dynamic tension . . . . . . . . . . . 147
6 Space Applications 155

6.1 Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155
6.2 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156
6.3 String magnetic gradiometer . . . . . . . . . . . . . . . . . . . . . . . 157
6.4 Thermal noise and the mechanical quality factor . . . . . . . . . . . . 160
6.5 Vibration noise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163
6.6 Operation in the natural space environment . . . . . . . . . . . . . . 164
6.7 Thermal radiation in space . . . . . . . . . . . . . . . . . . . . . . . . 164
6.8 25−70 Kelvin operation . . . . . . . . . . . . . . . . . . . . . . . . . 166
6.9 4−25 Kelvin operation . . . . . . . . . . . . . . . . . . . . . . . . . . 167
6.10 Low-frequency noise background . . . . . . . . . . . . . . . . . . . . . 168
6.11 Low-frequency noise improvements . . . . . . . . . . . . . . . . . . . 170
6.12 Readout . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170
6.13 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171
7 Conclusions and Future Work 179

7.1 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179
7.2 Future work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
7.2.1 Heading error . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
7.2.2 Other technology . . . . . . . . . . . . . . . . . . . . . . . . . 181
7.2.3 Present technology . . . . . . . . . . . . . . . . . . . . . . . . 182
A List of publications 185

List of Figures
1.1 The four main contributions to the total magnetic field . . . . . . . . 3

1.2 East-West magnetic survey passing over an underground dyke with
induced magnetisation . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.3 East-West magnetic survey passing over a pipe shaped ore body with
induced magnetisation . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.4 East-West magnetic survey passing over an spheroid shaped ore body
with induced magnetisation . . . . . . . . . . . . . . . . . . . . . . . 8
1.5 The combined gravity and magnetic gradiometer was comprised of 4
parallel channels: two for the measurements of gravity gradients Txz ,
Tyz and two for the measurement of magnetic gradients Bxz , Byz . . . 21
1.6 Concept diagram of the DSMG . . . . . . . . . . . . . . . . . . . . . 22
1.7 Components of the DSMG system . . . . . . . . . . . . . . . . . . . . 23
1.8 Isolation characteristics of the 4-stage mechancial isolation system . . 24
1.9 A brief extract of the data taken during airborne trials of the DSMG 26
1.10 Noise comparison of magnetic gradiometers . . . . . . . . . . . . . . . 27
1.11 Comparison of thermal noise with electronics noise in the DSMG . . . 29
2.1 Photograph of the direct string magnetic gradiometer (DSMG) during

a line survey at Shenton Park . . . . . . . . . . . . . . . . . . . . . . 41
2.2 Magnetic gradient response of a 28.7 cm3 iron sphere moved north-
south on a conveyor belt at three different positions below the gra-
diometer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
2.3 Magnetic tensor components along a 1000 m north-south survey line
past an ore body . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
2.4 Amplitude spectral distribution plots of the direct string magnetic gra-
diometer making a quiet recording in the laboratory . . . . . . . . . . 46
ix
2.5 Quiet recording of the direct string magnetic gradiometer in the labo-
ratory with a signal of amplitude 2 nT/m at frequency of 0.1 Hz . . . 47
2.6 The complete 6000 s quiet laboratory recording . . . . . . . . . . . . 47
2.7 Contour map of total magnetic intensity at the Shenton Park testing
site . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
2.8 Magnetic gradients along a survey line at 6 464 113 m east at Shenton
Park . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
3.1 Detailed concept diagram of DSMG . . . . . . . . . . . . . . . . . . . 53

3.2 Diagram of a single pickup coil and the string . . . . . . . . . . . . . 55
3.3 Mutual inductance between the string and a pickup coil as a function
of the distance separating them . . . . . . . . . . . . . . . . . . . . . 57
3.4 Diagram of the LC series resonant bridge tuned to 10 MHz pump signal
provided by a low phase noise oscillator. . . . . . . . . . . . . . . . . 60
3.5 Sketch of the ribbon and a pair of capacitive plates using an oblique
cabinet projection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
3.6 Voltage across the LC circuit versus frequency . . . . . . . . . . . . . 62
3.7 Diagram of the complete capacitive readout circuit . . . . . . . . . . 63
3.8 Diagram of the ribbon and the two capacitive plates with the left plate
tilted . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
3.9 Relative output of the frequency mixer versus frequency . . . . . . . . 69
3.10 Photograph of experimental setups A and B . . . . . . . . . . . . . . 71
3.11 Spectral density of differential displacements measured by an
accelerometer placed at the one quarter position on the sensor . . . . 72
3.12 Spectral density of mixer output between 900 Hz and 1100 Hz . . . . 73
3.13 Spectral density of mixer output between 0 Hz and 800 Hz . . . . . . 74
4.1 Heading error tests are completed by rotating the sensor through 360◦
in a magnetically benign area to check for self-induced gradient response. 83
4.2 DSMG frame with the string clamped rigidly at each end . . . . . . . 88
4.3 The susceptibility of brass rises rapidly with increasing iron concentration 95
4.4 Magnetisation of three non-magnetic materials that are used in the
construction of the DSMG . . . . . . . . . . . . . . . . . . . . . . . . 96
4.5 Magnetisation of Torlon samples from two different manufacturers . . 97
4.6 Cross section view of the string and nearby magnetically susceptible
parts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
4.7 Sketch of the gradiometer system including the sensor frame, vacuum
tank and mechanical isolator . . . . . . . . . . . . . . . . . . . . . . . 106
4.8 This graph shows the amplitude of two 2nd violin modes (in plane
motion and out of plain motion) for the ribbon inside the DSMG . . 109
5.1 DSMG frame with the string clamped rigidly at each end . . . . . . . 119
5.2 The string inside the DSMG is 0.25 m long and vibrates along the
x-axis. Its cross section can be either a flat ribbon or a round wire. . 123
5.3 This graph compares the measured loss angle with the theoretical gas
losses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
5.4 Plot of the ribbon’s Q factor as a function of tension in a high vacuum 129
5.5 Plot of the ribbon’s relaxation time as a function of the ribbon’s reso-
nant frequency in a high vacuum . . . . . . . . . . . . . . . . . . . . 130
5.6 Spectral distribution of DSMG noise whilst using a flat ribbon as the
sensing element . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
5.7 Quiet recordings of the DSMG in the laboratory whilst using a flat
ribbon as the sensing element . . . . . . . . . . . . . . . . . . . . . . 132
5.8 Spectral distribution of DSMG noise whilst using a round wire as the
sensing element . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
5.9 Sketch of Eq. 5.30. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
5.10 This graph shows measurements of the change in temperature when
alternating current is applied to the flat ribbon or round wire. . . . . 136
5.11 As more current is applied to the string, the noise level should decrease
and the temperature should increase . . . . . . . . . . . . . . . . . . 136
5.12 As the air pressure is decreased, the current must also be decreased to
avoid overheating . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
5.13 In this graph we consider the DSMG sensitivity in different air pressures
whilst under the constraint T < 373 K . . . . . . . . . . . . . . . . . 142
6.1 String magnetic gradiometer design sensitivities as a function of ambi-

ent temperature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165
6.2 The middle of the ribbon heats up to 50 K while the ends are in thermal
contact with a heat sink at 40 K . . . . . . . . . . . . . . . . . . . . . 166
6.3 Comparison of magnetic gradiometers . . . . . . . . . . . . . . . . . . 169
List of Tables
1.1 The Euler structural index, N, depends on the geometry of the source. 6
3.1 Phase noise of the 10 MHz quartz oscillator as a function of frequency 75
4.1 Concentration of ferromagnetic elements in each sample . . . . . . . . 89

4.2 The amount of mass lost by the rectangular metal samples in acid
increases with longer immersion times, uncertainty is ± 0.01g. . . . . 90
4.3 The initial magnetic volume susceptibility of metal samples decreases
after an acid wash, although most of the reduction occurs in the first
10 minutes of the acid wash . . . . . . . . . . . . . . . . . . . . . . . 91
4.4 Magnetic permeability of cutting tools . . . . . . . . . . . . . . . . . 92
4.5 Initial magnetic volume susceptibility of plastic and ceramic samples
before and after an acid wash . . . . . . . . . . . . . . . . . . . . . . 92
4.6 Intrinsic susceptibility of each material examined in this paper . . . . 94
5.1 Estimate of the residual losses of the ribbon and wire in a perfect vacuum126
5.2 Results from laboratory noise measurements . . . . . . . . . . . . . . 134
5.3 High temperatures degrade the mechanical properties of the aluminium
alloy 6061-T6511 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
5.4 Room temperature properties of various materials . . . . . . . . . . . 146
6.1 Stress relaxation becomes increasingly problematic at higher tempera-

tures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161
6.2 Prospective materials for space DSMG . . . . . . . . . . . . . . . . . 167
xiii
Chapter 1
Introduction
1.1 Tools for Geophysicists

The search for mineral wealth and the prospect of a large find has inspired many
generations of mining prospectors[1]. The demand for minerals increases every year
and securing the required resources is vital to the functioning of modern society.
However, most of the easy to find ore deposits that were located near the surface
have already been found. Geophysicists need every new tool they can get in order to
discover deeper and larger ore bodies[2].
The science of geophysics can be used when there are physical differences between
the properties of the target ore body and the properties of the country rock[2]. The
tools and techniques available to geophysicists for measuring physical properties in-
clude: gamma ray specrometers, grounded electrical methods, gravimeters, gravity
gradiometers, seismic, resistivity and polarisation, magnetics and electromagnetics.
Magnetics is the cheapest and most widely used tool by geophysicists in hard
rock environments. Magnetic surveys form an integral part of mineral exploration
programs and are also commonly part of geological mapping programs. A new growth
area has been the use of magnetometers in hydrocarbon exploration[3].
The first vector magnetometers were developed in the 1930s[4]. It was recognised
that vector measurements of the magnetic field inferred additional information about
the local geology. Further insight could be gained by combining multiple vector
magnetometers together in order to measure gradients. However, early magnetic
vector surveys saw limited success due to the difficulty in reducing the system noise
and sensor orientation errors to the levels needed to adequately measure the subtle
1
2 Chapter 1. INTRODUCTION
field changes[5].
In the 1960’s, total magnetic intensity sensors were developed that could measure
the magnitude of the magnetic field with unprecedented accuracy[5]. Total magnetic
intensity sensors can be rotated freely without introducing heading errors[6]. Most
geophysicists stopped measuring the vector and tensor magnetic field, instead using
multiple measurements of the total field to calculate estimates of these components[5].
Recently, there has been renewed interest in measuring the 1st order tensor com-
ponents of the magnetic field, also known as the magnetic gradient. Magnetic gradient
surveys retain the benefits of vector surveys without the disadvantage of extreme ori-
entation sensitivity. The data to be gained from magnetic gradient measurements
will open up a wide new range of data processing techniques for geophysicists[7].
Devices that measure the magnetic gradient are known as magnetic gradiometers.
A magnetic gradiometer based upon the principals of a vibrating string is the subject
of this thesis.
1.2 Magnetic field of the Earth
The magnetic field of the Earth resembles the field of a large bar magnet near its
centre or the field of a uniformly magnetised sphere. The field is thought to be
produced by currents of electrically charged particles in the outer part of the Earth’s
core, which is mostly composed of molten iron. As the earth slowly cools, the lighter
charged particles are rising whilst the the heavier iron condenses onto the base of
the hot inner core below. The rotation of the Earth tends to organise these electric
currents into loops aligned along the north-south polar axis.
Over 99 percent of the Earth’s magnetic field energy is confined entirely within
the core[8]. We only observe the small portion of the magnetic field that extends to
the surface and beyond. Fig. 1.1D shows the Earth’s magnetic field measured at a
altitude of 400 km above sea level, where it ranges from less than 30000 nT above
South America and South Africa to over 50000 nT above parts of Canada, Australia
and Siberia.
Surrounding the Earth is the magnetosphere. It extends 10 Earth radii on the
1.2. MAGNETIC FIELD OF THE EARTH 3
Figure 1.1: Total magnetic field intensity measured by a satellite 400 km above the Earth.
The four main contributions to the total magnetic field are: (a) spatially uniform mag-
netic noise produced by the magnetosphere, (b) variable magnetic noise produced by the
ionosphere, (c) shorter wavelength magnetic fields produced by magnetisation of the crust,
and (d) the long wave length magnetic field produced by the Earth’s outer core. This figure
was reproduced from a paper by Mandea and Purucker[9].
dayside, and more than 100 Earth radii on the nightside. The magnetosphere contains
belts of charged particles that constitute an electric current called the ring current[9].
The ring current induces a magnetic field in the Earth’s crust. Because the Earth is
much smaller than the magnetosphere, the induced magnetic field is nearly uniform
over the planet’s surface. Despite being spatially uniform, temporal variations occur
on the order of seconds, minutes and hours[10], with two harmonic overtones of periods
6 months and 1 year[9].
At altitudes of 50 km to 1000 km is the ionosphere. High energy photons such as

ultra violet light or x-rays, contain enough energy to ionize atoms. Electric currents
of ionized particles circulate in the ionosphere, generating magnetic fields[9]. The
amount of ionization depends on the amount of radiation received from the sun.
Thus, there is a diurnal variation in the magnetic field with a period of 1 day and
micropulsations with periods of seconds[10]. There are also magnetic storms which
occur several times a month, depending on solar activity.
The magnetic field at the surface of the earth can be roughly approximated by a
magnetic dipole, tilted 11◦ from its rotation axis. Such a dipole accounts for approxi-
mately 80% of the total magnetic field[11]. On top of this symmetric dipole, there are
numerous higher order poles and very large scale anomalous features produced by the
generating mechanism of the core. Lastly, but most relevant for the purposes of this
thesis, is the anomalous set of features in the Earth’s crust caused by concentrations
of magnetic minerals or other features of interest which distort the local magnetic
field[10].
Fig. 1.1 shows the relative contributions of the core magnetic field, crust magnetic
field, the magnetospheric noise and the ionospheric noise to the total field. During
a magnetic survey, a base station or reference station is often used. Magnetospheric
noise and ionosphere noise can be eliminated by subtracting the base station’s mag-
netic field reading from the survey’s magnetic field reading.
1.3. MAGNETIC MINERALS 5
1.3 Magnetic minerals

The application of the magnetic gradiometer discussed in this thesis is to measure
spatial changes in the Earth’s magnetic field. Spatial anomalies smaller than 1 km are
usually caused by a concentration of magnetic minerals which may be of geological
interest, or by iron artifacts which may be of archeological interest. Pure iron is
extremely magnetic. However, it does not occur naturally. The anomalies from
naturally occuring rocks are normally due to the presence of the very common mineral
magnetite (Fe3 O4 ) or its related minerals ulvospinel (Fe2 TiO4 ), maghemite (γFe2 O3 ),
ilmenite (FeTiO3 ) and pyrrhotite (Fe1−x S)[10; 12].
Many economic mineral deposits have significant similarities in terms of mineral
assemblages. Non-economic magnetic minerals may be associated with valuable non-
magnetic ores. Magnetic survey data may be used to recognise favourable locations
for ore bodies and, where possible, to directly detect ore bodies[13].
The magnetism inside magnetic minerals can be either induced or remanent. In-
duced magnetisation refers to the act of the Earth’s magnetic field inducing the mag-
netic field of the mineral to line up in the same direction. The magnetic field B inside
a mineral will equal:
B = μ0 H(1 + χ) (1.1)
where μ0 H is the magnetic field of the Earth and χ is the magnetic susceptibility
of the mineral. The susceptibility of magnetite is typically about 3 so the magnetic
field inside the mineral will normally be a multiple of the Earth’s field.
Remanent magnetisation refers to a mineral’s magnetic field always pointing in
the same direction. When heated, the magnetic field of remanent minerals will align
to the Earth’s field. However, after cooling below the Curie point, the magnetic field
becomes set and does not change.
The Earth’s magnetic field is slowly changing and has reversed direction many
times in the past. The time between magnetic reversals is sometimes as short as 10000
years and sometimes as long as 25 million years[9]. The magnetisation of remanent
minerals contains a record of the Earth’s magnetic field in the past. Remanence can
be recognised when a mineral’s magnetisation direction is different from the direction
Source N
Regional anomaly 0
Thin dyke or mineral bed 1
Long pipe 2
Compact ore body 3
Table 1.1: The Euler structural index, N , depends on the geometry of the source.
of the Earth’s present magnetic field.

The ratio of natural remanent magnetization to induced magnetization, called the
Koenigsberger ratio, can be used to characterise a magnetic ore body. Magnetite
has mostly induced magnetisation. Ulvospinel, maghemite and ilmenite have similar
amounts of induced and remanent magnetisation. Pyrrotite has mostly remanent
magnetisation[13].
The magnetic fields created by minerals are dipoles. In the space far from a
compact source, the field falls off as the inverse cube of the distance. Near more
realistic sources, the total magnetic intensity |B| will scale as a power law:
1
|B| ∝ (1.2)
rN
where r is distance and 0 < N < 3 is the Euler structural index. Typical values
of N are shown in Table 1.1. The Euler structural index can be used to calculate
the depth or the nature of a mineral deposit[14]. Because N > 0, the magnetic field
is stronger closer to the source and deeper sources have weaker signals. As a result,
magnetic survey aircraft try to fly as low as possible, normally only 60 − 150 m above
the ground[6].
1.4 What is a magnetic gradient?
Total magnetic intensity surveys measure the magnitude of the magnetic field |B|.
Vector magnetic surveys measure the magnitude and direction of the magnetic field.
There are 3 components of the vector magnetic field, one for each dimension:
1.5. EXAMPLE MAGNETIC GRADIENTS 7

Bx By Bz (1.3)
For each vector component, we can take an x, y or z derivative. Thus, there are
9 different magnetic gradients that make up the magnetic gradient tensor:
⎛ ⎞
∂Bx ∂Bx ∂Bx
⎜ ∂x ∂y ∂z ⎟
⎜ ⎟
⎜ ∂By ∂By ∂By ⎟
⎜ ⎟ (1.4)
⎜ ∂x ∂y ∂z ⎟
⎜ ⎟
⎝ ∂Bz ∂Bz ∂Bz ⎠
∂x ∂y ∂z
In practice, we only need to know 5 of the magnetic gradient components when
outside of a magnetic source[7]. Once 5 components have been measured, off diagonal
symmetry (eg ∂By /∂z = −∂Bz /∂y) and Laplace’s equation ∇ · B = 0, can be used to
calculate the other 4. Surveys that measure all of the magnetic gradient components
can produce more detailed and quantitatively interpretable maps and 3D models,
rather than the simple bump detection of total magnetic intensity surveys.
As an observer moves away from a source, the magnetic gradient decreases much
faster than the magnetic field. Taking the derivative of Eq. 1.2 gives:

∂Bz ∂ 1
∝ (1.5)
∂z ∂z r N
1
∝ (N +1) (1.6)
r
where r is distance and 0 > N > 3 is the Euler structural index. In the space
far from a compact source, the magnetic gradient falls off as the inverse 4th power of
the distance. Magnetic gradient measurements are thus insensitive to distant sources
that we are not interested in, such as the global magnetic field of the Earth.
1.5 Example magnetic gradients

In the last hundred years, considerable advances have been made in the collection
and organision of geophysical evidence. What once took weeks for a surface based ex-
ploration team can now be accomplished in hours with an airborne magnetometer[1].
|B|
Bxx
X (m)
0 1000 2000
2000 Bxy
Y (m)
Survey line Bxz
1000
-100 -100
-200 -200 Byy
-300 -300
0
Z (m) Byz
Bzz
0 1000 2000
X(m)
Figure 1.2: East-West magnetic survey passing over an underground dyke with induced
magnetisation. The dyke is ΔY = 500 m long, ΔX = 10 m thick, ΔZ = 200 m in depth
extent, and 100 m depth to top. The magnetic gradients are shown in arbitrary units.
|B|
Bxx
X (m)
0 1000 2000
2000 Bxy
Y (m)
Survey line Bxz
1000
-100 -100
-200 -200 Byy
-300 -300
0
Z (m) Byz
Bzz
0 1000 2000
X(m)
Figure 1.3: East-West magnetic survey passing over a pipe shaped ore body with induced
magnetisation. The pipe is ΔY = 500 m long, ΔZ = ΔX = 20 m diameter, and 100 m
depth to top.
|B|
Bxx
X (m)
0 1000 2000
2000 Bxy
Y (m)
Survey line Bxz
1000
-100 -100
-200 -200 Byy
-300 -300
0
Z (m) Byz
Bzz
0 1000 2000
X(m)
Figure 1.4: East-West magnetic survey passing over an spheroid shaped ore body with
induced magnetisation. The sphere is 100 m in diameter and 100 m depth to top.
1.5. EXAMPLE MAGNETIC GRADIENTS 9
Both types of magnetic exploration are still used and the principals are similar. The
target terrain is normally broken up into a 2 dimensional grid of survey lines. The
line spacing is normally 1 to 2 times the height above the source[15], which translates
to line spacings of 200-500 m for airborne surveys[6].
As a simple example, consider a single gridline from a hypothetical surface survey.

The sketches in Figs. 1.2, 1.3 and 1.4, each show a 2000 m long survey line passing
over an underground magnetic anomaly. To the right of the sketches is a set of graphs
showing the total magnetic intensity data and magnetic gradient data recorded during
the hypothetical surveys. The magnetic gradients are written in a commonly used
short form[7], eg Byz ≡ ∂By /∂z. The gradients are shown in arbitrary units. Large
iron ore deposits can have magnetic gradient signatures greater than 1000 nT when
the surveyor is directly over the deposit[13] whereas paramagnetic rock formations
buried 100 m below ground can have magnetic field signatures up to 0.5 nT/m[3].
In all three scenarios, the total magnetic intensity (the red curve) shows a large
bump at x = 1000 m. We know there is something there but there is no way to
tell the difference between the pipe shaped ore body, the underground dyke and the
spherical ore body.
Each component of the magnetic gradient tensor is sensitive to a different di-

mension. The pipe and dyke are elongated in the y direction. This causes the y
components of the magnetic gradient tensor Bxy , Byy and Byz to be slowly varying.
The plot of Bxz for the pipe, shows sharp dipolar peaks because the pipe is very thin
in the x and z directions.
Both the dyke and spherical ore body have the same depth below the surface (the
z direction) so their graphs of the Bzz tensor component look nearly identical. The
pipe is very thin in the z direction, which causes Bzz to oscillate near x = 1000 m.
All three graphs of the Bxx component are similar because all three ore bodies are
thin in the x direction.
Magnetic gradients can thus be used to constrain the shape of magnetic anom-
alies. In addition, the tensor components with sharp peaks allow the anomalies to be
resolved with greater spatial resolution. From the results of this simple hypothetical
survey, we can start to see some of the advantages of magnetic gradient measurements.
1.6 Advantages of measuring the magnetic gradient
1.6.1 Gradient measurements in general
Schmidt and Clarke[3] showed the enormous benefits that magnetic gradient mea-
surements could bring to airborne and surface geophysical surveys. Many of these
advantages apply to any gradient measurement, whether it be magnetic gradient or
gravity gradient[16]. The general advantages are:
• Higher resolution. Especially of shallow features (note the sharp peaks in

Figs. 1.2, 1.3, and 1.4). This will also allow discrimination of smaller features
and definition of discrete magnetic bodies located close together. In airborne
surveys, the higher resolution can be traded off against survey height, allowing
higher and hence safer flying[3; 17].
• Selectivity. Suppression of regional gravity or magnetic fields produced by

deep sources[16] (the regional anomalies shown in Fig 1.1C will produce gradi-
ents less than 0.01 nT/m[18]).
• Minimising aliasing. Aliasing is the effect of having greater data density

along flight lines (often less than 5 m) than between flight lines (often 200 m to
500 m apart)[6; 17]. With gradient data there is better interpolation between
survey lines. This can allow larger line spacings and hence cut survey costs[15].
It is also possible to determine on which side of a survey line that an anomaly
lies[3].
• Quantitive analysis. Better definition of structural features (in section 1.5 we

determined the shape of the dyke from the magnetic gradients). This includes
easier detection of long thin anomalies (eg the pipe in Fig. 1.3), determination
of the strike and dip directions of a mineral bed, and refined estimates of an
anomaly’s Euler structural index[3].
• Ease in interpretation. Inverse model solutions are more unique using gra-
dients. Tighter constraints are possible during modeling and better estimates
can be made regarding the extent of a mineral deposit[19].
1.6. ADVANTAGES OF MEASURING THE MAGNETIC GRADIENT 11
1.6.2 Magnetic gradients in particular
Because of the nature of the Earth’s magnetic field, there are advantages that apply
only to magnetic gradients:
• Insensitivity to rotation. Magnetic gradients surveys are insensitive to the

orientation errors that plague vector surveys[3]. Section 1.2 showed that the
Earth’s magnetic field is very strong (approximately ∼ 55000 nT in Western
Australia) whereas the signal from the local geology may be very weak (as small
as 10 nT). A tiny movement of a vector magnetometer can produce orientation
errors much larger than the geological signal you are trying to detect.
• Immunity to magnetic variations. By making a gradient measurement, time

varying magnetic noise from the distanct core of the Earth can be suppressed.
Furthermore, by measuring both the magnetic field and the magnetic gradient
on-board a moving vehicle it is possible to separate the spatial derivative ∂By /∂z
from the temporal derivative ∂By /∂t[20].
• Compensate errors. Measurements of the vertical gradient can help correct

errors introduced by altitude variations[15]. The magnetic gradient can also
be combined with vector data to separate the magnetic signal of the survey
aircraft from the ambient magnetic field; this technique was used on the Voyager
spacecraft[21].
• Source characterisation. Koenigsberger Ratio (remanent : induced) can

be computed[19]. Magnetic gradients also provide better mapping of steeply
dipping geological contacts in high-magnetic latitudes[22].
1.6.3 Magnetic gradient tensor versus Total field gradient
In addition to the above, there is extra information to be gained from the magnetic
gradient tensor (Eq. 1.4) that is not available from total field gradients (Eq. 1.7).
According to Schmidt and Clarke[3], the specific advantages of a full tensor magnetic
gradiometer include the following:
• The components of the magnetic gradient tensor are contravariant, with desir-
able mathematical properties, allowing magnetization mapping, rigorous con-
tinuation, reduction to the pole, depth slicing, invariants, etc.
• Redundant tensor components (there are 9 but only 5 are linearly independent)
give inherent error correction and noise estimates.
• Tensor components can be combined to make calculations unaffected by aliasing

across flight lines.
• Better information on magnetisation directions. The magnetic moment of com-

pact sources can be directly calculated.
• Direct determination of anomaly geometry which is irrespective of whether the

source is remanent or induced.
The ability of total field gradients to suppress global magnetic noise has been
questioned by Hogg[15]. According to Hogg, gradient surveys will replace diurnal
variations with other noises such as aircraft orientation error in crosswinds. However,
these orientation errors can be eliminated from magnetic gradient tensor surveys by
using invariant combinations of tensor components such as I1 = Bxx Byy + Byy Bzz +
2 2 2
Bzz Bxx − Bxy − Byz − Bzx [3].
1.6.4 Space applications
As listed in section 1.6.2, magnetic gradients can be used to suppress global mag-
netic noise. This can also be achieved by having a stationary magnetometer on the
ground to calibrate for changes in the global magnetic field, this is common practice
in surveying. However, this base station technique is not perfect because the mag-
netic noise at the base station may not be perfectly correlated with the magnetic
noise at the survey aircraft. This limitation is particularly true for global magnetic
surveys which are taken from satellites orbiting h = 200 km to h = 600 km above
the Earth’s surface. These satellites are too far away for the Earth’s magnetic noise
to be correlated. Magnetic gradiometers have the advantage of not requiring a base
station and are the ideal instrument for magnetic surveys in near Earth space[20].
1.7. LIMITATIONS OF MAGNETIC GRADIENT MEASUREMENTS 13
By measuring the 5 independent magnetic gradient components and the 3 vector

components, the exact location and distance of a dipole source can be calculated[20].
This is particularly useful in the environment of space where geological mapping
with 2 dimensional survey grids is impossible. Magnetic gradiometers have another
advantage in space, of being insensitive to solar magnetic noise.
1.6.5 Borehole applications
When taking measurements with a magnetic gradiometer in a borehole, magnetic

material adjacent to the borehole produce very localised spikes, easily distinguishable
from the smoothly varying signature of large off-hole sources[3]. The magnetic gradi-
ent data can also be inverted to provide key information on the depth distribution of
inductively magnetised bodies around the hole.
1.6.6 Archeological applications
Vertical magnetic gradients can be used to perform 3 dimensional inversions of arche-

ological sites. Gradient data provides higher resolution information about the shallow
subsurface than total field data. The magnetic gradient is influenced little by the cul-
tural and industrial noise that can be present at the dig site. Herwanger et al.[23]
showed that a magnetic gradiometer could be used to find information on the loca-
tions, 3-D geometries, and sizes of certain archeological features that had different
magnetic susceptibility from the surrounding sediments.
1.6.7 Other applications
Magnetic gradients can also be of use in tracking military vehicles[24], detecting

unexploded ordinances[25; 26] and locating underwater cables[27].
1.7 Limitations of magnetic gradient measurements

The previous subsections identified the additional information that magnetic gradients
could provide about targets. However, this information can only be utilised when
there is a sufficient signal to noise ratio.
At a distance of 100 m, the total magnetic intensity of a paramagnetic ore body

might be between 10 nT to 100 nT. A good magnetometer has sensitivity better than
0.01 nT so the signal to noise ratio is more than sufficient. The same ore body will
typically produce a magnetic gradient between 0.05 nT/m and 3 nT/m[7]. The best
existing gradiometers in section 1.9, have sensitivity ∼ 0.01 nT/m, which is adequate.
Smaller and/or deeper targets will be more difficult to detect and require higher
sensitivity. Nevertheless, magnetic gradiometers operating with this sensitivity can
be quite useful, and provide insight that is not available with total magnetic intensity
data. The best results are achieved by measuring the total field, magnetic vector, and
the magnetic gradient, together at the same time.
The magnetic gradient tensor can in principal be calculated from densely sampled
and accurate total field data[16]. However, in practice such data is seldom available
and using closer line spacings would increase survey costs. Low altitude flight data
could instead be upwards continued to estimate the magnetic gradient tensor at high
altitudes but this would significantly decrease the spatial resolution of the data[16].
Thus, the magnetic gradient tensor must be directly measured.
1.8 Common mode rejection
The simplest way to make a magnetic gradiometer is to use two identical magnetome-
ters, separate them by a fixed distance, and compare their output. More sophisticated
gradiometers may have two sensing elements inside a single device or have one elon-
gated sensing element but the principal is the same in all cases. Effectively, all existing
magnetic gradiometers operate by measuring the difference in the magnetic field at
two or more locations separated by a baseline.
The ability of magnetic gradiometers to subtract magnetic fields which are com-
mon to both sensing elements is called the common mode rejection ratio. A high
common mode rejection ratio is desirable in order to detect small magnetic gradients
and reject the Earth’s large uniform field. Subtraction can be either digital (software
gradiometer) or analogue (electronic gradiometer)[28]. Analogue subtraction requires
both the electronic circuits and the sensing element geometry to be almost perfectly
1.9. EXISTING MAGNETIC GRADIOMETERS 15
symmetric. Even after balancing is achieved, the gradiometer may be affected by

mechanical and thermal stresses, which vary over time and with exposure to extreme
environments. The common mode rejection ratio of some magnetic gradiometers
may decrease with time[29]. In addition, the inevitable non-linearity in each sensor’s
electronics are much more difficult to balance than linear errors[28].
Software gradiometers are easier to balance than electronic gradiometers[28]. Re-
balancing in real time is possible by combining magnetic gradient data with vector
data and gyroscopes. However, the dynamic range of the analogue to digital convert-
ers must be very high to detect small signal variations superimposed on the larger
field of the Earth.
If the common mode rejection ratio of a magnetic gradiometer is not sufficiently
high, then the device will detect the Earth’s magnetic field and be subject to the same
orientation errors suffered by vector magnetometers. On-board a moving platform
such as a survey aircraft, orientation errors may form a large part of the total noise
budget.
1.9 Existing magnetic gradiometers
A search of the literature did not find any comprehensive review articles of magnetic
gradiometers written in English. Breiner[30] has written a review of magnetometers
and gradiometers in geophysics. Lenz and Edelstein[31] have written a thorough
review of all magnetic sensors. Stolz et al.[32] have included a review of magnetic
gradiometers at the start of their paper. Clem et al.[28] have written a review of
SQuID magnetic gradiometers. Zhang[33] has written a review paper on magnetic
gradiometers in Chinese.
There are many research groups developing magnetic gradiometers based on differ-
ent technologies. The best magnetic gradiometers use SQuIDs to measure the entire
magnetic gradient tensor. Less sensitive fluxgate gradiometers are also used because
they can operate at room temperature. Total field magnetic gradiometers measure
total field gradients, eg ∂|B|/∂z, and are thus less susceptible to orientation errors.
A brief overview of these systems and technologies is presented here.
1.9.1 Fluxgate gradiometers
A temporal magnetic gradiometer for airborne surveying was developed by Wickerham

[34] in 1954. The gradiometer measured the time derivative of the magnetic field,
∂Bz /∂t, using a two-phase induction-type generator and was designed for use in
conjunction with a fluxgate magnetometer.
The first magnetic gradiometer to measure spatial gradients was reported by
Morris[35] in 1961 and used three collinear fluxgate magnetometers. The central
fluxgate measured the Earth’s magnetic field and fed its output into a feedback loop
to null the ambient field of the entire device. This technique alleviates many of the dif-
ficulties associated with common mode rejection discussed in section 1.8. The signal
from the outer pair of fluxgates was then subtracted to record the magnetic gradient.
Using a ∼ 4 m baseline between the sensors, Morris’ magnetic gradiometer had a
sensitivity of 5 nT/m[35]. More recently, Koch et al.[36] built a fluxgate magnetic
√
gradiometer using similar techniques and achieved sensitivity of 0.14 nT/m/ Hz rms
at 1 Hz.
Fluxgate magnetometers usually contain one or two parallel cores of highly per-
meable material. Alternating current is applied to a primary coil wound around the
cores, causing the cores’ magnetisation to saturate every half cycle. The permeability
fluctuates at twice the applied frequency. Therefore, the flux generated by external
magnetic fields will be modulated at the second harmonic frequency and such flux
is detected by a secondary coil around the cores. A balanced core arrangement is
commonly used to prevent the excitation signal from appearing in the secondary coil.
Fluxgate sensors are vector magnetometers and can measure any of the components
Bx , By and Bz . Modern fluxgate magnetometers have sensitivities on the order of
√
0.005 nT/m/ Hz root means square (rms) at 1 Hz when operating in the field[37].
Fluxgates gradiometers are absolute gradiometers and long term accuracy can
be better than 0.5 nT/m[41]. However, as an absolute device, fluxgate’s require a
large dynamic range to accurately subtract the Earth’s magnetic field. Cerman et
al.[38] discuss some of the issues and techniques for digital subtraction in fluxgate
gradiometers.
Single-core fluxgate gradiometric sensors have been developed that use gradiomet-
ric coil windings[39]. However, measuring the magnetic gradient with two separate
sensors and subtracting their reading gives better stability and lower noise[37].
Fluxgate gradiometers that compare the vector measurements from pairs of flux-
gates are availably commercially. One example, the Bartington Grad601 magnetic
gradiometer, is used in archaeological prospecting and surface magnetic surveys[40].
Using a 1 m baseline between the sensors, the sensitivity is 0.1 nT/m. Misalignment
issues are minimised by using a rigid linear sensor configuration that can only mea-
sure the diagonal components of the magnetic tensor Bxx , Byy or Bzz . Merayo et
al.[41] have developed a triaxial fluxgate magnetic gradiometer that can measure the
√
entire magnetic gradient tensor with sensitivity 0.03 nT/m/ Hz rms at 1 Hz. The
triaxial fluxgate[41] and Grad 601 are software gradiometers, and both are limited by
digitisation noise at high frequencies.
1.9.2 SQuID gradiometers
Superconducting quantum interferance devices (SQuIDs) are the most sensitive de-
tectors of magnetic flux that we know of[20]. Superconductivity is the phenomenon
whereby certain materials have exactly zero electrical resistance when they are cooled
below a critical temperature. Loops of superconducting wire respond almost instantly
to changes in the magnetic field. When a current is applied to a superconducting loop
which contains a thin insulating barrier (known as a Josephson junction), a voltage
develops across the barrier which is a function of the magnetic flux through the loop.
However, the zero value of the magnetic field (and hence zero magnetic flux) does
not manifest as special point in the voltage-flux curve so it cannot be distinguished.
Unless special techniques are used, SQUIDs only measure changes in the magnetic
field instead of the absolute magnetic field[42]. More information on SQuIDs can be
found in a book by Clarke and Braginski[43].
The intrinsic noise of SQuID magnetometers in the laboratory is very low, typically
√
less than 0.0001 nT/ Hz rms[26; 44; 32; 45; 46]. The external noise from terrestrial
magnetic and electromagnetic fields is considerably higher so the best performance
can only be achieved behind mumetal shields[46]. Noise can also be suppressed in
the laboratory by connecting SQuIDS in gradiometric fashion. However, on-board a
survey craft, SQuID gradiometers are exposed to high levels of vibration and noise
levels tend to be substantially higher. In addition, cryostats are required to keep
SQuIDs below their critical superconducting temperatures. Many developments of
SQuID gradiometers suffer from engineering problems that are yet to be solved[32].
Foley et al.[47] have developed a full tensor magnetic gradiometer which uses soft-
ware to subtract the output from pairs of high temperature DC SQuIDs. Additional
common mode rejection is achieved by rotating the gradiometer[48]. The sensitivity is
√ √
∼ 0.01 nT/m/ Hz rms at 1 Hz in the laboratory and ∼ 0.1 nT/m/ Hz rms during
deployment in the field[47].
Clem et al.[44] measure the full magnetic gradient tensor using three pairs of
orthogonal high temperature SQuIDs. A fluxgate magnetometer was placed between
each pair of SQuIDs and was used to null the Earth’s field using a feedback loop.
√
The sensitivity is 0.0008 nT/m/ Hz rms at 1 Hz whilst stationary in the laboratory.
√
Low frequency rotations increased the noise floor to 0.005 nT/m/ Hz rms.
The only SQuID gradiometer available commercially is JESSY star[49]. Magnetic
gradients are measured by 2 to 6 pairs of pickup coils, each pair connected differentially
to a single low temperature SQuID[32]. Intrinsic common mode rejection of 104 is
further enhanced by combining the magnetic gradient outputs with magnetometer
√
outputs during signal processing. The sensitivity is 0.00008 nT/m/ Hz rms at 1 Hz
√
in the laboratory and ∼ 0.01 nT/m/ Hz rms during an airborne trial[32].
1.9.3 Total field gradiometers
Total field gradiometers are constructed from pairs of total field magnetometers, which
may be either proton precession magnetometers, optically pumped magnetometers or
atomic field magnetometers. Generally, these devices utilise oscillators whose oscil-
lation frequency is proportional to the magnitude of the ambient magnetic field[30].
Various survey companies create a gradiometer by placing these total field magne-
tometers on both wing tips of an aeroplane[50]. Gradiometers can also be formed
from multiple total field magnetometers suspended below a helicopter. Hollyer and
Hrvoic[51] have developed a tri-directional helicopter gradiometer that can measure
magnetic gradients with precision better 0.001 nT/m using a base line of ∼ 3 m.
Individual total field magnetometers can have absolute accuracy better than 0.1 nT
so a pair can be accurately balanced with very high common mode rejection[52]. The
major limitation of total field gradiometers is that they measure total field gradients:

∂|B| ∂|B| ∂|B|
(1.7)
∂x ∂y ∂z
The above 3 total field gradients are not contravarient and contain less information
than the 9 components of the magnetic gradient tensor shown in Eq. 1.4. Section
1.6.3 lists the advantages of tensor gradiometers over total field gradiometers.
A scalar magnetometer can be converted into a vector magnetometer in a particu-

lar direction by applying a sufficiently large bias magnetic field in the same direction.
This technique was used by Affolderbach et al.[53] in an optically pumped caesium
vapour magnetometer. A uniform Bz = 500 nT magnetic field was applied to a single
caesium vapour cell, thus making the vapour preferentially sensitive to vector mag-
netic fields in the z direction. Two laser beams and photodiodes were then used to
measure the frequency of photon absorbtion/re-emission by the caesium vapour in
each half of the cell. The output of the photodiodes was demodulated and subtracted
to record the magnetic gradient. Using a baseline of 15 mm, the sensitivity of the
gradiometer is approximately ∼ 1 nT/m[53].
1.9.4 Other magnetic gradiometers
Magnetoresistive sensors are small low cost devices that measure the magnetic field
by detecting changes in the electrical resistance of thin films. The best sensors have
√
sensitivity on the order of 0.05 nT/ Hz rms at 1 Hz. Perry et al.[24] have developed
a magnetic gradiometer for military vehicle tracking which uses pairs of magnetore-
sistive sensors separated by a 0.145 m baseline.
Magnetic gradiometers have also been made using induction coils[54] and optical
fibres[55].
1.10 History of the Direct String Magnetic

Gradiometer
This thesis concerns the characterisation and improvement of a Direct String Magnetic
Gradiometer (DSMG). I present here a brief overview of this magnetic gradiometer
and the work done by Dr. Alexey Veryaskin and Dr. Wayne McRae before I started
my PhD in 2005.
1.10.1 The beginning 1997-2001
The Direct String Magnetic Gradiometer was first proposed by Dr. Alexey Veryaskin
in 1997[57]. It was first conceived as part of a combined magnetic and gravity gra-
diometer. The gradiometer took advantage of some specific dynamic properties of
a metallic current carrying string fixed at both ends. The string would vibrate at
750 Hz in response to magnetic gradients and vibrate at 10 Hz in response to grav-
ity gradients. The high frequency modulation was provided by alternating current
and the low frequency modulation was provided by periodically switching the string
from a state of high stiffness to a state of low stiffness. The entire gradiometer was
cooled down to 77 K using liquid nitrogen in order to reduce the thermal noise[58]. A
diagram of the combined magnetic and gravity system with applications in airborne
geophysics surveying is shown in Fig. 1.5.
Heavy strings are more sensitive to gravity gradients whereas light strings will
deflect further in the presence of magnetic gradients. As a result the gradiometer
was divided into two separate systems. The gravity gradiometer now uses a heavy
phosphorus-bronze string and has applications in bore holes.
The direct string magnetic gradiometer (DSMG) now uses a low density aluminium
6061 string and has primary applications in airborne surveying. On-board a survey
aircraft, the vibration noise is considerable so a mechanical isolator is required. Using
cryogenics in an aircaft is expensive and a cryogenic gradiometer would be competing
with SQuIDs. There are also occupation safety and health restrictions on where
cryogenic systems can be operated. Some countries do not allow cryogenic equipment
on board aircraft. For these reasons, the DSMG now operates at room temperature.
1.10. HISTORY OF THE DIRECT STRING MAGNETIC GRADIOMETER 21
Figure 1.5: The gravity and magnetic gradiometer was comprised of 4 parallel channels:
two for the measurements of gravity gradients Txz , Tyz and two for the measurement of
magnetic gradients Bxz , Byz . The entire system was placed on a stable table. The pitch,
roll and yaw rates (Ωx , Ωx , Ωx ) were measured to compensate for vibration noise. This
figure was reproduced from a paper by Veryaskin[58].
1.10.2 Principal of Operation
The basic operational principle of the DSMG is to measure the deflection of a current
carrying string in the presence of magnetic fields. In order to differentiate between the
uniform field of the Earth and the more useful gradient field, produced by minerals
local to the sensor, we ‘select’ for deflections caused by a varying magnetic field along
the length of the sensing element. This is achieved by driving the string using an
AC current at the second harmonic frequency of the string, nominally 750 Hz. The
force from a uniform magnetic field would deflect the string in its fundamental mode
(minima at the ends where the string is clamped and maximum at the centre). In
contrast, the force from the gradient field would deflect the string at its second order
mode, as shown in Fig 1.6.
By tuning the AC current to the second violin mode frequency, we get resonance.
Pickup Coils
Sensing element
Figure 1.6: Concept diagram of the DSMG. An AC current is tuned to the 2nd violin mode
of the string, causing it to vibrate in the presence of magnetic gradients. The vibration is
detected with a pair of pickup coils connected in differential mode to make a radio frequency
bridge. This figure was reproduced from a presentation by McRae et al.[59].
On resonance the string’s vibration amplitude will slowly build up over time in re-
sponse to magnetic gradients until the friction force equals the driving force. The off
resonance response to uniform magnetic fields will be weak. Thus the string provides
the first stage of common mode rejection.
The displacement of the string can be measured with two inductive pickup coils,
connected in differential mode, placed close to the string, one quarter and three quar-
ters along the length of the string. The differential readout provides the second stage
of common mode rejection. A pair of feedback coils are used to keep the magnetic
gradient constant, compensating any changes due to external gradient sources. The
feedback loop thus extends the dynamic range of the DSMG and ensures the string
signal is sufficiently strong for signal processing[59]. The string, pickup coils and
feedback are contained within in a sensor frame. The sensor has dimensions 300 mm
by 30 mm by 30 mm and weighs 300 g. The DSMG contains one or two sensors.
As shown in Fig. 1.6, the signal sent into each sensor is a combination of a high
frequency carrier signal and a low frequency drive signal. The pickup coils form a
Figure 1.7: The sensing element of the DSMG is a 250 mm long round string with diameter
125 μm. The string is enclosed in a 300 mm long sensor frame. The DSMG contains two
sensors, measuring Byz and Bxz respectively. Both sensors are housed in a 4 stage mechanical
isolator which is inside a vacuum tank. The electronic input and output of each sensor is
connected to a low noise pre-amplifier located about 500 mm away from the vacuum tank.
A cable connects the pre-amplifier unit to a digital signal processing unit (DSP) placed 5 m
from the sensors. The DSP demodulates the signal and transfers data to a nearby laptop[59].
differential RLC circuit, known as a radio frequency bridge. The carrier signal is
tuned to the resonance of the RLC circuit whereas the drive frequency is tuned to the
mechanical resonance of the string. The mutual inductance between the string and
the pickup coils changes as the string vibrates. Thus the carrier signal measured by
the pickup coils will contain sidebands which are a function of the string’s amplitude
(which is proportional to the magnetic gradient). The modulated carrier can be seen
in the output signal of Fig. 1.6. This signal is synchronously detected and digitized
with 16 bit analogue to digital converters. Over sampling increases the effective
resolution to 20 bits[59].
An overview of the DSMG system is presented in Fig 1.7. The diagram includes
a photograph of the original DSMG sensor frame. The thermal expansion coefficient
of the string and frame should match so they were both made of aluminium. Each
sensors in the DSMG can be positioned vertically to measure Byz or Bxz . Measuring
Bxy is more difficult because this requires suspending the DSMG horizontally, which
would degrade the performance of the mechanical isolator.
1.10.3 Vibration noise
To shield them from seismic noise, the two sensors are housed in a 4 stage mechanical
isolator which damps vibration by at least 80 dB in the range 400 Hz to 900 Hz. The
string frequency is nominally 750 Hz and is always within this range. The isolator
and sensors are placed inside a vacuum tank and evacuated to eliminate the effect of
external acoustic noise. A picture of the mechanical isolator undergoing performance
testing in 2003 is shown in Fig 1.8.
Figure 1.8: The mechanical isolator is shown in the foreground of the photograph inside a
clear acrylic vacuum tank. A high power shaker, shown in the background of the photograph,
is used to vibrate the entire vacuum tank at 0.5 ms−2 . The graph to the right shows
the calculated (black line) and measured (red line) isolation characteristics of the 4-stage
isolation system at frequencies from 5 Hz to 900 Hz. The measured transfer function at high
frequency (above 400 Hz as indicated by the arrow) is limited by the instrument noise floor.
This figure was reproduced from a paper by McRae et al.[59].
Laboratory testing showed that the DSMG was completely immune to accelera-
tions up to 0.25 ms−2 at any frequency. Acceleration in the range 0.25 ms−2 to 1 ms−2
at the string’s resonant frequency did generate excess noise but acceleration at other
frequencies did nothing. Accelerations higher than 1 ms−2 caused nonlinear vibra-
tions in the isolator and generated a substanital response in the DSMG output at
many different frequencies[56]. However, accelerations of this magnitude are unlikely
to occur on an aircraft during level flight[59].
1.10.4 Airborne trials 2003-2004
In 2003 and 2004, the DSMG was tested in airborne trials. A brief snapshot of the trial
results is presented in Fig 1.9. Fig 1.9A shows the output of the magnetic gradiometer
when it is flown over the same geological target three times. The repeatability of the
results proves that the DSMG can detect magnetic geological features from a survey
aircraft. The deviations in the MG output from leg to leg are related to particular
aircraft motions and can be treated as a ‘heading variation’[56].
In Fig 1.9B, the survey aircraft is undergoing a calibration square. The calibration
square involves a series of roll, pitch and yaw manoeuvers in each of the cardinal
compass directions. It is commonly used to quantify and then compensate heading
error from the aircraft. Heading error can be defined as a systematic error that
depends on the direction that the aircraft is pointing relative to the direction of the
Earth’s magnetic field. The peak to peak heading error of a magnetically clean aircraft
is typically on the order of 10 nT[60; 61]. The peak to peak gradient would be on the
order of 10 nT/m. However, the heading error shown in Fig 1.9B is much greater,
approximately 1000 nT/m peak to peak.
The excess heading error was attributed to magnetic material in the sensor frame
and mechanical isolator[59]. The magnetic field of the Earth induces material in the
DSMG to magnetise in the same orientation as the Earth’s field. If the DSMG is
placed onboard a moving vehicle then those parts with induced magnetisation will
produce a magnetic gradient which varies whenever the vehicle changes direction.
Fig 1.9C presents a recording taken during a level flight at an altitude of 1 km. At
such a distance from the ground, the magnetic gradient should be negligible and the
DSMG should only detect noise. The plot shows a long term variation from 30 nT/m
to 40 nT/m which is caused by heading error. When the measured BX component
of the magnetic field is overlaid on the Bxz magnetic gradiometer output, it is easy
to see the correlation between the two signals[59]. These results show that heading
error is causing significant amounts of excess noise in the DSMG during flight.
Figure 1.9: These three plots show a brief extract of the data taken during airborne trials
of the DSMG: (a) DSMG sensor trials over a geological target, (b) Heading error during a
calibration square manoeuver, and (c) Magnetic gradient recording during a level flight at
high altitude. This figure was reproduced from a report by McRae and Veryaskin[56].
1.11. TARGET SENSITIVITY 27
1.11 Target sensitivity
The collection of vertical total field gradients, i.e. ∂B/∂z, is now accepted as industry
standard for airborne magnetic surveys. The sensitivity is limited by magnetic noise
from the aircraft[62]. The peak to peak heading error of a survey aircraft is typically
on the order of 10 nT[60; 61], but this can be reduced to less than 0.5 nT using
software compensation[61]. Under normal turbulence conditions, magnetic parts in
the aircraft will create gradient noise of approximately 0.02 nT/m at a 0.5 s sampling
√
interval[18; 61], which is equivalent to 0.014 nT/m/ Hz.
Stationary gradiometers Mobile gradiometers
Figure 1.10: The left graph shows the noise levels of 5 different magnetic gradiometers
in the laboratory whereas the right graph shows the much higher noise levels of the same
gradiometers on-board a mobile platform. The gradiometers and their moving platforms
are: (a) Direct String Magnetic Gradiometer attached to the stinger of a survey aircraft[56],
(b) Electronic fluxgate gradiometer in a rotational shaker with amplitude ±3◦ and frequency
0.2 Hz[36], (c) Digital fluxgate gradiometer, stationary data only[41], (d) High temperature
SQuID gradiometer in the field and subject to high winds[47], and (e) Low temperature
SQuID gradiometer towed below a helicopter[32]. The dashed line is the target sensitivity
of the DSMG.
Fig. 1.10 presents the noise level of 5 different magnetic gradiometers, including
the DSMG. The pair of graphs show the large difference between the sensitivity of
these gradiometers in the laboratory and the sensitivity in the field. The excess noise
is normally caused by higher levels of vibration on-board moving platforms, magnetic
noise of the aircraft, and/or inadequate common mode rejection. In the DSMG, the
excess noise during flight is caused by heading error. Methods for reducing heading
error are discussed in Chapter 4.
The target sensitivity of the DSMG for airborne geophysical applications is
√
0.01 nT/m/ Hz, shown as a dashed line in Fig. 1.10. Achieving this target would
bring the noise floor of the DSMG below the magnetic noise of the typical survey
aircraft. The sensitivity would also be comparable to that of the SQuID magnetic
gradiometers plotted in the right hand graph of Fig. 1.10, in the frequency range of
interest for geophysical exploration 0.01-0.625 Hz, without the need for cryogenics.
Schmidt and Clark[3] have shown that a magnetic gradiometer with this sensitivity
would be of great use in geophysical exploration.
1.12 Noise
The sensitivity of the DSMG is limited by various noises which are indistinguishable
from magnetic gradients. These noises include electronic noise, vibration noise, ther-
mal noise, and magnetic noise from instrument components. Electronic noise comes
from the electronics used to measure the string’s displacement whereas the other three
noises make the string vibrate.
Fig. 1.11 compares the experimental displacement noise in the string with the the-
oretical noise contributions from thermal noise and electronic noise. Electronic noise
√
has the characteristics of white noise, with a noise level of (1.0 ± 0.1) × 10−11 m/ Hz
in both Fig. 1.11A and Fig. 1.11B. Thermal noise is concentrated at the string’s me-
chanical resonance. The sharpness of the thermal noise peak depends on the string’s
mechanical Q factor, note the difference between Fig. 1.11A and Fig. 1.11B.
The displacement levels shown in Fig. 1.11 were calibrated by measuring a known
magnetic gradient with the DSMG and converting to displacement using Eq. 5.7.
1.12. NOISE 29
A Electronic noise
Thermal noise
B
Total noise
Experiment
Displacement (m/√ Hz)
Displacement (m/√ Hz)

−11 −11
10 10
−12 −12
10 10
545 550 555 560 565 375 380 385 390 395
Frequency (Hz) Frequency (Hz)
Figure 1.11: Spectral distribution of DSMG noise at frequencies near the second violin
mode resonance. Electronic noise is flat whereas thermal noise depends on the string’s
mechanical Q factor. (a) Air pressure is 1 atmosphere and the string’s mechanical Q factor
is 100 ± 15. (b) Air pressure is 90 ± 9 Pa and the string’s mechanical Q factor is 300 ± 35.
The string is carrying current so it will heat up in high vacuum, thus lowering the resonant
frequency as shown.
The data in Fig. 1.11B was taken with the vacuum pump turned on, although
turning the pump off did not visibly change the DSMG’s noise level. The DSMG was
enclosed inside a vibration isolator which damps all mechanical vibration, including
vibration produced by the vacuum pump. I conclude that the contribution of vibration
noise to the total noise budget is negligible in these laboratory measurements.
1.13 Thesis outline

√
Fig. 1.10 shows that the target sensitivity, 0.01 nT/m/ Hz, is more than an order
√
of magnitude below the current noise floor of the DSMG, 0.18 nT/m/ Hz. In order
to reach this target, this thesis is a study of all noise sources affecting the DSMG.
Vibration noise has already been thoroughly investigated by McRae et al.[59; 56] and
a brief summary of their work is presented in section 1.10.3 of this introduction.
Chapter 2 introduces the magnetic gradiometer. I show the ability of the DSMG
to measure magnetic gradients of real targets in the laboratory and in the field. The
utility of the gradiometer for airborne surveying is evaluated.
The third chapter of this thesis addresses electronic noise in the inductive pickup
coils which are used to detect the position of the string. I also present a new capacitive
displacement readout which has higher sensitivity.
The fourth chapter addresses heading error in the DSMG from magnetic compo-
nents. I compare possible low susceptible materials and present strategies for min-
imisation of magnetic contamination.
The fifth and sixth chapters of this thesis consider the thermal noise floor of the
string (the sensing element of the DSMG). Chapter 5 shows how thermal noise can
be reduced by manipulating the string’s dimensions, air pressure, temperature, etc.
I then determine the optimal parameters for enhancing performance. In Chapter 6,
I consider the deployment of the DSMG in space. The DSMG is particularly suited
to the environment of space and significant increases in sensitivity are predicted.
Chapter 6 also considers the noise level at low frequencies and low temperatures.
The conclusion of the thesis contains a summary of the work contained therein.
I then present suggestions for future work and discuss some of the issues that will
be faced implementing these suggestions. Amongst other things, this thesis describes
what could be achieved with the direct string magnetic gradiometer.
BIBLIOGRAPHY 31
Bibliography
[1] L. Gilchrist, The lure of underground treasures, The astronomical society of
Canada 43, 97-105 (1949).
[2] R. Van Blaricom, Practical Geophysics for the Exploration Geologist, 1st edition,
Northwest Mining Association, Spokane, 1980.
[3] P. Schmidt, D.A. Clark, The magnetic gradient tensor: its properties and source
characterization, The Leading Edge 25, 75-78 (2006).
[4] R.C. Snare, A history of vector magnetometry in space, Geophysical Monograph

103, 101-114 (1998).
[5] T. McConnell, B. Dragoset, Introduction to this special section: magnetic gra-

diometry, The Leading Edge 25, 45 (2006).
[6] M.N. Nabighian, V.J.S. Grauch, R.O. Hansen, T.R. LaFehr, Y. Li, J. W. Peirce,
J. D. Phillips, M. E. Ruder, The historical development of the magnetic method
in exploration, Geophysics 70, 33ND-61ND (2005).
[7] P. Schmidt, D.A. Clark, Advantages of measuring the magnetic gradient tensor,
Preview 85, 26-30 (2000).
[8] G.A. Glatzmaier, What causes the periodic reversals of the earth’s magnetic
field? Have there been any successful attempts to model the phenomenon?,
Scientifc American, October 21, (1999).
[9] M. Mandea, M. Purucker, Observing, Modeling, and Interpreting Magnetic

Fields of the Solid Earth, Surveys in Geophysics 26, 415-459 (2005).
[10] S. Breiner, Practical Geophysics II for the Exploration Geologist, 2nd edition,
Edited by R. Van Blaricom, Northwest Mining Association, Spokane, 1992, Ch.
5.
[11] R.T. Merrill, M.W. McElhinny, P.K. McFadden, The magnetic field of the earth,
2nd edition reprint, Academic Press, 1998, p 20.
[12] D.A. Clark, Magnetic petrophysics and magnetic petrology: aids to geological
interpretation of magnetic surveys, AGSO Journal of Australian Geology & Geo-
physics 17(2), 83-103 (1997).
[13] P.J. Gunn, M.C Dentith, Magnetic responses associated with mineral deposits,
AGSO Journal of Australian Geology & Geophysics 17(2), 145-158 (1997).
[14] V.C.F. Barbosa, J.B.C. Silvaz, W.E. Medeiros, Stability analysis and improve-
ment of structural index estimation in euler deconvolution, Geophysics 64, 48-60
(1999).
[15] R.L.S. Hogg, Practicalities, pitfalls and new developments in airborne magnetic
gradiometry, First Break 22, No. 7 (2004).
[16] L.B. Pedersen, T.M. Rasmussen, The gradient tensor of potential field anomalies:
some implications on data collection and data processing of maps, Geophysics
55, 1558-1566 (1990).
[17] H. Golden, Advantages of tensor magnetic gradiometry, Internal report at WMC

Exploration, 2001.
[18] D.R. Cowan, M. Baigent, S. Cowan, Aeromagnetic gradiometers - a perspective,

Exploration Geophysics 26, 241-246 (1995).
[19] H. Golden, A Novel Magnetic Gradiometer: Description, Design Issues and Trial
Results, ASEG 2007 presentation.
http://gravitec.co.nz/pdfs
[20] R. Hastings, R. P. S. Mahler, R. S. Schneider Jr., J. H. Eraker, Cryogenic mag-

netic gradiometers for space applications, IEEE Transactions on Geoscience and
Remote Sensing GE-23, 552-561 (1985).
[21] K.W. Behannon, M.H. Acuna, L.F. Burlaga, R.P. Lepping, N.F. Ness, F.M.
Neubauer, Magnetic field experiment for voyagers 1 and 2, Space Science Reviews
21, 235-257 (1977).
[22] P. Hood, Gradient measurements in aeromagnetic surveying, Geophysics 30,

891-902 (1965).
BIBLIOGRAPHY 33
[23] J. Herwanger, H. Maurer, A.G. Green, and J. Leckebusch, 3-D inversions of mag-
netic gradiometer data in archeological prospecting: Possibilities and limitations,
Geophysics 65, 849-860 (2000).
[24] A.R. Perry, P.V. Czipott, Y. Dalichaouch, S. Kumar, H. Trammell, Standoff De-
tection and Tracking of Vehicles Using a Compact Magnetic Sensor, Presentation
at Mines, Demolition and Non-Lethal Conference and Exhibition, 2002.
www.dtic.mil/ndia/2002mines/kumar.pdf
[25] A. Salem, T, Hamada, J.K. Asahina, K. Ushijima, Detection of unexploded ord-

nance (UXO) using marine magnetic gradiometer data, Exploration Geophysics
36, 97-103 (2005).
[26] T.J. Gamey, Development and Evaluation of an Airborne Superconducting Quan-

tum Interference Device-Based Magnetic Gradiometer Tensor System for Detec-
tion, Characterization and Mapping of Unexploded Ordnance, United States
Army Technical Report ADA495604.
http://handle.dtic.mil/100.2/ADA495604
[27] N. Lawrence, Cable tracking in shallow water, Hydro International 11, No. 6
(2007).
[28] T.R. Clem, C.P. Foley, M.N. Keene, The SQUID Handbook Vol. II: Applications
of SQUIDs and SQUID Systems, J. Clarke & A.I. Braginski (Eds.), Wiley-VCH,
Weinheim, 2006, Ch. 14.
[29] M. Bick, K.E. Leslie, R. Binks, D.L. Tilbrook, S.K.H. Lam, S. Gnanarajan, J.
Du, C. P. Foley, Highly balanced long-baseline axial gradiometer based on high-
Tc superconducting tape, IEEE Transactions on Applied Superconductivity 15,
765-768 (2005).
[30] S. Breiner, Magnetometers for geophysical applications, SQuID Applications to

Geophysics, SEG workshop proceedings, 1980.
[31] J. Lenz, A.S. Edelstein, Magnetic Sensors and Their Applications, IEEE Sensors
Journal 6, 631-649 (2006).
[32] R. Stolz, V. Zakosarenko, M. Schulz, A. Chwala, L. Fritzsch, H.G. Meyer, E.O.

Kostlin, Magnetic full-tensor SQUID gradiometer system for geophysical appli-
cations, The Leading Edge 25, 178-180 (2006).
[33] C. Zhang, Airborne tensor magnetic gradiometry - the latest progress of airborne
magnetometric technology, Chinese Journal of Engineering Geophysics 3, 354-
361 (2006).
[34] W.E. Wickerham, The gulf airbourne magnetic gradiometer, Geophysics 19, 116-
123 (1954).
[35] R.M. Morris, B.O. Pedersen, Design of a second harmonic fluxgate magnetic field
gradiometer, Review of Scientific Instruments 32, 444-448 (1961).
[36] R.H. Koch, G.A. Keefe, G. Allen, Room temperature three sensor magnetic field
gradiometer, Review of Scientific Instruments 67, 230-235 (1996).
[37] P. Ripka, Advances in fluxgate sensors, Sensors and Actuators A 106, 8-14
(2003).
[38] A. Cermana, A. Kuna, P. Ripka, J.M.G. Merayo, Sensors and Actuators A 121,
421-429 (2005).
[39] P. Ripka, P. Navratil, Fluxgate sensor for magnetopneumometry, Sensors and

Aclualor A 60, 76-79 (1997).
[40] G. Bartington, C.E. Chapman, A high-stability fluxgate magnetic gradiometer

for shallow geophysical survey applications, Archaeological Prospection 11, 19-
34, (2004).
[41] J.M.G. Merayo, P. Brauer, F. Primdahl, Triaxial fluxgate gradiometer of high

stability and linearity, Sensors and Actuators A 120, 71-77 (2005).
[42] P. Carelli, M.G. Castellano, K. Flacoo, R. Leoni, G. Torrioli, An absolute mag-

netometer based on dc Superconducting QUantum Interference Devices, Euro-
physics Letters 39, 569-574 (1997).
BIBLIOGRAPHY 35
[43] The SQUID Handbook Vol. I: Fundamentals and Technology of SQUIDs and
SQUID Systems, J. Clarke & A.I. Braginski (Eds.), Wiley-VCH, Weinheim, 2006.
[44] T.R. Clem, D.J. Overway, J.W. Purpura, J.T. Bono, R.H. Koch, J.R. Rozen,
G.A. Keefe, S.Willen, R.A. Molding, High-T, SQUID Gradiometer for Mobile
Magnetic Anomaly Detection, IEEE Transacitons on applied superconductivity
11, 871-875 (2001)
[45] C.P. Foley, K.E. Leslie, R.A. Binks, S.H.K. Lam, J. Du, D.L Tilbrook, E.E.
Mitchell, J.C. Macfarlane, J.B. Lee, R. Turner, M. Downey, A. Maddever, Issues
relating to airborne applications of HTS SQUIDs, Superconductor Science and
Technology 15, 1641-1645 (2002).
[46] R.L. Fagaly, Superconducting quantum interference device instruments and ap-
plications, Review of Scientific Instruments 77, 101101 (2006).
[47] C.P. Foley, D.L Tilbrook, K.E. Leslie, R.A. Binks, G.B. Donaldson, J. Du, S.K.
Lam, P.W. Schmidt, D.A. Clark, Geophysical exploration using magnetic gra-
diometry based on HTS SQUIDs, IEEE Transactions on Applied Superconduc-
tivity 11, 1375-1378 (2001).
[48] K. Leslie, K. Blay, D. Clark, P. Schmidt, D. Tilbrook, M. Bick, C. Foley, R. Binks,

Helicopter trial of magnetic tensor gradiometer, ASEG Extended Abstracts 2007
(2007).
[49] H. Meyer, Oskar der internationalen bergbauindustrie für IPHT jena, Optik &
Photonik, Mar. 2008, No. 1, p 9 (2008).
[50] P. Killeen, Exploration trends & development in 2008, Supplement to The North-
ern Miner 95, No. 1, 6-15 (2009).
[51] G. Hollyer, I. Hrvoic, Development of a tri-directional helicopter gradiometer for

mineral exploration applications, First break 23, July, (2005).
[52] E.B. Alexandrov, Recent progress in optically pumped magnetometers, Physica

Scripta T105, 27-30 (2003).
[53] C. Affolderbach, M. Stähler, S. Knappe, R. Wynands, An all-optical, high sensi-

tivity magnetic gradiometer, Applied Physics B 75, 605-612 (2002).
[54] R.J. Prance, T. D. Clark, H. Prance, Compact broadband gradiometric induction

magnetometer system, Review of Scientific Instruments 76, 117-121 (1999).
[55] R.D. Rempt, Novel Nulling/Balancing Technique for Fiber Optic Magnetic Gra-
diometer, IEEE Photonics Technology Letters 1, 105-106 (1989).
[56] W. McRae, A. Veryaskin, Airborne magnetic gradiometer review, Internal report

at Gravitec Instruments, 2004.
[57] A.V. Veryaskin, Patent EP20000929523, Measurement of magnetic fields using

a string fixed at both ends, 2000.
[58] A. Veryaskin, A novel combined gravity & magnetic gradiometer system for
mobile applications, SEG Expanded Abstracts 19, 420-423 (2000).
[59] W. McRae, A.V. Veryaskin, L. Ju, D.G. Blair, E. Chin, J. Dumas, B. Lee, String
magnetic gradiometer system: recent airborne trials, SEG Expanded Abstracts
23, 790-793 (2004).
[60] C.V. Reeves, The role of airborne geophysical reconnaissance in exploration geo-
science, First break 19, 501-508 (2001).
[61] P.J. Hood, Aeromagnetic gradiometer program of the Geological Survey of

canada, Geophysics 54, 1012-1022 (1989).
[62] A. Wooldridge, Review of modern magnetic gradiometer surveys, SEG Expanded

Abstracts 23, 802-805 (2004).
Chapter 2
Characteristics of the DSMG
2.1 Preface
The paper which forms the bulk of this chapter, introduces the Direct String Magnetic
gradiometer (DSMG) and shows the string is able to detect magnetic gradients. The
paper then presents magnetic gradient measurements in the lab and the results of
a ground survey in the field. These trial measurements are used to characterise the
DSMG and evaluate its effectiveness for airborne surveying. A significant part of the
ground survey presented here, was performed by Mr. Howard Golden and Dr. Wayne
McRae, with the author helping. The other experiments in this Chapter were carried
out by the author.
Most of the measurements were conducted in 2006 and the results shown here
demonstrate the sensitivity of the DSMG at that time. The present DSMG uses
a ribbon with the same dimensions but achieves better sensitivity by using higher
current and a stronger string. The new DSMG uses a string made of Al 6061-T6511
instead of the Al 6061-TF string reported here.
The paper was published in Exploration Geophysics. My contribution to this pa-
per amounts to 80% of the experimental work and 80% of the manuscript preparation.
37
38 Chapter 2. CHARACTERISTICS OF THE DSMG
Exploration Geophysics 40 (2009) 222-226
Results from a novel direct magnetic gradiometer

Andrew Sunderland1 , Howard Golden1,2 , Wayne McRae1,2 , Alexey V Veryaskin1,2 , Li Ju1 ,
David G Blair1
1 School of Physics, University of Western Australia, 35 Stirling Highway, Crawley, WA 6009

2 Gravitec Instruments, School of Physics, University of Western Australia
Abstract
Development is continuing on a novel direct magnetic gradiometer that
uses a string as its single sensing element. A string driven by an AC
current and vibrating in its second order ‘S’ shaped mode is used to di-
rectly measure the magnetic gradient. The sensitivity is measured to be
√
0.4 nT/m/ Hz in the laboratory. Long-term drift is very low, and static
measurements are made in the field with precision 0.2 nT/m. Laboratory
measurements of moving iron spheres of various volumes demonstrate the
accuracy of the sensor. By combining the sphere measurements with for-
ward modelling of a hypothetical mineral deposit, we illustrate the util-
ity of the gradiometer for airborne surveying. A 20 m line survey on
the ground demonstrates the performance of the gradiometer in the field.
These experimental and numerical results indicate that such a system is
potentially viable for airborne and ground deployment and there is no
reason the sensor could not be used in other applications.
2.2 Introduction
Project AMATI, so named in honour of the famous violin maker Nicolo Amati, is
developing a direct string magnetic gradiometer (DSMG) capable of measuring off-
diagonal components of the magnetic gradient tensor. As reported by McConnell and
Dragoset (2006), vector and spatial gradients of the magnetic field have been used for
many years to interpret information about the magnetic mineral content of surveyed
terrain. From a review of recent magnetic gradient surveys by Wooldridge (2004),
the useful applications of gradient data include gradient enhanced gridding, diurnal
2.2. INTRODUCTION 39
free levelling, anomalous magnetic field products, strike estimation of 2D bodies, and
more accurate automated depth solutions. Schmidt and Clark (2000) described the
key advantages of conducting magnetic surveys that measure the magnetic gradient
tensor.
Veryaskin (2001) introduces the concept of a DSMG. The sensing element consists
of a single aluminium 6061 alloy string or wire, i.e. an object with transverse dimen-
sions much smaller than its longitudinal dimension. The string is held under tension
with its second violin mode or ‘S’ shaped mode at f0 = 850 Hz. This ‘S’ mode is
only sensitive to magnetic gradients, whereas the fundamental mode or ‘C’ shaped
mode couples with the uniform magnetic field. Alternating current with frequency f
is pumped along the string, setting it in motion with an Ampere force per unit length:
dF(z, t)
= [ez × B(z)]is sin(2πf t) (2.1)
dz
where ez is a unit vector along the Z direction chosen to point along the string’s
length, B is the magnetic field in the x − y plane, is sin(2πf t) is the AC drive current
with amplitude is = 0.3 A and frequency f = f0 = 850 Hz. We detect motion in
the X direction which is sensitive to the magnetic field component By (see Figure 2.1
for reference coordinates). The DSMG is insensitive to Bx and Bz . Integrating the
ampere force along the length of the string gives the X component of the total force:
l
2
2πz
Fx (t) = By is sin(2πf t) sin dz (2.2)
l
− 2l
Figure 2.1 shows a picture of the DSMG with the string aligned to the vertical
z-axis and vibrating in ‘S’ mode along the x-axis. Throughout this paper, the x-axis
is east, the y-axis is north and the z-axis is up. By tuning the AC current to the
second violin mode frequency, f = f0 , we get resonance. On resonance the string’s
vibration amplitude will slowly build up over time until the friction force equals the
ampere force. Using the time dependent strings equation from Bland (1960), the
amplitude of vibration produced by static magnetic field is:

Fx Q −t
x(t) = 2 2 1 − exp cos(2πf t)
2π mf τ
l

2
is Q −t 2πz
= 2 2 1 − exp cos(2πf t) By (z) sin dz (2.3)
2π mf τ l
− 2l
then using a Taylor expansion By (z) = zByz + 12 z 2 ∂ 2By /∂z 2 + 16 z 3 ∂ 3By /∂z 3 +
1 4
24
z ∂ 4By /∂z 4 ....

is Q −t Byz l2 ∂ 3 B (π 2 − 6)l4
x(t) = 2 2 1 − exp cos(2πf t) + · + · · · (2.4)
2π mf τ 2π ∂z 3 48π 3
where Q = πτ f is the mechanical quality factor of the string, τ is the string relaxation
time, l is the length of the string and m is the mass of the string. The Q factor can
be defined as the number of oscillations required for the oscillating system’s energy to
fall off to e−2π , or ∼ 1/535, of its original energy. The degree of signal amplification
at resonance is equal to the Q factor of the system. The magnetic gradient measured
by the DSMG is:
∂ 3 B (π 2 − 6)l2
BDSM G = Byz + · +··· (2.5)
∂z 3 24π 2
≈ Byz (2.6)
The higher order terms in the equation 2.5 can be ignored, provided that the distance
to the magnetic source is larger than the length of the string. In this far field case,
the DSMG can directly measure the local off-diagonal component of the magnetic
gradient tensor Byz with a single sensing element. The displacement of the string
along the x-axis is measured with two inductive pickup coils, connected in differential
√
mode. The pickup coils can measure displacement with an accuracy of 10−11 m/ Hz
√
corresponding to an equivalent magnetic gradient of 0.4 nT/m/ Hz.
The DSMG is normally suspended vertically to measure Byz . It can be rotated
azimuthally to measure Bxz . Measuring Bxy is more difficult because this requires
suspending the DSMG horizontally, which would degrade the performance of the
mechanical isolator. The string is also sensitive to gravity gradients, although the
gravitational force produces low frequency motion that is far from resonance. The
gravity gradient amplitude is attenuated relative to the magnetic gradient amplitude

by a factor of Q. The displacement due to gravity gradients is:
l
2
1 2πz
xgravity = 2 2 gx (z) sin dz
2π lf l
− 2l
gxz l
= (2.7)
4π 3 f 2
√
The displacement noise floor of 10−11 m/ Hz corresponds to an equivalent grav-
√
ity gradient of ∼ 106 Eötvös/ Hz and this means that the DSMG is insensitive to
gravity. In Golden et al. (2007), the authors present the development of string gravity
gradiometer using similar principles but with more massive string and lower mechan-
ical frequency. Golden et al.’s string gravity gradiometer has target sensitivity of
√
5 Eötvös/ Hz.
Figure 2.1: Photograph of the direct string magnetic gradiometer (DSMG) during a line
survey at Shenton Park showing the x-, y- and z-axis. The DSMG is installed on a mechan-
ical isolator inside a transparent acrylic vacuum tank. Also shown is a sketch of the string
indicating that it is aligned to the vertical z-axis and vibrates in ‘S’ mode along the x-axis.
The 20 buckets placed in a line mark the location of each measurement point.
The DSMG sensor is 300 mm by 30 mm by 30 mm and weighs 300 g. It operates

at room temperature in either atmospheric pressure or in a vacuum. McRae et al.
(2004) describe the effect of aircraft acoustic and vibration noise on the DSMG and the
noise reduction achieved from mounting the DSMG sensor inside a four-stage passive
isolator. The stages have resonant frequencies at 17 Hz, 40 Hz, 70 Hz and 150 Hz,
providing more than 80 dB of mechanical isolation above 200 Hz. The isolator is
attached to the inside of a cylindrical vacuum tank made of transparent Perspex with
a diameter of 230 mm and height of 700 mm. The sensor, isolator and vacuum tank
have combined weight of 9.1 kg.
2.3 Iron Sphere measurements
The introduction showed how the DSMG can measure magnetic gradients using a
single sensing element. As a practical demonstration of the DSMG sensor we per-
formed a series of experiments with three iron spheres of volume 8.5 cm3 , 14.0 cm3
and 28.7 cm3 . All three spheres have a volume magnetic susceptibility of k =
1.6 ± 0.3. Inside the laboratory, the magnetic field varied considerably from place
to place due to metal equipment. The magnetic field is: Bx = 0 nT ± 5000 nT,
By = 25 000 nT ± 5000 nT, Bz = 50 000 nT ± 5000 nT.
We suspended the magnetic gradiometer 1 m above the ground with its sensitive
axis facing magnetic north. In this alignment, the DSMG measures the Byz tensor
component. We placed the spheres on a conveyer belt 30 cm beneath the gradiometer
where they moved from 60 cm north of the sensor to 60 cm south of the sensor. A
second measurement was taken as the spheres moved back from 60 cm north to 60 cm
south. Figure 2.2 shows both measurements, from the difference of the two curves, we
estimate the error of the experiment to be on the order of 1 nT/m. The high signal
to noise ratio of the −30 cm position causes the two measured curves to overlap in
that graph. The round trip on the conveyor belt took 128 s during which data was
recorded at 10 points per second without averaging. For each of the three spheres, we
took recordings at three different vertical positions of −30 cm, −45 cm and −65 cm
below the centre of the string. We also performed test recordings with the conveyor
2.3. IRON SPHERE MEASUREMENTS 43
belt running below the DSMG but without iron spheres. Conveyor belt parts were
replaced with plastic and wood until the error from moving parts was reduced to
1 nT/m.
Iron sphere measurements

Byz DSMG theory DSMG measured
100
0
Magnetic gradient (nT/m)
−100 30 cm below gradiometer
20
0
5
0
−60 −40 −20 0 20 40 60
Northing (cm)
Figure 2.2: Magnetic gradient response of a 28.7 cm3 iron sphere moved north-south on
a conveyor belt at three different positions below the gradiometer. The thin grey curve
represents the theoretical magnetic gradient Byz according to equation 6, the thin black curve
shows the theoretical gradient BDSM G that the string is expected to measure according to
equation 5, the two dotted curves are the experimental data for each direction of the sphere’s
movement. To enable comparison with a flyby past a geological target, the northing shown
in this graph is the relative position of the DSMG with respect to the sphere, despite the
fact that it is the sphere which is moving.
Using the known magnetic moment of the spheres, we can calculate both the the-
oretical magnetic gradient Byz and the theoretical gradient BDSM G including higher
order terms that the string is expected to measure. Figure 2.2 compares these two
calculated gradients with the measured gradient for the 28.7 cm3 sphere. Examining
the −30 cm position, the measured data has a good fit with the theoretical gradient
BDSM G , whereas diverging from the first order gradient Byz . Because we are dealing
with a target distance comparable with the length of the string sensing element, the
far field approximation of equation 2.6 breaks down. This leads to a deviation be-
tween the theory for an ideal point like magnetic gradiometer and the integral for the
real string sensor element. At scaled distances appropriate for a geological target the
far field approximation would apply and the ideal magnetic gradiometer and string
magnetic gradiometer would agree. The close match between Byz and BDSM G for
the data of the −45 cm and −65 cm positions shows that the DSMG can accurately
measure magnetic gradients at these distances. The match is so close for the −65 cm
position that the curve representing BDSM G is superimposed on top of Byz .
Magnetic modelling of tensor components

10
0
−10 Bxx Byy Byz Bzz
0.01
0
−0.01 Bxy Bxz
0
Depth (m)
−100
Nickel deposit
−200
−300
−500 −400 −300 −200 −100 0 100 200 300 400 500
Northing (m)
Figure 2.3: Forward modelling by Graham Jenke using Potent. The top and middle graphs
show the magnetic tensor components along a 1000 m north-south survey line past an ore
body. For the model, we used the Earth’s magnetic field located at 30◦ S, 120◦ E, which is in
the middle of the Western Australian outback. The total magnetic intensity is 57 800 nT,
the inclination is 64◦ 14 and the declination is 0◦ 48 . Bxy and Bxz are very small and are
shown separately in the middle graph. The bottom graph shows an elevation view of the ore
body, facing west. The anomaly is a hypothetical magmatic nickel sulphide ore body with
magnetic volume susceptibility 1.0 SI. The body has an ellipsoid shape 500 m long striking
east-west, 200 m in depth extent, 20 m thick, 100 m depth to top.
2.4. LABORATORY NOISE PERFORMANCE 45
These results can be compared with a representative geological target consisting

of an ellipsoid shaped nickel deposit extending from 100 m down to 300 m below the
surface. Figure 2.3 shows the calculated six tensor components Bxx , Bxy , Bxz , Byy ,
Byz , Bzz that could be recorded by a generic magnetic gradiometer travelling on the
surface in a north-south direction from y = −500 m to y = 500 m (the anomaly is
located at y = 0 m). The DSMG can only measure the off diagonal components
Bxy , Bxz , Byz . The component Byz in Figure 2.3 shows a similar dipolar profile and
magnitude to that of the iron sphere at a depth of h = −65 cm. This demonstrates
that the sensitivity of the DSMG is sufficient to detect and interpret realistic geological
targets. The mismatch between BDSM G and Byz for the ellipsoid with 100 m depth
to top is ∼ 4 fT/m, an inaccuracy smaller than 1 ppm.
2.4 Laboratory noise performance
We conducted laboratory tests of the DSMG in an unshielded laboratory. The mea-

surements were conducted after midnight, and time intervals had to be chosen care-
fully to avoid the magnetic noise from nearby operation of an elevator. Our frequency
band of interest is between 0.01 Hz and 0.625 Hz for a survey in an aeroplane flying
at 50 m/s. This frequency range corresponds to geological target depths from 80 m to
5 km, as recommended by Schmidt and Clark (2000). Figure 2.4 shows the frequency
√
spectrum of the DSMG noise, which is flat at 0.4 nT/m/ Hz over this frequency
band. The graph shows a 2 nT/m signal with frequency 0.1 Hz, which we produced
for this recording so that we could calibrate the noise floor. Some additional low
frequency noise is present below 2.5 mHz with a 1/f spectrum, the spectral density
(units nT2 /m2 /Hz) is characterised by a 1/f 2 spectrum of random walk noise. Some
possible causes of low frequency noise in the DSMG are thermal drift, frequency
tracking errors, stress relaxation and changes in pressure.
The raw data in Figure 2.4 is the signal coming out of the analogue to digital
converter (ADC). The mechanical resonance of the string’s vibration acts as a me-
chanical filter with bandwidth Δf = f0 /2Q, where Q is the mechanical quality factor
of the string and f0 is the resonant frequency. At the input of the ADC there is
an analogue electronic low pass filter to suppress high frequency noise. The graph
shows the noise in the raw data tapering off at frequencies above 1 Hz. In order to
determine the signal to noise ratio at high frequencies we performed test recordings
with calibration signals from DC to 3 Hz in 0.1 Hz intervals. We then normalised
√
the raw data to show the calibrated gradient noise in units of nT/m/ Hz. Figure
2.4 shows that the calibrated gradient noise is flat for frequencies up to 2 Hz because
the low pass filters affect noise and signal equally. For frequencies above 2 Hz, the
quantization noise in the ADC increases the DSMG noise vis a vis the signal.
Spectrum of quiet magnetic gradiometer recording

100
Gradient noise (nT/m/ √Hz)
Raw data
Calibrated gradient
10
Band−pass filter
0.1
−4 −3 −2 −1 0 1
10 10 10 10 10 10
Frequency (Hz)
Figure 2.4: Amplitude spectral distribution plots of the direct string magnetic gradiometer
making a quiet recording in the laboratory. The calibrated gradient tries to correct the raw
data for the mechanical and analogue electronic filtering. The band-pass filter is uesd to
produce the data in Figure 2.5. The peak at 0.1 Hz is a 2nT/m signal.
We processed the DSMG data using a digital band-pass filter with brick wall cutoff
from 0.01 Hz to 0.625 Hz, as annotated to Figure 2.4. Figure 2.5 then presents a 200 s
sample of the digitally filtered data in time domain format. This filter accepts all data
pertinent to airborne surveys while minimising the noise. The 2 nT/m sinusoidal
signal with frequency 0.1 Hz can be seen in Figure 2.5 above a noisy background.
The signal to noise ratio of 6 gives a root mean square noise of 0.3 nT/m.
Figure 2.6 presents the complete quiet laboratory recording used for analysis in this
section. The DSMG output has drifted a fraction of a nT/m after 90 min of operation.
2.4. LABORATORY NOISE PERFORMANCE 47
2 nT/m peak signal

6

4
2
0
−2
−4
−6
0 50 100 150 200
Time (s)
Figure 2.5: Quiet recording of the direct string magnetic gradiometer in the laboratory
with a signal of amplitude 2 nT/m at frequency of 0.1 Hz. This graph shows a short 200 s
extract from the complete 6000 m recording.
We processed the raw data with a 10 s moving average filter. By averaging, we filtered
out the 0.1 Hz calibration signal so the graph shows only noise. The equivalent brick
wall bandwidth is reduced to 0.05 Hz and the root mean square noise is reduced
to 0.15 nT/m. This noise floor provides the basis for static magnetic gradiometer
applications.
Quiet recording of magnetic gradiometer

0.5
−0.5
0 1000 2000 3000 4000 5000 6000
Time (s)
Figure 2.6: The complete 6000 s quiet laboratory recording used in Figure 2.4 and 2.5.
This graph shows the remaining noise after applying a 10 s moving average to filter out the
0.1 Hz signal.
2.5 Field testing
We conducted field testing at the University of Western Australia field station at

Shenton Park, Perth, a magnetically quiet location away from buildings and power
lines. Latitude is −31◦ − 56 − 54 and longitude is 115◦ 47 42 . Before testing the
DSMG we took total magnetic intensity measurements with a proton precession mag-
netometer in a 20 m by 20 m grid at 1 m intervals. Figure 2.7 presents a contour map
of the data, showing a ΔB = 6000 nT magnetic anomaly at grid location 6 464 111 m
east, 386 607 m north. The graph also shows a 20 m long north-south survey line for
DSMG measurements at an Easting of 6 464 113 m, located 2 m to the east of the
anomaly.
In Figure 2.8 we present the magnetic gradient measured by the DSMG along
the survey line at 6 464 113 m east. We placed the magnetic gradiometer at 1 m
intervals over the 20 m survey line, taking five measurements of Byz with the DSMG
at each location. Each measurement placed the DSMG on level plastic bucket (see
Figure 2.1) and took 10 s. The gradient profile measured by the DSMG shows little
signal at the start of the line, large magnetic oscillations when the northing is between
386 605 m and 386 615 m, then finally little signal at the end of the line. The gradient
reaches a maximum at 386 607 m north, which is the closest approach along the survey
line to the anomaly. Figure 2.8 also shows the mean and standard deviation of Byz
measurements at each point. The standard deviation was typically ±0.2 nT/m except
at a northing of 386 607 m (near the anomaly). Figure 2.1 is a photograph of the
DSMG during the survey at Shenton Park.
This simple but effective field test shows that the DSMG can make reliable mea-
surements in a field environment and produce a predictable gradient signal over a
known anomaly with a steep gradient response. A more complex testing regime in-
volving testing of the immunity of the DSMG to vibration, tilt, temperature, and
other environmental factors will be necessary to prove the suitability of the sensor for
real-world measurements on the ground or in an aircraft.
2.5. FIELD TESTING 49
Total Magnetic Intensity survey
6464100 6464110 6464120
386620 386620
DSMG 58900 nT
survey line 58750
6464113E 58600
386615 386615 58450
Northing (m) 58300
58150
386610 386610 58000
57700
386605 386605
52500
386600 386600
6464100 6464110 6464120
Easting (M)
m
Figure 2.7: Contour map of total magnetic intensity at the Shenton Park testing site. Note
that the scale to the right is non-linear for field strengths below 58 000 nT. At grid location
6 464 111 m east, 386 607 m north, the magnetic field bottoms out at 52 500 nT. The graph
shows a 20 m long north-south survey line at an Easting of 6 464 113 m, where gradiometer
measurements are taken (shown in Figure 2.8).
Magnetic line survey
Gradiometer Byz
200
−200
2
Standard deviation in Byz
(nT/m)
0
386600 386605 386610 386615 386620
Northing (m)
Figure 2.8: Magnetic gradients along a survey line at 6 464 113 m east at Shenton Park.
The top graph shows the measured gradient Byz from the direct string magnetic gradiometer
(DSMG) at 1 m intervals. The bottom graph shows the standard deviation of the five DSMG
measurements made ar each position.
2.6 Conclusion
We have shown that the DSMG is capable of measuring magnetic gradients with
a precision of 0.2 nT/m. Work is progressing on lowering the noise floor for static
measurement down to 0.01 nT/m. The frequency spectrum of the DSMG noise is flat
√
at 0.4 nT/m/ Hz over the frequency band of interest for airborne applications. Our
measurements of the magnetic gradient from iron spheres show that the gradiometer
approximates a point gradiometer (far field approximation) for objects further than
45 cm from the sensor. By measuring the signal from the iron sphere, we demonstrate
sufficient sensitivity to detect a hypothetical ellipsoid nickel deposit 500 m long, 200 m
depth extent, and 20 m thick located 100 m below the surface. Project AMATI is
developing a magnetic gradiometer system for airborne use, but recent interest could
lead to the sensor being packaged for static ground and mineral borehole use.
Acknowledgements
We would like to thank Mr. Graham Jenke for forward geological modelling. The
authors would also like to thank Mr. Slawomir Gras for many useful discussions and
suggestions. Project AMATI is funded in part by a linkage grant from the Australian
Research Council.
BIBLIOGRAPHY 51
Bibliography
Bland, D. R., 1960, Vibrating strings: an introduction to the wave equation, 1st
edn. Routledge and Paul, 40-41.
Golden, H., McRae, W., and Veryaskin, A. V., 2007, Description of and Results
from a Novel Borehole Gravity Gradiometer: Extended Abstracts, 19th ASEG
Geophysical Conference and Exhibition, doi: 10.1071/ASEG2007ab047.
McConnell, T., and Dragoset, B., 2006, Introduction to this special section:
magnetic gradiometry: Leading Edge, 25, 45. doi: 10.1190/1.2164753
McRae, W., Veryaskin, A. V., Ju, L., Blair, D. G., Chin, E., Dumas, J., and
Lee, B., 2004, String magnetic gradiometer system: recent airborne trials: SEG
Technical Program Expanded Abstracts, 23, 790-793. doi: 10.1190/1.1845297
Schmidt, P. W., and Clark, D. A., 2000, Advantages of measuring the magnetic
gradient tensor: Preview, 85, 26-30.
Veryaskin, A. V., 2001, Magnetic gradiometery: a new method for magnetic gra-
dient measurements: Sensors and Actuators A, 91, 233-235. doi: 10.1016/S0924-
4247(01)00489-7
Wooldridge, A., 2004, Review of modern magnetic gradiometer surveys: SEG

Technical Program Expanded Abstracts, 23, 802-805. doi: 10.1190/1.1851323
Chapter 3
Displacement readout
3.1 Preface: Inductive readout
3.1.1 Principal of operation
The sensing element of the Direct String Magnetic Gradiometer (DSMG) is a 250 mm
long aluminium string, carrying alternating current. The string is essentially a driven
oscillator. The current sent into the string is a combination of a high frequency
∼ 3 MHz carrier signal and a low frequency ∼ 750 Hz drive signal. The drive current
interacts with the ambient magnetic field and sets the string into motion at ∼ 750 Hz.
The displacement of the string is measured with two inductive pickup coils. The
coils are placed at the one quarter and three quarter positions along the string and
connected differentially to detect second harmonic vibrations.
Figure 3.1: (a) Uniform magnetic fields cause the string to vibrate in its 1st violin mode
(b) Magnetic gradients cause the string to vibrate in its 2nd violin mode. The frequencies
shown above are nominal. The actual frequencies are typically within the range: 325 Hz <
f1 < 425 Hz and 650 Hz < f2 < 850 Hz.
53
54 Chapter 3. DISPLACEMENT READOUT
In Fig. 3.1a, the magnetic field is uniform and the string will vibrate in its 1st
violin mode as shown. The string is vibrating at a frequency far from its mechanical
resonance so the amplitude of vibration is very small. Near the string is a pair of
pickup coils used to form an RLC circuit. The mutual inductance between the string
and the pickup coils changes as the string vibrates. However, vibration in the 1st
violin mode will produce very little signal in the RLC circuit because the pickup coils
are connected in differential mode. The output in the diagram contains only the
∼ 3 MHz carrier.
In Fig. 3.1b, there is a magnetic gradient and the string will vibrate in its 2nd
violin mode as shown. The string is vibrating at resonance so the amplitude of
vibration is very large. Vibration in the 2nd violin mode will produce a strong signal
in the differential RLC circuit. The output in the diagram shows the ∼ 3 MHz
carrier being modulated by the ∼ 750 Hz string vibrations. Common mode rejection
is provided by both the string’s resonant response and the pickup coil geometry.
3.1.2 Mutual inductance
The inductive readout works because the mutual inductance between the string and
each pickup coils will change as the string moves. Each pickup coil consists of 40
turns of wire wrapped around a plastic rectangular prism, see Fig. 3.2. The mutual
inductance, M, between the string and one turn of wire is:
Lp xmax
μ0
M= dx dy (3.1)
2πx
0 xmin
μ0 Lp [log(xmax ) − log(xmin )]
= (3.2)
2π
where μ0 is the magnetic permeability of free space, xmax = xmin + Tp is the

maximum perpendicular distance of the turn to the string, xmin is the minimum
perpendicular distance of the turn to the string, Tp = 0.028 m is the thickness of the
rectangular prism, and Lp is the length of the prism parallel to the string.
When taking into account all 40 turns, where will be a lateral displacement yn for
each turn n. In this case, the total mutual inductance is:
3.1. PREFACE: INDUCTIVE READOUT 55
Figure 3.2: Diagram of a single pickup coil and the string. Lp = 0.006 m, 2Y = 0.054 m,
Tp = 0.028 m, and xmax = xmin + Tp . xmin and xmax vary as the string moves.

2 2 − log 2 2

40 μ L log
0 xmax + y n xmin + y n
M= (3.3)
n=1
2π
The length of the entire coil of turns is 0.0054 m so yn ranges from −0.0027 m to
+0.0027 m in steps of 0.000135 m. The above summation can be approximated by
the following integral:

N/2 μ0 L log xmax + (2Y /N) n − log
2 2 2 2 2
xmin + (2Y /N) n2
M= dn (3.4)
2π
−N/2

μ0 LN x2max + Y 2 xmax Y xmin Y
= log + tan−1 − tan−1
2π x2min + Y 2 Y xmax Y xmin
(3.5)
where N = 40 turns and Y = 0.0027 is the lateral displacement of the first and
last turns. Taking the x derivative of the above expression gives:

∂M μ0 LN −1 Y −1 Y
= tan − tan (3.6)
∂x 2πY xmax xmin
Setting values of xmin = 0.0005 m, xmax = xmin + 0.028 m = 0.0285 m, gives

the calculated theoretical value of ∂M/∂x = 1.28 × 10−4 H/m. While xmin << Y ,
the inverse tangent function is saturated at 12 π. As long as this condition holds, the
derivative ∂M/∂x will be approximately constant and the mutual inductance will be
linear.
More explicitly, the sensitivity of mutual inductance to displacement will be linear
in the region 0.000 m < xmin < 0.001 m where xmin is the distance of the string to
the pickup coils.
3.1.3 Dynamic range
To verify that the mutual inductance is indeed linear in this region, I measured the
mutual inductance with the string at different positions, see Fig. 3.2. An alternating
current with amplitude i and angular frequency ωRF was applied to the string. The
flux through one pickup coil is φ = iM. The electromotive force is then:
∂φ
VEM F = (3.7)
∂t
= ωiM (3.8)
where t is time. A pickup coil with 40 turns was attached to a vernier scale so that
it could be moved closer and further from the string. I then measured the voltage
VEM F across the pickup coil so that the mutual inductance could be calculated. Fig.
3.3 plots the mutual inductance, M, as a function of the pickup coil’s position, xmin .
The experimental data in Fig. 3.3 is in good agreement with the theoretical
data. The systematic error of 8% is attributed to instrumental error but does not
affect the results discussed here. The graph shows that the mutual inductance is
linear for distances up to 0.001 m. The pair of pickup coils are normally located only
0.0005 m away from the string so the inductive readout is linear over its entire region
of operation. The maximum magnetic gradient that can be detected by the string in
the linear region of operation is of order ∼ 0.01 T/m. Even larger magnetic gradients
could be accommodated with feedback. However, the dynamic range of the DSMG
is limited by the 20 bit digital to analogue converters used by the feedback loop.
3.1. PREFACE: INDUCTIVE READOUT 57
0.7
0.6 Experiment
Linear Fit
Mutual inductance (ȝH)

0.5 Theory
0.4
0.3
0.2
0.1
0
0 0.002 0.004 0.006 0.008 0.01
Distance of string to pickup coil (m)
Figure 3.3: Mutual inductance between the string and a pickup coil as a function of the
distance separating them. The experimental data points are calculated from the voltage
measured across the pickup coil, VEM F . The straight line is a linear fit to the first 11 data
points. The theoretical mutual inductance is from Eq. 3.5.
3.1.4 Sensitivity
√
The sensitivity of the inductive readout system is (1.0 ± 0.1) × 10−11 m/ Hz, see Fig.
1.11 in the introduction. The electronic noise level is lower than the thermal noise
level so the present inductive readout is adequate. However, the inductive electronics
may not be sensitive enough after other improvements in the DSMG are implemented.
The following paper describes a new capacitive displacement readout that I have
developed. This readout will improve the sensitivity of a proposed DSMG for space
applications (discussed in Chapter 6). The paper was published in Sensors and Ac-
tuators A. My contribution to this paper amounts to 95% of the experimental work,
and 80% of the manuscript preparation.
Sensors and Actuators A 153 (2009) 5-12
Differential readout for a magnetic gradiometer

Andrew Sunderland1 , Alexey V Veryaskin, Howard Golden1,2 , Wayne McRae1,2 , Li Ju1 ,
David G Blair1

Abstract
A magnetic gradiometer system being developed by the authors uses a
vibrating string as its sensing element. The next generation of sensor will
use a wide ribbon instead of a string. In order to precisely measure the
displacement of the ribbon we have designed a new differential capacitive
√
readout with displacement noise below 10−13 m/ Hz. Two capacitive
plates are placed at the one quarter and three quarter positions adjacent
to the ribbon. An inductor is added in series to make an LC series circuit.
Displacement is determined by measuring the phase. From a scan of
phase versus frequency, the linear dynamic range is measured to be ∼
360 nm. The LC series readout is highly immune to stray capacitance and
is designed for use with long cables in a magnetic gradiometer system.
3.2 Introduction
Capacitive sensing has been used for precision displacement measurement for many
applications. Transducers in gravitational wave detectors have achieved sensitivity of
√
∼ 10−18 m/ Hz using superconducting capacitors (with SQuID readouts) at DC[1]
and at 10 GHz using low noise microwave electronics[2]. Sensitivity is limited by the
Nyquist noise of both the mechanical system to be measured and the electronic read-
out system. In this paper we address the measurement of very small displacements
of a vibrating ribbon which enables the direct detection of magnetic field gradients.
√
In this case sensitivity of ∼ 10−13 m/ Hz is adequate.
The Direct String Magnetic Gradiometer (DSMG) was first proposed by one
of the authors[4], and subsequent prototypes have been built with sensitivity 4 ×
√
10−10 T/m/ Hz. Magnetic gradients exert a force on a current carrying ribbon and
cause it to vibrate at its second harmonic frequency, as described in section 3.7.
Sensing magnetic gradients requires the detections of antisymmetric second harmonic
vibrations. We achieve this with differential capacitive readout. The readout operates
at radio frequencies (RF) and the design is optimised for measuring the displacements
of small areas on a low mass ribbon.
Resonant RF circuits are widely utilised in phase-sensitive mechanical transducers

where displacements of a sensing element need to be measured[3]. Inductive and
capacitive components are normally connected in parallel to make a high impedance
LC bridge. A parallel bridge needs to be pumped by an external RF power source
with a high impedance output in order to preserve the high electronic Q factor of the
bridge. The capacitor separation distance and the LC brdige Q factor determine the
noise floor of such mechanical-displacement-to-voltage transducers[3].
Instead of using such a traditional parallel resonant circuit, we present a series LC

bridge that can be connected to standard off-the-shelf 50 Ω input-output integrated
circuits. The phase versus displacement characteristic of a series RF resonant bridge
is the same as in a parallel one. However, both the impedance and the output voltage
reach a minimum at resonance, effectively short circuiting any stray capacitance in
long cables. Long cables are required because magnetically susceptible components
will create spurious magnetic signals unless there are removed several metres from
the sensing element of the DSMG.
In our new readout system, capacitive plates are placed at one quarter and three
quarters positions along the ribbon, below and above it, to enable differential readout.
The small gaps between the ribbon and the two plates form a pair of capacitors. These
capacitors are connected in series with an inductor to make an LC series circuit as
shown in Fig. 3.4. Ribbon vibrations with respect to the fixed capacitive plates
produce phase changes that can be measured by down converting with a frequency
mixer.
Readout noise makes a significant contribution to the total noise budget of the
DSMG. The differential capacitive readout will lower the readout noise and increase
sensitivity. Here we report on the first laboratory results obtained with the new
readout.
10 MHz in
4:1 transformer
7.6 ȝH
55 pF ± 10 pF
ribbon
75 pF
± 10 pF
Figure 3.4: Diagram of the LC series resonant bridge tuned to 10 MHz pump signal
provided by a low phase noise oscillator.
3.3 Amplitude measurements
The readout described in this paper is designed to measure the location of an alu-
minium ribbon at the one quarter and three quarter points. This is achieved with
an LC series circuit. A capacitor is formed from the gap between the ribbon and
the capacitive plate at the one quarter position. A second capacitor is formed from
the gap between the ribbon and the capacitive plate at the three quarter position. A
coil of wire wound around an air core is the inductor of the LC circuit. The ribbon
has length l = 250 mm, width w = 5 mm and thickness t = 10 μm. A sketch of
the ribbon is shown in Fig. 3.5. In order to maximise the electronic Q factor of the
capacitors, the capacitive plates have a width of 5 mm which exactly matches that of
the ribbon. The length of the plates is 92 mm so the area per plate is A = 460 mm2 .
The circuit in Fig. 3.4 has two capacitors in series which were measured using
an LC meter to be C1 = 55 pF ± 10 pF and C2 = 75 pF ± 10 pF respectively. The
total capacitance is thus C = 32 pF ± 5 pF, whilst the inductance was measured at
7.6 μH ± 0.2 μH using the same LC meter. The LC series circuit acts as a band stop
filter with resonant frequency:
3.3. AMPLITUDE MEASUREMENTS 61
e
at
x
pl
92 mm
ve
iti
ac
z
ap
C
y
n
250 mm
bo
ib
R
e
at
pl
ve
iti
ac
10 ȝm
ap
C
5 mm
Figure 3.5: Sketch of the ribbon and a pair of capacitive plates using an oblique cabinet
projection. Not to scale.
1
f0 = √
2π LC
= 10.2 MHz ± 0.8 MHz (3.9)
Because of the poor precision of the LC meter, there is a large error in the pre-
dicted resonant frequency. The exact resonant frequency could only be determined
by experiment. The distance between the capacitive plates and the ribbon was ad-
justed until the measured resonant frequency was equal to 10 MHz. The frequency
dependent impedance of the LC circuit is:
j
Z = 2πf Lj − + RLC
2πf C

f f0
= RLC QLC − j+1 (3.10)
f0 f
where j is the imaginary unit, f is frequency, RLC is the resistance due to losses
in the capacitors and inductors, and QLC is the electronic quality factor of the LC
bridge:
2πf0 L
QLC = (3.11)
RLC
Near f → f0 we can define Δf = f − f0 and Eq. 3.10 can be approximated to:

Δf
Z ≈ RLC 2QLC j +1 (3.12)
f0
In order to test the resonance, signals with frequencies ranging from 9 MHz to
11 MHz with amplitude Vsig = 1.44 V were injected into the LC circuit. The signal
amplitude at the output of the amplifier was measured to be Vamp = 18 V peak. Fig.
3.6 shows the resonance manifesting itself as a voltage minimum at f0 = 10.000 MHz±
0.005 MHz.
4
3
Amplitude (V)
1
10 MHz signal only
All harmonics
0
9.8 9.85 9.9 9.95 10 10.05 10.1 10.15 10.2
Frequency (MHz)
Figure 3.6: This graph plots the voltage across the LC circuit versus frequency as measured
by an oscilloscope. Resonance is at f0 = 10.000 MHz ± 0.005 MHz.
The signal generator is a Stanford Research Systems DS345 signal generator which
has an output impedance of Rgen = 50 Ω. The output goes to a mixer with input
impedance Rout = 50 Ω. The 4:1 transformer shown in Fig. 3.4 increases the im-
pedance of the LC bridge by a factor of 4. There is also a variable resistor in parallel
which is used to tune the resonance. The full circuit diagram is shown in Fig. 3.7.
An oscilloscope was connected to measure the voltage difference between point A and
point B in the circuit. The three resistors and the amplifier output can be converted
into a Thevenin equivalent circuit[5] with input impedance of Rin = 11.5 Ω and input
voltage Vin = 0.23Vamp = 4.2 V peak. The voltage difference across the LC bridge
should be:
3.4. CAPACITANCE VERSUS DISTANCE 63
50 ȍ A
10 MHz Z 15ȍ 50 ȍ Mixer
Figure 3.7: Diagram of the complete circuit. The box marked Z is the LC series resonant
bridge shown in Fig. 3.4, the left 50 Ω resistor is the output impedance of the amplifier, the
right 50 Ω resistor is the input impedance of the mixer and the 15 Ω resistor is a variable
resistor used to tune the sharpness of the resonant peak.
4Z
Vout = Vin · (3.13)
11.5 Ω + 4Z
The series LC bridge acts as a band stop filter such that the output voltage will
be equal to the input voltage at every frequency except near resonance f = f0 . At
resonance the impedance Z will be purely resistive, limited only by the inductor and
capacitor losses:
4RLC
lim Vout = Vin · (3.14)
f →f0 11.5 Ω + 4RLC
Fig. 3.6 shows that the voltage at resonance is 1.6V peak which is 38% of the input
voltage. Substituting into the above equation gives the losses as RLC = 1.8 Ω ± 0.5 Ω
and the electrical quality factor of the LC bridge as per Eq. 3.11 is QLC = 270 ± 70.
At resonance the output appears highly distorted. This is because the 10 MHz
signal is attenuated by the LC bridge whereas the higher harmonics pass unaffected.
Fig. 3.6 shows that the measured level of harmonic distortion is 0.4 V at resonance.
3.4 Capacitance versus distance

Capacitive plates are placed at one quarter and three quarters positions along the
ribbon, below and above it. The capacitance changes with distance so differential
readout in the x direction is possible (see Fig. 3.5 for reference coordinates). Math-
ematically, the relative positions of the plates to the ribbon are x1 > 0, y1 = 0,
z1 = (l/4) and x2 < 0, y2 = 0, z2 = (3l/4). The capacitances of the two capacitive

plates next to the ribbon are:
A 0 A 0
C1 = , C2 = (3.15)
x1 −x2
where x1 and x2 are the relative positions of the capacitive plates to the ribbon
along the x-axis. A negative sign is placed in front of x2 in order to account for the
fact that the second capacitive plate is on the opposite side of the ribbon to the first
capacitive plate. The total capacitance in series is:
1
C = 1
+ C12
C1
A 0
= (3.16)
x1 − x2
By placing the capacitors on opposite sides of the ribbon, the readout is inherently
differential. If the initial distance |x1 | is much larger than |x2 | then the capacitance
C1 will be much larger than C2 however the readout will still be perfectly differential
by virtue of the way capacitance combines in series. Differential readout requires the
plates be exactly at the one quarter and three quarters positions along the ribbon.
The plates can be placed with accuracy 0.1 mm. The area of the plates are machined
with tolerance 0.01 mm along their length and width which will cause the areas of
the plates to differ with a relative error of 0.2%. This mismatch will cause a common
mode signal of −54 dB.
The average size of the two capacitive distances shall be defined:
|x1| + |x2|
x0 = (3.17)
2
Whilst the distance between the capacitive plates and the ribbon is too small
to measure directly, the average gap can be calculated using the ratio between the
measured value of the capacitance and plate area:
A 0
x0 =
C
= 60 μm ± 20 μm (3.18)
3.4. CAPACITANCE VERSUS DISTANCE 65
where 0 is the permittivity of free space. A serious misalignment problem occurs

if a capacitive plate is tilted as shown in Fig. 3.8. For the left-hand side plate shown
in the diagram, the majority of the electrical current will flow through the edge of the
plate closer to the ribbon. This causes the effective area of the left capacitive plate
to decrease. Also, the effective position of the left plate will no longer be at the one
quarter ribbon position but instead will measure the displacement closer to the end
of the ribbon. Each plate’s position along the x-axis can be adjusted with two screws,
one at each end of the plate. Each plate is relatively long and the gap between plate
and ribbon is very small so it is very easy for the capacitive plates to tilt with respect
to the ribbon. The screws were adjusted until the measured common mode rejection
ratio was 30 dB.
x
z
capacitive
plate 92 mm length
ribbon
0.06
mm
gap
Figure 3.8: Diagram of the ribbon and the two capacitive plates with the left plate tilted.
Each plate’s position along the x-axis can be adjusted with two screws, one at each end of
the plate. The anticipated misalignment could create a sensitivity to common mode signals.
From Eq. 3.16, the change in capacitance in response to small displacements is:
ΔC Δx2 − Δx1
=
C0 2x0
Δx
= (3.19)
x0
where the variable Δx = ((x1 − x2 )/2) is defined as the amount of differential
displacement. The change in capacitance will then shift the resonant frequency:
Δf Δx
= (3.20)
f0 2x0
If the frequency injected into the circuit remains unchanged at f = 10 MHz, and
the resonant frequency shifts, then the impedance and voltage will vary according
to Eqs. 3.12 and 3.13. Displacement measurements are possible by measuring the
voltage across the LC bridge. The complex voltage output Vout of Eq. 3.13 has its
greatest sensitivity at resonance however the real voltage amplitude |Vout | is flat. Fig.
3.6 shows that the amplitude |Vout | is indeed independent of frequency exactly at
f = f0 . At resonance the most sensitive measurements will be phase measurements.
3.5 Phase measurements

In order to make phase measurements, the output voltage is combined with the ref-
erence signal in a frequency mixer as shown in Fig. 3.7. The voltage at the input of
the mixer is simply the voltage as per Eq. 3.13. Using the approximation for Z given
in Eq. 3.12, this voltage expands to make:

Δf
4RLC 2QLC j f0 + 1
Vout = Vin · (3.21)
11.5 Ω + 4RLC + 8RLC QLC j Δf
f0
Dividing through by RLC and rearranging gives:
Vin 1 + 2QLC j Δf
f0
Vout = 11.5Ω
·
1 + 4RLC 1 + 11.5Ω + 4R QLC j Δf
8RLC
f0
LC
Δf
Vin 1 + 2QLC j f0
= · (3.22)
1
1 + k 1 + 1+k 2QLC j Δf
f0
with the definition k = (11.5 Ω/4RLC ) ≈ 1.6. Then using the complex conjugate
to make the denominator real:

2
4Q2LC Δf 2QLC k Δf
Vin 1 + 1+k f0
+ 1+k f0
j

2QLC 2 Δf 2
Vout = (3.23)
[1 + k] 1 + 1+k f0
If this voltage is converted into amplitude/phase form Vout = |Vout |e2πf0 t+φj , the
phase change φ due to ribbon movements is equal to:
3.5. PHASE MEASUREMENTS 67
⎡ ⎤
⎢ 2QLC k Δf ⎥
φ = tan−1 ⎣
f0
2 ⎦ (3.24)
Δf
1+k+ 4Q2LC f0
√ !
The phase increases from φ = 0 at Δf = 0 to a maximum of φ = tan−1 1/(2 k + 1)
√
at Δf = (f0 k + 1)/(2QLC ) before approaching zero as Δf → ∞. φ is an odd function
√
so there is a minima at Δf = −(f0 k + 1)/(2QLC ). At resonance the sensitivity of
phase to frequency shifts is:
dφ 2QLC
lim = (3.25)
f →f0 df f0 (1 + k)
and by Eq. 3.20, the sensitivity to displacement is:
dφ QLC
lim = (3.26)
f →f0 dx x0 (1 + k)
We shall define the electronic quality factor of the entire circuit to be:
f0
Qe =
Δfmaxima − Δfminima
QLC
= √ (3.27)
k+1
Using our previous calculated values of QLC = 270±70 and k = 1.6, the electronic
quality factor of the entire circuit should be Qe = 170 ± 50.
The output of the mixer is a combination of Vout and the reference signal Vin .
The velocity of propagation along a coaxial cable is two thirds of the speed of light.
At 10 MHz the wavelength is 20 m so both signals will have phase delays ϕ1 and ϕ2
depending on the length of cables. The mixer output is proportional to the product
of the real parts of the signals:

Vmixer ∝ Vout eϕ1 j Real Vin eϕ2 j Real

∝ |Vout |e2πf0 tj e(φ+ϕ1 )j Real |Vin |e2πf0 tj e(ϕ2 )j Real
∝ |Vout ||Vin | cos(2πf0 t + φ + ϕ1 ) cos(2πf0 t + ϕ2 ) (3.28)
Dividing by the constant factor 12 |Vin | and performing a trigonometric expansion

gives:
Vmixer ∝ |Vout | cos(φ + ϕ1 − ϕ2 ) + |Vout | cos(4πf0 t − φ − ϕ1 + ϕ2 ) (3.29)
The second term in the above expression has a frequency of 20 MHz and is filtered
out with a low pass filter. The sensitivity of the first term to ribbon displacement is:
dVmixer dφ ∂ cos(φ + ϕ1 − ϕ2 ) d|Vout |

∝ |Vout | + cos(φ + ϕ1 − ϕ2 ) (3.30)
dx dx ∂φ dx
|Vout | is insensitive to displacement at resonance so the second term in the above
expression goes to zero. By substituting in Eq. 3.26, the first term reduces to:
dVmixer −QLC sin(ϕ1 − ϕ2 )|Vout |

lim ∝ (3.31)
f →f0 dx x0 (1 + k)
From the above equation, maximum sensitivity will occur when ϕ1 − ϕ2 = (nπ/2)
where n is an odd integer. With a wavelength of 20 m at 10 MHz, a phase delay of
ϕ1 − ϕ2 = (π/2) can be achieved by making the cable running from the amplifier to
the LC bridge several metres long. This is consistent with the magnetic gradiome-
ter requirement to remove all magnetic parts at least 3 m from the sensor during
operation.
In order to perform phase measurements, signals with frequencies ranging from
9.96 MHz to 10.06 MHz were injected into the LC circuit. Fig. 3.9 shows the output
of the mixer as a function of frequency. The graph also shows the best theoretical fit
using Eq. 3.29 which has QLC = 500 ± 100 and ϕ1 − ϕ2 = 2.5 ± 0.4 radians. There is a
discrepancy between the quality factor of 500 ± 100 derived from phase measurements
and the quality factor of 270 ± 70 derived from amplitude measurements. A possible
cause is operating near the power limits of the amplifier and transformer. The dis-
crepancy narrowed when the signal power was decreased. Taking the harmonic mean
of the two quality factors gives QLC = 350±100. The fitted phase delay of 2.5 radians
is higher than the ideal value of (π/2). This causes the maxima to move closer to
resonance at f = 9.98 MHz and causes the minima to skew far from resonance at
f = 10.06 MHz.
The impedance of the LC bridge is not matched to the 50 Ω characteristic im-
pedance of the coaxial cables used in the experiment. Because the cable lengths are
3.5. PHASE MEASUREMENTS 69
200
Best fit
Measured
Mixer output (mV)

100
−100
−200
9.94 9.96 9.98 10 10.02 10.04 10.06
Frequency (MHz)
Figure 3.9: Setting the zero position to 10.002 MHz, the graph plots the relative output of
the frequency mixer versus frequency. The mixer output is proportional to the phase of the
signal across the LC circuit; the graph also shows the best theoretical fit with an electrical
Q factor of QLC = 500 ± 100 and phase delay ϕ1 − ϕ2 = 2.5 ± 0.4 radians using Eq. 3.29.
appreciable fractions of the wavelength, transmission line effects must be considered.

Running from the amplifier to the LC-bridge is a BNC terminated coaxial cable of
length 0.7 m and a Mogami W2490 twisted pair cable of length 0.5 m. Modelling the
transmission line effects of 50 Ω impedance cables up to lengths of 2 m showed that
there were only negligible changes to the quality factor Qe of the entire circuit. The
resonant frequency was calculated to shift about 0.01 MHz per metre of cable. From
experiment, changing between coaxial cables of length 0.3-3 m shifted the resonance
frequency by up to 0.05 MHz.
These results show that an impedance mismatch will shift the resonant frequency.
An impedance mismatch can also cause asymmetry in the frequency dependence of
the oscillator (shown in Fig. 3.6). When f < f0 , the LC bridge acts as a capacitor.
Bridge capacitance acts in parallel to the coaxial cable capacitance so the impedance
changes slowly. At resonance f = f0 , any stray capacitance of the cables is short
circuited by the low impedance of the LC bridge so there is no problem. When
f > f0 , the LC bridge acts as an inductor. This inductance can combine with the
parallel capacitance of the coaxial cables to cause a sharp rise in impedance.
The measured curve in Fig. 3.9 has a maximum slope in the mixer output of
15 μV per Hz at f = 9.997 MHz. Using Eq. 3.20, the sensitivity to displacement at
resonance is:
dVmixer f0
= 15 μV/Hz ×
dx 2x0
6
= 1.25 × 10 V/m (3.32)
The capacitive readout has a very high transfer coefficient of over 106 V/m, made
possible by a high electronic quality factor and a small distance of x0 = 60 μm
between the capacitive plates and the ribbon. A disadvantage of this small gap is
that the sensitivity to displacement will be linear over a only a very short range. From
Fig. 3.9, the linear region is from 9.99 MHz to 10.02 MHz which is equivalent to a
displacement dynamic range of 360 nm.
The phase measurements here establish the theoretical transfer coefficient based on
frequency shifts instead of actual displacements although in theory the two should be
equivalent. Another problem is that the distance x0 is not measured, only calculated.
A true calibration of mixer output versus displacement is performed in the next
section.
3.6 Noise measurements
For noise measurements, the DS345 signal generator was replaced with a SC10 10
MHz high-stability ovenized quartz oscillator. The oscillator produces a fixed fre-
quency output at 10 MHz with very low phase noise and harmonic signal levels below
−60 dBc. The SC10 unit has an output voltage of +13 dBm with output impedance
of 1 Ω. A variable resistor was placed in series and adjusted until the voltage across
the LC bridge was 1.6 V peak, consistent with the amplitude and phase measure-
ments.
A mechanical shaker and an 8630C5 PiezoBEAM accelerometer were used to cal-
ibrate the displacement versus mixer output. The sensor frame was placed on a pivot
so that all vibration would be differential, as shown in Fig. 3.10a. The shaker was po-
sitioned at one end to cause the entire frame to vibrate at 1032 Hz including clamps,
ribbon and capacitors. The ribbon ‘S’ mode has a low resonant frequency of 330 Hz
3.6. NOISE MEASUREMENTS 71
Figure 3.10: Photograph of experimental setups A and B. Setup A has the middle of
the frame placed on a pivot. The left-hand tip of the frame is then driven by a shaker to
produce differential vibration. Setup B has the pivot moved to the end of the frame to
increase stability. With a thickness of 10 μm, the ribbon is too thin to see clearly in these
photographs.
so the ribbon can be treated as a free mass that is isolated from vibration. The capac-
itive plates were clamped firmly so they vibrated in synch with the frame. Therefore,
the gap between the capacitive plates and ribbon will fluctuate with amplitude equal
to 90% of the frame vibration amplitude.
The accelerometer was placed on the sensor frame at the one quarter point, in the
same horizontal position as the left-hand capacitor. With the pivot, the measured ac-
celeration will equal the differential acceleration which can be converted to differential
displacement using the square of the angular frequency. With the shaker switched on,
the accelerometer recorded an amplitude of 1.2 × 10−12 m rms at 1032 Hz as shown
in Fig. 3.11a. Fig. 3.12a shows the mixer output recorded simultaneously including a
1.15 μV rms signal at 1032 Hz. Comparing the mixer voltage with the accelerometer
output gives a calibration coefficient of:
dVmixer
= 9.8 × 105 V/m (3.33)
dx
The noise level recorded by the accelerometer in Fig. 3.11 a varies as the inverse
√
square of the frequency with rms amplitude 7.9 × 10−7f −2 m/ Hz. The noise
spectrum is characteristic of seismic noise:
Differential displacement (m/√Hz)
−11
10
a
b
−12
10
−13
10
900 950 1000 1050 1100
Frequency (Hz)
Figure 3.11: Spectral density of differential displacements measured by an accelerometer

placed at the one quarter position on the sensor. The noise levels between 900 Hz and
1100 Hz are shown for both of the experimental setups shown in Fig. 3.10a and b. Calibration
signals of (a) 1.2 × 10−12 m rms and (b) 5.8 × 10−13 m rms at 1032 Hz show up as spectral
√ √
densities of 2.4 × 10−12 m/ Hz and 1.16 × 10−12 m/ Hz respectively because the frequency
resolution is 0.25 Hz.
100000
a
b
1032 Hz signals
Mixer output (nV/√Hz)

10000
1000
100
10
900 950 1000 1050 1100
Frequency (Hz)
Figure 3.12: Spectral density of mixer output between 900 Hz and 1100 Hz for the ex-
perimental setups shown in Fig. 3.10a and b. Note the 50 Hz mains power harmonics at
900 Hz, 950 Hz, 1000 Hz, 1050 Hz and 1100 Hz. Calibration signals of (a) 1150 nV rms and
√ √
(b) 530 nV rms at 1032 Hz show up as spectral densities of 2300 nV/ Hz and 1060 nV/ Hz
respectively because the frequency resolution is 0.25 Hz.
“Typically, seismic noise rolls off at 40dB/decade at frequencies above the 1Hz

√
microseismic peak of approximately 10−6 m/ Hz[6].”
In order to isolate the readout from seismic noise, sorbothane 50 was placed below
all experimental apparatus. Sorbothane 50 is a vibration damping material with a
loss tangent of 0.5 between 10 Hz and 30000 Hz. In addition, the pivot was moved
to the end of the sensor to increase stability, as shown in Fig. 3.10b. In this new
setup, the shaker vibration will be half differential and half common mode. The
accelerometer output shown in Fig. 3.11b has already been divided by 2 to reflect this.
√
The seismic noise level was reduced to 3.2×10−7f −2 m/ Hz rms. The accelerometer
also recorded a differential vibration signal of 5.8 × 10−13 m rms at 1032 Hz made by
the shaker. Fig. 3.12b shows the mixer output recorded simultaneously including a
0.53 μV rms signal at 1032 Hz. Comparing the mixer voltage with the accelerometer
output gives a calibration coefficient of:
dVmixer
= 9.2 × 105 V/m (3.34)
dx
Eqs. 3.32, 3.33 and 3.34 calculate the calibration coefficient of the readout using
three different methods. The average and uncertainty is:
Calibration coefficient = (1.05 ± 0.17) × 106 V/m (3.35)
Fig. 3.12b shows that the mixer output includes harmonics of 50 Hz mains power
at 900 Hz, 950 Hz, 1000 Hz, 1050 Hz and 1100 Hz. Fig. 3.13 presents the mixer
output over a wide band and shows that the entire spectrum is contaminated with
50 Hz mains power.
100000
Mixer output (nV/√Hz)
10000
1000
100
10
0 200 400 600 800
Frequency (Hz)
Figure 3.13: Spectral density of mixer output between 0 Hz and 800 Hz for the experi-
mental setups shown in Fig. 3.10b. Note the mains power harmonics at multiples of 50 Hz.
Both the amplifier and oscillator are powered with direct current (DC). However,
a switched mode power supply is used to convert 240 V, 50 Hz mains power into
DC. The line regulation is not perfect so the direct current contains ∼ 2 mV rms
noise at 50 Hz and higher harmonics. This noise feeds through to the mixer output.
This problem can be overcome with alternative power supplies. Mains power noise is
eliminated during field operation where batteries are used as the power source.
At present, sensitive readout is only possible at frequencies away from the mains
power harmonics. In the region around the 1032 Hz signal shown in Fig. 3.12b, the
√
noise floor between 1023 Hz and 1047 Hz is 53 nV/ Hz rms. This translates to a
√
displacement noise floor of 5.7 × 10−14 m/ Hz rms, using the calibration coefficient
3.7. APPLYING THE CAPACITIVE READOUT TO A DSMG 75
given in Eq. 3.34 for this setup.

The phase noise specifications of the oscillator are shown in Table 3.1. Phase
noise of −158 dBc/Hz at 1000 Hz on a carrier signal of 1.6 V peak is calculated to be
√
20 nV/ Hz rms. This theoretical noise level is very close to the mixer output noise
√
level of 53 nV/ Hz rms. There is more phase noise at low frequencies. Fig. 3.13
shows that the mixer output has a 1/f noise corner of 200 Hz.
Frequency Phase noise

(Hz) (dBc/Hz)
10 -120
100 -150
1000 -158
10000 -158
Table 3.1: Phase noise of the 10 MHz quartz oscillator as a function of frequency
3.7 Applying the capacitive readout to a magnetic

gradiometer
3.7.1 Background
The Direct String Magnetic Gradiometer (DSMG) uses a single aluminium “string” as
the sensing element[7]. The string is held under tension with its second violin mode
at f0 ≈ 800 Hz. Alternating current with frequency f is pumped along the string,
setting it into motion with a magnetic force per unit length:
∂F
∂z = is [ez × B(z)] sin(2πf t) (3.36)
(z, t)
length, B is the magnetic field in the x-y plane, is sin(2πf t) is the AC drive current
with amplitude is = 0.3 A and t is time. We can tune the AC current frequency to
the string’s second harmonic frequency to create resonance f = f0 . The string is a
stretched thin flat ribbon with length l = 0.25 m, width w = 0.125 mm and thickness
0.025 mm so the resonant vibration only occurs in one direction.
The second harmonic of mechanical vibration of the string can only be driven by
force gradients. The magnetic force in Eq. 3.36 can be substituted into the time-
dependent string equation[8]. We find that the displacement x is proportional to the
magnetic gradient Byz :
is Ql2 − τt

x(t) = 1 − e cos(ω0 t)Byz (3.37)
πmω02
where Q = πτ f is the mechanical quality factor of the string, ω0 = 2πf0 is the
angular resonant frequency, τ is the string relaxation time, l = 0.25 m is the length
of the string and m = 2 × 10−6 kg is the mass of the string.
The displacement of the string along the x-axis is measured with two inductive
pickup coils, connected in differential mode. The pickup coils can measure displace-
√
ment with a rms accuracy of 10−11 m/ Hz corresponding to an equivalent magnetic
√
gradient of 4 × 10−10 T/m/ Hz.
3.7.2 Thermal noise
The string will experience stochastic motion caused by thermal noise. The spectrum
of thermal noise is given by Saulson[9], and with a slight change in notation (k → mω02
and φ → 1/Q), the noise spectrum is:
2 8kT mω02 Q1
xth (f ) =
ω (mw02 − mω 2 )2 + m2 w04 Q12
4kT τ
≈ (3.38)
mω02 [τ 2 (ω − ω0 )2 + 1]
where k = 1.4×10−23 J/K is Boltzmann’s constant and T = 300 K is absolute tem-

perature. Whilst we aim to tune f → f0 , the magnetic force depends on cos(ω0 t)Byz ,
according to Eq. 3.37. This means that the magnetic gradient signal is carried in two
sidebands with frequencies:
f = f0 + fB
f = f0 − fB (3.39)
3.7. APPLYING THE CAPACITIVE READOUT TO A DSMG 77
where fB is the frequency of the magnetic gradient Byz . The spectral distribution
of thermal noise can be presented in the form:

4kT τ
xth (f ) = (3.40)
mω02 [τ 2 ωB2 + 1]
The string has a relaxation time of τ = 0.1 s so the thermal noise level is x(f ) =
√
6 × 10−12 m/ Hz rms in the band 0 Hz < fB < 1 Hz.
3.7.3 Rationale for the new readout
A recent paper by the authors[10] presents a plan to lower the noise floor of the
gradiomater. From Eq. 3.37, we can increase the sensitivity of the device to magnetic
gradients by increasing the current. However, if the current is too high then the string
will overheat. For this reason, it was proposed to replace the string with a wide thin
ribbon that is more efficient at radiating heat. The new design is incorporated in this
paper as a 5 mm wide ribbon, 10 μm thick and 0.25 m long, with its second violin
mode of oscillation at f0 ≈ 330 Hz.
Under the new plan, the ribbon relaxation time will be τ = 75 s. For a fB = 1 Hz
√
signal, the thermal noise level will be xth = 1.4 × 10−13 m/ Hz rms, according to Eq.
3.40. Measuring such small signals is beyond the capability of the present inductive
readout system. An inductive readout is more suitable for thin round strings whereas
capacitive readout is more suitable for wide ribbons. This is the reason that we have
designed a new capacative readout.
√
At 330 Hz, the capacitive readout’s noise floor is approximately 10−13 m/ Hz,
which will be about same level as the thermal noise under the new plan. By sub-
stituting these noises into Eq. 3.37, we find that the best possible magnetic gradient
√
sensitivity with the new readout is ΔByz ≈ 10−11 T/m/ Hz.
3.7.4 Common mode rejection ratio
The DSMG is designed to measure magnetic gradients (differential mode) and to reject
uniform magnetic fields (common mode). The optimum common mode rejection ratio
of the entire sensor K has been calculated previously to be[11]:
9Q2m k fB
K = when << Qm (3.41)
32 f0
3k f0 fB
= · when >> Qm (3.42)
16 fB f0
where Qm is the mechanical quality factor of the string, k is the common mode
rejection ratio of the inductive pickup coils, f0 is the resonant frequency of the second
violin mode of the string and fB is the frequency of the magnetic gradient signal.
The capacitive readout has a poor common mode rejection ratio of k = 30 dB
(see section 3.4). A typically moderate mechanical quality factor of Qm = 200[11],
should provide an additional 80 dB common mode attenuation according to Eq. 3.41.
The common mode rejection ratio of the entire sensor would then be K = 110 dB.
This is sufficient to attenuate the uniform magnetic field of the Earth from Bearth ≈
50000 nT down to a differential signal of 0.2 nT. A magnetic field difference of 0.2 nT
is equivalent to a magnetic gradient of 0.8 nT/m across a 0.25 m baseline.
3.8 Conclusion
We have presented a capacitive readout with a very high transfer coefficient of

106 V/m, made possible by a high electronic quality factor and a small distance of
60 μm between the capacitive plates and the ribbon. The linear region of the readout
is only 360 nm. As part of a magnetic gradiometer, magnetic force feedback can be
used to keep the readout at the zero point. We have shown that coaxial cables up
to 3 m in length can be used effectively in the capacitive readout. Most of magnetic
gradiometer’s hardware can now be placed 3 m away from the sensing element. The
amount of false magnetic signals coming from the magnetically susceptible hardware
(also known as heading error in geophysics) is therefore reduced to negligible levels.
The readout is able to precisely measure the differential displacement of a low mass
√
ribbon. Laboratory results show that the noise floor is below 10−13 m/ Hz over a
wide band. In principle this should be sufficient to measure magnetic gradients with
√
sensitivity 10−11 T/m/ Hz in high vacuum environments such as space applications.
3.8. CONCLUSION 79
Acknowledgements
Work on the DSMG project is funded in part by a linkage grant from the Australian
Research Council.
3.9 Postscript: Space applications

The capacative readout described in the preceding paper will be used in a proposed
DSMG for space applications (described in chapter 6). The readout was tested on
a 5 mm wide, 10 μm thick, and 0.25 m long ribbon. The capacitive readout’s noise
√
floor is approximately 10−13 m/ Hz. Assuming a ribbon relaxation time of 75 s, the
paper states that the best possible magnetic gradient sensitivity with the new readout
√
is ΔByz ≈ 10−11 T/m/ Hz. If a 1 m long ribbon was used, then a magnetic gradient
√
sensitivity of ΔByz ∼ 10−13 T/m/ Hz should be possible with the new readout.
However, the ribbon used in the preceding paper has a very low mechanical quality
factor, Q ≈ 500 ± 100. The Q factor was measured in a high vacuum of 0.016 Pa so
gas damping was negligible. The likely reason for the low Q factor is the material;
the ribbon was made from off the shelf aluminium 8011-O foil which has a very low
yield strength of 50 MPa[12]. Using a stronger alloy such as aluminium 6061-T6511
should increase the Q significantly. Fig. 5.4 in Chapter 5 shows that a aluminium
6061-T6511 ribbon with width 125 μm and thickness 25 μm, can have a Q factor of
∼ 20000 and relaxation time of ∼ 70 s.
BIBLIOGRAPHY 81
Bibliography
[1] M Bassan, P Carelli, V. Fafone, Y. Minenkov, G.V Pallottino, A. Rocchi1, F.
Sanjust, G. Torrioli, A new capacitive read-out for EXPLORER and NAUTILUS,
Journal of Physics: Conference Series 32 (2006) 89-93.
[2] M.E. Tobar, E.N. Ivanov, D.G. Blair, Parametric transducers for the advanced
cryogenic resonant-mass gravitational wave detectors, General Relativity and
Gravitation 32 (2000) 1799-1821.
[3] V.B. Braginsky, A.B Manukin, Measurement of weak forces in physics experi-
ments, Chicago University Press, Chicago, 1977, p. 11.
[4] A.V. Veryaskin, Patent EP20000929523, Measurement of magnetic fields using

a string fixed at both ends, 2000.
[5] P. Horowitz, W. Hill, The art of electronics, 2nd Edition, Cambridge University
Press, Cambridge, 1989, p. 11.
[6] B.H. Lee, Phd Thesis, University of Western Australia, 2007, p. 132,
http://www.gravity.uwa.edu.au/docs/PhDThesis/Ben_Thesis.pdf
[7] A.V. Veryaskin, Magnetic gradiometry: a new method for magnetic gradient
measurements, Sensors and Actuators A 91 (2001) 233-235.
[8] D.R. Bland, Vibrating strings: an introduction to the wave equation, 1st Edition,
Routledge and Kegan Paul, London, 1960, p. 41.
[9] P.R. Saulson, Thermal noise in mechanical experiments, Physics Review D 42

(1990) 2437-2445.
[10] A. Sunderland, A.V. Veryaskin, W. McRae, L. Ju, D.G. Blair, Direct string
magnetic gradiometer for space applications, Sensors and Actuators A 147 (2008)
529Ű535.
[11] A.V. Veryaskin, Theory of operation of direct string magnetic gradiometer with
proportional and integral feedback, International Journal of Applied Electromag-
netics and Mechanics, accepted for publication.
[12] Z.P. Xing, S.B. Kang, H.W. Kim, Softening behaviour of 8011 alloy produced by
accumulative roll bonding process, Scripta Materialia 45 (2001) 597-604.
Chapter 4
Heading error
4.1 Preface
Heading error is a systematic error that depends on the direction that a magnetic
gradiometer is pointing relative to the direction of the Earth’s magnetic field. The
airborne trials in section 1.10.4 showed a peak to peak heading error of ∼ 1000 nT/m
for the DSMG. It was hypothesised that components with induced magnetisation were
aligning in the Earth’s field and thus were causing directional dependant error.
A heading error of this magnitude is unacceptable so the aluminium sensor frame
(shown in figure 1.7) was replaced with a Torlon 4301 frame. A plastic such as Torlon
should be less magnetic than a metal. With the new Torlon frame, the heading error
from a 360◦ rotation was reduced to 225 ± 5 nT/m peak to peak, as shown in Fig 4.1.
150
Heading error
100
Magnetic Gradient (nT/m)
50 Raw data
Fitted
0
-50
-100
-150
0 90 180 270 360
Azimuth
Figure 4.1: Heading error tests are completed by rotating the sensor through 360◦ in a
magnetically benign area to check for self-induced gradient response.
83
84 Chapter 4. HEADING ERROR
The following paper describes an analysis of the magnetic susceptibility of different

materials that might be used in the DSMG, for the purposes of reducing heading error.
The paper has been published in Smart Material and Structures. My contribution
to this paper amounts to 95% of the experimental work, and 80% of the manuscript
preparation.
Smart Materials and Structures 18 (2009) 095038
Low magnetic susceptible materials

and applications to magnetic gradiometry
Andrew Sunderland1 , Li Ju1 , David G Blair1 , Wayne McRae1,2 , Howard Golden1,2

Abstract
Magnetic gradiometers are powerful tools for mineral exploration. To
avoid the problem of heading errors, gradiometers must be made from low
susceptibility materials. We compare possible low susceptible materials
for gradiometers and strategies for minimisation of magnetic contamina-
tion. In particular we present the favourable magnetic properties of the
advanced engineering plastic Torlon.
4.2 Introduction
Newly introduced magnetic gradiometers could revolutionise airborne magnetic sur-
veys. The gradient of the magnetic field can be used to interpret geological targets
at depths up to several kilometers[1]. However, measurements would be biased by
the presence of any magnetic material inside the gradiometer itself. Whereas mag-
netic fields from far field dipole sources scale as the inverse third power of distance,
magnetic gradients scale as the inverse fourth power and are particularly sensitive to
close objects.
There are many research groups developing magnetic gradiometers with different
technologies: GETMAG[2] measures the magnetic gradient using a pair of SQuIDs,
JESSY STAR[3] uses one SQuID that reads from two coils, Merayo et al.[4] use a
pair of fluxgate magnetometers, and various survey companies create a gradiometer
by placing total field magnetometers on both wing tips of an aeroplane[5; 6]. The
gradient is obtained by measuring the magnetic field at two locations and taking the
difference.
A Direct String Magnetic Gradiometer (DSMG) has been developed that employs
a single aluminium "string" as the sensing element[7]. Many DSMG instrument com-
ponents are located near the long thin string where even small levels of magnetic
contamination can lead to false signatures much larger than the signal from a distant
geological target. These false signatures will manifest themselves as heading error.
McFee et al.[8] define heading error as a systematic error that depends on the
direction that the magnetic gradiometer is pointing relative to the direction of the
Earth’s magnetic field.
The magnetic field of the Earth induces material in the gradiometer to magnetise
in the same orientation as the Earth’s field. If the gradiometer is placed onboard a
moving vehicle then those parts with induced magnetisation will produce a magnetic
gradient which varies whenever the vehicle changes direction. This variable error is a
form of heading error.
For small magnetic fields, the induced magnetisation μ0 M of a material rises lin-
early with the applied magnetic field μ0 H = BEarth . When the vehicle carrying the
gradiometer changes direction, the direction of μ0 H will rotate in the magnetometer’s
reference frame but the magnitude of μ0 |H| will stay constant. The induced magneti-
∂M
sation of each part can be predicted from the magnetic susceptibility χ = ∂H
and
the heading of the vehicle. Using parts with low magnetic susceptibility will keep the
total amount of heading error small.
Pure iron, nickel or cobalt materials are an obvious source of magnetic contami-
nation. Other nonferrous metals of commercial grade are often less than 100% pure
and usually contain iron as an impurity. Care must be taken when metal parts are
used in a magnetic gradiometer because the susceptibility varies from alloy to alloy.
For these reasons, we report on the magnetic susceptibility of some common ma-
terials at room temperature. Section 4.3 describes the DSMG sensor and why it
is particulary vulnerable to heading error. In section 4.4 we examine the materials
used in this paper and why Torlon is important. Section 4.5 investigates the mag-
netic contamination caused by machining parts. We measure the susceptibility of
each material before and after surface treatment. The magnetic properties of an ad-
vanced engineering plastic Torlon are presented in detail (Torlon is used throughout
4.3. DIRECT STRING MAGNETIC GRADIOMETER 87
the DSMG). Section 4.6 presents a study of the measurement uncertainties in this
work. In conclusion, we aim to determine which materials are more suitable for use
in the DMSG. This work should also be of interest to all researchers working in the
design and development of sensitive magnetic instrumentation because these devices
require low magnetic susceptibility materials.
4.3 Direct String Magnetic Gradiometer

The DSMG detects magnetic gradients with only one sensing element, a long thin
aluminium string. The string is clamped at both ends under tension with its second
harmonic frequency approximately 800 Hz. An AC current sets the string into a
resonant motion due to the magnetic force per unit length:
∂f
= is [ez × B] sin(ωt) (4.1)
∂z
length, is is the amplitude of the AC drive current, ω is the string’s drive angular
frequency, t is time and B is the magnetic induction vector.
We can tune the AC current frequency to the string’s second harmonic frequency
to create resonance. The string is a stretched thin flat ribbon so the resonant vibration
only occurs in one direction. We can model the ribbon as a one-dimensional harmonic
oscillator with an infinite number of resonant modes[9]:

d 2 d 2 2 n 2l dBy is
Xn + X n + ω n Xn = (1 − (−1) ) By − (−1)
n
sin(ωt) (4.2)
dt2 τ dt πn πn dz η
From the above equation, the magnetic gradient term dBy /dz of the driving force
couples to all resonant modes, while the conventional magnetic field term By is only
coupled to the odd ones. Figure 4.2 shows the first two resonant modes of the DSMG.
This method of detection is insensitive to uniform magnetic fields. The ’S’ shape of
the second harmonic couples strongly to gradients but weakly to uniform fields. The
resonance at the second harmonic will amplify magnetic gradient signals whilst the
frequency is well off the first harmonic frequency that couples to the uniform magnetic
Resonant modes of magnetic gradiometer

Torlon Frame
1st violin mode
Torlon Frame
2nd violin mode
Figure 4.2: The top diagram shows the DSMG frame with the string clamped rigidly at
each end. An AC current is running along the string such that its frequency is the same as
the 1st violin mode of the string. Resonance will occur. The ’C’ mode string vibrations will
be sensitive to uniform magnetic fields (the magnetic field is shown going into the page). In
the bottom diagram, the current has a frequency equal to the 2nd violin mode of the string.
The ’S’ mode string vibrations are sensitive to magnetic gradients.
field. Also, the mechanical displacement detection is designed to preferentially detect

the second harmonic oscillations and suppress the fundamental mode oscillations.
The high Q factor resonant structure of the string leads to a high common mode
rejection that can be achieved without stringent balancing requirements[10].
Other magnetic gradiometers should not detect uniform magnetic fields however
there is always a small but finite imbalance in the two sensing elements. Erroneous
detection of uniform fields is another form of heading error. Careful calibration of
the two sensing elements can be used to cancel out the heading error from magnetic
parts. The DSMG is a single element device so this remedy is not available. The
DSMG is particularly vulnerable to heading error.
4.4 Choice of low susceptibility materials
In this work we report on the laboratory materials: Torlon 4301, G10 Epoxy, machin-
able ceramic(Macor), Teflon (PTFE) and Polyethylene terephthalate(PET). These
4.4. CHOICE OF LOW SUSCEPTIBILITY MATERIALS 89
Nominal concentrations Measured concentrations

Material (ppm) (ppm)
Fe Ni Co Fe Ni Co
Leaded Brass <3500 <4000 <5000 680. 180. 0.7
Yellow Brass <5000 <5000 <5000 1700. 1500. 7.5
Aluminium 6061 <7000 <5000 <5000 1700. 45. 0.1
OFHC Copper 2. <1. <1. <5. 1.3 <0.1
Torlon 4301:
- Quadrant EPP 820. 4. <1.
- Ensinger 720. 5. <1.
G10 26. <0.2 <0.1
Macor 700. 0. 590.
PTFE <0.1 <0.1 <0.1 5. <0.2 <0.1
PET <0.1 <0.1 <0.1 <5. <0.2 .4
Table 4.1: Nominal concentrations were taken from material datasheets. The concentration
of ferromagnetic elements in each sample were measured by the Chemistry Centre of Western
Australia using a Inductively Coupled Plasma - Atomic Emission Spectrometer (ICP-AES).
are all nonmagnetic materials that are suitable for magnetic detectors. The material
Torlon 4301 is used throughout the latest DSMG. Macor, G10 and PTFE have been
used in previous DSMGs.
The DSMG frame is made of Torlon 4301 (shown in figure 4.2) because Tor-
lon’s expansion coefficient is approximately the same as Aluminium 6061. An in-
crease/decrease in temperature will loosen/stretch the string. In turn, the resonant
frequency will shift and the level of 1/f noise will rise. This can be remedied by
matching the thermal expansion coefficient of the frame to the aluminium string. A
literature search by the authors found no magnetic susceptibility data for Torlon.
Most parts in magnetic instruments are made from non-magnetic plastics. Metals
are potentially magnetic so they should be avoided where possible. However, some
metal parts are required for their strength, electrical conductivity and thermal prop-
erties. In this work, we investigated aluminium 6061 because the DSMG’s string is
made of this material. We investigated oxygen free highly conductive copper (OFHC)
because it is an excellent non-magnetic electrical conductor. We also investigated two
brass alloys: yellow brass C27400 and leaded brass C33200. Although brass is known
to be undependable in its magnetic properties[11], brass parts are still used in the
construction of many magnetometers, eg [12; 13; 14].
Mass (g)
As After 10 min After 70 min After 24 hour
Metal machined acid wash acid wash acid wash
Leaded Brass 24.68 24.68 24.68 24.67
Aluminium 8.05 8.05 8.04 7.46
Table 4.2: The amount of mass lost by the rectangular metal samples in acid increases
with longer immersion times, uncertainty is ± 0.01g.
We measured the susceptibility of all these materials with a view to reducing

the heading error of the DSMG. Table 6.2 shows the concentration of ferromagnetic
elements in each of the materials. The presence of iron impurities may increase the
permeability of some of the materials. The results below show that this is the case,
and also that surface contamination is a critical issue.
4.5 Comparison of materials and contamination re-

duction
4.5.1 Surface contamination
Ideally, the induced magnetisation of a material is proportional to volume. However,

there can also be a surface contamination from machining. Steel cutting tools will
deposit iron onto the surface of the samples. This risk has been recognised in previous
work[15; 16; 17; 18], but has not been studied quantitatively. Here we do explore this
problem quantitatively, before addressing the intrinsic properties of the samples.
Soft tools wear out quicker so they should emit more iron particles. We hypothesise
that soft tools with high magnetic permeability should produce the most surface
magnetic contamination. To investigate our hypothesis, 14 rectangular prism samples
of 7 different materials were cut to dimensions 12 mm x 16 mm x 16 mm with a high
carbon steel sawblade. Another 12 cylindrical samples of 6 different materials were
machined with a high speed steel tool bit on the lathe to make cylinders of diameter
5 mm and length 5 mm. We drilled a 3 mm diameter hole into each cylinder and
tapped the holes with a M3 tool steel tap. By testing both prisms and cylinders, we
can compared the magnetic contamination from two different cutting motions.
4.5. COMPARISON OF MATERIALS AND CONTAMINATION REDUCTION91
5
Susceptibility, Ȥ (10í )
As machined 10 min 70 min 24 hour
Metal Samples acid wash acid wash acid wash
Prisms
Leaded Brass 2000 ± 300 180 ± 50 110 ± 50
Aluminium 6061 1400 ± 400 250 ± 20 230 ± 20 230 ± 20
Cylinders
Leaded Brass 21 ± 5 9 ±3
Yellow Brass 20000 ± 3000 20000 ± 3000
Aluminium 6061 460 ± 100 17 ± 1
OFHC Copper 22 ± 4 2±2
Table 4.3: The initial magnetic volume susceptibility of metal samples decreases after an
acid wash, although most of the reduction occurs in the first 10 minutes of the acid wash. The
prism samples were measured with a Bartington MS2b susceptibility meter. The cylindrical
samples were measured using an Aerosonic 3001 Vibrating Sample Magnetometer (VSM).
To remove any surface magnetism, the metal prisms were immersed in 3% hy-
drochloric acid for durations of 10 minutes, 70 minutes, and 24 hours. Corrosion
of the samples prevented us from using a stronger acid. Table 4.2 shows that the
aluminium sample lost 7% of its mass after being immersed in acid for 24 hours.
Table 4.3 presents the susceptibility of the metal samples before and after an acid
wash. The DSMG operates at room temperature so measurements were taken at
room temperature. All susceptibility values in this paper are volume susceptibility in
SI (MKS) units. A 24 hour acid wash reduced the susceptibility of the metal prisms
by an order of magnitude, although most of this reduction occurred in the first ten
minutes. The magnetic susceptibility that remains after 24 hours could be produced
by larger sawblade fragments that a weak acid can not remove. We conclude that a
10 minute acid wash is an optimal compromise between contamination reduction and
sample corrosion.
Washing the yellow brass sample in acid had no effect because the intrinsic sus-
ceptibility is so high that surface magnetic contamination is irrelevant.
The unwashed aluminium and leaded brass cylinders have susceptibilities that are
much lower than their rectangular prism counterparts. This result can be explained
by table 4.4, the relative permeability of the sawblade used to machine the rectangular
prisms is much higher than the relative permeability of lathe tool bit and M3 tap used
Tool Materal Relative Hardness

Permeability Rockwell C
Lathe tool bit High speed steel 1.4 64 to 65
M3 Tap Tool steel 3.7 61 to 62
Hacksaw blade High carbon steel 11. 59
Table 4.4: The permeability measurements were made with a Bartington MS2b suscepti-
bility meter. The permeability of the tools varied ±30% when rotated. Hardness values are
from a book by ASM International[33].
5
Susceptibility, Ȥ (10í )
Plastic and ceramic samples As machined 24 hour acid wash
Prisms
Torlon 4301 (Quadrant EPP) í1.7 ± 0.4 í1.9 ± 0.4
G10 0.8 ± 0.2 í0.1 ± 0.1
Macor í0.6 ± 0.2 í0.8 ± 0.2
PTFE í0.4 ± 0.1 í0.6 ± 0.2
PET í0.5 ± 0.1 í0.7 ± 0.2
Cylinders
Torlon 4301 (Quadrant EPP) í1.1 ± 0.3 í2.1 ± 0.1
Torlon 4301 (Ensinger) 4.5 ± 0.5 0.57± 0.05
Table 4.5: Initial magnetic volume susceptibility of plastic and ceramic samples before
and after an acid wash. The prism samples were measured with a ZH Instruments SM-30
susceptibility meter. The cylindrical torlon samples were measured with a Quantum Design
MPSM-7 SQuID.
to machine the cylinders. The sawblade is also softer than the other tools so it should
produce greater magnetic contamination according to our hypothesis.
We also investigated the magnetic contamination caused by steel scissors. After
cutting 50 mm long, 0.125 mm diameter enameled copper wire with steel scissors,
the volume susceptibility was 2.9 × 10−5 ± 10−6 . When the same wire was cut with
a titanium side cutters, the volume susceptibility was only 5 × 10−6 ± 10−6 .
Table 4.5 shows the magnetic susceptibility of the plastic and ceramic samples
before and after a 24 hours acid wash. In all cases the susceptibility was very low.
The plastic and ceramic samples have much lower magnetic surface contamination
than the metal samples. This result can be explained by the following background
theory on machining.
In Atkin’s and Liu’s[19] model of ductile machining, the required cutting force
will be greater for materials with a high toughness to strength ratio. Both macor
and G10 have very low toughness and relatively high strength so these materials are
not cut but ground during machining. The brittle material will break in segmented
chips[20] with an exceptional surface finish that is unlikely to contain impurities.
The three plastics Torlon 4301, PTFE and PET are soft materials with low co-
efficients of friction so the cutting force will be low according to Atkin’s and Liu’s
model[19]. Both surface damage and tool wear will be minimal so contamination
should be low.
Special precautions must be taken when machining OFHC copper in regards to
cleanliness because of the great toughness of the material. Surface impurities may be
plowed under and entrapped during spinning operations on copper[21].
Work hardening of the surface during machining could also contribute to the
change in susceptibility. Hutchison and Reekie[22] found that this was the case in
copper and aluminium samples and that normal susceptibility could be restored by
washing the samples in a weak nitric acid.
4.5.2 Intrinsic susceptibility of materials
The susceptibility of a sample after an acid wash should reflect the intrinsic suscepti-
bility of the material. Table 4.6 presents the intrinsic susceptibility of every material
in this paper. For aluminium and leaded brass, we measured the susceptibility of both
cylinders and prisms, but only report the intrinsic susceptibility of the cylinders. The
cylinder measurements are more reliable because they were exposed to less magnetic
contamination. We were able to measure the susceptibility of Torlon EPP accurately
for both prisms and cylinders, so the table reports an average of these two values.
We find that the non-metals are the least magnetic materials, all having magnetic
susceptibilities on the order of 10−5. G10 has the closest susceptibility to zero. In
any case, the heading error caused by the plastic and ceramic parts will be negligible.
Our results for macor and PTFE are in good agreement with the results of Keyser
and Jefferts[23] whilst our result for PET agrees with Tanimoto et al.[24]. Our mea-
surement of G10’s susceptibility is of the same order as Bossi et al’s[25] who found
traces of ferromagnetic impurities in commercial G10.
Material Intrinsic susceptibility Instrument

Ȥ (10í5) used
Metals
Leaded Brass 9. ± 3 VSM
Yellow Brass 20000. ± 3000 VSM
Aluminium 6061 17. ± 1 VSM
OFHC Copper 2. ± 2 VSM
Plastics
Torlon 4301 (Quadrant EPP) í2.0 ± 0.1 SQuID/SM-30
Torlon 4301 (Ensinger) 0.57 ± 0.05 SQuID
G10 í0.1 ± 0.1 SM-30
Macor í0.8 ± 0.2 SM-30
PTFE í0.6 ± 0.2 SM-30
PET í0.7 ± 0.2 SM-30
DSMG components
Solder (60% Tin, 40% Lead) 0.05 ± 0.02 SQuID
Aluminised mylar tape í2. ± 1 VSM
Printed circuit board 0. ± 1 VSM
Enameled copper wire 0.5 ± 0.1 SQuID
Twist pair wire (Mogami 2490) í0.2 ± 0.1 SQuID
Solid core wire (Brand Rex GT951007) 0.22 ± 0.05 SQuID
Table 4.6: We present the intrinsic susceptibility of each material examined in this paper.
All machined samples have been washed in acid. The right hand column notes the instrument
used to make the susceptibility measurement.
The susceptibility of OFHC copper is very low so it can be used for magnetic
gradiometer parts that need to conduct electricity. There is however, a discrepancy
between the magnetic susceptibility of OFHC copper χ ≈ (2 ± 2) × 10−5 measured
using the VSM and the textbook value of χ = −1 × 10−5 [26]. This could be due to
an imperfect acid wash or the limited precision of the VSM measurement.
The leaded brass, yellow brass and aluminium 6061 samples are intrinsically more
magnetic than the other materials. These three metals are unsuitable for use in
a magnetic gradiometer. There is a discrepancy between intrinsic susceptibility of
aluminium 6061 χ ≈ (17 ± 1) × 10−5 measured using the VSM and the susceptibility
value reported by Keyser and Jefferts[23] χ = 1.9×10−5. This could be due to different
amounts of magnetic contamination between samples or an erroneous response of the
Bartington MS2 to electrical conductivity (for example see Benech and Marmet[27]).
In any case, aluminium 6061 is no longer used in the magnetic gradiometer (except
0
Susceptibility of Brass
10
Yellow Brass
-1
10 Barker et al
Butts et al
Susceptibility, F
ASM
-2 this work
10
Leaded Brass
-3
Fickett et al
10 this work
-4
10
-5
10
10 100 1,000 10,000
Fe concentration, X (ppm)
Figure 4.3: The susceptibility of brass rises rapidly with increasing iron concentration.
The data is compiled from a paper by Barker et al.[32], a paper by Butts et al.[28], a book
by ASM International[26], a paper by Fickett et al.[11] and measurements performed by the
authors on yellow and leaded brass samples. Barker and Butts measured the concentration
of iron in their samples. For the data of ASM and Fickett, the graph plots the nominal
concentration according to the UNS alloy number. The black line is a best fit for yellow
brass at low Fe concentrations using a square law χ ∝ X 2 . The grey line is a best fit for
leaded brass.
for the string) because we wish to avoid any possible source of heading error.
We compare the intrinsic susceptibility of the brass samples with the work of
other researchers in figure 4.3. The graph plots volume magnetic susceptibility vs.
iron concentration for yellow brass (61-67% Cu, 33-39% Zn) and leaded brass (58-65%
Cu, 33-40% Zn, 1-3% Pb). The results clearly show that the magnetic susceptibility
χ of brass depends strongly on iron concentration X, the relationship is somewhere
between χ ∝ X 2 and χ ∝ X 3 for low concentrations of iron. This leads to the general
expectation that sample purity will often lead to variations in susceptibility in other
copper alloys. However, there are other factors that affect the susceptibility of copper
such as heat treatment[28], concentration of the iron impurity in small clumps[29],
cold working and oxygen concentration[30]. The graph presents the susceptibility of
Hysteresis curves of three samples

0.1
Magnetisation, P0M (PT)

0.08
0.06
0.04
0.02
0
-0.02
-0.04
-0.06 Solder
Solid core wire
-0.08 Twisted pair wire
-0.1
-0.02 -0.015 -0.01 -0.005 0 0.005 0.01 0.015 0.02
Applied Field, P0H (T)
Figure 4.4: Magnetisation of three non-magnetic materials that are used in the construction
of the DSMG. The hysteresis loops start at μ0 H = 0.02 T, go down to μ0 H = −0.02 T and
back to μ0 H = 0.02 T. Data taken using a Quantum Design MPSM-7 SQuID. Every time
the hysteresis loops crossed μ0 H = 0, the MPSM’s magnet was heated and reset in order
to ensure there was no vortex trapping (vortex trapping could potentially cause a large H
offset).
brass as cast or as rolled because relying on a heat treatment to lower the susceptibility
of a magnetic gradiometer is not sufficiently robust for all environments.
We also performed susceptibility measurements on a few non-standard wires and
parts that are used in the DSMG, see figure 4.4. The bottom of table 4.6 shows that
all of these materials have negligible magnetic susceptibility.
4.5.3 Torlon
Torlon 4301 is used thoughout the DSMG so we did a thorough examination of its
magnetic properties and compared two different manufacturers: Quadrant EPP and
Ensinger. Figure 4.5 shows the magnetisation μ0 M of the acid washed Torlon cylin-
ders as a function of the applied field μ0 H. The initial magnetic susceptibility of the
Ensinger Torlon sample is positive χ = ∂M
∂H
= 6 × 10−6 . When the applied magnetic
field exceeds μ0 H > 0.1 T, the magnetisation has saturated and the volume suscep-
tibility becomes negative χ = −1.3 × 10−5 . We conclude that Torlon is diamagnetic
Hysteresis curve of Torlon

8
Magnetisation, P M (PT)
P0M (PT)
250
6 0
4 -250
-7 0 7
0
P0H (T)
2
0
-2
-4
Quadrant EPP Torlon
-6 Ensinger Torlon
-8
-0.5 -0.4 -0.3 -0.2 -0.1 -0 0.1 0.2 0.3 0.4 0.5
Applied Field, P0H (T)
Figure 4.5: Magnetisation of Torlon samples from two different manufacturers, taken after
washing in acid for 24 hours. The hysteresis loops start at μ0 H = 0, go up to μ0 H = 7 T
(shown in insert), down to −7 T, and then back to 7 T. Data taken using a Quantum Design
MPSM-7 SQuID.
with some ferromagnetic impurites.
The entire magnetisation curve of Ensinger Torlon exhibits hysteresis. The curve
deviates from a straight line by Δμ0 M = 2.4 μT. In comparison, pure iron saturates
at a magnetisation of μ0 M = 2.15 T[31]. Assuming the ferromagnetic impurities are
entirely Fe, the Ensinger and Quadrant EPP Torlon samples should contain 8 and 6
parts per million (ppm) Fe by mass respectively. Elemental analysis of both Torlon
samples (shown in table 6.2) reveal a much higher concentration of iron, ∼ 800 ppm
Fe. This discrepancy could be due to the size and distribution of the iron particles
affecting their magnetic ordering.
The SQuID results in table 4.6 show that the magnetic susceptibility of Torlon
varies between the manufacturers. The cause of this variance is not clear since both
samples have similar amounts of ferromagnetic impurities. Nevertheless, the initial
magnetic susceptibility of the Quadrant EPP torlon cylindrical samples χ = −2.1 ×
10−5 are in good agreement with the Quadrant EPP rectangular samples χ = −1.9 ×
10−5 .
4.6 Error analysis

For each of the materials in this work, there is two samples per material. The differ-
ence between the two samples is a good estimate of the relative error. There is also
core uncertainty contributions in the susceptibility apparatus that create absolute
error. In this work, we used 4 different instruments to measure volume magnetic
susceptibility:
• Bartington MS2b susceptibilty meter. This instrument was used to mea-
sure the susceptibility of the 3 cm3 metal prims. The MS2b is designed for
10 cm3 samples so the raw measurements were corrected using a formula by
Dearing[34]. The susceptibility precision is ±3 × 10−5 for samples of this size.
However, the metal prisms were highly contaminated and the difference between
the 2 samples of the each material was greater than the error of the instrument
in all cases.
• ZH Instruments SM-30 susceptibility meter. This was used to measure

the susceptibility of the 3 cm3 plastic and ceramic prims. The SM-30 is designed
for much larger samples so the raw measurements were corrected using a formula
derived by Gattacceca et al.[35]. The susceptibility precision of the instrument
is ±10−6 for samples of this size. We could not detect any difference between
the 2 samples of the each material. Gattacceca’s formula has an uncertainty of
±20% and this is the main source of error for this instrument.
• Aerosonic 3001 Vibrating Sample Magnetometer (VSM). This instru-

ment was used to measure the susceptibility of the metal cylinders, mylar tape
and printed circuit board. The background susceptibility signal without any
samples was (1 ± 1) × 10−5 . For the copper and brass samples, there was a
significant difference between the susceptibility of the 2 samples of the same
material. We added the sample error to the background error to get the uncer-
tainty of these measurements.
• Quantum Design MPSM-7 SQuID. This was used to measured the suscep-
tibility of the torlon cylinders and three different types of wire. The suscepti-
bility was calculated by taking a linear fit of 11 evenly spaced magnetisation
4.7. CONCLUSION 99
measurements between μ0 H = −0.01 T and μ0 H = 0.01 T. The magnetisation

precision is Δμ0 M = ±10 nT. The error in the linear fit was added to the ±5%
inaccuracy of the SQuID in order to get the total uncertainty of these mea-
surements. There was also additional uncertainty in the susceptibility of the
unwashed torlon cylinders because of the variation in the susceptibility between
each pair of identical samples.
In order to predict the heading error of a magnetic gradiometer operating in the

Earth’s magnetic field, all of our susceptibility measurements were conducted in as
low magnetic fields as possible. Weak signals make these measurements less precise
than the results of other researchers. We could improve our precision by taking
measurements in stronger magnetic fields and then extrapolating down to small fields,
however extrapolating can be inaccurate because of hysteresis. The hysteresis loops
in figure 4.4 show that hysteresis is a problem in commercial materials, even for
materials that are only weakly magnetic.
The SQuID is the most accurate instrument out of the 4. However, the suscep-
tibility meters have the advantage of taking measurements using very low applied
fields, μ0 H << 0.0001 T. The MS2b and VSM have relatively low precision but
they are sufficient to measure the susceptibility of the contaminated metal samples
with better than 10% accuracy. The ZH Instruments SM-30 susceptibility meter is
extremely precise for measuring the susceptibility of large samples such as the plastic
prisms. However, the SM-30 is not suitable for measuring metal samples which may
be conductive.
The VSM utilised some custom built sample holders. We found that it had a
small background susceptibility signal of (1 ± 1) × 10−5 . The other instruments were
standard commercial devices and thus any background signals were either calibrated
for or below the limits of detectibility.
4.7 Conclusion
We have investigated materials for use in high sensitivity magnetic gradiometers. We

have shown that the material Torlon, which is a practical substitute for aluminium,
has advantageous magnetic properties. The intrinsic susceptibility of Torlon 4301

varies between χ = 5.7 × 10−6 and χ = −2.1 × 10−5 , depending on the manufacturer.
Other non-metals were shown to have low magnetic susceptibility χ ∼ 10−5 and
are suitable for use in sensitive magnetic instruments. OFHC copper also has low
susceptibility, it can be used to make parts where the strength or conductivity of a
metal is required.
Surface contamination had a large impact on the susceptibility of metal samples.
Aluminium 6061 and leaded brass both have intrinsic susceptibilities on the order of
χ ∼ 10−4 . Cutting the same materials with steel tools increased their susceptibility.
Cutting with saw blades increased the susceptibility approximately one hundred fold
to χ ∼ 10−2 , while machining with a lathe caused lesser contamination.
We found that washing metal parts in 3% hydrochloric acid for 10 minutes to
be the optimal compromise between contamination reduction and sample corrosion.
Washing the torlon cylinders in acid reduced the variation in susceptibility between
samples. Washing the other plastic and ceramic samples in acid had negligible effect
on their susceptibility which was already low.
Acknowledgements
It is hoped that these results will prove useful in the design and construction of
magnetic gradiometers. The authors would like to thank Dr. Alexey Veryaskin of
Gravitec Instruments for many useful discussions and suggestions, A/Prof. Tim St
Pierre of the BioMagnetics and Iron BioMineralisation group at UWA for the use
of the wet acid laboratory, Dr. Robert Woodward and Mr. Matt Carroll of the
Biomagnetics Group at UWA for use of a MPSM-7 SQuID, Prof. Li of the Tectonic
Special Research Center at UWA for the use of a MS2b susceptibility meter, Mr.
Barry Price of the Chemistry Centre of WA for elemental analysis and Mr. Mads
Toft of Alpha Geoscience for the use of a SM-30 susceptibility meter. Work on the
DSMG project is funded in part by a linkage grant from the Australian Research
Council.
4.8. POSTSCRIPT: HEADING ERROR CALCULATION 101
4.8 Postscript: Heading error calculation

Here I describe a magnetic model of the DSMG which is used to estimate the amount
of heading error. The model utilises the susceptibility measurements in the preced-
ing paper. Most plastic parts inside the DSMG are made from torlon because of
its suitable thermal and mechanical properties. All electrical wires are made of oxy-
gen free highly conductive copper. Most other metallic parts are aluminised mylar,
phosphorous bronze and red brass.
4.8.1 The frame
The sensing element is a thin string 250 mm long, 0.125 mm wide and 0.025 mm
thick. The string is enclosed in a torlon frame of dimensions 280 mm by 30 mm by
30 mm. Fig. 4.6 shows a close up view of the cross section of the string. At nearest
approach, copper wiring and sections of the frame are only 0.4 mm away from the
string. The close proximity makes the the sensor vulnerable to heading error from
magnetic parts.
The magnetic field along at the string can be estimated by integrating the magnetic
contribution of each part of the sensor. During integration, we divided the frame into
infinitesimally small cubes. The Earth’s magnetic field is simulated by a uniform field
of 50000 nT in the y direction. The uniform field excites a magnetisation μ0 M =
Bearth χ in each cube that is proportional to its magnetic susceptibility χ.
Each small cube can be approximated to a perfect dipole. The contribution of
this cube to the magnetic field at position r is then[36]:
μ0 [3(m · ) − m 2 ]
dBdip (r) = (4.3)
4π 5
where m = M(r )dV is the magnetic moment of the cube at position r , dV is the
volume of the cube and = r − r . The magnetic field is an integration over all cubes:

μ0 [3(M(r ) · ) − M(r ) 2 ]
Bdip (r) = dV (4.4)
volume 4π 5
The above integration in three dimensions is computationally complex. Before
computation, the expression must be reduced analytically. The magnetic field in the
6
0.8 mm
4
String
0
−2
X−4 Torlon Copper
−6
−10 −8
Y −6 −4 −2 0 2 4 6 8 10
Figure 4.6: Cross section view of the string and nearby magnetically susceptible parts.
The arrows show the X and Y components of the magnetic field produced by these nearby
parts. The longest arrow shows a magnetic field strength of 0.94 nT and the shortest arrow
shows 0.22 nT.
space outside of a magnetised object is the same as would be produced by a surface

current K = M × n̂ on the boundary[36]. n̂ is the unit vector that is normal to the
surface. The magnetic field can then be calculated using Biot-Savart law:
"
μ0 K×
B(r) = dA (4.5)
4π surf ace 3
The surface current can be broken into lines. It is a simple problem to integrate the
Biot-Savart equation over the length of a wire which is carrying a steady current[36]:
μ0 I
Bwire (r) = [sin(θ1 ) − sin(θ1 )]ø̂ (4.6)
4π
where I is the current and ø̂ is the unit azimuthal vector around the wire. Inte-
grating over all lines to make a surface gives the following in coordinate free form:
" # #
μ0 K 2 (K × 1 ) #K · 1 K · 2 #
B(r) = · ## − # dl (4.7)
4π |K × 1 |2 K 1 K 2 #
where 1 = r − r1 is the vector at the start of each line and 2 = r − r2 is the
vector at the end of each line. The integral above can be reduced into a primitive
expression containing an elliptic integral of the first kind. The elliptic integral cannot
be evaluated explicitly but equation 4.7 can be efficiently integrated numerically.
Fig. 4.6 shows the magnetic field in the vicinity of the string as calculated nu-
merically. The vector plot shows the inhomogeneous magnetic field produced by:
the magnetised torlon frame, aluminised mylar coating the frame and parallel copper
wires. The longest arrow in the plot shows a magnetic field strength of 0.94 nT and
the shortest arrow shows 0.22 nT. The z component of the magnetic field is not shown
in the plot but it is less than 0.005 nT at all locations.
The maximal sensitivity of the DSMG occurs at the one quarter and three quarter
points along the ribbon. These two points are 0.125 m apart and magnetic field
errors can be converted into magnetic gradient error by dividing by this baseline.
The magnetic field component By might vary as much as 0.6 nT, giving a maximum
heading error of 5 nT/m.
4.8.2 Parts near the one quarter and three quarter points
along the string
In addition to the magnetic contribution from the highly symmetrical frame mod-
elled above, the sensor contains several hundred small parts. Some of these parts
are placed asymmetrically about the string. Some parts may also have anisotropic
magnetisation. For this reason the heading error generated by each small part has
been calculated individually.
There is a significant number of parts clustered around the one quarter and three
quarter points along the string. Considering parts at location:
r (x , y , z ) = (x , y , L4 ) (4.8)
where L = 0.25 m is the length of the string. (x , y , L4 ) is the three quarter
point of maximum DSMG sensitivity. Small parts can be approximated as perfect
dipoles. The magnetic field in the space surrounding the dipole can be calculated
using equation 4.3. The point in the string closest to the parts is r(x, y, z) = (0, 0, L4 ).
The y component of the magnetic field inside the string near this point is:
μ0 my [2y 2 − x2 − (Δz)2 ]

By (0, 0, L4 + Δz) = 5 (4.9)
4π x2 + y 2 + (Δz)2
The DSMG system measures the gradient by integrating the force along the string.
∂
For far field objects, the gradient is simply Byz = B .
∂z y
A far field dipole that
produces a magnetic field scaling as the inverse third power of distance will produce
a magnetic gradient scaling as the inverse fourth power of distance. However for
near field objects, the exact formula must be used with the force calculated at each
position along the string. The magnetic gradient measured by the DSMG is[37]:
l
2π 2 2πz
Byz = 2 By (z) sin dz (4.10)
l − 2l l
Trying to evaluate this integral using the magnetic field from equation 4.9 gives a
non-analytic solution involving sine integrals. If the parts are very close to the string,
x << l and y << l, then the string can be treated as an infinite wire. In this case:

2πz
lim sin →1 (4.11)
z→ L
4
l
so equation 4.10 can then be approximated:

2π +∞
Byz ≈ 2 By (z)dz (4.12)
l −∞
μ0 my (y 2 − x2 )
≈ (4.13)
l(y 2 + x2 )2
μ0 my
< 2 2 (4.14)
l (y + x2 )
The above equation shows that the magnetic gradient measured by the DSMG
will scale as the inverse second power of distance for near field objects. A part near
the one quarter position along the string will produce a magnetic gradient equal and
opposite to an equivalent part at the three quarter position.
4.8.3 Parts near the end points of the string
There is another cluster of parts next to the end points of the string. Considering
parts at location:
r (x , y, z ) = (0, 0, L2 + Δz ) (4.15)
The magnetic field along the string is:
μ0 my
By (0, 0, z) = − (4.16)
4π(z − L2 − Δz )3
Using equation 4.10, the magnetic gradient measured by the DSMG is:
l
2π 2 μ0 my 2πz
Byz =− 2 sin dz (4.17)
l − 2l 4π(z − L2 − Δz )3 l
The above expression has no analytic solution. A good approximation can be
achieved by replacing the sine function with a triangle wave. The magnetic gradient
is then:
3πμ0 my
Byz = − (4.18)
2Δz (l + Δz )(l
+ 4Δz )(3l + 4Δz )
In the limit Δz << l:
πμ0 my
Byz ≈ − (4.19)
2Δz l3
The above equation shows that the magnetic gradient measured by the DSMG
will scale as the inverse distance for objects near the ends of the string.
4.8.4 Heading error calculation
All parts of the DSMG have been grouped into regions as shown in Fig. 4.7. The
parts in region A are closer to the string so they will induce larger magnetic gradients
compared to the parts in region G. For each part, the induced heading error was
calculated using equation 4.13 or equation 4.18. The table at the bottom of Fig. 4.7
displays the total magnetic moment and total heading error of each region.
The high symmetry of the frame, mylar tape and copper wires in the immediate
vicinity of the string has capped the induced magnetic gradient at 5 nT/m. Despite
being further removed from the string, regions A and E induce large magnetic gra-
dients because we can not assume that the hundreds of small parts in these regions
will also cancel out. We calculate that the amount of heading error which will be
produced by all magnetic parts is 28nT/m peak to peak.
The magnetic susceptibility measurements contained in the preceding paper and

the act of washing each part in acid, provide us with confidence that the figure
of 28nT/m is an upper bound on heading error. If the induced magnetism of the
hundreds of parts cancels out to any extent then the actual heading error will be
lower than predicted. In addition, the distances to the string shown in Fig. 4.7 are
conservative, parts further away should have less heading error than predicted.
G
G Vacuum Tank
F D C B C
A
Plastic Frame
String
Isolator E
Region Distance Plastic OFHC Other Magnetic Peak to peak

to string parts copper metal Moment heading error
(m) parts parts (10-8A·m2) (nT/m)
Symmetrical plastic frame 0.0004 7 6 5 42 5.0
(includes pickup and feedback coils)
A – other parts inside frame 0.005 122 6 3 2 8.0
B – parts attached to outside of frame 0.012 47 8 0 1 1.5
C – hardware at ends of string 0.03 13 0 7 4 0.4
D – 4th (sorbethane) stage of isolator 0.05 14 0 0 2 0.1
E – 1st,2nd,3rd stages of isolator 0.05 32 0 250 334 12.1
F – Inner flask and silicone plugs 0.09 5 0 4 61 0.3
G – vacuum tank, lid, connectors, valve 0.16 12 0 7 625 0.7
TOTAL 252 20 276 1070 28.1
Figure 4.7: Sketch of the gradiometer system (not to scale) including the sensor frame,
vacuum tank and mechanical isolator. The picture shows parts grouped into regions with
dotted lines. For each region, the table shows the magnetic moment and heading error of all
parts in that region. A cross section of the symmetric plastic frame is shown in figure 4.6.
4.8.5 Discussion
I have calculated the maximum heading error from magnetic parts to, be 28nT/m
peak to peak. This maximum is less than the amount of heading error actually
measured in figure 4.1, which is 225 ± 5 nT/m peak to peak. Something else besides
magnetic parts is creating a directional dependant error.
The common mode rejection ratio of the DSMG may not be high enough. In
this case, the DSMG will erronously detect the Earth’s uniform magnetic field as
small magnetic gradient. The DSMG would be subject to the same orientation errors
suffered by vector magnetometers.
The theoretical common mode rejection ratio of the DSMG should be very high
according to Veryaskin[10]. However, Veryaskin’s paper contains the assumption:
“When the string is a stretched thin flat wire the resonant motion is strictly
one dimensional with its sensitivity axis pointing perpendicular to the
plane of motion. In this case the string represents a one-dimensional me-
chanical harmonic oscillator having an infinite number of resonant (violin)
modes.”
For a full analysis, vibration in two directions should be considered.
4.8.6 Out of plane vibration
Here I will consider vibration in two directions. The string inside the DSMG is a
flat ribbon with width a = 125 μm in the X direction, thickness b = 25 μm in the
Y direction and length L = 250 mm in the Z direction. The resonant frequencies for
transverse vibration in the X direction will differ from the resonant frequencies in the
Y direction because the stiffness EI is different. The stiffness EI depends on the
moment of area, which is
ab3 a3 b
Ix = Iy = (4.20)
12 12
= 1.6 × 10−19 m4 = 4.1 × 10−18 m4 (4.21)
E ≈ 69 GPa is the Young’s modulus of aluminium 6061. The ribbon is long

and thin so the resonant frequency depends mostly on the axial tension F . I will

also introduce a non-dimensional parameter, ξ = L F/(EI), that characterises the
relative importance of the ribbon’s stiffness. Using a formula by Zui et al.[38], the
resonant angular frequency for the second violin mode is:

2π F ξ
ω2 ≈ × (4.22)
L η ξ − 2.2
where η = 8.4 × 10−6 kg/m is the ribbon’s mass per unit length. Zui et al’s formula
is valid for ξ > 60.
In order to distinguish between modes which are nearby, a high mechanical Q is
needed. The DSMG was placed in a 4.6 Pa vacuum. The z axis of the ribbon was
vertical so gravity would have negligible affect on transverse vibration. An alternating
current with amplitude 0.3 A was applied to the ribbon. This heated the ribbon and
reduced the 2nd violin mode resonance frequency to approximately 440 Hz. I then
did a quick frequency sweep from 430 Hz to 450 Hz with the alternating current.
The alternating current frequency was then switched to 1000 Hz, sufficiently high
so that it not longer affected the ribbon. The sweep had excited two modes with
resonant frequencies ω2x = 438.75 ± 0.25 Hz and ω2y = 444.25 ± 0.25 Hz, which are
the in plane and out of plane second violin modes. Figure 4.8 shows both modes of
vibration slowly ringing down.
The tension of the ribbon is approximately T = 0.1 N. Using equation 4.22, the
theoretical second violin mode frequencies are: ω2x = 438.75 Hz and ω2y = 444.01 Hz.
The difference is 5.26 Hz between the theoretical resonances and the difference is
5.5 Hz between the measured resonances.
During normal operation of the DSMG, the in plane second violin mode is excited
at resonance and used to detect magnetic gradients. The nearby out of plane mode will
also be excited but the excitation will be off resonance. This off resonance response
will create a phase error that will decrease the common mode rejection ratio of the
string.
The pickup coils used to measure ribbon displacement are balanced for in plane
motion and thus have a reasonable common mode rejection. The pickup coils should
1
t=0s
t=8s
0.1 t = 16 s
t = 24 s
0.01
Voltage (V)
0.001
0.0001
0.00001
430 432 434 436 438 440 442 444 446 448 450
Frequency (Hz)
Figure 4.8: This graph shows the amplitude of two 2nd violin modes (in plane motion and
out of plain motion) for the ribbon inside the DSMG. Both modes are ringing down slowly.
not detect out of plane motion but figure 4.8 shows that they are (albeit with a 40
dB attenuation). The pickup coils may not be balanced for out of plane motion and
may therefore detect common mode vibrations.
Out of plane motion may thus degrade the common mode rejection ratio of both
the string and pickup coils. Some preliminary experiments with the DSMG support
this hypothesis and suggest that out of plane vibration could be the cause of heading
error.
Bibliography
[1] Schmidt P W and Clark D A 2000 Advantages of measuring the magnetic gra-
dient tensor Preview 85 26-30
[2] Schmidt P, Clark D, Leslie K, Bick M, Tilbrook D and Foley C 2004 GETMAG - a
SQUID magnetic tensor gradiometer for mineral and oil exploration Exploration
Geophysics 35 297-305
[3] Stolz R, Zakosarenko V , Schulz M, Chwala A, Fritzsch L, Meyer H G and

Kostlin E O 2006 Magnetic full-tensor SQUID gradiometer system for geophysical
applications The Leading Edge 25 178-180
[4] Merayo J M G, Brauer P and Primdahl F 2001 Triaxial fluxgate gradiometer of

high stability and linearity Sensors and Actuators A 120 71-77
[5] Killen P G 2006 Airborne geophysical surveying Supplement to The Northern

Miner 92 No. 2, 5-12
[6] Killen P G 2007 Airborne geophysical surveying Supplement to The Northern

Miner 93 No. 1, 6-15
[7] McRae W, Veryaskin A V, Ju L, Blair D G, Chin E, Dumas J and Lee B 2004

String Magnetic gradiometer system: Recent airborne trials SEG Expanded Ab-
stracts 23 790-793
[8] McFee J E, Bell M, Dempsey B, Chesney R H and Das Y 1985 A magnetosta-

tic signature measurement and analysis system Journal of Physics E: Scientific
Instruments 18 54-60
[9] Veryaskin A V 2001 Magnetic gradiometry: a new method for magnetic Sensors
and Actuators A 91 233-235
[10] Veryaskin A V 2009 Theory of operation of direct string magnetic gradiometer

with proportional and integral feedback International Journal of Applied Elec-
tromagnetics and Mechanics 29 197-215
BIBLIOGRAPHY 111
[11] Fickett F R 1992 Low Temperature magnetic behavior of “nonmagnetic” mate-

rials Advances in Cryogenic Engineering 38 1191-1197
[12] Kobayashi K and Uchikawa Y 2003 Development of a high spatial resolution

SQUID magnetometer for biomagnetic measurement, IEEE Transactions on
magnetics 39 3378-3380
[13] Morello A, Angenent W G J, Frossati G and de Jongh L J 2005 Automated and

versatile SQUID magnetometer for the measurement of materials properties at
millikelvin temperatures Review of Scientific Instruments 76 023902
[14] Bartington Instruments, Mag-03MRN Three-Axis Magnetic Field Sensor

http://www.bartington.com/products/Mag-03mrnthreeaxismagneticfieldsensor.cfm
[15] Spencer J F and John M E 1927 The magnetic susceptibility of some binary
alloys Proceedings of the Royal Society of London. Series A 116 61-72
[16] Constant F W and Formwalt J M 1939 Investigation of Ferromagnetic Impurities

I Physical Review 56 373-377
[17] Matsubayashi K, Maki M, Tsuzuki T, Nishioka T and Sato N K 2002 Magnetic

properties (communication arising): parasitic ferromagnetism in a hexaboride?
Nature 420 143-144
[18] Wang W, Hong Y, Yu M, Rout B, Glass G A and Tang J 2006 Structure and
magnetic properties of pure and Gd-doped HfO2 thin films Journal of Applied
Physics 99 08M117
[19] Atkins A G and Liu J H 2007 Toughness and the transition between cutting and
rubbing in abrasive contacts Wear 262 146-159
[20] Society of Manufacturing Engineers 1996 Materials finishing and coating 4th
Edition (Tool and Manufacturing Engineers Handbook vol 3) ed C Wick and R F
Veilleux (Dearborn, Michigan: Society of Manufacturing Engineers) section 16-6
[21] Kohl W H 1995 Handbook of Materials and Techniques for Vacuum Devices
(Woodbury, New York: America Institute of Physics) p 199
[22] Hutchison T S and Reekie J 1948 Strain Sensitivity of Magnetic Susceptibility

Physical Review 73 517 - 518
[23] Keyser P T and Jefferts S R 1989 Magnetic susceptibility of some materials

used for apparatus construction (at 295 K) Review of Scientific Instruments 60
2711-2714
[24] Tanimoto Y , Fujiwara M, Sueda M, Sueda K, Inoue K and Akita M 2005 Mag-
netic levitation of plastic chips: applications for magnetic susceptibility measure-
ment and magnetic separation Japanese Journal of Applied Physics 44 6801-6803
[25] Bossi I, Dilley N R, OŠBrien J R and Spagna S 2004 Optimization of sample

holder materials for sensitive magnetometry measurements at low temperatures
Materials Research Society Symposium Proceedings 825E G5.5
[26] ASM International 2001 ASM Specialty Handbook: Copper and Copper Alloys ed
J R Davis (Materials Park: ASM International) p 487
[27] Benech C and Marmet E 1999 Optimum depth of investigation and conductivity
response rejection of the different electromagnetic devices measuring apparent
magnetic susceptibility Archaeological Prospection 6 31-45
[28] Butts A 1954 Copper: the science and technology of the metal / its alloys and
compounds (New York: Reinhold Publishing) p 504
[29] Huck F B, Savage W R and Schweiter J W 1973 Magnetic susceptibility of α-

phase Cu-Al and dilute magnetic Cu-Al(Fe) alloys Physical Review B 8 5213-5220
[30] Ficket F R and Sullivan D B 1974 Magnetic studies of oxidized impurities in pure
copper using a SQUID system Journal of Physics F Metal Physics 4 900-905
[31] Zhou S 1999 Electrodynamics of Solids and Microwave Superconductivity (New

York: John Wiley & Sons) p 90
[32] Barker J R 1948 The testing of brass and other constructional materials for
ferromagnetic impurities Journal of Scientific Instruments 25 363-364
BIBLIOGRAPHY 113
[33] ASM International 1982 Engineering Properties of Steel ed P Havrey (Materials

Park: ASM International)
[34] Dearing J A, 1999 Environmental magnetic susceptibility: Using the Bartington

MS2 System, 2nd edition (Kenilworth: Chi publishing)
[35] Gattacceca J, Eisenlohr P and Rochette P 2004 Calibration of in situ magnetic

susceptibility measurements Geophysical Journal International 158 42-49
[36] Griffiths D J 1999 Introduction to Electrodynamics, 3rd edition (Upper Saddke

River: Prentice-Hall International) p 217, p 246, p 264
[37] Sunderland A, Golden H, McRae W, Veryaskin A V, Ju L and Blair D 2009

Results from a novel direct magnetic gradiometer Exploration Geophysics 40
222-226
[38] Zui H, Shinke T and Namita Y 1996 Practical formulas for estimation of cable
tension by vibration method Journal of Structural Engineering 122 651-656
Chapter 5
Sensitivity and optimation
5.1 Preface
The paper which forms the bulk of this chapter gives a full theoretical analysis of the
Direct String Magnetic Gradiometer (DSMG). The paper presents the equations of
motion that govern the dynamics of the string. The noises affecting the string and
DSMG are then discussed. Of special importance, is the string’s mechanical quality
factor which is a measure of how underdamped an oscillator is. The mechanical quality
factor also determines the coupling of the string with sources of thermal noise.
The paper then presents some noise measurements which compare the sensitivity
of a DSMG using either a round string or a flat string. Noise measurements are also
taken in different air pressures.
In the paper, I present an in depth study of the most critical parameters of the
magnetic gradiometer: temperature, pressure, dimensions, and material. I first de-
termine the optimal string temperature. I then show that the sensitivity should be
independent of air pressure if the temperature is kept constant. I then calculate the
optimum string diameter and show that aluminium 6061 is the best material to make
the string out of.
The paper has been published in Review of Scientific instruments. My contribution
to this paper amounts to 90% of the experimental work, 95% of the modelling, and
80% of the manuscript preparation. Following comments by the referees of this thesis,
section 5.7.1 has been substantially rewritten and differs from the same section of the
published paper.
115
116 Chapter 5. SENSITIVITY AND OPTIMISATION
Review of Scientific Instruments 80 (2009) 104705
Optimising a Direct String Magnetic Gradiometer

for Geophysical Exploration
Andrew Sunderland1 , Li Ju1 , David G Blair1 , Wayne McRae1,2 , Alexey V Veryaskin1,2

Abstract
Magnetic gradiometers are tools for geophysical exploration. The mag-
netic gradient is normally calculated by subtracting the outputs of two
total field magnetometers which are separated by a baseline. Here we
present a unique device that directly measures magnetic gradients using
only a single string as its sensing element. The main advantage of a di-
rect string magnetic gradiometer is that only gradients can induce second
harmonic string vibrations. A high common mode rejection ratio is thus
naturally achieved without any balancing technique. Performance de-
pends on the ability to dissipate heat whilst minimising air damping. By
combining high current, an elevated temperature and low pressure, we can
√
easily achieve sensitivity of 0.18 nT/m/ Hz. Further increases in sensi-
tivity can be attained by optimising the sensing element. In this paper we
present an in depth study of the most critical parameters of the magnetic
gradiometer. We describe the design for the next generation of sensor,
√
which will reach the required sensitivity of 0.01 nT/m/ Hz using only
1 watt of power. By combining a few single-axis magnetic gradiometer
modules, it will be possible to deploy a full tensor magnetic gradiometer
with more than sufficient sensitivity for airborne geophysical applications.
5.2 Introduction
Magnetic gradiometry is a technology that measures spatial variations of magnetic
fields. There are two distinct areas and relevant techniques in relation to magnetic
gradiometry, namely - total field gradiometry and vector magnetic gradiometry. The
first technique uses two total field (i.e. independent on the orientation of the magnetic
field measured) magnetometers separated by a baseline. This strategy has been suc-
cessfully implemented by various survey companies[1]. The second technique, vector
magnetometry, measures the spatial variation of both the magnitude and direction
of the magnetic field. Thus, 9 components of the magnetic gradient tensor can be
measured, of which 5 are linearly independent. Vector magnetic gradiometry can
provide useful information for geophysical exploration[2], especially in regions of high
magnetic intensity such as over banded iron formations.
There are several types of vector magnetic gradiometers: Superconducting Quan-

tum Interference Device (SQuID) based gradiometers (both Low Tc [3] and High Tc [4]),
induction coil gradiometers[5], and fluxgate gradiometers[6], etc. All these techniques
are based upon a separation of two vector magnetometers (or flux meters) and process-
ing a differential signal as a measure of a spatial derivative of the magnetic field vector.
The best sensitivities are achieved with SQuID based systems. However, SQuIDs re-
quire a cryogenic environment and suffer from thermal cycles of cooling the system
down to operational temperature.
In this paper we present a Direct String Magnetic Gradiometer (DSMG) which

can detect gradients using a string as its single sensing element. A high common mode
rejection ratio is naturally achieved without any balancing technique. Moreover, by
optimising the string’s dimensions and material, it is theoretically possible to achieve
sensitivities similar to that of SQuID based magnetic gradiometers.
The DSMG consists of a vibrating metal string clamped at both ends which is
carrying alternating current. Its frequency is tuned to the second violin mode of the
string, which is coupled only to a magnetic gradient[7]. The sensor operates as an up-
converter (or a mixer) converting a quasi-DC magnetic gradient into an audio band
mechanical oscillation of the string. There are no restrictions on what technology
could be used to measure the displacement of the string. The DSMG is flexible
enough to use optical interferometers, amplitude and phase sensitive RF resonant
read-outs and potentially quantum noise limited optoacoustical tranducers[8; 9].
In this paper, we present the latest results using strings with rectangular cross
section and round cross section. The noise floor of the DSMG is shown to be in good
agreement with theory. We briefly consider the DSMG operation principle and, in
greater details, the most critical parameters which allows us to enhance the achieved
performance even further.
5.3 Dynamics of a bounded current carrying string

Here we consider a string that has a highly uniform mass per unit length distributed
along its long dimension z. The string inside the DSMG is L = 0.25 m long and is
clamped at both ends with sandwich clamps. The boundary conditions are thus:
∂ ∂
x(0, t) = 0 x(L, t) = 0 x(0, t) = 0 x(L, t) = 0 (5.1)
∂z ∂z
The clamps are moved apart slightly so that the string is stretched to a length
L, which is slightly longer than the string’s natural length L0 . This keeps the string
under tension:
L − L0
F = EA (5.2)
L0
where F is tension, E = 69 GPa is the Young’s modulus of aluminium 6061 and
A is the cross sectional area of the string.
AC current is applied to the string causing it to interact with the external magnetic
field. The string’s dynamics can then be described by the following force balance
equation[7; 10]:
∂2 ∂ ∂2 ∂4
η x(z, t) + h x(z, t) − F x(z, t) + EI x(z, t)
∂t2 ∂t ∂z 2 ∂z 4
= i sin(ωt) [By (0, t) + Byz (0, t)z] − ηg cos(θ) + noise (5.3)
where η is mass per unit length, t is time, h is the friction coefficient per unit
length, I is the moment of area (see Eq. 5.42), i is the alternating current amplitude,
ω is the angular frequency of that current, By (0, t) is the y component of uniform mag-
netic field, Byz (0, t) is the first order magnetic gradient, g = 9.8 ms−2 is acceleration
due to gravity and θ is the string’s inclination relative to the horizontal.
5.3. DYNAMICS OF A BOUNDED CURRENT CARRYING STRING 119
Resonant modes of magnetic gradiometer

x
y DSMG Frame
Clamp
z
1st violin mode
2nd violin mode
L = 0.25 m
Figure 5.1: The top diagram shows the DSMG frame with the string clamped rigidly at
each end. An AC current is tuned to the 1st violin mode of the string causing the string
to vibrate in the presence of magnetic fields. The ’C’ shape of the 1st harmonic is sensitive
to uniform magnetic fields (the magnetic field is shown going into the page). In the bottom
diagram, the AC current is tuned to the 2nd violin mode of the string. The ’S’ shape of the
2nd harmonic is sensitive to magnetic gradients.
The string has a sag parameter λ, which characterises the importance of gravity.
The sag parameter is approximately[11]:
$
ηgL cos(θ) EA
λ= (5.4)
F F
During normal operation of the DSMG, the tension is normally high enough so
that we can neglect gravity (λ << 1). If the tension is also much greater than the
stiffness (F >> EIL−2 ), then the violin mode angular frequencies approach that of
a perfect string[12]:

nπ F
ωn ≈ (5.5)
L η
with mode shape
nπz
x(z, t) = X sin cos(ωt) (5.6)
L
where X is the amplitude of vibration and n is the mode number. Fig 5.1 presents
a diagram of the string inside the DSMG. The diagram shows how the string’s second
violin mode is highly sensitive to magnetic gradients.
By tuning the AC current to the second violin mode frequency of the string,
ω = ω2 , the string’s vibration amplitude will slowly build up over time until the
friction force equals the driving force[13]. We find that the displacement of the string
x is proportional to the magnetic gradient Byz in the following equation of motion:

is QL2 −t 2πz
x(z, t) = 1 − exp cos(ω0 t) sin Byz (5.7)
πmω02 τ L
where ω0 = ω2 is the resonant frequency of the 2nd violin mode and will be used
henceforth in this paper, Q = τ ω0 /2 is the mechanical quality factor of the string,
τ = 2η/h is the string relaxation time and m is the mass of the string. The Q factor
is a measure of how underdamped an oscillator is, or more explicitly:
Energy stored
Q = 2π × (5.8)
Energy dissipated per cycle
5.4 Noise
The sensitivity of the DSMG is limited by various noises which are indistinguishable
from magnetic gradients. These noises include electronic noise, vibration noise, ther-
mal noise, and magnetic noise from instrument components. Electronic noise comes
from the electronics used to measure the string’s displacement whereas the other three
noises make the string vibrate. In this section, we discuss the contribution of each
noise to the total noise budget of the DSMG. Noise from magnetic components inside
the DSMG is discussed in another paper[14].
5.4.1 Electronic noise
The displacement of the string along the X axis is measured with two inductive
pickup coils, connected in differential mode. Each pickup coil has a displacement
sensitivity of approximately ∼ 400 V/m and experiences electronic noise on the order
5.4. NOISE 121
√
of ∼ 3nV/ Hz. From experiment, we have determined that a pair of pickup coils can
√
measure differential displacement with an rms accuracy of xel = 10−11 m/ Hz.
We can use Eq. 5.7 to convert displacement sensitivity (xel ) into magnetic gradient
√
sensitivity. Thus, the electronic noise floor of the DSMG in units of T/m/ Hz is:
√
2 2πηω0xel
Byz = (5.9)
iτ L
where i is the peak AC current applied to the string, η is the mass per unit length,
L is the length of the string, ω0 is the resonant frequency of the string’s 2nd violin
mode and τ is the string relaxation time.
The electronic noise level is lower than the thermal noise level so the present
inductive readout is adequate. In section 5.6, we show that electronic noise contributes
somewhere between 15% and 50% of the total noise budget. The DSMG readout
system is designed for maximum dynamic range and linearity during field operation.
5.4.2 Vibration noise
The DSMG is designed for airborne geophysical surveys. Onboard a survey air-
craft, the string will be exposed to engine vibration noise of approximately 3 ×
√
10−2 m/s2 / Hz[15]. Vibration noise at the same frequency as the string’s resonant
frequency will be amplified by a factor of Q , where Q is the string’s mechanical
quality factor.
McRae et al.[15] describe the noise reduction achieved from mounting the DSMG
sensor inside a four-stage passive isolator. The stages have resonant frequencies at
17 Hz, 40 Hz, 70 Hz and 150 Hz, providing more than 80 dB of mechanical isolation
at frequencies above 400 Hz. The isolator is attached to the inside of a small vacuum
tank made of transparent Perspex. We find that with a moderate vacuum of 1000 Pa
or better, the acoustic coupling can be reduced to the required level.
5.4.3 Thermal noise
The string will experience stochastic motion caused by thermal noise. The spectrum of
thermal noise is given by Saulson[16], and with a slight change in notation (k → mω02
and φ → 1/Q), the noise spectral density is:
2 4ψn2 (z)kT mω02 Q1

xth (ω, z) = (5.10)
2 4 1
ω (mw0 − mω ) + m w0 Q2
2 2 2
where k = 1.38 × 10−23 JK−1 is Boltzmann’s constant, T is the temperature of

√
the string and ψn (z) = 2 sin (2πz/L) is the normalised second violin mode. At
resonance, ω = ω0 , the thermal noise level is:

8kT Q 2πz
xth (z) = sin (5.11)
mω03 L
The thermal noise in the above expression can be compared to the magnetic
gradient signal in Eq. 5.7. We then find that the thermal noise limited sensitivity of
√
the DSMG in units of T/m/ Hz is:

4π ηkT ω0
Byz = (5.12)
i L3 Q
$
4π 2ηkT
= (5.13)
i L3 τ
where i is the peak AC current applied to the string, η is the mass per unit length,
T is the absolute temperature of the string, L is the length of the string, Q is the
string’s mechanical quality factor and τ is the string relaxation time. The string’s
resonance acts as a mechanical filter that affects both magnetic gradient signals and
thermal noise equally. For this reason, thermal noise acts like white noise and is
independent of frequency. In sections 5.5 to 5.9, we will discuss the effect of different
factors on the thermal noise level.
5.5 The loss angle

The amount of thermal noise in Eq. 5.12 depends on the mechanical Q factor of the
string. This follows from the fluctuation dissipation theorem[17]. A high Q string is
weakly coupled to the thermal reservoir thus it will experience less thermal noise. It
is convenient to express the mechanical quality factor Q in terms of the loss angle
1/Q. The loss angle is the sum of all mechanisms that dissipate the vibration energy
of the string.
5.5. THE LOSS ANGLE 123
Flat ribbon Round wire

cross section cross section
b = 25 ȝm
x
a = 125 ȝm
y
d = 125 ȝm
Figure 5.2: The string inside the DSMG is 0.25 m long and vibrates along the x-axis. Its
cross section can be either a flat ribbon or a round wire.
5.5.1 Air damping
Thin strings have a large surface to volume ratio so the losses from air damping can
be substantial. These air damping losses include both the dissipation of energy as
sound waves at high air pressure and the dissipation of energy in noninteracting gas
molecules colliding with the wire at low air pressure.
For simplicity we will only consider the noninteracting gas molecule model. The
molecular mean free path should be much larger than the string diameter in order to
prevent the reflection of air molecules back onto the wire[18]. Here we show that the
model is a good fit to the data at low pressures.
Inside the DSMG, we use strings with two different cross sections: a round wire
of diameter d = 125 μm and a flat ribbon with width a = 125 μm and thickness
b = 25 μm. The cross sections are shown in Fig. 5.2.
If the flat ribbon is moving with velocity v, then a pressure difference arises be-
tween the front and back of the ribbon[19]:
$
2M
P front − P back = 4v P (5.14)
πRTair
where P is air pressure, R = 8.314 JK−1 mol−1 is the universal gas constant,
M = 0.02897 kg mol−1 is the molar mass of air, and Tair = 300 K is ambient air
temperature. The damping force per unit length is then:
$
∂F 2M
= 4vaP (5.15)
∂L πRTair
where a = 125 μm is the width of the ribbon. The friction coefficient per unit
length is then:
1 ∂F
hribbon = × (5.16)
v $∂L
2M
= 4aP (5.17)
πRTair
The damping effect of gas molecules on a flat ribbon is a factor of 4/π times
stronger than damping effect on a round wire of the same diameter[19]. Therefore:
$
2πM
hwire = P d (5.18)
RTair
where d = 125 μm is the diameter of the wire. The mass per unit length for both
shapes is:
πd2 ρ
ηribbon = abρ ηwire = (5.19)
4
where b = 25 μm is the thickness of the ribbon and ρ = 2690 kgm−3 is the density
of aluminium 6061. The relaxation time of an oscillator is equal to the ratio of mass
to damping force, τ = 2η/h. The attenuation rates for the vibrating ribbon and
round wire are thus:
$ $
1 2P 2M 1 2P 2M
= =
τribbon ρb πRTair τwire ρd πRTair
= 0.081P = 0.0162P (5.20)
Both the ribbon and round wire will be vibrating at approximately 750 Hz inside
the DSMG. Therefore, using Q = τ ω0 /2, the loss angles are:
1 1
= 3.4 × 10−5 P +
Qribbon Qother
1 1
= 6.9 × 10−6 P + (5.21)
Qwire Qother
We measured the loss angle of the ribbon and round wire at various pressures
between 0.002 Pa and 105 Pa (1 atmosphere). Dynamic friction between the string
and the clamps at each end of the string could be a source of excess losses. Therefore,
−2
10
Ribbon with Al clamps
Ribbon with torlon clamps
Wire with Al clamps
Wire with torlon clamps
Wire with teflon clamps
−3 Ribbon (theory)
Loss angle, 1/Q

10
Wire (theory)
Teflon clamps
−4
10
−5 Torlon clamps Al clamps

10 −3 −2 −1 0 1 2 3 4 5
10 10 10 10 10 10 10 10 10
Pressure (Pa)
Figure 5.3: This graph compares the measured loss angle with the theoretical gas losses in
Eq. 5.21. The value of 1/Qother in that equation has been fit to the measured data.
we tested clamps made from Teflon PTFE, Torlon 4301 and Aluminium 6061 to
investigate these losses. The second violin mode frequency was set to 750 ± 50 Hz.
The strings were excited at resonance and then the loss angle was determined from
the relaxation time of the vibration ringdown. Fig. 5.3 compares our results with the
theory in Eq. 5.21.
The air damping losses of both ribbon and wire are in good agreement with theory
when the pressure is below P < 60 Pa. Above this pressure, the losses are lower than
predicted because the assumptions made for the non-interacting gas model are no
longer valid.
5.5.2 Clamp losses
Once gas damping has been eliminated, other excess loss mechanisms start to show
up[20]. As P → 0, the loss angles of the ribbon and the wire asymptotically approach
constant values. We have taken the asymptotes from Fig. 5.3 and presented them in
Table 5.1 as the residual losses in a perfect vacuum. The table shows that the residual
Form Clamp Loss angle Loss angle of Ref.

of string clamp material
Ribbon aluminium (2.1 ± 0.2) × 10−5 4 × 10−6 [21]
torlon (2.4 ± 0.9) × 10−5 2 × 10−2 [22]
Wire aluminium (6 ± 5) × 10−6 4× 10−6 [21]
torlon (2.2 ± 0.6) × 10−5 2× 10−2 [22]
teflon (9.3 ± 0.6) × 10−5 5 × 10−2 [23]
Table 5.1: This table shows an estimate of the residual losses of the ribbon and wire in a
perfect vacuum, based upon an extrapolation of the data in Fig. 5.3. The loss angle of the
round wire with aluminium clamps is based upon only one measurement at 1.4 Pa. At this
pressure, the loss angle of the round wire should be dominated by air damping so we are
very uncertain about estimating the loss angle in a perfect vacuum.
loss angle of the round wire is highly dependant on the choice of clamp material.
As the string oscillates, the clamps at each end of the string will recoil slightly
in response to the oscillations, since the clamp structure cannot be perfectly rigid.
Therefore, some of the kinetic energy will be stored in the clamps. If the clamps are
lossier than the string itself, a significant part of the energy loss may take place in
the clamps. The amount of clamp recoil loss is[16]:
1 1 1 mb ωa ωb
= + × × (5.22)
Qb,recoil Qb Qa ma (ωa2 − ωb2 )2
where 1/Qb,recoil is the total loss angle of the string including clamp recoil loss,
1/Qb is the intrinsic loss angle of the string by itself , 1/Qa is the loss angle of the
clamp material, ωb is the resonant angular frequency of the string and ωa is the
resonant angular frequency of the clamp. The above equation shows that the amount
of clamp losses depends on the ratio of the string mass to the clamp mass, mb /ma .
This means that the thinner ribbon should have less clamp losses. Table 5.1 shows
that this is true in practice.
The losses in the round wire are higher when the clamps are made from Teflon
PTFE or Torlon 4301. Clamps were made from these materials because they were
known to be non-magnetic and we wished to avoid heading error noise from magnetic
contamination[14]. However, the high loss angles of these materials may preclude us
from using them as clamps in future work.
The loss angle of the flat ribbon did not change when switching from aluminium
6061 clamps to torlon 4301 clamps. This indicates that the losses are instrinsic to the
ribbon itself. These intrinsic losses are investigated in section 5.5.3.
5.5.3 Violin mode losses
The intrinsic losses of the string take place where there are large changes in stress.
The distribution of stress along the string depends on the mode shape so each mode
of vibration will have a different loss angle. The theory in this subsection is mainly
based upon the work of Saulson and Gonzalez[10], Irvine[11], Gretarsson[20] and Tri-
antafyllou and Grinfogel[24]. We have combined their work to calculate two different
types of violin mode losses.
The string vibrates transversely in its second violin mode under tension. Each
oscillation modulates that tension infinitesimally. Thus, the restoring force is due
almost entirely to the equilibrium tension F and only a very small amount is due to
dynamic tension ΔF . Dynamic tension is subject to dissipation whereas the restoring
force due to equilibrium tension is dissipation free[20]. The dynamic tension is stored
as elastic energy Velastic whereas the equilibrium tension is stored as gravitational
potential energy Vgrav [11].
There is however, a third restoring force. This is the restoring force due to bending
of the string, which is a restoring force due to differential strain across the string
diameter. Bending tension is small except near the clamping points, where rapid
changes in curvature occur. Bending tension is an elastic restoring force so it is also
subject to dissipation[20]. The total energy stored in the string is[10; 20]:
Vtotal = Vgrav + Velastic + Vbending (5.23)
The fraction of energy spent in dissipative processes determines the loss angle for
violin modes[20]:
1 Velastic 1 Vbending 1
= × + × (5.24)
Q violin Vtotal Q elastic Vtotal Q bending
The fraction of energy stored as elastic energy is calculated in Appendix A, section
5.11. The fraction of energy stored as bending energy depends on the stiffness EI of
the string[10]:
$ $
Vbending 2 EI (nπ)2 EI
= 1+ (5.25)
Vbending + Vgrav L F 2L F
where n is the mode number, E is Young’s modulus and I is the string’s moment
of area. The violin mode loss angle is then:
$ $
1 2 EI (nπ)2 EI 1
= 1+ (5.26)
Q violin L F 2L F Qbending
where 1/Qbending is the material loss angle of the aluminium string measured under
zero tension[20].
We measured the loss angle of the ribbon in a high vacuum, P = 0.002 Pa, so that
air damping would be negligible. To measure the loss angle under different amounts
of tension, we heated the ribbon.
Initially, the string was stretched with tension F = 0.23±0.02 N. Electrical power
was then used to heat up the ribbon inside the vacuum tank. This causes the natural
length of the ribbon, L0 , to expand in proportion to aluminium’s thermal expansion
coefficient[25]. At the same time, the distance between the clamps, L, stays constant
so the tension will vary as per Eq. 5.2.
Our experimental results are presented in Fig 5.4 which shows the loss angle as
a function of tension. The loss angle was determined by exciting the ribbon in its
second violin mode and measuring the ringdown. The amount of tension in the ribbon
was determined from the value of the resonant frequency using formulae by Zui et
al.[12]. The temperature was calculated from the changes in the ribbon’s resistivity.
The experimental loss shown in Fig. 5.4 is in good agreement with the theoretical
violin mode loss angles from Eqs. 5.26 and 5.60. The value of the material loss
angle 1/Qbending = 1/Qelastic = 0.0025 ± 0.0015 was fit to the experimental data. The
loss angle should be temperature dependent so this value applies only at 430 ± 40 K.
There is a peak in the observed loss angle at 64 Hz. This peak is predicted by the
sharp increase in the elastic energy at 55 ± 5 Hz, calculated in Appendix A, section
5.11.
Eq. 5.26 predicts that the violin mode loss angle should continue to shrink as
the tension is increased. However, the experimental loss angle stops decreasing for
Tension (N)
−4 −3 −2 −1
−2 10 10 10 10
10
Ribbon, theory
500K Ribbon, experiment
Loss angle, 1/Q −3

10
430K
430K
430K
430K
−4
10
420K
420K
300K
340K
−5
10
10 100 1000
Second violin mode frequency (Hz)
Figure 5.4: This is a plot of the ribbon’s Q factor as a function of tension in a high vacuum.
The peak near 10−3 N is predicted by Eq. 5.60 and the general downward slope is in good
agreement with the theory in Eq. 5.26. The graph also shows the resonant frequency and
temperature for each measurement. The absolute uncertainty in temperature is ±40 K but
the relative uncertainty between adjacent data points is much less.
tensions above F > 0.01 N. Other sources of loss (such as clamp or bulk loss) are
increasing the total losses of the ribbon above the value predicted from intrinsic losses.
Our material loss angle (1/Q = 0.0025 ± 0.0015 at 430 K) is significantly above
the loss angle measured by Duffy[21] (1/Q = 4 × 10−6 at room temperature) but is in
good agreement with the high temperature results of Szenes and Zsambok[26]. The
intrinsic loss angle of aluminium 6061 increases exponentially at temperatures above
T > 200 K[21; 27; 28]. Dislocation damping causes a large peak in the loss angle
around 500 K[26; 29; 30] followed by another peak around 900 K[30]. The exact
temperature of the peaks depends on the annealing temperature and the frequency[29;
30]. Fig. 5.4 shows a large increase in the loss at 500 K. Also, the loss angle of our
ribbon (thickness b = 25 μm) should be higher than the loss angle of the bulk metal
because of surface losses[20].
100
Relaxation time (s)
10
Ribbon, theory
Ribbon, experiment
1
10 100 1000
Second violin mode frequency (Hz)
Figure 5.5: This graph plots the ribbon’s relaxation time as a function of the ribbon’s
resonant frequency, whilst in a high vacuum. This graph has been adapted from Fig 5.4
using the formula Q = τ ω0 /2.
The sensitivity of the DSMG does not depend on the loss angle directly but instead
depends on the relaxation time τ , see Eq. 5.13. We have used the formula Q = τ ω0 /2
to transform the results of Fig. 5.4 into Fig 5.5. Our experimental results in Fig 5.5
clearly favour a frequency of ≈ 80 Hz (high relaxation time equals low thermal noise)
but the main conclusion is that relaxation time is roughly independent of frequency.
5.6 Noise measurements
To test the performance of the DSMG, we conducted a series of noise measurements

in the laboratory. For the first experiment we used the flat ribbon as the sensing
element of the DSMG. Each recording was of duration 200 s. The main source of
noise in the DSMG is thermal noise. To vary the amount of thermal noise we varied
the Q factor of the ribbon (see Eq. 5.12). And to vary the Q factor we varied the air
pressure (see Eq. 5.21).
Fig. 5.6 shows the frequency spectrum of DSMG noise between 0.1 Hz and 1 Hz.
The graph shows a 0.88nT/m signal with frequency 0.25 Hz, which we produced for
this recording so that we could calibrate the noise floor. Measurements were taken
10
0.88 nT/m signal
Gradient noise (nT/m/√ Hz)
1
1 atmosphere
1700 Pa
500 Pa
90 Pa
34 Pa
29 Pa
0.1
0.1 0.2 0.3 0.4 0.5 1.0
Frequency (Hz)
Figure 5.6: Spectral distribution of DSMG noise whilst using a flat ribbon as the sensing
element. The graph shows that the noise level depends on the air pressure. A calibration
√
signal of 0.88 nT/m manifests as a spectral density of 8.8 nT/m/ Hz rms because the
frequency resolution is 0.005 Hz.
at 6 different air pressures between 29 Pa and atmospheric pressure. By lowering the

pressure we achieved a significant reduction in the noise.
A low pass analogue filter was used to suppress noise at frequencies above 0.5 Hz.
Fig. 5.7 then presents the 200 s long recordings in time domain format. With the
DSMG operating at atmospheric pressure, the 0.88 nT/m sine wave calibration signal
is swamped by the noise. In a 29 Pa vacuum, the thermal noise is lower so the sine
wave signal can be seen clearly.
There was no control system for the vacuum pump. All vacuum valves were
Ribbon at atmosphere
2
Magnetic gradient (nT/m) 0
−2
Ribbon in 29 Pa vacuum
2
−2
0 50 100 150 200
Time (s)
Figure 5.7: Quiet recordings of the DSMG in the laboratory whilst using a flat ribbon as
the sensing element. Both graphs include a 0.88 nT/m signal at a frequency of 0.25 Hz. The
noise level of the magnetic gradiometer is much lower in vacuum.
manually operated. Because of this, the pressure tended to drift slowly during the
course of our measurements. This introduces some additional low frequency noise
which can be seen in Fig. 5.6 at frequencies below 0.13 Hz.
In our second experiment we used the round wire as the sensing element inside
the DSMG. The vacuum pump was turned off and the DSMG was left running for 3
hours. Only after the pressure and magnetic gradient drift had stabilised, did we take
any measurements. The noise spectrum displayed in Fig. 5.8 has the characteristics
of white noise.
A current of 0.3 A was applied to both the flat ribbon and the round wire, during
all noise measurements. This causes the strings to heat up as they thermally dissipate
electrical power. We calculated the temperature of a string by measuring changes in
its resistivity. The exact dependance of aluminium resistivity on temperature is well
documented by Desai et al.[31], whereas the dependance of resistivity on impurities
is accounted for using Matthiessen’s rule[31].
We found that the ribbon grew hotter as the air pressure decreased. This caused
the ribbon resonant frequency to drop in a similar way to that described in section
5.5.3. The DSMG software is designed for resonant frequencies between 400 Hz and
900 Hz. In air pressures below 29 Pa, the ribbon frequency dropped below 400 Hz so
5.7. OPTIMISING THE MAGNETIC GRADIOMETER 133
Gradient noise (nT/m/√Hz)

0.5
0.4
0.3
0.2
0.1
0.1 0.2 0.3 0.4 0.5 1
Frequency (Hz)
Figure 5.8: Spectral distribution of DSMG noise whilst using a round wire as the sensing
element. The DSMG is in a 10000 Pa vacuum.
no more measurements could be taken.

At each of the 6 different air pressures, we recorded the ribbon’s temperature, Q
factor and relaxation time. These values were then used to calculate the thermal noise
limited sensitivity of the ribbon using Eq. 5.13. The electronic noise was calculated by
√
inserting the readout sensitivity of 10−11 m/ Hz into Eq. 5.9. These two theoretical
noise levels are presented in columns 5 and 6 of Table 5.2. The experimental noise
(from Fig. 5.6 and Fig 5.8) is presented in column 7. The experimental noise values
are approximately 20% to 40% higher than the theoretical noise values.
Thermal noise is the main source of noise at all pressures, for both ribbon and wire.
Electronic noise makes a measurable contribution to the total noise at atmospheric
pressure but becomes negligible at low pressures. This is because thermal noise scales
√
proportional to 1/ Q whereas electronic noise scales proportional to 1/Q.
5.7 Optimising the magnetic gradiometer
We have shown that the sensitivity of the DSMG can improved by reducing the
thermal noise. The equation which specifies the thermal noise limited sensitivity, Eq.
5.12, has many parameters that can be adjusted to improve performance. However,
we will show that there is a trade off between temperature, current and the Q factor
Form Pressure Relaxation String Thermal Electronic Measured

time temperature noise noise total noise
√ √ √
P (Pa) τ (s) T (K) (nT/m/ Hz) (nT/m/ Hz) (nT/m/ Hz)
Ribbon 1 atmosphere 0.06 ± 0.01 325 ± 10 0.39 ± 0.03 0.34 ± 0.05 0.65 ± 0.10
1700 ± 800 0.08 ± 0.03 326 ± 10 0.34 ± 0.06 0.33 ± 0.13 0.60 ± 0.10
500 ± 90 0.12 ± 0.03 332 ± 10 0.28 ± 0.04 0.16 ± 0.04 0.42 ± 0.05
90 ± 9 0.25 ± 0.03 357 ± 20 0.20 ± 0.02 0.06 ± 0.01 0.30 ± 0.03
34 ± 3 0.44 ± 0.01 420 ± 50 0.17 ± 0.02 0.036 ± 0.002 0.26 ± 0.03
29 ± 3 0.68 ± 0.08 440 ± 50 0.14 ± 0.03 0.029 ± 0.004 0.18 ± 0.02
Wire 10000 ± 6000 0.38 ± 0.02 310 ± 10 0.30 ± 0.02 0.29 ± 0.03 0.65 ± 0.10
Table 5.2: Results from laboratory noise measurements. The first 4 columns show the
operating conditions. The 5th column shows the thermal noise limited sensitivity calculated
from Eq. 5.13 whereas the 6th column shows the electronic noise calculated using Eq. 5.9.
The last column shows the measured noise level from Figs. 5.6 and 5.8. The pressure was
measured with a Pirani gauge which is less accurate at high pressures.
of the string.
5.7.1 Optimal temperature
Thermal noise creates an rms displacement of the string which is proportional to the
square root of the string’s temperature[16]. In section 5.4.3, we compared the thermal
noise level to the magnetic gradient signal and found that the thermal noise limited
sensitivity is inversely proportion to current.
√
T
Byz ∝ (5.27)
i
Initially, the thermal noise limited sensitivity will improve as more current is
applied to the string. However, as greater amounts of current are applied to the
string, the string’s temperature will rise above room temperature. Ultimately, too
much current in the string will cause an excessive temperature rise and the sensitivity
will degrade.
The optimal level of current in the string will depend on the string’s dimensions
and material. Instead, it is more convenient to specify the optimal string temperature
which only depends on the string material. High currents dissipate larger amounts of
power so large temperature gradients are required. Thus the full dependance of the
thermal noise limited sensitivity Byz upon the string temperature T is:
50
Noise (arbitrary units)

√
T
Byz ∝ (5.28)
$i
T 40
∝ (5.29)
T − Tair
T
∝√ (5.30)
T − 300 30
300 600 900
Temperature (K)
Figure 5.9: Sketch of Eq. 5.30.
where = 4.0 × 10−8 + 1.13 × 10−10(T − 300) is the resistivity of aluminium 6061-
T6511, T is the string temperature and Tair ≈ 300 K is the air temperature. Eq.
5.30 is only an approximation because it assumes Q is independent of T and ignores
thermal radiation. Both intrinsic material losses and radiation are more important
in high vacuum, although the temperature dependence of these two factors tend to
cancel out.
Fig. 5.9 is a sketch of Eq. 5.30 in arbitrary units. The sketch indicates that
the thermal noise limited sensitivity should be optimal when the current has raised
the string to a temperature of ≈ 600 K. Once this temperature has been reached,
applying more current will no longer help the sensitivity. The sensitivity will still be
near its optimal value when the string temperature is between 370 K and 900 K.
We measured the temperature of both the flat ribbon and round wire as a function
of current, at atmospheric pressure and in a 1.4 Pa vacuum. As discussed previously
in section 5.6, we determined the temperature of the string by measuring changes in
its resistivity. The straightforward results presented in Fig 5.10 show that the tem-
perature increases quadratically with current. The measured values of temperature,
T and current, i, together with the Q factors from Fig. 5.3, were inserted into Eq.
5.12 to calculate the thermal noise limited sensitivity. The theoretical value of the
sensitivity is then presented in Fig. 5.11.
Fig. 5.11 plots the indirect relationship between the thermal noise limited sensitiv-
ity Byz and the string temperature T , determined from experiment. The conclusion
from this graph is that we should apply sufficient current to the string so that it
900
Ribbon in 1.4 Pa vacuum
Temperature of String (K) 800 Wire at atmosphere
Wire in 1.4 Pa vacuum
700
600
500
400
300
0 0.5 1 1.5 2 2.5 3
Peak AC Current (A)
Figure 5.10: This graph shows measurements of the change in temperature when alternat-
ing current is applied to the flat ribbon or round wire.
Calculated gradient noise (nT/m/√Hz)
1
Ribbon in 1.4 Pa vacuum
Wire at atmosphere
Wire in 1.4 Pa vacuum
0.1
0.01
300 350 400 450 500
Temperature of Wire (K)
Figure 5.11: The data points in this graph have been adapted from Fig 5.10. As more
current is applied to the string, the noise level should decrease (see Eq. 5.12) and the
temperature should increase (see Fig. 5.10). This graph plots the indirect relationship
between temperature and noise.
heats to a temperature of at least 370 K. The amount of noise reduction that could
be achieved by using even higher currents is less than 15%. Fig. 5.11 and Eq. 5.30
show that this temperature minimum is true for both flat ribbon and round wire, in
air and in vacuum.
5.7.2 Maximum temperature
If the temperature is too high then it will degrade the mechanical properties of the
string. The amount of stress/tension in the string sets the resonant frequency of the
magnetic gradiometer (see Eq. 5.5). Low levels of stress relaxation are desirable to
ensure a long operating life for the DSMG. Table 5.3 indicates that the temperature
should therefore be as low as possible.
Temperature Loss in stress Yield strength Maximum

after 10000 after 10000 string length
(K) hours (%) hours (MPa) (m)
298 5 275 0.427
373 16 270 0.422
423 43 215 0.377
450 62 150 0.315
478 67 95 0.251
Table 5.3: This table shows that high temperatures degrade the mechanical properties
of the aluminium alloy 6061-T6511. The data in the second and third columns are taken
from Kaufman[32]. Electrical currents may also introduce additional stress relief above and
beyond the values shown here due to the electroplastic effect[33].
The problem of stress relaxation could be overcome by replacing one of the clamps
with a spring that would keep the string permanently at the same tension. However,
the same relaxation process that causes stress relaxation will also cause the Q factor
of the string to decrease[34][35] and hence increase the thermal noise.
Table 5.3 shows that long exposure to high temperatures will lower the yield
strength of the aluminium 6061 alloy. This will reduce the maximum frequency of
the string. The resonant frequency formula, Eq. 5.5, can be rewritten in the form:
$
nπ σ
ω0 = (5.31)
L ρ

nπ Y
< (5.32)
L ρ
where ω0 is the angular resonant frequency, n = 2 is the mode number, L is the

length of the string, σ is the stress in the string, ρ = 2690 kgm−3 is the density of
aluminium 6061 and Y is the yield strength. To isolate the string from vibration
noise, the string resonant frequency should be in the range 400 Hz to 900 Hz (see
section 5.4.2). For robust operation, the string should be strong enough to handle a
resonant frequency of 750 Hz. The maximum string length is then:

nπ Y
L< (5.33)
ω0 ρ
√
< 2.6 × 10−5 × Y (5.34)
The above formula is independent of the string’s thickness and can be used for
both the flat ribbon and the round wire. We have calculated the maximum string
length for 5 different yield strengths (the yield strength depends on the temperature)
and presented the results in the last column of Table 5.3.
According to Eq. 5.12, the thermal noise limited sensitivity Byz depends very
strongly on the string length, Byz ∝ L−3/2 . Therefore, the string length should be
as long as possible. Table 5.3 shows that the temperature cannot be increased above
400 K without significantly reducing the string length.
A string temperature between 370 K and 400 K is optimum and is a fair compro-
mise between high current (see Fig 5.10) and high strength (see Table 5.3).
5.7.3 Optimum pressure
In section 5.6, we showed that by reducing the pressure, we can reduce the thermal
noise of the DSMG. The low air pressure also causes the temperature of the string to
rise. However, in section 5.7.2, we found that the optimum string temperature was
between 370 K and 400 K. In order to keep the temperature within this range, the
current must be reduced when the air pressure is low.
The electrical power in the string can be thermally dissipated in three different
ways: conduction by colliding air molecules Pair , conduction along the axis of the
metal string Pmetal , or thermal radiation Pradiation . Approximate formulae are given
by Ubisch[18]:
Pair = α P S(T − Tair )

Pmetal = 20 i2 κ(T − Tair )
Pradiation = ςS(T 4 − Tair

4
) (5.35)
where α = 0.96 ± 0.01 is the thermal accommodation coefficient of air on ma-

√
chined aluminium[36], = (Cv + 12 R)/( 2πRMTair ) = 1.17 ms−1 K−1 is the mole-
cular conductivity of air, P is air pressure, S is the surface area of the string, Cv =
20.85 JK−1 mol−1 is the specific heat of air at constant volume, R = 8.314 JK−1 mol−1
is the universal gas constant, M = 0.02897 kg mol−1 is the molar mass of air, 0 =
4.0 × 10−8 Ωm is the resistivity of aluminium 6061-T6511 at room temperature,
κ = 185WK−1 m−1 is the conductivity of aluminium 6061-T6511, i is the peak AC
current, is the emissivity of the string, ς = 5.67 × 10−8 Wm−2 K−4 is the Stefan
Boltzmann constant, T is string temperature and Tair = 300 K is the ambient air
temperature. The formula for air conduction assumes that the the mean free path
of air molecules is longer than the string. The formula for metal conduction is only
valid for Pmetal << Pair + Pradiation .
We fit the ’in vacuum’ data from Fig. 5.10 into the above formulae. The fitted
emissivity values for the flat ribbon and round wire are:
ribbon = 0.065 ± 0.03 (5.36)
wire = 0.16 ± 0.06 (5.37)
These emissivity values are consistent with the work of Haugh[37] who measured
the emissivity of rolled aluminium 6061 to be 0.066 at 533 K. Haugh found that
the emissivity varies considerably with surface condition and temperature (∂ /∂T =
4.4 × 10−5 K−1 ). Sandblasting or heavily oxidising the surface of the string would
increase the emissivity[37] and thus allow the the string to carry greater amounts of
electrical power. However, a string with a very rough surface is likely to have a lower
Q factor[38]. These two factors tend to cancel out (see Eq. 5.12) so the affect of
surface roughness on sensitivity is unpredictable.
We applied current to the flat ribbon and round wire at different air pressures
between 1 Pa and atmospheric pressure. At each pressure, we adjusted the current
until the temperature was 373 K (calculated by electrical resistivity measurements).
Fig. 5.12 compares our results with the theory in Eq. 5.35. The right hand axis of

the graph shows the peak AC current calculated using the formula i = 2P/R where
P is total electrical power and R is the string’s resistance.
The electrical power dissipated by both ribbon and wire are in good agreement
with theory when the pressure is below P < 60 Pa. However, above 60 Pa, the power
dissipated is lower than predicted because the assumptions made for the noninter-
acting gas model are no longer valid. At high pressure, the conductivity of air is
independent of pressure[18].
The thermal dissipation results in Fig. 5.12 have the same shape as the vibration
dissipation results in Fig. 5.3. This is because the dissipation process of colliding
with air molecules, is the same in both cases. The data from these two graphs were
inserted into Eq. 5.12 to calculate the thermal noise limited sensitivity each pressure.
The theoretical value of the sensitivity is then presented in Fig. 5.13.
Fig. 5.13 shows that if we keep to the temperature constraint T < 373 K, then
the thermal noise limited sensitivity will be independent of pressure. The sensitivity
will continue to be independent of temperature as long as the loss angle is dominated
by air damping and the thermal dissipation is dominated by air conduction. The
situation changes at very low pressures. Fig. 5.13 shows the round wire’s thermal
noise level is highly dependant on the clamp material in high vacuum.
In Table 5.2, we showed that the contribution of thermal noise to the total noise
budget for a 0.25 m long ribbon could be improved by reducing the pressure and
maintaining a steady current of 0.3 A. The ribbon’s relaxation time is higher at low
Power dissipation in Wire (T < 373 K)

1
Air conduction
Metal conduction 1
Radiation
Current (A)
Power (W)
0.1 Total theory
Experiment
0.3
0.01
0.1
0.001 −3 −2 −1 0 1 2 3 4 5
10 10 10 10 10 10 10 10 10
Pressure (Pa)
Power dissipation in Ribbon (T < 373 K)
1
Air conduction
Metal conduction
Radiation
0.3
Current (A)
Power (W)
0.1 Total theory

Experiment
0.1
0.01
0.03
0.001 −3 −2 −1 0 1 2 3 4 5
10 10 10 10 10 10 10 10 10
Pressure (Pa)
Figure 5.12: As the air pressure is decreased, the current must also be decreased to avoid
overheating. We experimentally adjusted the current at each pressure until the temperature
was 373 K. These experimental results are compared with the theoretical heat dissipation
in Eq. 5.35.
pressures and this will improve the sensitivity, according to Eq. 5.13. However, by
reducing the pressure to 29 Pa, we caused the ribbon’s temperature to increase to
440 ± 50 K and decreased the ribbon’s strength. As shown in Table 5.3 and dis-
cussed in section 5.7.2, we should maintain the string’s temperature between 370 K
and 400 K as a balance between string strength and DSMG sensitivity. Accordingly,
Fig. 5.12 shows that the current should be reduced at low pressures and increased
at high pressures. By taking into account the change in current, Fig 5.13 shows that
the thermal noise limited sensitivity should independent of pressure. Therefore, the
1
Ribbon
Calculated gradient noise (nT/m/√Hz)

Wire with Al clamps
Wire with torlon clamps
Wire with teflon clamps
Ribbon (theory)
Wire (theory)
Teflon clamps
0.1 Torlon clamps
Al clamps
0.01 −3 −2 −1 0 1 2 3 4 5
10 10 10 10 10 10 10 10 10
Pressure (Pa)
Figure 5.13: In this graph we consider the DSMG sensitivity in different air pressures
whilst under the constraint T < 373 K. The data points shown here are calculated from the
data points in Figs. 5.3 and 5.12. The graph shows that the noise floor of the DSMG is
almost independent of pressure. At very low pressures, the loss angle depends on the clamp
material and this affects the sensitivity.
improvements in sensitivity that we reported in Table 5.2, could also be achieved

by operating the DSMG at 1000 Pa, for example, and maintaining the ribbon tem-
perature at 373 K. This would allow us to increase the current from 0.3 A to 0.5 A
(according to Fig. 5.12) and increase the ribbon’s length from 0.25 m to 0.42 m (ac-
cording to Table 5.3). Thus, despite the initial results in Table 5.2, the conclusion is
that a DSMG operating under the constraints of a maximum temperature, will have
a sensitivity which is independent of pressure.
Operating in high vacuum is technically difficult, especially if the DSMG is de-

ployed onboard a survey aircraft. The only remaining reason to operate in vacuum is
to acoustically isolate the DSMG from aircraft engine noise. A moderate vacuum of
1000 Pa is adequate.
5.7.4 Optimum wire diameter
In our laboratory noise measurements in Table 5.2, we found that the round wire and
flat ribbon had similar noise levels at high pressure. However, Fig. 5.13 shows that
if the current is increased until the temperature is 373 K then the round wire should
have better sensitivity. Thicker wires should have better sensitivity because they can
carry more current.
Assuming that all heat is conducted by colliding air molecules (see Eq. 5.35), the
maximum current in a round wire is:

α P π 2 d3 (T − Tair )
i= (5.38)
4
∝ d3/2 (5.39)
where d is the wire diameter. The proportionality shown above is consistent with
Preece’s equation[39]. The density per unit length of a wire is given in Eq. 5.19
whereas the relaxation time of a wire slowed down by air damping is given in Eq.
5.20. By inserting these values into Eq. 5.13, we can present the thermal noise limited
sensitivity of a round wire in the form:
1/4
8 2πM kT
Byz = (5.40)
d RTair L3 α(T − Tair )

2.6 × 10−12 T
≈ (5.41)
d L3 (T − Tair )
The above equation shows that the sensitivity is inversely proportional to the wire
diameter as long as the loss angle is dominated by air damping. However, very thick
wires will be very stiff. The stiffness EI depends on the moment of area, which is[20]:
ab3 πd4
Iribbon = Iwire = (5.42)
12 64
The tension in the wire F = d2 ρL2 ω02/(16π) is substituted into Eq. 5.26. The
violin mode and air damping losses at pressure P = 1000 Pa can then be written in
the form:

1 πd E 1
≈ 2 × (5.43)
Qviolin L ω0 ρ Qmaterial
1 0.95
≈ (5.44)
Qair ρω0 d
The wire diameter will be optimum when the above two losses are equal. If the
wire is L = 0.42 m long, then the optimum wire diameter is:
1/4
Q2material
d ≈ 0.55L (5.45)
Eρ
≈ 0.0012 ± 0.0004 m (5.46)
where E ≈ 66 GPa[40] is the Young’s modulus of aluminium 6061 at 373 K,

ρ ≈ 2680 kgm−3 [25] is the density of aluminium 6061 at 373 K and Qmaterial = 400 is
the intrinsic loss angle of the string measured in section 5.5.3.
By extrapolating Fig. 5.12 with Eq. 5.39, a wire with diameter 1.2 mm should
be able to handle a current of i ≈ 40 A in a 1000 Pa vacuum. The Q factor of the
wire would be approximately Q ≈ 8000 and the thermal noise limited sensitivity of
√
the DSMG could be as good as Byz ≈ 0.005 nT/m/ Hz rms.
Instead of a round wire, we could use a ribbon with the same thickness b = 1.2 mm
but much larger width a >> 1.2 mm. A wider ribbon could carry larger currents and
achieve better sensitivity. However, a peak AC current of i = 40 A is already too
high. The ultimate sensitivity of the magnetic gradiometer may depend on how much
power is feasible during deployment in the field.
5.8 Sensitivity as a Function of the Dissipated Power
In a practical application (i.e. airborne, surface and borehole deployment), the power
to the DSMG will be limited by the available supply. Taking this into account, we
can derive a new equation for the thermal noise limit. By inserting the formula

i = 2P/R = (2P η)/(Lρ) into equation 5.12, we can present the thermal noise
limited sensitivity in the form:
5.9. COMPARISON OF MATERIALS 145

4π ρkT ω0
Byz = (5.47)
L 2P Q
If the length of the wire is limited by the yield strength (see Eq. 5.33) and the wire
diameter’s is set to optimum (see Eq. 5.45), then the thermal noise limited sensitivity
is:
1/8 $
Eρ7 4 f03 T
Byz ≈ 6.1 × 10−11 (5.48)
Y 6 Q2material P
1/8
2.4 × 10−5 Eρ7 4
≈ √ (5.49)
P Y 6 Q2material
where k = 1.38 × 10−23 JK−1 is Boltzmann’s constant and f0 = 750 Hz is the

resonant frequency of the second violin mode. All the variables inside the brackets in
Eq. 5.49 are material specific properties; this allows us to conduct a comparison of
materials.
5.9 Comparison of materials
Until now, we have assumed that the string must be made of aluminium 6061-T6511.
Here we compare this aluminium alloy with other possible materials for the DSMG.
Eq. 5.49 shows that the thermal noise limited sensitivity depends on the following
material parameters: Young’s modulus E, density ρ, resistivity , yield strength Y
and the mechanical quality factor Qmaterial . We have presented the values of these
parameters for 9 different materials in Table 5.4. These parameters are then substi-
tuted into Eq. 5.49 to calculate the sensitivity of a DSMG using only 1 watt of power.
If the power is kept low then we can assume T ∼ 300 K. The results are shown in
the last column of the table.
The noise levels in the table compare favourably with the sensitivity of 0.01 nT/m
recommended by Clarke and Schmidt[2] for airborne geophysical applications. Alu-
minium 6061 is the best material for the DSMG because of its low density and low
resistivity. Pure copper and aluminium also have low resistivity but the yield strength
is too low so the string would have to be very short.
Material E ρ Y Qmaterial Reference Byz for 1 watt

√
(GPa) (kgm−3 ) (10−8 Ωm) (MPa) () (nT/m/ Hz)
Al 6061-T913 69 2690 4.0 455 1000 [26; 41; 42] 0.006
BeCu Hard 130 8360 7.8 1000 20000 [43; 44] 0.007
Al 6061-T6 69 2690 4.0 275 1000 [26; 41; 42] 0.009
Beryllium 300 1850 4.1 180 1000 [52; 42] 0.011
Niobium 105 8600 15.1 240 180000 [45; 46] 0.016
Marval 18 Steel 186 8000 60. 1500 10000 [45; 43; 47] 0.016
Tungsten 411 19300 5.4 550 6000 [43; 48] 0.028
Pure Cu 100 8900 1.7 25 480 [49; 50] 0.058
Titanium 120 4540 54. 140 1000 [45; 43; 48] 0.094
Pure Al 33 2700 2.7 10 150 [51; 49] 0.130
Table 5.4: Room temperature properties of various materials and the references for those
properties. The last column shows what the sensitivity of the DSMG would be if this
material was used to make the string and 1 watt of power was consumed. For Beryllium and
Be/Cu, the proportional limit is shown instead of the yield strength. The Q factors shown
here apply for wires with diameters less than 1 mm.
Applying a T913 heat treatment[41] to the aluminium 6061 alloy would strengthen
the metal and allow longer strings to be used. Table 6.2 shows that this could improve
the sensitivity by 30%.
5.10 Conclusion
The DSMG is able to detect magnetic gradients using only a single string as the
√
sensing element. We have shown that the noise floor of the device is 0.18 nT/m/ Hz
rms in laboratory tests. We then showed that the sensitivity can be further improved
by using a longer string to increase the baseline of the gradiometer or by applying a
greater current to increase the signal to noise ratio. Sensitivity is ultimately limited
by how much current can be applied to the string without overheating the metal or
degrading the mechanical properties.
According to our results, the DSMG works best at temperatures between 370 K
and 400 K. A vacuum of 1000 Pa is required for the mechanical isolator but the
DSMG’s performance is otherwise independent of pressure. The optimum string
thickness is approximately 1.2 mm, the optimum current is 40 A and the optimum
length is 0.42 m. The string is made of aluminium 6061 and we have shown that
5.11. APPENDIX A: ELASTIC ENERGY AND DYNAMIC TENSION 147
this is indeed the best material for the gradiometer. Theoretically, these improve-
ments will allow the string to detect magnetic gradients with sensitivity better than
√
0.01 nT/m/ Hz rms per watt of power, which is more than sufficient for geophysical
applications.
5.11 Appendix A: Elastic energy and Dynamic ten-

sion
For low amplitudes of vibration, the string will behave linearly and the dynamic
tension will be very low. However, a phenomenon known as modal crossover can
create very large amounts of dynamic tension[11]. Whether or not crossover occurs
depends on the sag parameter defined in Eq. 5.4.
When the sag parameter is small, λ → 0, the violin modes will be that of a perfect
taut string. That is, the frequency of the first antisymmetric mode will be twice that
of the first symmetric mode. However, when λ > 2π, the first antisymmetric mode
will have the lowest resonant frequency. And at λ = 2π, the frequency of the first
symmetric mode and the frequency of the first antisymmetric mode are equal. Modal
crossover occurs at λ = 2π[11].
At crossover, both the dynamic tension and the elastic energy will be large. A
formula by Irvine[11] gives the fraction of energy that can go into elastic energy for
the first symmetric mode:
2
Veleastic 3
=
2 (5.50)
Velastic + Vgrav λ2 tan( 12 kL)
1+ 1
12 2
kL

where k = ω η/F is the wavenumber of the string, η is the mass per unit length
and ω is the angular frequency of vibration. The above equation reaches a maximum
at 2/3 when λ = 2π.
Irvine’s[11] result showed that dynamic tension exists only for symmetric modes.
However, Triantafyllou and Grinfogel[24] found that a small amount of dynamic ten-
sion can also exist for anti-symmetric modes of vibration. The dynamic tension can
be expressed in the form:
ηω 2L
ΔF (t) = 2 sin2 ( 12 kL) + 7
4
sin(θ) A0 exp(iωt) (5.51)
cos(θ)
where = ηgL/F is the ratio of weight to tension, F is tension, η is the mass per

unit length, ω is the angular frequency of vibration, k = ω η/F is the wavenumber
of the string, L is the length of the string, g is acceleration due to gravity and θ is
the string’s inclination relative to the horizontal. A0 is the same arbitrary amplitude
that Triantafyllou and Grinfogel used to express mode shape:
%
7
x(z, t) = A0 exp(iωt) sin(θ)(cos(kz) − 1)
4

2 2 &
k L
− sin(ks) kL − 1 + sin(kL) (5.52)
λ2
The above expression is correct to zeroth order[53]. Both Eq. 5.11 and Eq. 5.52
can be simplified using the fact that kL ≈ 2π for the first antisymmetric mode. The
gravitational potential energy is equal to:
L 2
F ∂x
Vgrav = dz (5.53)
2 0 ∂z

2 2 2
π2F 4π 49 2
sin (θ)
= 4π 2 1 − 2 + (5.54)
L λ 16
whereas the elastic energy depends on the dynamic tension squared:
(ΔF )2 L
Velastic = (5.55)
2EA
49n2 ω 4 L3 tan2 (θ)
= (5.56)
32EA
where E is Young’s modulus and A is the string’s cross sectional area. We can
compare equations 16 and 17 of Gonzalez and Saulson’s paper[10] to show that the
tension, frequency of vibration and the ratio of bending energy to gravitational po-
tential energy are all interrelated:

F nπ Vbending
ω≈ 1+ (5.57)
η L Vgrav
5.11. APPENDIX A: ELASTIC ENERGY AND DYNAMIC TENSION 149
Eqs. 5.53,5.55, and 5.57, can then be combined to determine the fraction of energy
stored as elastic energy:
Veleastic 8π 2
≈ 2 (5.58)
Vtotal 64π 2 λ2 4π 2
1− 2 + 8π 2 + λ2
49 2 sin2 (θ) λ
The elastic energy fraction shown above will be tiny except when λ ≈ 2π. We
can neglect the cases when the elastic loss angle is small because other losses such as
bending loss will dominate. If we restrict ourselves to the case λ ∼ 2π, then Eq. 5.58
will simply to:
Veleastic 2
≈
2 (5.59)
Vtotal 8(λ − 2π)
3+
7 sin(θ)
The above equation shows that the elastic energy of the first anti-symmetric mode
will reach a maximum of 2/3 when λ = 2π, similar to Irvine’s result for symmetric
modes[11]. The violin mode loss angle is:
1 2 1
=
2 × (5.60)
Qviolin 8(λ − 2π) Qelastic
3+
7 sin(θ)
where 1/Qelastic is the material loss angle of the aluminium string measured under
zero tension[20].
At 430 K, the Young’s modulus of aluminium 6061-T6 is approximately E ≈ 63 ±
3 GPa[40]. The mass per unit length of the ribbon is (8.4±0.8)×10−6 kgm−1 , angle of
inclination is θ = 10 degrees and the cross sectional area is A = (3.2 ± 0.3) × 10−9 m2 .
These values were substituted into Eq. 5.4 and evaluated using formulae by Zui et
al.[12] for the second violin mode. We find that the resonant frequency of the ribbon
is 55 ± 5 Hz when λ = 2π. At this frequency, there should be a peak in the violin
mode loss angle.
Acknowledgements
The authors would like to thank Mr. Howard Golden, Mr. Slawomir Gras and Ms.
Xu Chen for many useful discussions and suggestions. Work on the DSMG project is
funded in part by a linkage grant from the Australian Research Council.
Bibliography
[1] P. Killeen, Exploration trends & development in 2008, Supplement to The North-
ern Miner 95, No. 1, 6-15 (2009).
[2] P. Schmidt, D.A. Clark, The magnetic gradient tensor: its properties and source
characterization, The Leading Edge 25, 75-78 (2006).
[3] K. Leslie, K. Blay, D. Clark, P. Schmidt, D. Tilbrook, M. Bick, C. Foley, R.

Binks, Helicopter trial of magnetic tensor gradiometer, ASEG Extended Ab-
stracts, 2007.
[4] R. Stolz, V. Zakosarenko, M Schulz, A. Chwala, L. Fritzsch, H.G. Meyer, E.O.

[5] R.J. Prance, T.D. Clark, H.Prance, Compact broadband gradiometric induction
magnetometer system, Sensors and Actuators A 76, 117-121 (1999).

stability and linearity, Sensors and Actuators A 120, 71-77 (2005).
[7] A. V. Veryaskin, Magnetic gradiometry: a new method for magnetic gradient

measurements, Sensors and Actuators A 91, 233-235 (2001).
proportional and integral feedback, International Journal of Applied Electromag-
netics and Mechanics 29, 197-215, (2009).
[9] C. Zhao, L. Ju, H. Miao, S. Gras, Y. Fan, and D.G. Blair, Three-mode op-
toacoustic parametric amplifier: a tool for macroscopic quantum experiments,
Physical Review Letters 102, 243902 (2009).
[10] G.I. Gonzalez, P.R. Saulson, Brownian motion of a mass suspended by an anelas-
tic wire, Journal of the Acoustical Society of America 96, 207-212 (1994).
[11] H.M. Irvine, Energy relations for a suspended cable, The Quarterly Journal of
Mechanics and Applied Mathematics 33, 227-234 (1980).
BIBLIOGRAPHY 151
[12] H. Zui, T. Shinke, Y. Namita, Practical formulas for estimation of cable tension
by vibration method, Journal of Structural Engineering 122, 651-656 (1996).
[13] D.R. Bland, Vibrating strings: an introduction to the wave equation, 1st Edition
(Routledge and Kegan Paul, London, 1960), p. 41.
[14] A. Sunderland, L. Ju, D.G. Blair, W. McRae, H. Golden, Low magnetic suscep-
tible materials and applications to magnetic gradiometry, Smart materials and
Structures 18, 095038 (2009).
23, 790-793 (2004).
[16] P.R. Saulson, Thermal noise in mechanical experiments, Physical Review D 42,
2437-2445 (1990).
[17] H.B. Callen, R.F. Green, On a theorem of irreversible thermodynamics, Physical

Review 86, 702-710 (1952).
[18] H. Von Ubisch, On the conduction of heat in rarefied gases and its manometric
application I, Appliced Scientific Research A 2, 364-402 (1951).
[19] R.G. Christian, The theory of oscillating-vane vacuum gauges, Vacuum 16, 175-
178 (1966).
[20] A.M. Gretarson, PhD thesis, “Thermal noise in low loss flexures”, Syracuse Uni-
versity, 2002, p. 43.
[21] W. Duffy Jnr, Acoustic quality factor of aluminium and selected alloys from
50mK to 300K, Cryogenics 42, 245-251 (2002).
[22] M. Kochi, S. Isoda, R. Yokota, H. Kambe, Molecular aggregation of solid aro-

matic polymers. IV. Dynamic mechanical properties of aromatic polyamideimide
film, Journal of polymer science 24, 1619-1623 (1986).
[23] N.G. McCrum, B.E. Read, G.Williams, Anelastic and dielectric effects in poly-
meric solids (Wiley, New York, 1967), p. 453.
[24] M.S. Triantafyllou, L. Grinfogel, Natural frequencies and modes of inclined ca-
bles, Journal of Structural Engineering 112, 139-148 (1986).
[25] A.J.C. Wilson, The thermal expansion of aluminium from 0 to 650 C, Proceedings
of the Physical Society 53, 235-244 (1941).
[26] G. Szenes, D Zsambok, Grain boundary relaxation in Al-Mg-Si alloys, Physica

Status Solidi (a) 21, K105-K107 (1974).
[27] T. Suzuki, K Tsubono, H. Hirakawa, Quality factor of vibration of aluminum

alloy disks, Physics Letters A 67, 2-4 (1978).
[28] X. Liu, E. Thompson, B.E. White, Jr., R.O. Pohl, Low-temperature internal fric-
tion in metal films and in plastically deformed bulk aluminum, Physical Review
B 59, 11767-11776 (1999).
[29] E. Carreno-Morelli, S.E. Urreta, A.A. Ghilarducci, High temperature damping

in Al-Mg-Si industrial alloys, Physica Status Solidi A 158, 449-462 (1996).
[30] C. Belamri, S. Belhas, A. Riviere, Damping of a high-purity aluminum single

crystal at high temperatures, Materials Science and Engineering A 442, 142-146
(2006).
[31] P.D. Desai, H.M. James, C.Y. Ho, Electrical resistivity of aluminum and man-
ganese, Journal of Physical and Chemical Reference Data 13, 1131-1172 (1984).
[32] J.G. Kaufman, Tensile, creep and fatigue data at high and low temperatures (The
Aluminum Association, Washington DC, 1999), p. 166.
[33] M. Braunovic, V.V Konchits, N.K. Myshkin, Electrical Contacts: Fundamentals,

Applications and Technology (CRC Press, Boca Raton, 2007), p. 242.
[34] T. Ke, Experimental evidence of the viscous behavior of grain boundaries in

metals, Physical review 71, 533-546 (1947).
[35] Q.P. Kong, H. Zhou, X.M. Li, X. Wang, Dynamic internal friction of mono-
crystalline aluminum during high temperature creep, Scripta Metallurgica et
Materialia 32, 1043-1048 (1995).
BIBLIOGRAPHY 153
[36] M.L Wiedmann, P.R. Trumpler, Thermal accommodation coefficients, Transac-

tion of ASME 68, 57-64 (1946).
[37] M.J. Haugh, Radiation Thermometry, edited by D.P. DeWitt and G.D. Nutter
(Wiley, New York, 1988), Ch. 17.
[38] V.B. Braginsky, V.P. Mitrofanov, V.I. Panov, Systems with small dissipation
(University of Chicago Press, Chicago, 1985), p. 21.
[39] W.H. Preece, On the heating effects of electric currents, Proceedings of the Royal
Society of London 36, 464-471 (1883).
[40] F. Augereau, D. Laux, L. Allais, M. Mottot, C. Caes, Ultrasonic measurement of

anisotropy and temperature dependence of elastic parameters by a dry coupling
method applied to a 6061-T6 alloy, Ultrasonics 46, 34-41 (2007).
[41] Aluminum Standards and Data (The Aluminum Association, Washington DC,
2000).
[42] M. Levine, C. White, Material damping experiments at cryogenic temperatures,

Proceedings of SPIE 5179, 165-176 (2003).
[43] G. Cagnoli, L. Gammaitoni, J. Kovalik, F. Marchesoni, M. Punturo, Low-

frequency internal friction in clamped-free thin wires, Physics Letters A 255,
230-235 (1999).
[44] ASM Specialty Handbook: Copper and Copper Alloys, edited by J.R. Davis
(ASM International, Materials Park, 2001), p. 473.
[45] M. Baker, L. Ju, D.G. Blair, The Q-factor of flexure membranes, Measurement
Science and Technology 12, 1666-1671 (2001).
[46] K. D. Maglić, N.Lj. Perović, G.S. Vuković, Lj.P. Zeković, Specific heat and elec-
trical resistivity of niobium measured by subsecond calorimetric technique, In-
ternational Journal of Thermophysics 15, 963-972 (1994).
[47] Tool materials, edited J.R. David (ASM International, Materials Park, 1995), p.
151.
[48] Goodfellow Cambridge Limited, material properties.

https://www.goodfellow.com
[49] K. Kasahara, K. Yamamoto, T. Uchiyama, S. Miyoki, M. Ohashi, K. Kuroda,

T. Tomaru, T. Suzuki, T. Shintomi, Study of heat links for a cryogenic laser
interferometric gravitational wave detector, Proceedings of the 28th International
Cosmic Ray Conference, Trukuba, Japan, 3115-3118 (2003).
[50] B. N. Singh, D.J. Edwardsb, P. Tofta, Effect of neutron irradiation and post-
irradiation annealing on microstructure and mechanical properties of OFHC-
copper, Journal of Nuclear Materials 299, 205-218 (2001).
[51] W. Wasserbäch, S. Abens, S. Sahling, R.O. Pohl, E. Thompson, Low-

temperature acoustic and thermal properties of plastically deformed, high-purity
polycrystalline aluminum, Physica Status Solidi (b) 228, 799-823 (2001).
[52] N.J. Petch and E. Wright, The plasticity and cleavage of polycrystalline beryl-
lium. I. Yield and Flow Stresses, Proceedings of the Royal Society of London A
370, 17-27 (1980).
[53] L. Grinfogel, PhD thesis, “Dynamics of elastic taut inclined cables” Massachusetts
Institute of Technology, 1984, p. 42.
Chapter 6
Space Applications
6.1 Preface
This chapter is based upon a proposed version of the Direct String Magnetic gra-
dioemter that would be deployed in space. The initial modelling featured extremely
high currents and impossibly long strings. Dr. Alexey Veryaskin was responsible for
helping to choose more realistic parameters for the space environment.
The bulk of this chapter is a paper about the space DSMG, published in Sensors
and Actuators A. The paper describes the basic background and principles of the
DSMG with sufficient references to fill in the gaps for readers who wish to study
the subject in depth. Fundamental limits are described rather than derived, with
particular reference to space applications. I go on to describe in theoretical terms how
a gradiometer for space applications could be developed from the existing terrestrial
and airborne DSMG. I show that a space borne gradiometer should in principle, rival
or exceed the sensitivity of offered by SQUID magnetometers, without the use of
cryogenics. My contribution to this paper amounts to 80% of the modelling and 80%
of the manuscript preparation.
155
156 Chapter 6. SPACE APPLICATIONS
Sensors and Actuators A 147 (2008) 529-535
Direct string magnetic gradiometer for space

applications
Andrew Sunderland1 , Alexey V Veryaskin1,2 , Wayne McRae1,2 , Li Ju1 , David G Blair1

Abstract
Recently, a novel direct string magnetic gradiometer (DSMG) has been
developed, where a vibrating wire, driven by an AC current, is used as a
single sensitive element. It is designed to directly measure the local off-
diagonal components of the magnetic gradient tensor, Bxz , Byz and Bxy ,
provided the distance to an object creating magnetic anomalies is much
larger than the length of the string. This requirement is well satisfied
in space, if the sensor is deployed from a satellite platform orbiting near
the planet under surveillance. Current instruments operating at 1 kPa
√
pressure achieve sensitivity of 4 × 10−10 T/(m· Hz) in the band 0.0025-
0.3 Hz. In this paper we show that proposed modifications to the current
gradiometer design, specifically aimed at the deployment in space, could
√
have a magnetic gradient sensitivity better than 10−13 T/(m· Hz) in the
frequency range of interest for specific missions both for fundamental re-
search and for such applications as geophysical exploration on Mars and
other solar system planets. Also, by combining a few single-axis mag-
netic gradiometer modules, it is possible to deploy a full tensor magnetic
gradiometer.
6.2 Introduction
Magnetic gradiometry, as a powerful tool for magnetic anomaly mapping, have been
discussed in literature in conjunction with future planetary and deep space missions
(see for example Alves and Madeira[1; 2]). Although conventional magnetometers are
6.3. STRING MAGNETIC GRADIOMETER 157
more commonly deployed on satellites, interest is growing in the use of magnetic gra-
diometers to extract data that cannot be obtained from magnetic field measurements
alone. Hastings et al.[3] described several advantages of using a magnetic gradiometer
to directly measure magnetic gradients in space. In their paper, some cryogenically
cooled SQUID-based magnetic gradiometer designs have been considered. SQUID-
based magnetic gradiometers are currently under development mainly for airborne
geophysical reconnaissance purposes [4; 5]. On their own, SQUID gradiometers pro-
vide a very high sensitivity to magnetic gradients in the laboratory environment[6],
and require some additional auxiliary equipment and a compensation technique when
deployed from a moving platform[4]. Due to logistical difficulties, the use of SQUIDs
in space has been limited until recently[2][7].
Many fluxgate magnetometers have been previously used in space with sensitivities
√ √
ranging from 10−12 T / Hz to 10−11 T / Hz[8]. They do not require any cryogenic
environment, and fall into a medium range on the sensitivity scale compared to the
SQUID-based magnetometers. In the past few years, fluxgate gradiometers have been
proposed for space missions. The best performance reported to date is 9.3×10−11 T/m
in the band 0.01 − 10 Hz[9].
Existing magnetic gradiometer technology is not sensitive enough for all space
applications. The ST5 mission calculated the magnetic gradient at altitudes from
300 km to 800 km above the Earth by comparing the output of a pair of fluxgate
magnetometers situated on satellites 400 km apart. At an altitude of 400 km, the cal-
culated magnetic gradient varied between −10−13 T/m and +10−13 T/m[10]. Tehro
et al. estimated that chondrite asteroids of diameter 10 km would have magnetic
signatures of approximately 15 nT during flybys at a distance of 200 km[11]. This
corresponds to a magnetic gradient of 3 × 10−13 T/m. The magnetic anomalies on
the surface of Mars are expected to vary between 100 nT and 1000 nT[1; 11].
6.3 String magnetic gradiometer
Recently, a novel Direct String Magnetic Gradiometer (DSMG) has been developed
[12; 13; 14]. It consists of a single string or wire (i.e. an object with transverse dimen-
sions much smaller than its longitudinal dimension) made from the alloy aluminium-
6061. The string is held under tension with its second harmonic oscillation mode at
f0 ≈ 850 Hz. An AC current tuned to the second harmonic is used to drive the string.
This sets the string into a resonant motion due to the Ampere force per unit length:
∂f
= is [ez × B] sin(ωt) (6.1)
∂z
length, is is the amplitude of the AC drive current, ω is the string’s drive angular
frequency, t is time and B is the magnetic induction vector. The physics of the DSMG
is very similar to that of MEMS vibrating wire magnetometers[15]. As per Eq. 6.4,
the DSMG’s longer base length makes it significantly more sensitive to magnetic
gradients compared to the micro-machined devices.
When the string is a stretched thin flat ribbon[14], the resonant motion is strictly
one-dimensional with its sensitivity axis pointing perpendicular to the plane of mo-
tion. In this case, the ribbon represents a one-dimensional mechanical harmonic
oscillator having an infinite number of resonant modes[12]:
d 2d
2
Xn + Xn + ωn2 Xn =
dt
τ dt
2 n 2l dBy is
(1 − (−1) ) By − (−1)
n
sin(ωt) (6.2)
πn πn dz η
It is assumed that the ribbon vibrates in the XOZ plane of its local reference frame
the origin of which is coincident with a lower clamp point. The upper clamp point
determines the ribbon’s length l. Xn (t) is the amplitude of a n-mode mechanical
displacement of the ribbon from its unperturbed position aligned with Z axis. It
is also assumed that all non-linear terms can be ignored as, in fact, the maximum
possible mechanical displacements do not exceed the nanometer scale[16]. In Eq. 6.2,
η is the ribbon’s mass per unit length, and τ is its mechanical relaxation time, which
is the same for all resonant modes of the ribbon. Nn (t) represents the fundamental
thermal noise source (in terms of acceleration noise) which sets an absolute limit on
the sensitivity of DSMGs. It has the following correlation function in the white noise
area[12]:
6.3. STRING MAGNETIC GRADIOMETER 159
8kT

Nn (t1 )Nm (t1 ) = δnm δ(t1 − t2 ) (6.3)
ηlτ
where k = 1.4 × 10−23 J/K is the Boltzmann’s constant and T is the absolute tem-
perature.
As it follows from Eq. 6.2, the magnetic gradient term of the driving force couples
to all resonant modes, while the conventional magnetic field term is only coupled to
the odd ones.
During operation the string oscillator is an integral part of a dynamic feedback
loop[17]. The dynamic properties of such a complex system are different from those
of a stand-alone mechanical oscillator described in previous work[12]. In particular,
there are a number of additional parameters that can play a crucial role in creat-
ing an optimised DSMG with a possibility to implement effective noise suppressing
algorithms, such as electronic cooling[18].
By its very nature, DSMG is a modulation-demodulation device. Like fluxgate
magnetic gradiometers, it is capable of detecting the quasi DC magnetic gradients
both in relative and in absolute units. The detection method provides strong im-
munity to the uniform magnetic field. Firstly the AC drive current does not couple
strongly to the uniform field. The second harmonic drive frequency naturally couples
to the magnetic gradient but is well off the drive frequency that couples to the uni-
form magnetic field. Secondly, the mechanical displacement detection is designed to
preferentially detect the second harmonic oscillations and suppress the fundamental
mode oscillations. A common mode rejection ratio of the order of 107 is naturally
achieved without any balancing technique. The mechanical Q factor of the ribbon
provides first stage amplification of the signal to the level where an instrumental
read-out noise is lower than the fundamental thermal noise of the ribbon. The latter
determines the fundamental rms noise floor of a DSMG[12]:
$
dBy 4π 2ηkT ∗
= (6.4)
dz RM S is l3 τ τ ∗
where, for the current DSMG design, l = 0.25 m is the length of the ribbon, η =
8 × 10−6 kg/m is the mass per unit length, τ ≈ 0.1 s is the relaxation time at a
pressure of 1 kPa, τ ∗ = 5 s is the measurement time and T ∗ is the effective noise
temperature. A complete theory of operation of DSMGs is presented in another

paper[17].
A DSMG designed to date, operates typically at T = 300 K in a 1 kPa vacuum.
An inductive read-out system has been developed in order to detect the gradient
√
driven displacements of the ribbon at a level of 6 × 10−13 m/ Hz. The physics of
the read-out system and its possible improvements are discussed in Section 11 below.
√
The achieved sensitivity of the DSMG is 4 × 10−10 T/(m· Hz) in an unshielded
environment.
Below we consider some possible ways of greatly reducing the DSMG’s thermal
noise limiting factor by using some advantages of the natural space-borne environ-
ment. We show that a DSMG specifically designed for space-borne applications can
be as sensitive as SQUID-based devices without the requirement of using cryogenic
equipment. Also, by their nature, DSMGs should exhibit very low 1/f noise, allowing
measurements within the 0.0002 − 0.17 Hz band specified by Hastings et al.[3].
6.4 Thermal noise and the mechanical quality factor
During deployment in space, the DSMG would operate at pressures between 10−4 Pa
at an altitude of 230 km[19] and 10−8 Pa on the Moon[20]. This means that despite
the large surface area to mass ratio of the ribbon, gas friction damping is negligi-
ble. Measurements of the intrinsic mechanical Q factor of aluminium 6061-T6511 by
Duffy[21] with a cylinder of diameter 6mm give Q factors as high as 2.8 × 105 at
300 K. Increasing the Q factor of the ribbon is one way to reduce thermal noise.
The aluminium ribbon is strained and clamped at both ends. Initially the amount
of stress in tension is proportional to the amount of strain. The tension sets the
resonant frequency of the magnetic gradiometer. Low levels of stress relaxation are
desirable to ensure a long operating life for the DSMG. Table 6.1 shows that the
alloy aluminium-6061 must be tempered before use in order to avoid excessive levels
of stress relaxation. Fully annealed aluminium has lower resistivity but the amount
of stress relaxation is prohibitive. The aluminium alloy 6061-T651 was chosen as
the ribbon material to strike a compromise between low mass-density, low electrical-
6.4. THERMAL NOISE AND THE MECHANICAL QUALITY FACTOR 161
resistivity and high strength.
Temperature Loss in stress after 1000 hours Loss in stress after 10000 hours
Al 6061-O Al 6061-T6 Al 6061-O Al 6061-T6
298 K 18% 4% 27% 5%
373 K 43% 14% 62% 16%
423 K 72% 28% 100% 43%
450 K 100% 47% 100% 62%
Table 6.1: The results in this table are taken from Kaufman[46]. Stress relaxation becomes
increasingly problematic at higher temperatures. Tempering the aluminium (Al 6061-T6)
can reduce stress relaxation compared with fully annealed aluminium (Al 6061-O).
For space deployment it is proposed to use a wide thin ribbon. This would max-
imise the surface area available for thermal radiation which would allow a higher
current to be pumped along the ribbon and yet minimise the mass per unit length.
The high current and low mass would increase the sensitivity as per Eq. 6.4.
The mechanical Q factor of a very thin ribbon is lower than the Q factor of the
bulk metal because of surface losses. Gretarsson et al. give an expression relating
Qribbon to Qbulk [22]:

1 1 ds
= 1+μ (6.5)
Qribbon Qbulk V /S
where V /S is the volume to surface ratio, ds is the dissipation depth of surface loss
and μ is a measure of the fraction of elastic energy attributable to strains at the
surface of the sample. Preliminary experiments by the authors on aluminum-6061
with a ribbon of width 0.125 mm and thickness 0.025 mm in vacuum have measured
a ribbon Q factor of 200 ± 20. No tension was applied to the ribbon which vibrated in
a second order cantilever mode with a resonant frequency of 14 Hz. The experiments
were conducted in a high vacuum of 10−2 Pa to keep gas damping losses to a minimum,
1/Qgas ≈ 10−6 . In such a thin ribbon the surface loss is dominant. Fitting Qribbon =
200 to Eq. 6.5 gives:
1 1 2μds
≈ (6.6)
Qribbon Qbulk t
Using the ribbon thickness of t = 0.025 mm and assuming Duffy’s result of Qbulk =
280, 000, we can estimate that μds = 16 mm.
When tension is applied to the ribbon, the mode of vibration changes from can-
tilever mode to violin mode. This can lead to a higher Q factor, also known as an
enhanced Q factor. Using a formula in Gonzalez et al.[23], the enhanced Q factor is:
$
1 2
1 IE
= (6.7)
Qviolin Qribbon l T
where l is the length of the ribbon, I = (t3 w/12) is the moment of area, w is the
width of the ribbon, E is the young’s modulus of the aluminum alloy and T is the
tension. Substituting in the angular resonant frequency of the second violin mode

ω = (nπ/l) (T /ρtw) into Eq. 6.7 gives:

1 14nπt 3E
= (6.8)
Qviolin Qribbon ωl2 ρ
where ρ = 2690 kgm−3 is the density of aluminum-6061 and n = 2 is the mode

number. The violin mode of Gonzalez et al. refers to a ribbon suspending a test mass
in a gravitational wave detector. The physical explanation of the Q enhancement
factor is that the gravitational potential energy of the test mass is lossless. Violin
mode enhanced Q factors have also been observed by Dawid and Kawamura in a
ribbon clamped at both ends to a fixed aluminium block[24].
In another preliminary experiment, tension was applied to the previously men-
tioned ribbon of dimensions 0.025 mm × 0.125 mm × 0.25 m. The tension was such
that the resonant frequency of the second violin mode was f = 76 Hz. The result
was Qviolin = 18, 000 ± 2000, a proof in principal that the mechanical Q factor of a
very thin aluminum ribbon can be enhanced by applying tension. A further increase
in the tension to make the resonant frequency f = 295 Hz resulted in an even higher
Q factor of Qviolin = 70, 000 ± 10, 000. During the Q factor measurements, 0.16 A
of current was pumped along the ribbon which raised the temperature of the ribbon
to 365 K ± 10 K. This temperature is consistent with the proposed conditions for
deployment in space.
When expressed as a multiple of the bulk Q factor, the violin Q factor is:

1 1 8nπμds 3E
= (6.9)
Qviolin Qbulk ωl2 ρ
6.5. VIBRATION NOISE 163
The proposed ribbon dimensions for the space DSMG are 0.025 × 20 mm × 1 m.
The proposed ribbon thickness of 0.025 mm is a compromise between the difficulty of
machining a thin ribbon and the difficulty of supplying enough current to offset the
increased mass per unit length of a thick ribbon. The width of 20 mm and length
of 1 m are the maximum dimensions that are feasible for a magnetic gradiometer
deployed in space. The DSMG system also requires supporting mechanical structure
and electronics so the dimensions of the entire gradiometer are approximately 0.04 m×
0.04 m × 1.1 m.
The sensitivity of the DSMG does not directly depend on the Q factor but on the
relaxation time τ = (2Q/ω) of the ribbon. Using Eq. 6.9, the relaxation time is:
$
l2 ρ
τ = Qbulk (6.10)
4nπμds 3E
Eq. 6.10 shows that τ is independent of ω. Nevertheless it is proposed to lower
the resonant frequency of the ribbon from 850 Hz to 80 Hz in order to keep the
required Q factor to a manageable level. Extrapolating the results of the preliminary
experiment, the increased length of 1 m should allow a ribbon Q factor of the order
Qribbon ≈ 300, 000 (Q varies as the length of the ribbon squared). The relaxation time
would then be approximately τ ≈ 1200 s which is much greater than its current value
of 0.1 s. The large increase in the relaxation time of the ribbon should produce a
significant decrease in the thermal noise as per Eq. 6.4.
Whilst Q factor enhancements are well understood in the field of gravitational
waves, there is still an uncertainty in extrapolating the preliminary result to the final
dimensions of the ribbon. The authors intend to investigate the physics of longer
ribbons in future work.
6.5 Vibration noise
The existing magnetic gradiometer is designed for airborne deployment. The high
frequency of 850 Hz allows a mechanical isolator to dampen vibration noise from
the aircraft by 120dB. In contrast, the proposed DSMG for space deployment would
operate at 80 Hz, a frequency which would short circuit the isolator.
The acceleration noise from solar irradiance fluctuations in near Earth space is
√ √
on the order of 10−10 m/(s2 · Hz) at 1 mHz[25] and less than 10−12 m/(s2 · Hz)
√
at 80 Hz[26]. This compares with seismic noise of 3 × 10−6 m/(s2 · Hz) on the
√
ground[27] and engine noise of 3 × 10−2 m/(s2 · Hz) on a survey aircraft[13].
The reduction of vibration noise by 9 orders of magnitude more than compensates
for the greater vibration at lower frequencies. No vibration isolation is required for
the proposed DSMG.
6.6 Operation in the natural space environment

The high current pumped through the ribbon used to detect gradients generates a
significant amount of heat. In a high vacuum the dominant method of dissipating this
heat is thermal radiation. Table 6.1 shows that if the ribbon is allowed to heat up
to 423 K then the amount of stress relaxation becomes unacceptable. It is therefore
proposed that the ribbon not be allow to heat up past 373 K and that the ribbon
be stretched for 1000 hours before deployment to stabilise the stress relaxation. The
amount of stress relaxation during the next 9000 hours of operation at 373 K is only
2%. For ribbon dimensions as discussed above this limits the maximum current to
4 A (current density of 8 × 106 Am−2 ). From Eq. 6.4 it is easy to show that the
√
design sensitivity in space is 8 × 10−14 T/(m· Hz).
The entire ribbon would heat up to 373 K and dissipate approximately 0.8 W of
power irrespective of the ambient temperature surrounding the ribbon. The T 4 power
dependence of thermal radiation means that the DSMG sensitivity depends strongly
on the ribbon temperature yet weakly on the environment temperature. Fig. 6.1
shows that there is little change in sensitivity for environment temperatures ranging
from 80 to 300 K.
6.7 Thermal radiation in space

In near Earth space while shielded from the Sun, scientific instruments radiating
their heat into space can reach cryogenic temperatures between 30 K and 50 K[28].
The advanced sun shields that have been manufactured for the James Webb Space
6.7. THERMAL RADIATION IN SPACE 165
Pluto Saturn
−12 Uranus Jupiter
10
Gradient noise (T/m/√Hz)

−13
10
Al 6061
5N Al + 2.0%Ni
−14 5N Al + 0.1%Ni
10
5N Al + 0.2%Ce
5N Al
6N Al
−15 7N Al
10
1 10 100 1000
James Webb Space Telescope International Space Station
Temperature (K)
Figure 6.1: String magnetic gradiometer design sensitivities as a function of ambient tem-
perature. There is very little advantage in cooling the DSMG down to a temperature of
70 K. Further decreases in temperature below 70 K can however, deliver large increases
in sensitivity. The ambient temperatures of electronics exposed to sunlight near the outer
planets are shown together with the temperature of some satellites in near Earth orbits
which are shielded from the Sun[47].
Telescope can reduce 1370 Wm−2 of sunlight impacting on the front of the shield
down to a mere 105μWm−2 behind the shield where the ambient temperature is
approximately 7 K[28; 29]. A proposed ribbon made from low-resistivity aluminium
alloys (described in the next section) would dissipate a tiny 30 mW of power despite
a high current of 10 A. High-current audio amplifiers are available commercially
with output impedances as low as 0.005 Ω. Heat dissipated from power supplies and
support electronics could be screened from the ribbon by highly reflective mirrors or
by displacing all heat producing DSMG modules a meter from the ribbon.
These figures show that passive cooling of the DSMG system in space is feasible.
A radiator area of 0.2 m2 would cool the ribbon down to 40 K, whilst a radiator area
of 1 m2 would cool the system down to 27 K.
6.8 25−70 Kelvin operation

At 40 K the electrical resistivity of high purity 99.999% 5N aluminium is very low
ρ ≈ 1.80 × 10−10 Ωm and the thermal conductivity very high κ ≈ 2000 W/(m·K).
The thermal conductivity values are from an empirical fit by Woodcraft[30] and the
temperature dependence of resistivity are from Hashimoto et al.[31]. These properties
make thermal conduction along the axis of the ribbon the dominant form of heat
dissipation below 70 K. Fig. 6.2 shows the temperature profile along the ribbon for
an ambient temperature of 40 K.
40K magnetic gradiometer:
Temperature profile along the ribbon
52
50
Temperature (K)
48
46
44
42
40
0 0.2 0.4 0.6 0.8 1
Position (m)
Figure 6.2: 0.03 W of power is dissipated by thermal conduction along the 1 m length of
the ribbon. The middle of the ribbon heats up to 50 K while the ends are in thermal contact
with a heat sink at 40 K.
At temperatures below 10% of the melting point of aluminium, work hardening

reduces creep to zero[32] so annealed alloys with low resistivity can be used. Switching
the ribbon material from aluminium-6061 to pure aluminium may be unfeasible since
the yield strength of 5N aluminium is only 5 MPa[33]. Instead it is proposed to use
special highly conductive high-strength alloys such as those developed for the Atlas
Project[34; 35] as shown in table 6.2. One alloy of 5N aluminium with 0.1% nickel
has a greatly increased yield strength of 80 MPa at a cost of only moderately higher
electrical resistivity (ρ ≈ 2.2 × 10−10 Ωm at 40 K).
The high thermal and electrical conductivities mean that the current could be
increased to 10 A and the ribbon width reduced to 18 mm. The current density
could be as high as 2 × 107 Am−2 . Eq. 6.4 shows that high currents, low ribbon
6.9. 4−25 KELVIN OPERATION 167
Alloy Temper Yield strength Q at 300K Q at 10K RRR

Al 6061 T6511 275MPa 280000 3700000 3
5N Al + 2% Ni Cold worked 120MPa 170
5N Al + 0.1% Ni Cold worked 80MPa 590
5N Al + 0.2% Ce Cold worked 46MPa 1400
5N Al Annealed 5MPa 45000 220000 6000
6N Al Annealed 5MPa 45000
7N Al Single crystal 0.7MPa > 150000
Table 6.2: At progressively lower temperatures, aluminium alloys[46; 34; 33; 36] with
lower and lower values of electrical resistance are proposed. The low resistance comes at
the expense of a reduced yield strength. RRR (residual resistivity ratio) is the ratio of the
electrical resistivity at room temperature to the residual resistivity at 4 K.
mass and low temperature increase the signal to noise ratio. The design sensitivity is
√
approximately 10−14 T/(m· Hz) at 40 K.
6.9 4−25 Kelvin operation

Further increases in sensitivity are possible if the temperature could be lowered below
25 K by utilising very large thermal radiators in conjunction with sun shields or by
using liquid helium. As shown in table 6.2, first 5N, then 6N[33] and eventually 7N
single-crystal aluminium[36] are necessary to exploit these low temperatures.
For pure metal films, the resistance drops with decreasing temperatures until the
mean free path l of conducting electrons becomes larger than the film thickness t. In
the limit l >> t the ribbon resistance will increase by a factor of[37]:
Rribbon 4l
= (6.11)
Rbulk 3t log(l/t)
Using the empirical relationship lρ = 8.2×10−16 Ωm2 [38], the mean free path of 7N
aluminium with a resistivity of ρ = 1.5 × 10−13 Ωm is l = 5.4 mm. Magnetoresistance
reduces the mean free path to 4.7 mm[39]. This value of the mean free path is 187
times larger than the thickness of the ribbon. The resistance of the ribbon would then
increase by a factor of 47 due to this size effect. Size effects continue to be significant
for temperatures up to 25 K.
The low electrical resistivity at liquid helium temperatures can not be fully realised
due to size effects. Other difficulties with exploiting a low-temperature environment
include magnetoresistance, self-inductance and the low strength of pure alloys. Fig.
6.1 shows that the magnetic gradient sensitivity tapers off below 10K. The design
√
sensitivity at 4 K is approximately 2 × 10−15 T/(m· Hz) using a current of 10 A
and a current density of 2 × 108 Am−2 . The small improvement in sensitivity does
not justify the difficulty of achieving liquid helium temperatures. DSMGs are only
feasible for temperatures between 25 and 300 K.
6.10 Low-frequency noise background
“The phenomenon of 1/f noise, with spectral density scaling inversely with frequency
is common to virtually all devices”[40]. The typical frequencies of interest in for global
magnetic surveys range from 200 μHz to 0.17 Hz[3]. Deep space missions such as the
voyager measure the slowly varying interplanetary magnetic field with frequencies
ranging from 50 nHz up to 1 Hz[41]. 1/f flicker noise and random walk noise are
expected to be significant at these frequencies. The low-frequency noise of the DSMG
is compared with some existing devices in this section.
Fluxgate magnetometers have been used in space for more than 30 years. The
noise power spectral density of a high-performance fluxgate magnetometer is typical
of shot noise devices and characterised by a 1/f spectrum with a typical value of
√
3 × 10−12 T / Hz at 1 Hz for space research-grade instruments[8]. The source of the
noise is attributed to Barkhausen-like mechanisms that affect the motion of domains
in ferromagnetic material in the sensor cores[2]. One fluxgate magnetic gradiometer
√
built for use in space has a sensitivity of 3 × 10−11 T/(m· Hz) at 1 Hz[9].
Low-Tc SQUIDs were proposed as a highly sensitive magnetic gradiometer for
space-borne magnetic investigations more than 20 years ago[3]. SQUIDs have the
best sensitivity on offer for terrestrial operations. One low-Tc SQUID gradiometer
√
has a sensitivity of 6 × 10−14 T/(m· Hz)[4] in the laboratory with a 1/f noise corner
at 0.3 Hz. 1/f flicker noise is more severe in high-Tc gradiometers which have a typical
1/f noise corner of 10Hz[42; 5]. The two major sources of 1/f noise in dc SQUIDs are
fluctuations of the critical current in the Josephson junctions and motion of flux lines
(vortices) trapped in the body of the SQUID[6]. The frequency ranges of interest in
6.10. LOW-FREQUENCY NOISE BACKGROUND 169
−7
10
−8
10
Gradient noise (T/m/√Hz)

−9
10 (a)
−10
10 (b)
−11
10 (c)
(e)
−12
10
(d)
−13
10
−14
10 −6 −5 −4 −3 −2 −1 0 1 2
10 10 10 10 10 10 10 10 10
Frequency (Hz)
Figure 6.3: The magnetic gradient system noise of (a) an existing DSMG operating in a
moderate vacuum of 1 kPa in unshielded environment, (b) a fluxgate gradiometer developed
for magnetic surveying from a satellite[9], (c) a high-temperature SQUID gradiometer de-
veloped for magnetocardiography[42], (d) a low-temperature SQUID gradiometer developed
for airborne mineral exploration[4], and (e) the proposed space DSMG in high vacuum.
space exploration are entirely within the 1/f noise region of SQUIDs[3].
The noise performance of magnetometers is strongly affected by mechanical and
thermal stresses, which vary over time and with exposure to extreme environments.
Even state of the art magnetometers experience drift on the order of 10−10 T/year[2].
These mechanical and thermal effects introduce a random walk noise with power
scaling 1/f 2.
Because the DSMG is a single-element device, any drifts associated with non-
stationary unbalance of two different sensors (as in the case of multisensor-based
gradiometers) are absent. Also, it is a modulation-demodulation device such that
most of the 1/f noise in the DC region is suppressed compared to static gradiometers.
The low-frequency noise of the DSMG is characterised by a 1/f 2 spectrum of
random walk noise. In the closed loop operation[17], the ground-based DSMG has
a 1/f 2 noise corner of approximately 2.5 mHz (see Fig. 6.3). With the feedback
loop turned off, the frequency of the noise corner depends on the magnitude of the
external gradient; noise corners as high as 0.2 Hz have been measured when the
DSMG is exposed to gradients higher than 10, 000 nT/m.
6.11 Low-frequency noise improvements
The white noise floor of the space DSMG is lower than the ground based DSMG
by several orders of magnitude. If the level of random walk noise remains the same
then the frequency of the 1/f 2 noise corner would increase. The origins of the low-
frequency noise in the DSMG are not known at present although several models have
been proposed. In this section several arguments are presented which suggest that
the level of low-frequency noise can be reduced step in step with the white noise.
Any mismatch between the drive frequency of the AC current and the resonant
ribbon frequency creates low-frequency noise. The space DSMG’s low level of white
noise would allow more accurate tracking of the ribbon resonant frequency. In addi-
tion, the amplitude of ribbon vibrations during closed loop operation could be reduced
step in step with the white noise. By using the feedback of closed loop operation to
reduce the signal[17], any amplitude dependant low-frequency noise will be reduced
accordingly.
The very high mechanical Q factor of the space DSMG could be utilised for off-
resonance operation. The very large signal at resonance could be used exclusively to
identify the exact resonant frequency of the ribbon. The level of signal at resonance
could be ignored. The difference between the AC current drive frequency and the
resonant frequency would be known exactly. Amplitude fluctuations from frequency
drift could be fully compensated while any small variance in the mechanical Q would
produce only negligible noise during off-resonance operation.
For the reasons outlined above, the noise spectrum of the proposed DSMG for
space applications is expected to be flat for frequencies down to 2.5 mHz. Fig. 6.3
compares the expected performance of the DSMG with other technologies and shows
predominance of 1/f noise and 1/f 2 noise at low frequencies.
6.12 Readout
In addition to the thermal noise in the ribbon, there is also measurement noise in the
apparatus used to measure the position of the ribbon. Because of the DSMG’s high
Q factor, part of the frequency band of interest for space exploration will consist of
6.13. CONCLUSION 171
frequencies large enough such that response is not limited by dissipation but by the
“off-resonance” response of the oscillator. In the limit Q → ∞, the rms displacement
x of the ribbon produced by a frequency dependent magnetic gradient is:
dBy li
x= √ (6.12)
3
dz nπ 32ηf0 f
where n = 2 is the mode number, l is the length of the ribbon, η is the mass per unit
length, i is the peak current, f0 is the resonant frequency of the second violin mode
of the ribbon and f is the frequency of the magnetic gradient signal. For ground
√
based operations, the smallest detectable signal of 4 × 10−10 T/(m· Hz) produces a
√
displacement of 10−11 m/ Hz over a bandwidth of 1 Hz.
The present method of measuring ribbon deflections is pumping the ribbon with
a small radio frequency current of i = 0.2 A. The radio frequency current generates
a radio frequency flux around the ribbon. Two pickup coils connected in differential
mode measure the modulation in flux as the ribbon moves. Each pickup coil is capable
√
of measuring ribbon deflections of size 6 × 10−13 m/ Hz which is sufficient to detect
the minimum signal.
If the sensitivity improvements suggested in this paper are implemented then the
√
minimum signal amplitude will be only 9×10−15 m/ Hz for a 1 Hz signal. In order to
improve the readout sensitivity enough to detect a displacement this small, the pickup
coils could be replaced with a microwave cavity readout developed for gravitational
wave antennas[43]. Other solutions include a low-noise SQUID[44] or optical readout
using a Fabry Perot cavity[45].
6.13 Conclusion
√
The proposed space DSMG has a sensitivity of 8 × 10−14 T/(m· Hz) using only the
natural space environment. Even higher sensitivity is possible with passive cooling of
the DSMG in space. The size of the DSMG and the power consumption requirements
are comparable with existing magnetometers used in space missions. It is also possible
to deploy a full tensor gradiometer by combining several single-axis DSMGs.
Acknowledgements
The authors would like to thank Mr. Howard Golden of Gravitec Instruments for
many useful discussions and suggestions. Work on the DSMG project is funded in
part by a linkage grant from the Australian Research Council.
BIBLIOGRAPHY 173
Bibliography
[1] E.I. Alves, V.M.C. Madeira, Rationale for the deployment of a magnetic gra-
diometer on Mars, in: Proceedings of the Sixth International Conference on
Mars, 2003, paper 3014.
[2] M.H. Acuna, Space-based magnetometers, Review of Scientific Instruments 73

(2002) 3717-3736.
[3] R. Hastings, R.P.S. Mahler, R.S. Schneider Jr., J.H. Eraker, Cryogenic mag-
netic gradiometers for space applications, IEEE Transactions on Geoscience and
Remote Sensing GE-23 (1985) 552-561.

cations, The Leading Edge 25 (2006) 178-180.
[5] D. Tilbrook et al., Design and development of a SQUID tensor gradiometer, in:
Applied Superconductivity Conference, Houston, 2002.
[6] R.L. Fagaly, Superconducting quantum interference device instruments and ap-
plications, Review of Scientific Instruments 77 (2006) 101101.
[7] M. Klinger, J.H. Hinken, S.S. Tinchev, First space test of high-Tc SQUIDs, IEEE
Transactions on Applied Superconductivity 5 (1995) 2759-2761.
[8] F. Primdahl, J.R. Petersen, G. Marklund, N. Olsen, P. Brauer, T. Risbo, A.

Ranta, MAIDS: Magnetic Investigation in Deep Space, Science instrument pro-
posal to ESA for the SMART-1 Mission, Danish Space Research Institute, Tech-
nical University of Denmark, 1998.

stability and linearity, Sensors and Actuators A 120 (2005) 71-77.
[10] M. Purucker, T. Sabaka, G. Le, J.A. Slavin, R.J. Strangeway, C. Busby, Magnetic
field gradients from the ST-5 constellation: improving magnetic and thermal
models of the lithosphere, Geophysical Research Letters 34 (2007) L24306.
[11] M. Terho, L.J. Pesonen, I.T. Kukkonen, Magnetic properties of asteroids from
meteorite data - implications for magnetic anomaly detections, Earth, Moon,
and Planets 71 (1995) 225-231.
[12] A.V. Veryaskin, Magnetic gradiometry: a new method for magnetic gradient
measurements, Sensors and Actuators A 91 (2001) 233-235.
23 (2004) 790-793.
[14] H. Golden, W. McRae, A. Sunderland, A.V. Veryaskin, D.G. Blair, L. Ju, A novel
magnetic gradiometer: description, design issues, and trial results, in: Australian
Institute of Physics 17th National Congress, Brisbane, 2006.
[15] D.K. Wickenden, T.J. Kistenmacher, R.B Givens, J.C. Murphy, Development
of miniature magnetometers, Johns Hopkins APL Technical Digest 18 (1997)
271-278.
[16] G.V. Anand, Nonlinear resonance in stretched strings with viscous damping, The
Journal of the Acoustical Society of America 50 (1966) 1517-1528.
proportional and integral feedback, arXiv:0712.0908[physics,inst-det.],
2007.
[18] R.C. Ritter, G.T. Gillies, Classical limit of mechanical thermal noise reduction
by feedback, Physical Review A 31 (1985) 995-1000.
[19] National Aeronautics and Space Administration, Earth Atmospheric Model,

http://www.grc.nasa.gov/WWW/K-12/airplane/atmosmet.html
[20] G.A. Landis, Degradation of the lunar vacuum by a Moon base, Acta Astronau-
tica 21 (1990) 183-187.
[21] W. Duffy Jr., Acoustic quality factor of aluminium and selected aluminium alloys
from 50 mK to 300 K, Cryogenics 42 (2002) 245-251.
BIBLIOGRAPHY 175
[22] A.M. Gretarsson, G.M. Harry, Dissipation of mechanical energy in fused silica
fibers, Review of Scientific Instruments 70 (1999) 4081-4087.
[23] G.I. Gonzalez, P.R. Saulson, Brownian motion of a mass suspended by an anelas-
tic wire, The Journal of the Acoustical Society of America 96 (1994) 207-212.
[24] D.J. Dawid, S. Kawamura, Investigation of violin mode Q for wires of various
materials, Review of Scientific Instruments 68 (1997) 4600-4603.
[25] B.L. Schumaker, Disturbance reduction requirements for LISA, Classical and
Quantum Gravity 20 (2003) S239-S253.
[26] J. Pap, M. Anklin, C. Frohlich, C. Wehrli, F. Varadi, L. Floyd, Variations in

total solar and spectral irradiance as measured by the VIRGO experiment on
SOHO, Advances in Space Research 24 (1999) 215-224.
[27] D. Coward, J. Turner, D.G. Blair, Characterizing seismic noise in the 2-20 Hz
band at a gravitational wave observatory, Review of scientific instruments 76
(2005) 044501.
[28] Space Telescope Science Institute, JWST Design: Sunshield,

http://www.stsci.edu/jwst/overview/design/sunshade.html
[29] M.J. Amato, D.J. Benford, Harvey S. Moseley, J. Roman, An engineering concept
and enabling technologies for a large single aperture far-infrared observatory
(SAFIR), Proceedings of SPIE 4850 (2003) 1120-1131.
[30] A. Woodcraft, Predicting the thermal conductivity of aluminium alloys in the

cryogenic to room temperature range, Cryogenics 45 (2005) 421-432.
[31] E. Hashimoto, Y. Ueda, Zone refining of high-purity aluminum, Materials Trans-

actions JIM 35 (1994) 262-265.
[32] H.J. Frost, M.F. Ashby, Deformation Mechanisms Maps: The Plasticity and
Creep of Metals and Ceramics, Pergamon Press, New York, New York USA:
Pergamon Press, 1982, ch. 4.
[33] R.P. Reed, Aluminum. 2. A review of deformation properties of high purity

aluminium and dilute aluminium alloys, Cryogenics 12 (1972) 259-291.
[34] K. Wada et al., Development of high-strength and high-RRR aluminum-

stabilized superconductor for the ATLAS thin solenoid, IEEE Transactions on
applied superconductivity 10 (2000) 373-376.
[35] K. Wada, S. Meguro, H. Sakamoto, A. Yamamoto, Y. Makida, High-strength and

high-RRR ALNi alloy for aluminum-stabilized superconductor, IEEE Transac-
tions on applied superconductivity 10 (2000) 1012-1015.
[36] E. Hashimoto, Y. Ueda, T. Kino, Purification of ultra-high purity aluminum,

Journal de Physique IV 5 (1995) 153-157.
[37] E.H. Sondheimer, The mean free path of electrons in metals, Advances in Physics
1 (1952) 1-42.
[38] J.R Sambles, K.C. Elsom, G. Sharp-Dent, The effect of sample thickness on the
resistivity of aluminium, Journal of Physics F: Metal Physics 11 (1981) 1075-
1092.
[39] J.P. Egan, R.W. Boom, Measurement of the electrical resistivity and thermal
conductivity of high purity aluminium in magnetic fields, Advances in Cryogenic
Engineering (Materials) 36A (1990) 679-686.
[40] R.H. Koch, D.P. DiVincenzo, J. Clarke, Model for 1/f flux noise in SQUIDs and
qubits, Physical Review Letters 98 (2007) 267003.
[41] L.F. Burlaga, C. Wang, J.D. Richardson, Large-scale magnetic field sluctuations
and development of the 1999–2000 global merged interaction region: 1-60 AU,
The Astrophysical Journal 585 (2003) 1158-1168.
[42] Y. Zhang et al., A HTS SQUID gradiometer using superconducting coplanar

resonators for operation in unshielded environment, Chinese Journal of Physics
38 (2000) 330-338.
BIBLIOGRAPHY 177
[43] E.N. Ivanov, M.E. Tobar, P.J. Turner, D.G. Blair, Noncontacting microwave cou-
pling to a cryogenic gravitational wave antenna, Review of Scientific Instruments
64 (1993) 1905-1909.
[44] P. Carelli, M.G. Castellano, G. Torrioli, R. Leoni, Low noise multiwasher super-
conducting interferometer, Applied Physics Letters 72 (1998) 115-117.
[45] L. Conti, M. De Rosa, F. Marin, L. Taffarello, M. Cerdonio, Room temperature

GW bar detector with opto-mechanical readout, Journal of Applied Physics 93
(2003) 3589-3595.
[46] J.G. Kaufman, Tensile, Creep and Fatigue Data at High and Low Temperatures,
The Aluminum Association, Washington, DC, 1999, p. 166.
[47] National Aeronautics and Space Administration, Technology Readiness

Overview Mixed Signal Devices for Very Low Temperatures Background,
http://nepp.nasa.gov/index_nasa.cfm/898/
Chapter 7
Conclusions and Future Work
In this final chapter, I present a summary of my dissertation work, emphasising

conclusions drawn, and relevant future work that could be done.
7.1 Conclusions
This thesis has shown that the Direct String Magnetic Gradiometer (DSMG) can
accurately detect magnetic gradients with second harmonic vibrations of a current
carrying string. The string sensing element approximates a point gradiometer (far
field approximation) for objects further than 0.45 m from the sensor. At closer dis-
tances, high order gradients become significant and the error is approximately 10%
for objects 0.3 m away from the string.
The string is made of alumnium 6061 and I have shown that this is indeed the
best material for the gradiometer. Applying a T913 heat treatment to the aluminium
6061 alloy would strengthen the metal and allow longer strings to be used, thus
increasing the sensitivity to gradients. The best material to make the DSMG frame
out of is Torlon 4301, because Torlon’s expansion coefficient is approximately the
same as aluminium 6061. This thesis shows that Torlon 4301 is non-magnetic and
will not cause heading error. Torlon’s intrinsic magnetic susceptibility varies between
χ = 5.7 × 10−6 and χ = −2.1 × 10−5 , depending on the manufacturer.
The primary application of the DSMG is airborne magnetic surveying with a
√
target sensitivity of 0.01 nT/m/ Hz rms. This thesis has shown that the DSMG’s
performance depends on its ability to dissipate heat whilst minimising air damping.
179
180 Chapter 7. CONCLUSIONS AND FUTURE WORK
By combining a high current, an elevated temperature and low pressure, the ther-
√
mal noise level of the DSMG has been decreased from 0.65 nT/m/ Hz rms down
√ √
to 0.18 nT/m/ Hz rms. I have then shown that the 0.01 nT/m/ Hz rms target
sensitivity level, can be achieved by optimising the string dimensions and power con-
sumption, and materials. By measuring the signal from the moving iron spheres at a
distance of 0.6 m, I have demonstrated that the present DSMG system has sufficient
sensitivity to detect a hypothetical ellipsoid nickel deposit 500 m long, 200 m depth
extent, and 20 m thick located 100 m below the surface.
The present level of heading error is unacceptable in airborne geophysics applica-

tions. The DSMG detects a small 225 ± 5 nT/m peak to peak signal when rotated in
the Earth’s uniform magnetic field. I conducted a thorough investigation of magnetic
susceptible parts in the DSMG. I found that the maximum heading error from mag-
netic parts is 28nT/m peak to peak, which is only ∼ 10% of the measured heading
error. The remaining heading error is likely to be caused by inadequate common
mode rejection. The present version of the DSMG has a thin ribbon cross section.
Consequently, the second harmonic frequencies for out of plain and in plain motion,
differ by 5.5 ± 0.3 Hz. Some preliminary experiments indicate that common mode
signals from out of plane vibration, is the cause of heading error.
Deployment in space could be the most viable application of the DSMG because
of the ease of operation and enhancement of sensitivity. In the natural space envi-
ronment, I have calculated that a mechanical quality factor of Q ∼ 300000 is possible
√
and that the thermal noise floor could be reduced below 10−13 T/m/ Hz. A space
borne DSMG should be able to match the white noise level of SQuID gradiometers
and have a 1/f corner that is two orders of magnitude lower. Sun shields can be
used to passively reduce the temperature down to ∼ 40 K where it may be possible
√
to detect magnetic gradients with sensitivity 10−14 T/m/ Hz.
The proposed space DSMG will use a 20 mm wide ribbon as its sensing element.
However, the present DSMG uses an inductive readout which is not suitable for
measuring the displacement of such a wide ribbon. Therefore, I have developed a
√
new capacitive readout with displacement sensitivity ∼ 10−13 m/ Hz rms.
The capacitve readout has a very high transfer coefficient of 106 V/m, made pos-
7.2. FUTURE WORK 181
sible by a high electronic quality factor and a small distance of 60 μm between the
capacitive plates and the ribbon. The linear region of the readout is only 360 nm.
As part of a magnetic gradiometer, magnetic force feedback can be used to keep the
readout at the zero point. In principle, the readout should be sufficient to measure
√
magnetic gradients with sensitivity 10−13 T/m/ Hz rms using a 1 m long ribbon in
a high vacuum environment.
7.2 Future work
7.2.1 Heading error
Heading error remains are problem with the DSMG. Dr. Alexey Veryaskin is cur-
rently developing the next generation of inductive readout that uses 3 pickup coils to
measuring the ribbons displacement. The middle pickup coil would measure the By
magnetic field component in the vicinity of the string. Using magnetic field data to
compensate for magnetic gradient misalignments is a standard technique that is used
in airborne magentic gradient surveys, see Stolz et al.[1] for example.
The second harmonic frequencies for out of plain and in plain motion, differ by
5.5 ± 0.3 Hz and could be a source of heading error. This problem could be remedied
by using a round string so that both resonant frequencies would be equal and thus
there would be no phase error. Another solution is to use a much wider ribbon so
that the out of plain vibration mode would have a resonant frequency of several kHz
and could be ignored.
7.2.2 Other technology
The sensitivity of the DSMG may greatly enhanced by using a superconducting string.
Type I Superconductors break down in the presence of weak magnetic fields on the
order of 0.05 T. Type II Superconductors can stand higher magnetic fields however
most of the Type II Superconducting materials have very low Q factors, the exception
is pure Niobium. A niobium superconducting string could carry current densities of
order 1010 Am−2 [2]. Even higher current densities may be possible by using carbon
nanotubes with one group reporting current densities better than 1013 Am−2 [3]. How-
ever, at present, nanotubes cannot be made long enough for a magnetic gradiometer.
Both the nanotube and the niobium string would have to be very thin to achieve
such current densities and both would require extremely high vacuum to avoid air
damping.
7.2.3 Present technology
The sensitivity of the DSMG can be improved significantly without the need of exotic
technology. The modelling in Chapter 5 of this thesis shows the parameters which
√
are required to achieve a sensitivity of ≈ 0.01 nT/m/ Hz rms or better. The string
temperature should be kept between 370 K and 400 K. The optimum string length
for airborne applications is 0.56 m using aluminium 6061-T913 and the optimum
diameter is 0.0012 ± 0.0004 m. However, there are still some unresolved questions
before the target sensitivity can be realised in practice.
The sensitivity of the DSMG can be increased by carrying more current in the
string. However, ultra high currents may introduce new sources of noise into the
DSMG system. There is likely to be an optimum power level and this should be
determined in future work.
Heavily oxisided or anodised aluminium can have a thermal radiation emissivity
of ∼ 0.9, compared to ∼ 0.03 for highly polished aluminium. A DSMG using an
anodised aluminium string could have a very high sensitivity because such a string
can carry more current. In order to evaluate the feasibility of utilising anodised
strings, the affect of heavy oxidiation on the aluminium string’s Q factor should be
measured.
I have predicted in Chapter 5 of this thesis that aluminium strings with diameter
0.0012 ±0.0004 m will have the highest violin mode Q factors. This should be verified
by experiment. A DSMG using the optimum wire diameter could then be constructed.
Chapter 6 predicts that a 1 m long string should have a relaxation time that is 16
times higher than that of a 0.25 m long string. This relationship between relaxation
time and length should be verified by experiment because the high sensitivity of the
space DSMG relies on it.
The space DSMG should be developed because it can achieve much higher sen-
7.2. FUTURE WORK 183
sitivity than an airborne DSMG, whilst using much the same technology. Because
of the Space DSMG’s high Q factor, part of the frequency band of interest for space
exploration will consist of frequencies large enough such that response is not limited
by dissipation but by the “off-resonance” response of the oscillator. At 1 Hz from
√
resonance, the thermal noise amplitude will be ∼ 10−14 m/ Hz. This noise level is
far below the noise floor of the DSMG’s inductive readout. The new capacative read-
out’s noise floor is also too high. However, the DSMG’s present displacement readout
is adequate for detecting DC magnetic gradients which are on resonance where both
signal and noise will have high displacement amplitudes.
A space DSMG prototype could easily be built using a ribbon with the proposed
dimensions: 0.025 mm thick, 20 mm wide 1 m long. Preliminary experiments using
off the shelf aluminium 8011-O showed that this material was inadequate and that
a professional ribbon of made from aluminium 6061-T913 should be manufactured.
Using existing laboratory equipment at UWA, it should be easy to measure the Q
factor and noise floor of a DSMG utilising such a ribbon to detect DC magnetic
gradients. Locking the drive current frequency to the resonant frequency of an ultra
high Q string, could prove problematic with the present setup. The frequency tracking
of the DSMG could be improved by using a fixed weight, spring or piezoactuators to
apply tension to the string.
Bibliography
[2] R.P. Huebener, R.T. Kampwirth, R.L. Martin, T.W. Barbee, Jr., R.B. Zubeck,
Critical current density in superconducting niobium films, Journal of Low Tem-
perature Physics 19, 344-346 (1975).
[3] B.Q. Wei, R. Vajtai, P.M. Ajayan, Reliability and current carrying capacity of
carbon nanotubes, Applied Physics Letters 79, 1172-1174 (2001).
Appendix A
List of publications
Publications presented in this thesis

[1] A. Sunderland, H. Golden, W. McRae, A.V. Veryaskin, L. Ju, D.G. Blair, Results
from a novel direct magnetic gradiometer, Exploration Geophysics 40, 222-226,
2009 (Chapter 2).
[2] A. Sunderland, A.V Veryaskin, H. Golden, W. McRae, L. Ju, D.G. Blair, Differ-
ential readout for a magnetic gradiometer, Sensors and Actuators A 153, 5-12,
2009 (Chapter 3).
[3] A. Sunderland, L. Ju, D.G. Blair, W. McRae, H. Golden, Low magnetic suscep-
tible materials and applications to magnetic gradiometry, Smart Materials and
Structures 18, 095038, 2009 (Chapter 4).
[4] A. Sunderland, L. Ju, D.G. Blair, W. McRae, A.V Veryaskin, Optimising a Direct
String Magnetic Gradiometer for Geophysical Exploration, Review of Scientific
Instruments 80, 104705, 2009 (Chapter 5).
[5] A. Sunderland, A.V. Veryaskin, W. McRae, L. Ju, D.G. Blair, Direct string
magnetic gradiometer for space applications, Sensors and Actuators A 147, 529-
535, 2008 (Chapter 6).
185

Sunderland Andrew 2009

Caricato da

Informazioni sul documento

Titolo originale

Copyright

Formati disponibili

Condividi questo documento

Condividi o incorpora il documento

Opzioni di condivisione

Hai trovato utile questo documento?

Questo contenuto è inappropriato?

Copyright:

Formati disponibili

Sunderland Andrew 2009

Caricato da

Copyright:

Formati disponibili

Characterising and improving a magnetic gradiometer for

Andrew Sunderland, B.Sc.(Hons)

Submitted in partial fulﬁlment of

Characterising and improving a magnetic gradiometer for geophysical

by Andrew Sunderland, B.Sc.(Hons)

Supervisors: Prof. D.G. Blair, Dr L. Ju, and Dr A.V Veryaskin

2 Characteristics of the DSMG 37

5 Sensitivity and optimisation 115

6 Space Applications 155

7 Conclusions and Future Work 179

A List of publications 185

1.1 The four main contributions to the total magnetic ﬁeld . . . . . . . . 3

2.1 Photograph of the direct string magnetic gradiometer (DSMG) during

3.1 Detailed concept diagram of DSMG . . . . . . . . . . . . . . . . . . . 53

6.1 String magnetic gradiometer design sensitivities as a function of ambi-

3.1 Phase noise of the 10 MHz quartz oscillator as a function of frequency 75

4.1 Concentration of ferromagnetic elements in each sample . . . . . . . . 89

6.1 Stress relaxation becomes increasingly problematic at higher tempera-

1.1 Tools for Geophysicists

1.2 Magnetic ﬁeld of the Earth

At altitudes of 50 km to 1000 km is the ionosphere. High energy photons such as

1.3 Magnetic minerals

of the Earth’s present magnetic ﬁeld.

1.4 What is a magnetic gradient?

1.5 Example magnetic gradients

As a simple example, consider a single gridline from a hypothetical surface survey.

Each component of the magnetic gradient tensor is sensitive to a diﬀerent di-

1.6 Advantages of measuring the magnetic gradient

1.6.1 Gradient measurements in general

• Higher resolution. Especially of shallow features (note the sharp peaks in

• Selectivity. Suppression of regional gravity or magnetic ﬁelds produced by

• Minimising aliasing. Aliasing is the eﬀect of having greater data density

• Quantitive analysis. Better deﬁnition of structural features (in section 1.5 we

1.6.2 Magnetic gradients in particular

• Insensitivity to rotation. Magnetic gradients surveys are insensitive to the

• Immunity to magnetic variations. By making a gradient measurement, time

• Compensate errors. Measurements of the vertical gradient can help correct

• Source characterisation. Koenigsberger Ratio (remanent : induced) can

1.6.3 Magnetic gradient tensor versus Total field gradient

• Tensor components can be combined to make calculations unaﬀected by aliasing

• Better information on magnetisation directions. The magnetic moment of com-

• Direct determination of anomaly geometry which is irrespective of whether the

1.6.4 Space applications

By measuring the 5 independent magnetic gradient components and the 3 vector

1.6.5 Borehole applications

When taking measurements with a magnetic gradiometer in a borehole, magnetic

1.6.6 Archeological applications

Vertical magnetic gradients can be used to perform 3 dimensional inversions of arche-

1.6.7 Other applications

Magnetic gradients can also be of use in tracking military vehicles[24], detecting

1.7 Limitations of magnetic gradient measurements

At a distance of 100 m, the total magnetic intensity of a paramagnetic ore body

1.8 Common mode rejection

symmetric. Even after balancing is achieved, the gradiometer may be aﬀected by

1.9 Existing magnetic gradiometers

1.9.1 Fluxgate gradiometers

A temporal magnetic gradiometer for airborne surveying was developed by Wickerham

1.9.2 SQuID gradiometers

1.9.3 Total field gradiometers

A scalar magnetometer can be converted into a vector magnetometer in a particu-

1.9.4 Other magnetic gradiometers

1.10 History of the Direct String Magnetic

1.10.1 The beginning 1997-2001