EFMRI User Guide V1.0 PDF

i
COPYRIGHT © 2005 magritek Limited

This manual is COPYRIGHT © 2005 by magritek Limited. All rights reserved. This manual may not
be reproduced in electronic or printed form, in whole or in part, without the express permission of
magritek Limited. No use of this manual may be made for resale or for any other commercial purpose
whatsoever without prior permission of magritek Limited. The information in this manual is subject to
change without notice. Magritek Limited reserves the right to revise this manual and to make changes
from time to time in the content hereof without obligation to notify any person or persons of such
revisions or changes.
Caution and Safety Considerations

A DC current of up to 6A is pulsed through the outermost polarizing coil of this instrument. When using
large currents for long time periods, the coil will become warmer. Caution must be taken not to allow
the coil to become too hot. Under normal operating conditions the temperature of the coil should not
exceed 40°C. If under extreme operating conditions, the surface temperature of the coil does exceed
40°C, halt use of the apparatus until the temperature of the coil returns to normal.
The probe also generates a small magnetic field when in operation. Although this field is small, it could
potentially damage credit cards or watches, or other magnetically sensitive devices if they are left close
to the probe. It is suggested that magnetically sensitive devices are kept at a distance of at least 3m from
the probe.
ii
Identify System Components
Welcome to the User’s Manual for the Terranova Earth’s Field Nuclear Magnetic Resonance
(EFNMR) and Earth’s Field Magnetic Resonance Imaging (EFMRI) instrument. First, please identify
all of the system components in the following list:
• The Terranova MRI Probe which consists of three concentric tubes each wound with a
separate coil namely
o B1 coil (this coil is hidden but a gold SMA socket it visible)
o Gradient coils (these coils are hidden but an 8 pin socket is visible)
o Polarization coil
• The Terranova MRI Spectrometer
• Three connection cables between the Spectrometer and Probe
• Power Cable to connect a 24V DC source. Brown wire is positive; blue wire is negative.
• A USB cable
• A Magna-probe™ 3D compass
• Prospa Software setup CD
• Terranova MRI Software setup CD
• User Manuals
• ½ litre sample bottle
In order to use the Terranova MRI system, you will require the following:
• A 24 V power supply capable of sourcing 6.5 A continuous and 10 A pulsed DC current. This
can be achieved in one of three ways:
o Two deep cycle 12V car batteries connected in series, or
o A 24 V DC power supply. We have used a 150 W switchmode power supply from
Jaycar Electronics (www.jaycar.com.au), product number MP3114, and found it
more than adequate. The specifications of this power supply are as follows:
Output Voltage: 24 VDC
Tolerance: +/- 1%
Ripple & Noise: 240 mV
Output Current: 6.5 A
Efficiency: 85%
Input Frequency: 47~63 Hz
Cold Start Inrush Current: 35 A, or
o Purchase a power supply from magritek.
• A personal Computer running Windows 2000 or XP with
o at least 100 MB free disk space
o at least 64 MB RAM
o at least a 200 MHz Processor
o a USB interface
iii
Preface
The purpose of this user’s manual is to provide a guide to the setup and operation of the Terranova-
MRI Earth’s Field Nuclear Magnetic Resonance (EFNMR) system. It is divided into four main parts.
The first section provides a brief introduction to the theory of EFNMR and describes the hardware
components of the EFNMR apparatus. This section is intended to provide some background
information for users of the system.
The second and third sections instruct the user on how to setup the system and optimise the operating
parameters. It is intended as a step-by-step guide to getting started and running the first basic EFNMR
experiments. An effort is also made in this section to provide troubleshooting tips for signal
optimisation.
The fourth section of the manual describes the basic NMR experiments which are supported by the
system. This section is intended as a guide for how to successfully run the experiments and interpret the
results. Some background theory is provided for each experiment but prior knowledge of the basic
principles of NMR is assumed.
The fifth section of the manual describes the core MRI experiments which are supported by the system
which allow images to be acquired in 1, 2 or 3 dimensions.
A separate manual is also included for the software package PROSPA which provides a more detailed
description of the underlying way the software works. The user is encouraged to read through the
PROSPA manual in addition to this manual. Some of the more advanced image processing
functionality shown in this manual is described in the PROSPA manual.
iv
v
Terranova-MRI EFNMR User’s Manual

Table of Contents
Identify System Components................................................................................................................. ii
Preface....................................................................................................................................................iii
Table of Contents ................................................................................................................................... v
1 Introduction ...............................................................................................................................1-1
1.1 Theory ...........................................................................................................................1-1
1.2 EFNMR Apparatus Description and Operation.............................................................1-1
1.2.1 Polarization .....................................................................................................1-2
1.2.2 The PGSE Gradient Coil.................................................................................1-2
1.2.3 The shim and imaging gradient coils ..............................................................1-3
1.2.4 Excitation and Detection .................................................................................1-3
1.3 The Spectrometer ..........................................................................................................1-4
2 Instrument Setup .......................................................................................................................2-1
2.1 Connection ....................................................................................................................2-1
2.2 Software Installation......................................................................................................2-1
2.2.1 To install the Prospa software .........................................................................2-1
2.2.2 To install the Terranova-MRI software...........................................................2-2
2.2.3 To install the USB device driver .....................................................................2-3
2.3 Running Prospa .............................................................................................................2-3
2.4 Preliminary experiments................................................................................................2-4
2.4.1 Coarsely tuning the probe ...............................................................................2-4
2.4.2 Select an initial location for the EFNMR apparatus........................................2-6
2.4.3 Select initial parameters ..................................................................................2-9
2.4.4 Using a spin-echo to find a signal .................................................................2-11
2.4.5 Shimming......................................................................................................2-13
2.5 Optimising the FID......................................................................................................2-15
2.5.1 Fine tuning the Probe ....................................................................................2-15
2.5.2 Setting the Resonance Frequency .................................................................2-16
2.5.3 Setting the B1 pulse duration.........................................................................2-17
3 Common processing features....................................................................................................3-1
3.1 Setting the output file location ......................................................................................3-1
3.2 Common NMR parameters............................................................................................3-1
3.3 Shim parameters ............................................................................................................3-1
3.4 Signal averaging time domain data ...............................................................................3-2
3.5 Filtering time domain data.............................................................................................3-3
3.6 Using the magnitude option...........................................................................................3-3
3.7 Integrating spectral peaks ..............................................................................................3-4
3.8 Phase cycling the 180 º pulse ........................................................................................3-4
3.9 B1 coil ring-down ..........................................................................................................3-4
3.10 Stopping an experiment.................................................................................................3-5
3.11 Loading previous parameter sets ...................................................................................3-5
3.12 Loading previous data sets ............................................................................................3-5
4 Earth’s Field NMR Experiments..............................................................................................4-1
4.1 The pulse and collect NMR Experiment .......................................................................4-1
4.2 Spin-spin relaxation time...............................................................................................4-2
4.3 Spin-lattice relaxation time............................................................................................4-5
4.3.1 T1 in the Polarizing field .................................................................................4-5
4.3.2 T1 in the Earth’s field ......................................................................................4-6
4.4 The Pulsed Gradient Spin Echo (PGSE) Experiment ....................................................4-7
vi
4.5 The Carr-Purcell-Meiboom-Gill (CPMG) Experiment............................................... 4-10

4.5.1 The CPMG Pulse Sequence.......................................................................... 4-10
4.5.2 The Terranova EFNMR CPMG Experiment ................................................ 4-11
4.5.3 Sample CPMG results .................................................................................. 4-12
5 Earth’s Field MRI Experiments .............................................................................................. 5-1
5.1 k-space .......................................................................................................................... 5-1
5.1.1 Introduction .................................................................................................... 5-1
5.1.2 A graphical representation of k-space ............................................................ 5-2
5.1.3 Other ways of moving about k-space ............................................................. 5-3
5.2 Preliminary experiments ............................................................................................... 5-3
5.2.1 Viewing common parameters ......................................................................... 5-4
5.3 Gradient Echo Imaging ................................................................................................. 5-4
5.3.1 The confirmation window .............................................................................. 5-6
5.3.2 1D imaging ..................................................................................................... 5-7
5.3.3 2D imaging ..................................................................................................... 5-8
5.3.4 3D imaging ..................................................................................................... 5-8
5.3.5 Description of each gradient echo macro parameter....................................... 5-9
5.3.6 Parameter limits.............................................................................................. 5-9
5.3.7 Example Images ........................................................................................... 5-10
5.4 Spin Echo Imaging...................................................................................................... 5-12
5.4.1 1D imaging ................................................................................................... 5-14
5.4.2 The confirmation window ............................................................................ 5-14
5.4.3 2D imaging ................................................................................................... 5-16
5.4.4 3D imaging ................................................................................................... 5-16
5.4.5 A description of each spin-echo macro parameter........................................ 5-19
5.4.6 Parameter limits............................................................................................ 5-20
5.4.7 Example images............................................................................................ 5-20
5.5 Filtered back-projection imaging ................................................................................ 5-22
5.5.1 The confirmation window ............................................................................ 5-23
5.5.2 2D imaging ................................................................................................... 5-24
5.5.3 Reprocessing FBP data................................................................................. 5-24
5.5.4 A description of each FBP macro parameter ................................................ 5-25
5.5.5 Parameter limits............................................................................................ 5-26
A.1 Signal-to-noise calculation macro ........................................................................................... A-1
A.2 A macro for calculating the signal producing efficiency of the B1 and Bp coils. ................. A-3
Introduction 1-1
1 Introduction
1.1 Theory
The Terranova-MRI Earth's Field Nuclear Magnetic Resonance (EFNMR) apparatus provides a low
cost alternative to high field instruments for the demonstration of pulsed NMR/MRI techniques or for
the determination of NMR properties of samples outside of the laboratory.
BE
z
x y
Figure 1.1. The Earth's Field NMR Probe and axis conventions
Conventional high field NMR instruments require an extremely uniform magnetic field for polarization
of nuclear spins and the detection of the resultant magnetization. It is difficult and expensive to make a
magnetic field of the required homogeneity (much less than 1 ppm in some cases), and so sample sizes
are usually quite small. However because the polarization and detection fields are large, the resulting
signal-to-noise ratio can also be large. EFNMR works because detection takes place in the highly
uniform Earth's magnetic field. This means that large samples can be used - partially compensating for
the very low detection field. To improve the initial magnetization, a relatively crude electromagnet with
a field about 350 times larger than the Earth's field is used to polarize the sample. Note that the Earth’s
magnetic field differs, both in direction and magnitude, according to your location on Earth. Typically
the field direction is close to vertical in mid to high latitudes and has a magnitude on the order of 50
µT. Because of these dependencies, the EFNMR receiver coil will need to be specially tuned to your
location.
1.2 EFNMR Apparatus Description and Operation

The Terranova-MRI EFNMR apparatus (Figure 1.1 and Figure 1.2) consists of 2 coaxial solenoids and
a third which is a collection 4 coils generating x, y and z gradients (there are two x gradients, one for
imaging and shimming, and one for pulsed gradient NMR). The probe is designed such that the three
coils slot together in a bayonet configuration. This configuration allows for easy access to the three
constituent coils. However, the probe is not designed to be dismantled and therefore should not be
taken apart. The outermost, or polarizing (Bp), coil is used to provide an initial polarizing field via the
application of a large current - establishing a nuclear magnetisation in a sample placed inside the
apparatus. The middle, or gradient coils are used to provide linear magnetic field gradients across the
sample. The inner most, or B1, coil is used to excite and detect the precessing magnetisation.
Table 1-1. Typical coil parameters
Coil Parameter Polarizing Coil Gradient B1 Coil

Coils (pgse,x,y,z)
Avg. coil diameter (mm) 170 105 84
Calc. mag. Field (mT/A) 3.13 N/A 30
Field gradient (mT/m/A) N/A 2.38,0.31,0.28,0.28 N/A
Avg. Resistance (Ω) incl. cable 2.8 3.1/2.0/1.5/1.5 325
Tuning cap for 2300 Hz (nF) N/A N/A 9.7
1-2 Introduction
Figure 1.2. A view of the three EFNMR/MRI coils
1.2.1 Polarization
The polarization coil is used to generate a magnetic field many times stronger than the Earth’s field, for
the purpose of increasing the bulk nuclear magnetization of the sample. With the present apparatus, a
maximum current of 6.0 A (depending on available power supply), may be applied to the polarizing
coil, producing a 18.8 mT field (Bp) directed along the x-axis of the coil (for comparison the Earth's
field is about 50 µT). Typically this current is applied for about 4 or 5 seconds (several T1 periods) to
maximise the bulk nuclear magnetization in the sample. The current is then switched off adiabatically
(in other words gradually on the timescale of the precessing magnetization), so that the enhanced
magnetization is left lying along the Earth's field direction. (Ideally parallel to the z-axis of the
apparatus).
While the maximum available polarization current is 6.0 A, in practice it is also limited by the resistive
heating of the polarization coil. This heating is potentially dangerous because it can lead to heating of
both the apparatus and the sample. Power is proportional to the square of the current, P = I2R, and
therefore high currents passed through the polarization coil for long time periods, at a high duty cycle,
will cause significant resistive heating. For example, a current of 6 A applied for 4 s with a duty cycle
of 50% (i.e. 4 s on and 4 s off) results in approximately 48 W of power dissipation, 1-2 W of which
may be deposited into the sample. We suggest that a power dissipation of no greater than this be
employed. If the outside of the probe gets warmer than 40 °C, switch off the system and wait for the
probe to cool down.
1.2.2 The PGSE Gradient Coil
The PGSE gradient coil is designed to provide a linear

magnetic field gradient along the longitudinal x-axis of
the probe. The current gradient probe can produce a field
gradient per unit current of 2.38 mTm-1A-1. Using the
maximum allowable current of 2.0 A the coil produces a
magnetic field gradient of 4.76 mT/m. The gradient is
employed in the pulsed gradient spin echo (PGSE)
experiment to change the resonance or Larmor
frequency of the sample as a function of spin position.
Like the polarization coil, the maximum allowable
gradient current is determined by resistive heating
Figure 1.3 - The PGSE gradient coil
concerns and the same power dissipation restrictions
apply. However, the duty cycle for the gradient pulses is
typically very low.
Introduction 1-3
For example, in an experiment with two 50 ms gradient pulses and a repetition time of 4 s, the duty
cycle is only 2.5% and therefore the power dissipation, with the maximum current of 2 A, would be
only 0.3 W. It is important to note that in the case of a more severe duty cycle, the problem of where
the power is deposited becomes a significant issue because the gradient coil is entirely enclosed within
the apparatus.
1.2.3 The shim and imaging gradient coils
Covering the PGSE gradient coil are a set of 3 gradient

coils etched onto a flexible PCB. These provide 3
orthogonal gradients which are used for both removing
background gradients from the Earth’s field
(shimming), and for enabling magnetic resonance
imaging (MRI) experiments. These take a maximum
current of 200 mA and correspond to a gradient of
55/63 µT/m (z,y/x). Because of short duty cycles, low
currents and resistances, overheating should not be an
issue with these coils.
Figure 1.4. The 3 axis imaging gradients
1.2.4 Excitation and Detection
The B1 coil is used for both the excitation and detection of the NMR signal. In the simplest NMR
experiment - the 90° pulse and collect - a short pulse of electromagnetic energy is applied to the inner
B1 coil which surrounds the sample. This is applied at the resonant (Larmor) frequency of the nuclei in
the local Earth's field (typically between 2 and 2.5 kHz). This tips the bulk magnetization from the
Earth's field direction (assumed to be along the z axis) back to the orthogonal (x,y) plane.
The x component of the magnetization will then be detected by the B1 coil as an induced EMF, since
the magnetization will precess about the Earth's magnetic field at the Larmor frequency ω = γB E . To
increase sensitivity and reject external noise sources, the coil is resonated with an external tuning
capacitor to the Larmor frequency. Because the B1 coil is a very effective detector of external (i.e.
unwanted) fields, screening is important and is provided by shorting the polarizing coil during the
detection phase. The large inductance and small resistance of this coil means that it effectively cancels
all fields within it, as long as they are relatively uniform over the volume of the coil and axial in
direction. Note that the signal level is also reduced by this action, but only by about 20% since the
signal field is largely localised within the B1 coil which has a significantly smaller diameter than the BP
coil. On the other hand the noise reduction factor is observed to be about 20, giving a signal-to-noise
ratio improvement of about 15 times over a non-shorted polarizing coil.
Polarizing pulse 90o B1 pulse Applied pulses
Adiabatic transitions
BE
z z z z z z
ω = γ BE Magnetization evolution
y y y y y y
x x x x x
x
Equilibrium
Magnetization Realigned After 90
magnetization Precessing in BE
increasing with BE degree pulse
due to BE
y vs tv
Y component of
magnetization
Figure 1.5 The polarize, pulse and collect experiment pulse diagram
1-4 Introduction
The 1H NMR signal from the sample has a peak amplitude of about 100 µV for a 570 ml water sample
(using a 6 A, 4 s pulse) and is amplified by a series of low noise operational amplifiers and is then
digitised by a 16-bit converter. This data is then transferred to the controlling computer for
accumulation and display.
1.3 The Spectrometer

The EFNMR experiment is controlled by a digital signal processor (DSP), which is itself controlled by
a personal computer (PC) running the data Processing package Prospa. In a typical experiment Prospa
sends a precompiled DSP pulse program and all necessary parameters to the DSP unit via the PC's USB
port. It then starts this program running.
This program controls the complete NMR experiment, sending pulses to the BP, gradient and B1 coils
and then acquiring the NMR data at the appropriate time. While the pulse program is running on the
DSP, Prospa waits for a signal from the DSP to indicate that data is available for uploading. Once this
data has been collected, it is displayed, any required analysis is performed, and then the experiment is
ended, or if desired, repeated, perhaps with new parameters being sent to the DSP.
The entire EFNMR system is pictured in Figure 1.6; the various connections between components are
shown and labelled. The DSP is contained within the central box of electronics, called the
spectrometer. The spectrometer is connected to the PC via a USB interface. Additional cables connect
the spectrometer to its power source and to the B1, gradient and polarization coils. Figure 1.7 shows the
front panel of the EFNMR spectrometer.
Figure 1.6. The entire EFNMR system: Spectrometer and Probe
Figure 1.7. Front Panel of the EFNMR Spectrometer

Instrument Setup 2-1
2 Instrument Setup
To obtain best performance from the EFNMR apparatus it is important to choose the location of the
instrument carefully. The instrument should not be used within a couple metres of any ferrous objects
(e.g. chairs and tables with metal parts). Switch off any source of low frequency electrical noise in the
immediate vicinity (e.g. some fluorescent lights and CRT computer monitors). For a more detailed
description of how to find the optimal location and orientation for the instrument see section 2.4.2.
2.1 Connection
To use the apparatus, start by connecting the EFNMR probe to the electronics unit. Three cables are
required - the polarizing cable, the excitation/detection (B1) cable and the gradient cable. The
connections on the EFNMR probe are shown and labelled in Figure 2.1. The corresponding
connections on the spectrometer are shown and labelled in Figure 1.6. The 24V 6A DC power supply
should be connected to the spectrometer using the power cable as shown in Figure 1.6. It is very
important that the polarisation of the power supply is correct, positive is connected to the brown wire
and negative (or 0V) connected to the blue wire. (An incorrect connection will blow the fuse).
Turn on the power switch on the spectrometer,
located on the top right hand side (see Figure 1.7).
A red power light should illuminate below the
USB connection on the spectrometer. If the power
light does not turn on, check the power
connections between the power supply and the
spectrometer and then the fuse located directly
below the power switch. Now connect the
spectrometer to the PC with a USB cable. A green
data light should appear directly above the power
light when both the spectrometer and the
computer are on and are connected. Note: you
will need to install the USB device driver the first
Figure 2.1. The cable connections for the three EFNMR time you connect the spectrometer to the PC as
coils
described in section 2.2.3.
2.2 Software Installation

2.2.1 To install the Prospa software
Place the Prospa 2.1 setup CD into the CD drive. A setup wizard should appear automatically (Figure
2.2). If this wizard does not appear, open the CD-ROM directory and run the program “setup.exe”.
Click “Next” in the setup wizard window to begin the installation wizard. Follow the directions in the
setup wizard to choose a location for the program on your computer and select a destination for the
program under the start menu (Figure 2.2 to Figure 2.5). The wizard will also give you the option of
putting a Prospa icon on your desktop. When you reach the screen depicted in Figure 2.5 check the
details listed and click “Install”. This will load the Prospa software onto your computer.
Figure 2.2. The Prospa software setup wizard Figure 2.3. The license agreement
2-2 Instrument Setup
Figure 2.4. Select a destination for the Prospa software Figure 2.5. Ready to install the Prospa software
2.2.2 To install the Terranova-MRI software
Once you have installed Prospa v2.1, you then install the Terranova-MRI software from the Terranova-
MRI CD. This will add addition menus and a DLL to Prospa allowing it to communicate with the
spectrometer. The installation process proceeds as above except that you should make sure you choose
the same destination folder as that chosen for Prospa.
Figure 2.6. The Terranova-MRI software setup wizard Figure 2.7. The license agreement
Figure 2.8. Select a destination for the Terranova-MRI Figure 2.9. Ready to install the Terranova-MRI software
software (make sure it is the same as the location of the
Prospa package previously installed)
2.2.3 To install the USB device driver
Firstly connect the Earth’s field system to your computer using a USB cable (type A to type B). Power
up the Earth’s field system and wait for your computer to prompt you for a device driver. Then browse
to the driver contained on the Terranova-MRI CD in the folder called ‘USB drivers’ and let your
computer install it.
2.3 Running Prospa

To start Prospa, double click on the shortcut on your desktop or run Prospa from the start menu. The
Prospa window, pictured in Figure 2.10 will appear. At the top of the window you will find a list of
menus.
Figure 2.10. The layout of the Prospa program showing all available windows.
Under the main “File” menu (in the top-most window), you can find options for loading, saving,
importing and exporting data, setting the current folder, setting preferences and running a macro from a
file. (A macro is an interpreted program which provides similar functionality to Visual Basic™ or
Matlab™)
Under the “Windows” menu, the user can choose to display the command line interface (CLI), 1D, 2D,
3D or macro-editor windows. Upon launching the Prospa program the default 1D, CLI and macro-
editor windows will automatically appear.
In Prospa, experiments are run through the use of macros. These macros are organised into a series of
user definable menus. The menus supplied with Prospa and Terranova-MRI have the following
organisation:
• The “1D” menu contains macros for processing displayed 1D data sets.
• The “2D” menu contains macros for processing displayed 2D data sets.
• The “3D” menu contains macros for processing 3D data sets stored in matrices.
• The “NMRI” menu contains macros for loading and processing NMR data obtained directly
from a commercial spectrometers and also from the Terranova MRI experiments.
• The “NNLS” menu provides a macro for processing 1D relaxation and diffusion data which
has a range of relaxation or diffusion times.
• The “GUI” menu allows the design of new macro interfaces.
• The “Demo” menu runs a series of macros showing the functionality of the Prospa software
for processing various types of data.
• The “Layout” menu provides the user the option of choosing 1D, 2D or 3D window layout
schemes. The default is the 1D screen layout.
• The “EFNMR” menu contains macros for performing the basic (non-imaging) NMR
experiments.
• The “MRI” menu contains macros for performing the NMR imaging experiments.
In this manual we shall only discuss the EFNMR/MRI menus and experiments (the last 2 menus). The
remaining menus are discussed in the supplied Prospa documentation.
2.4 Preliminary experiments

Obtaining the first free-induction decay (FID) and spectrum is typically the most challenging step in
the set-up process for the EFNMR system. Many interdependent parameters and factors contribute to
the quality of the acquired EFNMR signal. Therefore it is difficult to provide the user with a step by
step process which can guarantee success. The following sections of the manual (2.4 and 2.5) contain a
description of the important factors and parameters which typically influence the quality of the
observed NMR signal. Using the suggestions found here, it should be possible for the user to obtain a
good quality signal. However, this process will typically take some time, up to several hours, and a
reasonable amount of trial and error is necessary. Therefore, it is highly recommended that the first
time user read sections 2.4 and 2.5 in their entirety before going back and attempting the first
experiment.
2.4.1 Coarsely tuning the probe
The first step once the system is running is to characterise the B1 coil
so that you can later tune it to the observed resonant frequency. This
will help towards maximising the observed signal.
The tuning macro can be found in the EFNMR menu under the name
AnalyseCoil. Run this macro and you will see the window shown in
Figure 2.11:
Figure 2.11. The AnalyseCoil macro window

The first thing to do here (and this applies to all the macros), is to set the working directory. Click on
the button labelled “…” and navigate to a folder where you would like your data to be stored. In the
following case this has been named “NMR data”. Move into this folder and press the button “Select
folder”.
Figure 2.12. The select folder window
The output of the AnalyseCoil macro will now be stored in a folder called “AnalyseCoil” which will be
a subfolder of “NMR data”.
Figure 2.13. The AnalyseCoil macro window with working directory selected
Next press the button called “Analyse” in the above dialog and the experiment will begin. All this
experiment does is apply an impulse to the B1 coil and look at the response as a function of the tuning
capacitance (which is under computer control). After a few minutes you should obtain a curve like that
shown on the right of Figure 2.14.
FID data (Acc: 1) Magnitude spectrum (Acc: 1) Resonance frequency vs. Capacitance
×10 2
15
3000
10 6
2800
Amplitude (µV/Hz)
5
Frequency (Hz)
2600
Signal (µV)
4
0
2400
-5 2200
2
-10 2000
-15 1800
0
0 0.02 0.04 0.06 0.08 0.1 1000 2000 3000 4000 5000 6 8 10 12 14 16
time (s) frequency (Hz) Capacitance (nF)
Figure 2.14. Sample output from the AnalyseCoil macro.
The three subplots, from left to right, represent the time domain response to the impulse, the frequency
domain response, and the peak of the response versus tuning capacitance curve. From this curve the
program is able to calculate the inductance and parasitic capacitance of the B1 coil. These are reported
in the command line interface window. The ratio of the spectral amplitude at the resonant frequency to
that at zero frequency gives the coil’s quality factor or Q – about 20 in this case.
In the pulse and collect experiment (to be described in section 4.1) you will determine the local NMR
resonant frequency. With this information you will be able to use the frequency vs. capacitance curve to
optimise the signal level.
2.4.2 Select an initial location for the EFNMR apparatus
Due to the location sensitive nature of the EFNMR apparatus, it is very probable that no signal will be
observed as a result of your first attempted measurement. There are two location dependent factors
which can greatly affect the quality of the observed signal: local magnetic field inhomogeneity, and
noise caused by interference from electronic equipment in the vicinity of the EFNMR probe.
The most common problem encountered, which results in either no signal being observed or the
observed signal being very short (i.e. persisting in time for less than 200 ms), is the inhomogeneity of
the local Earth’s magnetic field. The NMR signal arises from the phase coherence of an ensemble of
precessing nuclear spins. The exponential decay of the signal is a consequence of the loss of phase
coherence between the spins. One source of phase coherence loss is the relaxation process known as
spin-spin relaxation (characterized by the T2 relaxation time constant). Spin-spin relaxation is caused
by the magnetic dipole coupling between two neighbouring spins. In the presence of a completely
homogeneous field, spin-spin relaxation is the dominant source of phase coherence loss. For a bulk
water sample, the T2 time constant is typically on the order of two seconds and therefore the NMR
signal from such a sample should, ideally, persist for several seconds. However, in practice, the
magnetic field is not entirely homogeneous and therefore the loss of phase coherence is not solely due
to spin-spin relaxation, but is rather a combined effect of spin-spin relaxation and magnetic field
inhomogeneity. To describe the combined relaxation effects, an effective spin-spin relaxation time
constant, T2* (equation 2-1) is defined, where γ is the gyromagnetic ratio of the observed nucleus and
∆B0 is the magnetic field inhomogeneity.
1 1
*
= + γ∆B0 [2-1]
T2 T 2
The observed signal at a time t is given by:
S (t ) = S 0 exp(t T2* )cos(ω 0 t + θ ) , [2-2]
where S0 is the initial signal magnitude at t = 0, ω 0 the (angular) Larmor frequency and θ some
receiver dependent phase shift. From these equations it can be seen that a highly inhomogeneous
magnetic field (i.e. large ∆B0), will greatly reduce the observed signal magnitude and in many cases,
for typical observation times t > 20 ms, no signal will be observed. Therefore, it is very important to
find a location where the local Earth’s magnetic field is highly homogeneous over the volume of the
sample. A linewidth in the NMR spectrum on the order of a few Hz is desired. For an NMR signal at
2000 Hz, this requires a magnetic field homogeneity on the order of 0.1%.
At this point, it is useful to note that using a Hall probe to find an area of homogeneous Earth’s
magnetic field is difficult because typically such instruments only have an accuracy of about 1%, which
is insufficient for this application. The only reliable way to find an appropriate homogeneous location
is through trial and error with the EFNMR probe.
A given location within a lab may have a highly inhomogeneous Earth’s magnetic field due to the
proximity of ferrous objects including, but not limited to, steel benches, steel materials in the walls,
screw drivers and other such tools.
Proximity to an NMR lab or other labs employing strong magnetic

fields can also be a significant problem. It is often beneficial to place
the EFNMR probe on a wooden bench or stand (see Figure 2.15).
Also, the centre of the room (as far from the floor, ceiling and walls as
possible) is frequently a location of good homogeneity. Although these
are all useful hints for finding a good signal, it is important to note that
the most homogeneous spot within a room or building is not always
obvious. It is entirely possible for the most homogeneous area to be
near a steel object. So again we stress the need to try multiple locations
until an acceptable signal can be achieved.
Fortunately it is not always necessary to find a region of very high

homogeneity. The Terranova MRI system includes a set of x, y and z
shim coils which can be used to remove gross (linear) variations in the
field. Typically it is possible to reduce the linewidth of the NMR
signal from say 20 Hz to 1 Hz. If you start in a region of good
homogeneity then line-widths of less than 0.5 Hz are possible.
The orientation, as well as the location, of the probe is important for

MRI and can also influence the amount of noise detected. Firstly
measure the Earth's field direction using a conventional compass,
rotate the probe so that its longitudinal (x) axis is orthogonal to this
direction (this optimises the action of the 90º and 180º pulses,
increases the signal level slightly and is important for performing the
imaging experiments). In addition, the arrow marked on the end of the
probe (indicating the axis of the z-imaging gradient) should be directed
Figure 2.15
along the Earth’s field axis, this can be determined using the supplied
A stand and cradle for the
Magnaprobe™ 3 axis compass (see Figure 2.16). This may necessitate Terranova probe.
tilting the EFNMR probe slightly or in severe cases mounting the
probe on a cradle so it can be rotated easily (as shown in Figure 2.15)
Figure 2.16. Photos of the EFNMR probe showing the correct orientation to optimised imaging results.
The second factor which can reduce the quality of the observed signal is excessive noise caused by
interference from nearby electrical equipment. 50 or 60 Hz mains interference (as well as harmonics at
multiples of this frequency) are frequently observed in the NMR spectrum. An experiment called
MonitorNoise in the EFNMR menu has been supplied with the instrument to determine the best
location. Just select this menu option and observe the rms (root-mean squared) noise value displayed
above the left 1D plot. Following is an example.
Figure 2.17. The monitor noise macro
RMS noise = 3.2 µV Noise magnitude spectrum

20 1.0
Amplitude (µV/Hz)
0.8
10
Amplitude (µV)
0.6
0
0.4
-10 0.2
0.0
-20
0 0.1 0.2 0.3 0.4 0.5 2140 2160 2180 2200 2220
time (s) frequency (Hz)
Figure 2.18. A typical output from the MonitorNoise macro in a region of low noise
A typically acceptable noise level for MRI is less than 10 µV rms, anything under 5 µV is good and
under 3 µV is excellent. Noise levels between 10 and 100 µV will make basic NMR measurements
difficult and MRI almost impossible. If the noise level appears to be quite high, try moving and rotating
the probe to reduce the noise, remembering that the orientation of the probe should be at right angles to
the compass direction.
Experimentally, it is possible to reduce the observed noise on the FID and in the spectrum, through the
use of signal averaging (refer to Section 3.4). As the number of signal averages employed increases (try
a large number such as 32 or 64), the noise peaks should be reduced while any sample peaks should
increase in magnitude.
To summarise, the steps you should take to set up the EFNMR system are as follows:
1. Choose a suitable location well away from large or moving magnetic objects or obvious
electrical noise sources. (e.g. cars moving 5-10 metres away can shift the resonant frequency
by several Hz.)
2. Mount the EFNMR probe on a wooden or non-magnetic (this includes screws or bolts), table
or stand about 1 metre above the floor. Keep the rest of the apparatus (especially the power
supply) as far away from the probe as possible.
3. Determine the N-S direction using a compass and orient the probe so its longitudinal (x) axis
is orthogonal to this direction.
4. With the probe connected to the spectrometer and the MonitorNoise experiment running,
move the probe around the room while observing the noise level. At all times keep the probe’s
longitudinal axis pointing at right angles to the N-S direction. Find the location with the
minimum noise level.
5. Mount the probe in this position and then run the AutoShim macro (discussed below) to
optimise the field homogeneity.
2.4.3 Select initial parameters
Once a suitable low noise position has been found you can now search for an NMR signal. Initially
insert a large sample bottle filled with tap water into the probe. The supplied 73 mm diameter, 570 ml
bottle is an ideal sample container.
Figure 2.19. Pulse and Collect dialog.
The pulse and collect experiment is the simplest NMR pulse sequence, consisting of a polarizing pulse
and an excitation, or B1, pulse followed by the detection of the signal. Choose this experiment from the
EFNMR menu and the above dialog will then appear.
The polarization controls dictate the length and strength of the initial polarizing pulse in any EFNMR
experiment. For tap water the length of the polarizing pulse is typically set to 4000 ms (several T1 time
periods) in order to maximise the signal enhancement. The polarizing current parameter takes values
from 0 to 6.0 A. A suggested initial value is 6.0 A, although some power supplies may work better with
5.5A. If your power supply has any problems, you should reduce the polarising current appropriately.
The B1 frequency should be set to the Larmor frequency of the sample. The Larmor frequency is
dependent on the gyromagnetic ratio of the observed nucleus and the strength of the Earth’s magnetic
field. Typically this value will fall in the range 1800 to 2500 Hz. Section 2.5.2 covers how to correctly
set the B1 frequency. For an initial value choose something in the range given above.
The “Pulse duration” parameter (i.e. the length of the B1 pulse), controls the tip angle of the excitation.
An initial value of 1.50 ms is suggested, however it is important to remember to find the best 90° pulse
before running other experiments, as this can significantly improve the signal-to-noise ratio of the
spectrum.
The “Acquisition time” and the “Number of data points” determine the number of, and time between
adjacently sampled points. For initial parameters it is suggested that 16384 points and an acquisition
time of one second be employed.
The “Acquisition delay” is the time delay between the excitation of the sample and the detection of the
signal. The minimum value for this delay is limited by the ring down of the B1 coil (see section 3.9). A
value of 25 ms is suggested.
The “Average” function is discussed in Section 3.4. A single scan can be used for the initial
experiment. However, if difficulties arise in finding a strong signal it may be beneficial to employ
many averages and thus improve the signal-to-noise ratio of the acquired FID and spectrum.
The “Display range” controls the range of frequencies which will be displayed in the spectral window.
For example for the parameters in the dialog shown in Figure 2.19 the frequencies from 2138 to 2338
Hz will be displayed. After the experiment has been run, the spectrum can be zoomed out or in, as
desired. The display range parameter simply controls the initial range of frequencies displayed. At the
outset of experimentation it is worth choosing a B1 of 2200 Hz and a display range of 400 Hz.
The “Transmit B1 gain” parameter determines the amplitude of the B1 excitation pulse (units are volts).
The strength of the pulse, along with its duration, dictates the tip angle of the excitation. Therefore if
the transmit gain setting is decreased, the required length of the 90° and 180° pulse will increase. An
initial value of 2.5 is suggested. This can be increased or decreased (in the range from 0 to 10) to
change the length of the 90° and 180° pulses.
The “Capacitance” parameter determines the resonant frequency of the B1 coil. This field must be set
with a value between 4.4 and 17.15 nF. For now select an initial value of 12 nF for the capacitance. See
Sections 2.4.1 and 2.5.1 for directions on how to tune the coil.
The “Receive gain” value controls the degree of amplification of the received signal. An initial value
of 2 is suggested. This value has little effect on observed signal amplitude unless it is very large or
small since the display automatically compensates for different receiver gain values. If clipping occurs
on the FID signal, i.e. if the top of the FID appears to be cut off, reduce the receive gain. The receive
gain can take values between 0 and 10.
Table 2-1. Suggested Initial Parameter Values

Parameter Suggested Initial
Value
Capacitance 12 nF
B1 Frequency 2200 Hz
B1 Pulse Length 1.5 ms
Polarization Time 4000 ms
Polarization Current 6A
Transmit Gain 2.5
Receive Gain 2
Number of Points 16384
Acquisition Time 1s
90-Acquisition Delay 25 ms
Display Range 400 Hz
Number of Scans 1
After choosing some initial parameters, click the “Run” button to begin the experiment. (Note that the
very first experiment run after the spectrometer is switched on sometimes gives a false result and you
may need to run the experiment again before sensible results are obtained – this is a problem which will
be fixed in future software upgrades). Once started, you should hear relays clicking in the spectrometer
and after a few seconds a free induction decay (FID) signal and spectrum will hopefully appear in the
1D plot window (refer to Figure 2.20). The FID, on the left, should appear as an exponentially
decaying sinusoid. The spectrum, on the right, which is the Fourier Transform of the FID, should
display a noticeably large peak, hopefully like the one shown below in Figure 2.20.
FID data (Acc: 1) Magnitude spectrum (Acc: 1)

150
60
100
50
50
Signal (µV)
40
Amplitude
0 30
-50 20
10
-100
0
0 0.2 0.4 0.6 0.8 1 2100 2200 2300 2400

Figure 2.20. A typical FID and spectrum obtained with the pulse and collect experiment
(as acquired in a region of good field homogeneity).
As mentioned earlier, it is however very likely that no signal will be observed or that the observed
signal will be very short (i.e. persist in time for less than 200 ms). The problem is likely to be local
magnetic field inhomogeneity; however, it is beneficial at this point to do a quick check of the
parameters which you have used. Remember that these parameters should be checked and optimised at
each attempted location.
Make sure the resonant frequency is set to an appropriate estimate of the Larmor frequency. It is
possible to calculate a reasonable value for the Larmor frequency, given by ω 0 = γBE , from an estimate
of the local Earth’s field strength and the gyromagnetic ratio for 1H. A useful website for obtaining a
estimate of your local Earth’s magnetic field strength is:
http://nssdc.gsfc.nasa.gov/space/model/models/igrf.html.
See Section 2.5.2 for more information on setting the resonant frequency.
It is possible that the probe is incorrectly tuned. A poor tune will decrease the signal magnitude
observed but, except in the most extreme cases, it should not eliminate the signal entirely. Refer to
Section 2.4.1 for more specific information on tuning the probe. It is difficult to accurately tune the
probe to the Larmor frequency of the sample in the absence of a strong signal since the exact Larmor
frequency of the sample will then be unknown. However, as stated above, it is possible to calculate a
reasonable value for the Larmor frequency from an estimate of the local Earth’s field strength. Tune the
probe to this resonant frequency and repeat the NMR experiment. If no signal is observed, decrease (or
increase) the tuning capacitance and repeat the experiment. If the poor signal problem persists after
both increasing and decreasing the tuning capacitance then the tuning of the probe is unlikely to be the
most significant problem.
A poor choice of excitation pulse duration (B1 pulse length) can significantly reduce the amplitude of
the observed signal, but will not decrease the signal entirely. Therefore some signal should be found for
all values of this parameter. However, it may be worth while to increase and decrease this parameter by
small amounts and observe the effect on the FID. See Section 2.5.3 for more information about the B1
pulse length.
If a signal has been found you should proceed to section 2.4.5 and learn how to optimise the magnetic
field homogeneity by shimming.
2.4.4 Using a spin-echo to find a signal
It is difficult to optimise the location and experimental parameters of the EFNMR apparatus without an
observable signal. Therefore, in order to effectively troubleshoot the setup of the instrument it may be
useful to employ a slightly more sophisticated NMR method; the “spin-echo” experiment.
The signal magnitude obtained with a simple pulse and collect experiment is weighted by the effective
spin-spin relaxation time constant T2*. Therefore in a highly inhomogeneous field the signal magnitude
will be very low, making it very difficult to obtain a signal. In such cases, the spin-echo experiment can
be employed to reverse the effects of the magnetic field inhomogeneity and win back some of the
signal magnitude. The spin-echo sequence manipulates the phase coherence of the ensemble of excited
spins in the sample. As with all of the EFNMR experiments, the first step is the polarization of the
sample, with a polarizing pulse from the polarizing coil. This step is not shown in the spin echo pulse
sequence diagram (Figure 2.21).
90o 180o
x y B1 Pulses
τE τE
z z z z z z z
Magnetization evolution
r r
M M r
M
y y y y y y y
x x x x x x x
Equilibrium 90°-x pulse Dephasing 180°-y pulse Rephasing Echo maximum Dephasing
y vs tv
1.0 Sample from here

Y component of
magnetization
0.5
y
0.0
time
-0.5
-1.0
Figure 2.21. Spin echo pulse sequence diagram
The second step in the spin-echo experiment is to excite the sample with a 90° pulse. This results in the
rotation of the bulk magnetic field vector into the transverse plane. In the subsequent delay time period,
τΕ, the magnetization de-phases from the combined effects of spin-spin relaxation and magnetic field
inhomogeneity. The former, caused by the random motion of the spins, is essentially irreversible
whereas the latter can be reversed. Reversing the de-phasing due to magnetic field inhomogeneities is
the goal of the spin-echo experiment. At a time, τΕ, following the first 90° pulse, an 180° pulse is
applied to the sample. This pulse flips the magnetic field vectors about the y-axis. During the
subsequent time period, τΕ, the magnetization re-phases and forms what is known as a spin-echo. The
centre of the spin-echo occurs at a time 2τE after the initial 90° pulse. Only the de-phasing that occurred
as a result of magnetic field inhomogeneity will be re-focused.
Figure 2.22. The spin-echo experiment interface
The echo centre amplitude (as sampled at a time τΕ after the 180° pulse), will be weighted by the T2
decay time constant, not T2*. In an area of significant magnetic field inhomogeneity, it is possible to
refocus much of the NMR signal, de-phased by these local magnetic field inhomogeneities. Therefore,
employing this sequence, it may be possible to acquire a reasonably strong NMR signal in a location
where the simple pulse and collect experiment yields only noise.
Figure 2.23 and Figure 2.24 demonstrate the utility of the spin echo sequence in regions of high
magnetic field inhomogeneity. The FID and spectrum acquired using the simple pulse and collect
experiment (see Figure 2.23) show only noise (although a signal at very short acquisition times can just
be seen). However, the spin echo results in Figure 2.24 (obtained using the macro displayed in Figure
2.22), also acquired with a single scan, shows a significant NMR signal in the FID window and a broad
peak in the spectrum (peak width ~ 20 Hz). Despite the significant width of the spectral peak, this
signal provides the user with some very useful information. First, the presence of the signal confirms
that the apparatus is functioning correctly and that the initial parameters are reasonable. Second, the
approximate centre of this peak can be used as a good estimate of the local Larmor frequency.
60 1.6
1.4
40
1.2
20
Signal (µV)
1.0
Amplitude
0 0.8
0.6
-20
0.4
-40 0.2
0.0
0 0.2 0.4 0.6 0.8 1 2160 2180 2200 2220 2240
Figure 2.23. Results of a pulse and collect experiment in a region of magnetic field inhomogeneity.
50
Signal (µV)
4
Amplitude
-50
0 0.2 0.4 0.6 0.8 1 2160 2180 2200 2220 2240

Figure 2.24. Results of a spin echo experiment acquired at the same location as Figure 2.23.
Once a signal has been found, the next step is to improve the field homogeneity by moving the probe
about the room while running the spin-echo experiment (use a large number of scans but turn averaging
off). Choose a location with the narrowest line consistent with a low noise level.
2.4.5 Shimming
Once the narrowest spectral line has been found by physically moving the probe, the next step is to
improve the homogeneity using the experiment AutoShim. This macro uses an iterative procedure to
modify the current through the x, y and z gradients until the peak height in the frequency domain is
maximised. Since the peak height in the frequency domain is proportional to the echo width in the time
domain, maximising the peak height will decrease the T2* value and hence improve the field
homogeneity.
The AutoShim macro shown below uses the pulse and collect pulse sequence.
Figure 2.25. The Autoshim dialog
This macro will optimise the largest peak that it can see in the frequency range of width “Display
range” centred about the B1 frequency. The 3 sliders at the top represent the value of the shim currents
while the search range indicates the maximum range of currents which will be searched, centred on the
initial values. Typically this macro will require 5-20 minutes to optimise the field (the time depends on
the initial search range and stop precision). Apart from the B1 frequency, which will be location
dependent, use the parameters indicated above (they are not optimised but should still work). Before
and after plots and the output of this macro are shown in Figure 2.26 and Figure 2.27.
Note that the macro is based on the PulseAndCollect experiment and is only suitable for those cases
where a resonant peak can be seen using this macro. If this is not the case then the spin-echo version of
the AutoShim experiment should be used instead (AutoShimSE). If you are still unable to shim the
spectral line to less than 2 Hz then you should consider moving to a new location.
4
50
3
Signal (µV)
Amplitude
0
2
-50 1
0
0 0.2 0.4 0.6 0.8 1 2160 2180 2200 2220 2240
Figure 2.26. A pulse and collect experiment made before the shimming process was started.
The output of the AutoShim experiment shows the improvement in the linewidth as a function of
iteration. In this case the linewidth has decreased from about 25 Hz to 0.6 Hz. One important point to
note with this experiment is that it will not work unless the resonant peak is the largest peak in the
display range. This may require that you limit the range to avoid any significant noise peaks, or
alternatively signal average so that the resonant peak is the largest.
FID data (Acc: 1) Magnitude spectrum (Acc: 1) x = 2.86 y = -13.6 z = 12.89
100
80
60
50
Max amplitude
60
Signal (µV)
Amplitude
40
0
40
20
-50 20
-100 0 0
0 0.5 1 1.5 2 2160 2180 2200 2220 2240 0 20 40 60 80

time (s) Frequency (Hz) Iteration
Figure 2.27. At the end of the shimming process. The plot on the right shows the history of the shimming process,
the amplitude represents the height of the peak in the central plot.
2.5 Optimising the FID

Once a location of good magnetic field inhomogeneity is located and a good quality signal is obtained,
it is important to optimise the FID by tuning the coil, setting the B1 frequency, calibrating the B1 pulse
lengths and minimising the noise as much as possible.
2.5.1 Fine tuning the Probe
In order to maximise the signal-to-noise ratio of the EFNMR probe, it is important for it to be correctly
tuned to the Larmor frequency of the sample. Using the AnalyseProbe experiment discussed earlier it is
possible to set the tuning reasonably accurately. However with the pulse and collect experiment you
can fine-tune the probe to maximise the signal and ensure that images are not distorted because of
asymmetrical B1 sensitivity (since image data will contain a range of B1 frequencies).
The EFNMR probe can be accurately tuned using the basic pulse and collect experiment in a few easy
steps. In order to tune the probe, a pulse and collect experiment will be run, but with only a short delay
between the excitation pulse and the signal acquisition. Choosing a “Acquisition Delay” time of 2 ms
rerun the PulseAndCollect experiment (it will now use the shimmed gradient currents). By thus
reducing the delay between the sample excitation and signal detection the coil ring down is observed in
the acquired signal. This results in a sinc-like function superimposed on the signal in the spectral
window, as shown in Figure 2.28.
Magnitude spectrum (Acc: 1)
60
50
Amplitude
40
30
20
10
0
2000 2100 2200

frequency (Hz)
Figure 2.28. An example of the spectrum of a large water sample acquired by an incorrectly-tuned coil with a 90-
Acquisition delay of 2ms.
The broad sinc-type peak is the coil response to the short B1 impulse, while the narrow spike is the
NMR signal. If the sample peak is very strong, it is possible that the sinc function, caused by the ring
down of the coil, will be difficult to distinguish. If this is the case, reduce the polarization duration.
This will reduce the intensity of the sample peak.
The capacitance of the system is controlled by the Prospa software and is entered by the user into a text
field in many of the experiment dialogs. If the resonant frequency of the coil is too low (as depicted in
Figure 2.28), decrease the capacitance and re-run the pulse and collect experiment, maintaining the 2
ms 90-Acquisition delay. Conversely, if the resonant frequency of the coil is too high increase the
capacitance of the system and repeat the pulse and collect experiment. Continue to adjust the
capacitance value and repeat the pulse and collect experiment until the maximum of the sinc function
and the centre of the sample peak coincide, as shown in below. (In this case the capacitance was
reduced from the 11.7 to 10.5 nF).
Magnitude spectrum (Acc: 1)
80
Amplitude
60
40
20
2000 2100 2200 2300 2400

frequency (Hz)
Figure 2.29. An example of the spectrum of a large water sample acquired by a correctly tuned coil using a 90-
Acquisition Delay of 2ms.
Note: in the case of a very weak (or indistinguishable) sample peak, the coil can still be tuned to a
particular resonance frequency by aligning the centre of the sinc function to the desired frequency
value.
Repeating the pulse and collect value with a 90-acqusition delay of 25 ms will now result in an
enhanced signal amplitude:
FID data (Acc: 1) Complex spectrum (Acc: 1)
100 80
Amplitude (µV/Hz)
60
50
Signal (µV)
40
0 20
0
-50
-20
-100 -40
0 1 2 3 4 5 2196 2198 2200 2202 2204 2206 2208
Figure 2.30. An FID collected for 5 seconds to show the enhanced amplitude and linewidth obtained following
shimming and tuning. (Note that the data has been collected in complex rather than magnitude mode – this enables
the linewidth to be correctly determined).
2.5.2 Setting the Resonance Frequency
The amplitude of the NMR signal can be increased if the excitation frequency is set to the correct
resonance (Larmor) frequency of the sample. This can be done by changing the B1 frequency
parameter. The Larmor frequency can be determined from the spectrum as follows.
Acquire an FID and a spectrum using the Pulse and Collect experiment in the manner used to obtain
Figure 2.30. Using the “Allow region selection” button in the 1D plot, select a narrow frequency
region which contains the positive sample peak. Zoom in on this region by clicking the zoom-in tool
. Now select the “Allow data display” button . In the spectrum window, click and hold the left
mouse button; a set of axes should appear when you move the mouse. While holding the left mouse
button, drag the axes over the centre of the peak. The frequency and amplitude of the peak will be
displayed in the data display field located in the bottom left corner of the 1D plot window (see Figure
2.31). Enter this frequency value into the B1 frequency field in the pulse and collect dialog and either
run or exit this macro. This parameter will now be automatically entered into the B1 frequency field of
each macro you subsequently run.
Figure 2.31. Determining the NMR resonant frequency. Note the smaller noise peaks.
2.5.3 Setting the B1 pulse duration
The number of cycles for the 90° and 180° excitation pulses are determined by repeating the pulse and
collect experiment with different B1 pulse lengths (i.e. different numbers of cycles) and observing the
resultant signal amplitudes. The signal amplitude is determined by integrating the spectral peak over a
range of frequency values specified by the user. This is all controlled by the B1Duration macro which
can be accessed from the EFNMR menu.
Figure 2.32. The B1 Duration dialog
In the B1Duration dialog box (Figure 2.32), the B1 pulse parameters and the acquisition parameters are
the same as those defined for the simple pulse and collect experiment described above and should be
chosen accordingly. The B1 parameters define the array of B1 pulse lengths to be used for the
experiment. 10 steps ranging from the minimum of 0.227 ms (1/2 a cycle at f0) up to 3.5 ms in 15 steps
of 0.227 ms is a good range for the first attempt to locate the 90º and 180º pulse lengths. The
integration parameters define the range of frequency values over which the spectrum is integrated to
determine the signal value for each B1 pulse length. This interval can be chosen by observing the
frequency width of the peak from a single pulse and collect experiment. Typically it is sufficient to
integrate 5 to 10 Hz from either side of the B1 frequency.
After selecting the initial parameters, run the experiment. Three graphs will appear in the 1D plot
window: the FID, the spectrum and the output plot. The FID and the spectral plots are replaced after
each pulse and collect experiment. The third plot tracks the change in signal amplitude (as measured by
integrating the spectral peak) and is updated after each acquisition.
The signal intensity in the output plot should pass through a maximum and a minimum as is depicted in
Figure 2.33. The 90º pulse length is taken from the first maximum. The 180º pulse length is taken from
the first minimum after the 90. In the example of Figure 2.33, the 90º and 180º pulse lengths are 1.6
and 3 ms respectively.
FID amplitude vs B1 length (Acc: 1)
200
FID amplitude
150
100
50
0.5 1 1.5 2 2.5 3

B1 length (ms)
Figure 2.33. The output of the B1 Duration experiment.
If no minimum is reached over the range of times employed, repeat the experiment with a larger
number of steps or try a higher transmit gain. It may also be useful to use a smaller step size to better
determine the 90º and 180º pulses.
Record the 90° and 180° pulse lengths. All experiments require the 90º pulse time and any which use a
spin-echo require the 180° pulse length as well.
Once you have determined the best 90° and 180° pulse values, repeat the spin-echo and pulse and
collect experiments to verify that the signal levels are larger than before (assuming they were
incorrectly set).
Note that you can sometimes get quite strange looking B1 duration results which seem to oscillate about
the expect curve. This seems to be due to the effect of ring-down in the B1 coil which also contributes
to the tip-angle (see section 3.9)
Common processing features 3-1
3 Common processing features

Many of the macros in the NMR and MRI suite have common interface features. These will be
described in this section.
3.1 Setting the output file location
Figure 3.1. The output location part of each dialog
All macros save their data in much the same way. The user defines an output folder called the working
directory which will contain all his or her experimental results. Then in this directory the data from
each experiment will be stored in a subdirectory which has the name of that experiment.
By default the “Working directory” field will be blank when you first run Prospa, but once defined this
variable is remembered between Prospa sessions. This means that when you start a new experiment it
will default to the current working directory.
3.2 Common NMR parameters

A number of the NMR/MRI parameters are shared between the macros in the menus EFNMR and
MRI. If you change one of these parameters in one macro it will be used the next time you run one of
the other macros. These are parameters which don’t need to be changed very often once the system has
been set up. The parameters are shown below:
b1freq …...…. the NMR resonant frequency (in Hz)

polzI ……….. the polarizing current (in A)
cap …………. the B1 coil tuning capacitance (in nF)
duration90 .… the length of the 90 degree pulse (in ms)
duration 180... the length of the 180 degree pulse (in ms)
txGain ……… the amplitude of the B1 pulse (in Volts)
rxGain ……... the gain factor for the receiver (no units)
These parameters are stored in a text file called tnNMR.par (tn ≡ Terranova) which is placed in the user
preferences path, typically:
C:\Documents and Settings\username\Application Data\Prospa\Preferences V2.0\Other Macros
3.3 Shim parameters

Each macro has a button labelled “Shims” which provides access to the current shim settings (these are
the currents applied to the x, y and z shim coils which surround the B1 coil). Pressing this button
displays the following dialog window:
Figure 3.2. The shims window

3-2 Common processing features
The shim parameters are 3 numbers which are saved in the text file shim.par stored in the user
preferences folder
C:\Documents and Settings\username\Application Data\Prospa\Preferences V2.0\Other Macros
They have the variable names : x_shim, y_shim and z_shim. Currents are expressed in mA.
The textbox entries under the “Saved” label in Figure 3.2 show the currently saved shim values, while
those in the “Current” textboxes reflect the values which will be applied to the current experiment. You
can either modify these values by adjusting the sliders or by typing a number into the textbox and
pressing the enter key. Note that the slider updates to show this new value. The sliders have a limited
range of ±20 mA so larger values can only be entered via the textbox.
If you would like to save the new values you have entered, just press the “Save” button to update the
shim.par file.
To restore the values currently stored in the shim.par file press the “Reset” button.
Shim currents are also set by the auto-shim macros.
3.4 Signal averaging time domain data

One method for improving the observed signal-to-noise ratio is to average free-induction decays. The
average function repeats the basic experiment a number of times (as specified by the user), and the
resultant FIDs are averaged. To average FIDs, check the “Average” box in a macro dialog and select an
appropriate number of scans – 4, 8 or 16 for example. (We typically use a number of averages which is
a multiple of 2.) Run the experiment and observe that the signal-to-noise ratio improves. The signal
level is proportional to the number of averages, but because the noise is (ideally) random, the noise
level will increase with the square root of the number of averages. So 4 averages will increase the
signal-to-noise ratio by a factor of 2, 16 by a factor of 4 and 64 by a factor of 8.
20
8
10
Signal (µV)
6
Amplitude
0
4
-10
2
-20
0
0 0.2 0.4 0.6 0.8 1 2180 2200 2220 2240

A single scan
20 10
8
10
Signal (µV)
Amplitude
6
0
4
-10
2
-20 0
0 0.2 0.4 0.6 0.8 1 2180 2200 2220 2240

4 scans
15
8
10
5 6
Signal (µV)
Amplitude
0
4
-5
-10 2
-15
0
0 0.2 0.4 0.6 0.8 1 2180 2200 2220 2240

64 scans
Figure 3.3. The effect of signal averaging on the signal-to-noise ratio of FID and spectrum.
(Note that the amplitude is scaled by the number of scans to keep the calibrated amplitude constant).
3.5 Filtering time domain data

In many cases the signal-to-noise ratio of collected data can be improved, at the expense of some loss
in resolution, by applying a matched filter. Many of the NMR and MRI macros provide a filter
checkbox. Selecting this will multiply the incoming FID by an exponential filter. This is indicated in
the time domain data by a red curve. Note that the time domain data displayed in yellow has already
had the filter applied.
FID data (Acc: 1)
50
Signal (µV)
-50
-100
0 0.2 0.4 0.6 0.8 1
Time (s)
Figure 3.4. An FID after being multiplied by the exponential filter (displayed in red).
The severity of the filter can be adjusted by running the filters macro in the 1D menu and adjusting the
exponential parameter p2. (Have a look at the filters help file for more details).
3.6 Using the magnitude option

By default, most experiments display and analyse complex data in the frequency domain. However this
often requires that a phase shift is applied to bring the NMR spectral peak into absorption mode. This
phase shift is determined automatically, but if the signal-to-noise ratio in the frequency domain is poor
or there are overlapping peaks, this routine can fail, producing a peak with dispersive components. This
can give rise to erroneous values in the integration routines. (See section 3.7).
To combat this there is a magnitude option. In this case the magnitude of the spectrum is taken before
integration occurs. This has the disadvantage of producing broad, less well defined, peaks, but they will
always have a positive integral. If you use this option, make sure you enlarge the integration range to
enclose the broader peak.
Note that the amplitude of the frequency domain data is no longer correctly calibrated in the magnitude
mode since there is no longer a well defined relationship between the initial FID amplitude and the
integral of the magnitude peak.
3.7 Integrating spectral peaks

In many experiments it is important to determined the initial amplitude of the FID. Generally this is
more precisely determined in the frequency domain since the area under the real part of a peak, if
phased for absorption, is equal to the initial amplitude of the associated FID in the time domain. The
integration width parameter in many of the macros determines the number of data points which will be
included in the integration step. (The range will be b1freq ± integwidth /2 ). Make sure this is wide
enough to enclose the resonance peak without including any spurious noise peaks. The comments in
section 3.6 relating to the integration of magnitude peaks should also be considered.
3.8 Phase cycling the 180 º pulse

Any pulse sequence which includes a spin-echo is susceptible to the problem of imprecise 180º pulses.
Because the B1 coil is a short solenoid there will be a variation in the strength of the B1 field along the
axis of the B1 coil. Although an imprecise 90 º pulse will result in a loss of signal, an imprecise 180 º
pulse is a much bigger issue. Consider the case of the T2 experiment. Here a spin-echo is formed in
which the amplitude of the echo should reflect the true T2 decay constant. However an imprecise 180º
pulse can be viewed as having two components, one which is at 90º and one at 180º. The 180
component will cause some of the spins to refocus while the 90º will produce an independent FID. If
the T2* of the probe is long enough, then the FID and the echo can interfere producing an incorrect
amplitude – which often oscillates about the true amplitude as the echo time is varied.
One solution to this problem is to ensure that T2* is significantly shorter than the echo-time. This can be
achieved by deshimming the coil. Alternatively (or in addition) we can try and correct the error by
phase cycling. This works by performing two scans with different relative phases between the 90 and
180 pulses. The effect of this is to cancel out the “90º” component from the “180º” pulse. In reality
because of coil-ring down problems this technique is far from perfect in the EFNMR system and often
you will need to resort to the reducing T2* to combat this problem.
3.9 B1 coil ring-down

Because the B1 coil and associated tuning
capacitance form a resonant system there will
always be energy left in this circuit after the
application of a 90 or 180 degree pulse. The
amplitude of the voltage across the coil will
decay away by a factor of exp(−0.5) in
Q /(2π ) cycles. So for a 300 mV signal (as
seen after the 90 degree pulse in Figure 3.5)
to decay down to the noise level (say < 10
µV) will take:
1  10 µV  Q
n= ln  = 60 cycles
− 0.5  300 mV  (2π ) Figure 3.5. Oscilloscope trace showing ring-down after a 90º
pulse (for the first 1ms after the pulse the receiver input is
overloading)
For a Q of 17 (typical for the EFNMR B1 coils) at 2200 Hz, this corresponds to 25 ms. This means that
following such a pulse we must always wait for this length of time before collecting data. (Remember
that this signal will be very broad in the frequency domain and so will only be a problem if the
amplitude is quite large).
In addition this ring-down also adds to the effect of the 90 or 180 degree pulses making it difficult to
define a simple relationship between the length of the pulse and its action.
Ring-down can also be introduced if the current in one of coils which surrounds the B1 coil changes
suddenly. This includes the polarizing coil and the gradient coils. These transients can also act like B1
pulses.
Many of these effects also appear in high-field NMR, but what makes EFNMR unique is the low
resonant frequency, which means the ring-down time is much longer than would be seen at say 100
MHz (where the ring-down time will be measured in microseconds). Since the basic NMR parameters
are much the same at 2200 Hz as they are at 100 MHz, this places limits on some of the experiments
we can perform. In particular we are not able to investigate materials with T1 and T2s less than about
100 ms. (Or any other investigations which require echo-times less than 100 ms).
3.10 Stopping an experiment

There are two ways to halt a running experiment. To stop it immediately, press the escape key on the
keyboard. This sends a reset signal to the USB interface and the current experiment will abort. You will
be able to use Prospa immediately, but will have to wait a few seconds before you can use the
spectrometer again (look for the green light on the spectrometer or listen for the USB connection
sound).
The second method is to press the “Stop” button which is present on each macro dialog. This will cause
the current scan to complete and then the experiment will exit gracefully. You will be able to run
another experiment as soon as the old one finishes.
In either case there will be some delay before you can restart the experiment, so which method you use
is up to you.
3.11 Loading previous parameter sets

When you run an imaging or MRI macro it loads the last set of parameters used by this macro, along
with the last set of common parameters used (see section 3.2). However it is also possible to load
parameters used when performing a previous experiment of the same type. Simply click on the “Load”
button and navigate to the folder containing the data set previously collected. There should be a file
called acqu.par there. Select this file to load the parameters used when generating the data set stored in
this folder.
3.12 Loading previous data sets

All experiments save some sort of data. It can be in the form of a plot file (plot.pt1 or plot.pt2) which
can be loaded directly into a 1D or 2D plot, or it can in the form of a Prospa data file (data.1d, data.2d
or data.3d). The latter data sets can be loaded using the Prospa “File->Load” data menu command or
via the NMRI menu importData. See Prospa help files for more information on these possibilities.
Earth’s Field NMR Experiments 4-1
4 Earth’s Field NMR Experiments

4.1 The pulse and collect NMR Experiment
The most basic of the NMR experiments supported by the EFNMR system is the Pulse and Collect
experiment which was used to set up the system parameters. In this section this experiment will be
described in more detail.
Polarizing pulse 90o B1 pulse Applied pulses
Adiabatic transitions
BE
z z z z z z
ω = γ BE Magnetization evolution
y y y y y y
x x x x x
x
Equilibrium
Magnetization Realigned After 90
magnetization Precessing in BE
increasing with BE degree pulse
due to BE
y vs tv
Y component of
magnetization
Figure 4.1. The pulse and collect pulse sequence
The Pulse and Collect experiment is depicted in Figure 4.1. It consists of a polarization pulse and a B1
excitation pulse followed by detection of the NMR signal.
A theoretical expression for the signal detected by the EFNMR probe is given by:
E (t ) = γB E (B1 I )V S M 0 [1 − exp(−τ P / T1 )] exp(−t / T2* ) cos(γB E t ) sin(θ tip ) [4-1]
Here γ is the gyro-magnetic ratio of the nuclei being studied (2.675×108 s-1T-1 for hydrogen), BE is the
magnitude of the Earth's magnetic field (56 µT at our location), B1 I is the magnetic field produced by
the B1 coil when unit current is applied (~30 mT/A - this determines the efficacy of detection). VS is
the sample volume (e.g. 570 ml). M 0 is the magnetization produced in the sample by the application of
a polarizing pulse of sufficient length to produce equilibrium:
N S γ 2 h 2 (B P I )
M0 = IP [4-2]
4kTS
where I P is the current applied to the polarizing coil (typically about 6 A). This gives a value of M 0 =
6.29×10-5 Am-1 for water. The term [1 − exp(−τ P / T1 )] allows for shorter polarizing pulses (of length
τ P ) which will cause a reduction of the signal level because of the non-zero longitudinal relaxation
time T1 of the sample. (Note that a τ P of 10 s will give 99% of the possible signal for a water sample).
exp(−t / T2* ) represents the decay of the signal in the x-y plane due to field inhomogeneities and the
intrinsic spin-spin relaxation time of the sample, while the cos(γB E t ) term describes the oscillation of
the signal due to the precession of the magnetization vector in the x-y plane. Finally the sin(θ tip ) term
controls the fraction of the signal which appears in the x-y plane after a B1 pulse is applied with a tip
angle θ tip .
4-2 Earth’s Field NMR Experiments
Combining these results we expect a maximum signal amplitude of about 300 µV (appendix 1).
However several factors reduce this value. First the macro assumes the coils are long solenoids, which
produce uniform fields throughout their lengths. However the EFNMR probe uses short solenoids so as
to maximise the available polarizing current. This means that M0, the tip angle, and the detection
efficiency (as determined by B1/I), will vary with position. The macro in appendix 2 calculates these
effects. The results are plotted below along with their product. This will have a mean value equal to the
relative signal level detected.
Figure 4.2. A determination of the relative efficacy of the current probe compared with a system based on long
solenoids.
For the 570 ml sample (140 mm long) the signal level is only 0.75 of the long coil geometry. In
addition, shorting the BP coil during data acquisition to minimise the noise pickup reduces the detected
magnetisation by a further factor of 0.78. These two factors suggest a total signal level reduction factor
of 0.59 or a predicted signal level of 175 µV. Finally we normally only polarize the sample for about 4
seconds. This, (from equation 4-1), reduces the signal by a further factor of 0.8 to 140 µV. We typically
observe FID amplitudes from the 0.5 litre bottle of between 110 and 100 uV and allowing for some
reduction in the number of turns on the production versions of the B1 and Bp coils, the 140 µV estimate
would seem quite reasonable.
You can verify several aspects of equation 4-1. First by adjusting the amplitude of the BP pulses you
will see that E ∝ BP . Secondly, smaller samples will produce less signal - ideally in proportion to their
volume, although, as mentioned above, due to homogeneities in the B1 and BP fields, larger samples
will produce less signal than expected. It is difficult to vary BE without moving to a different part of the
Earth, however the gyromagnetic ratio can be changed by using a sample with a different NMR
sensitive nucleus. Fluorine (19F) is a good nucleus to try.
4.2 Spin-spin relaxation time

The decay of the FID is ideally exponential and is determined by the transverse or spin-spin relaxation
time T2. However inhomogeneities present in the local field will always produce a more rapid decay
rate, with a time constant represented by the symbol T2* . The application of a spin-echo pulse sequence
can be used to refocus the de-phasing of the magnetization caused by the inhomogeneous detection
field, allowing the spin-spin relaxation time T2 to be measured accurately.
In this next experiment, pictured in Figure 4.3, we will measure the spin-spin relaxation time of the
water sample using a spin-echo pulse sequence. This sequence applies two B1 excitation pulses. The
first is a 90° pulse that rotates the magnetization about the x-axis by 90° so that it lies along the y-axis.
A delay τ E is then inserted, (the echo time), in which the spins dephase due to inhomogeneities and
spin-spin relaxation effects. A 180° pulse is then applied along the y-axis which reflects the
magnetization components about the y-z plane. Another echo time-delay then occurs during which the
magnetization components refocus, producing an echo which has an amplitude reduced by spin-spin
relaxation effects alone. Data is collected from the peak of the echo onwards.
90o 180o
x y B1 Pulses
τE τE
z z z z z z z
Magnetization evolution
r r
M M r
M
y y y y y y y
x x x x x x x
Equilibrium 90°-x pulse Dephasing 180°-y pulse Rephasing Echo maximum Dephasing
y vs tv
1.0 Sample from here

Y component of
magnetization
0.5
y
0.0
time
-0.5
-1.0
Figure 4.3. Pulse sequence for measuring T2 relaxation (polarizing pulse not shown).
This experiment is repeated a number of times with different echo values. The relative echo amplitude
is determined by integrating the area under the spectral peak. To determine a suitable integration range,
you will first need to perform a trial spin-echo experiment.
If the echo time is particularly short or

T2* long, the excited signal from the 90°
pulse may not relax to zero before the
application of the 180° pulse. This may
result in unexpected results in the T2
plot. This can be avoided by purposely
spoiling the homogeneity of local
magnetic field.
Deshimming the probe is a good way to

accomplish this. Use the T2 experiment Figure 4.4. The set shims dialog with the y shim adjusted away
and click on the “Shims” button. In the from the optimal position.
dialog which appears (Figure 4.4) move
one of the sliders (x, y or z) and then
without saving the new settings, start the spin-echo experiment by pressing the “Run” button. Repeat
this process to find a suitable setting, such that the acquired echo is fast decaying but does not decay so
quickly that the peak in the spectrum is extremely broad. An echo which persists for about 100 ms is
suggested.

100 10
8
50
Signal (µV)
Amplitude
6
0
4
-50 2
0 0.1 0.2 0.3 0.4 0.5 2180 2200 2220 2240

Figure 4.5. A spin-echo obtained by deshimming the probe (echo time is 200 ms and acquisition delay is 25 ms).
Note that the shape of the spectral peak is actually a projection of the spin-density onto the y axis.
The minimum allowable echo time is 25 ms to allow for the ring-down of the coil. A suggested initial
range of echo times for the T2 experiment is 10 steps of 50 ms starting at 50 ms. After completing your
first T2 experiment you can adjust the echo time range to suit the T2 decay of your sample. You may
also choose to use the average function to improve the signal-to-noise ratio of your results.
Figure 4.6. The T2 dialog
From equation 4-1 the echo amplitude can be written E = E0 exp(−2τ E / T2 ) . As the relative echo
amplitude is determined for each τ value, it is plotted in the 1D plot window. In Figure 4.7 lines of best
fit has been displayed along with the extracted T2 value and calculated uncertainties.
Measuring T2 is not a trivial exercise, because of the problem of imperfect 180 degree pulses. The
following 4 graphs show how this effects T2 data and how to correct for it. With no phase cycling and a
long FID the data is far from exponential. Phase cycling helps a bit but there is a significant amount of
variation in the data. Putting on 10 Hz of broadening by adjusting the y-shim improves the result and
adding phase cycling to this produces smallest uncertainty.
T2 = (2300 ± 300) ms T2 = (1640 ± 90) ms
0.9
0.9
0.8
Attenuation (E/E0)
Attenuation (E/E0)
0.8
0.7
0.7
0.6
0.6
0.5
0.5 0.4
0.4 0.3
500 1000 1500 2000 500 1000 1500 2000

t (ms) t (ms)
Shimmed Shimmed with phase cycle

T2 = (1760 ± 40) ms T2 = (1750 ± 20) ms
0.9 0.9
0.8 0.8
Attenuation (E/E0)
Attenuation (E/E0)
0.7 0.7
0.6 0.6
0.5 0.5
0.4 0.4
0.3 0.3
500 1000 1500 2000 500 1000 1500 2000

t (ms) t (ms)
Deshimmed Deshimmed with phase cycle

Figure 4.7. T2 results showing the effects of phase cycling and deshimming
4.3 Spin-lattice relaxation time

The EFNMR apparatus is unique when compared to conventional high field NMR systems because the
polarizing field is separated from the detection field. In the context of T1 measurements this means that
two T1 parameters can be measured: T1 in the polarizing field and T1 in the Earth’s field.
4.3.1 T1 in the Polarizing field
The longitudinal or spin-lattice relaxation time, T1p, in the polarizing field can be measured by
performing an array of pulse and collect experiments with variable polarizing pulse times, τp. The pulse
sequence diagram for this experiment is depicted in Figure 4.8.
The relative signal amplitude (or what is equivalent, the area under the spectral peak) is then plotted as
a function of this time. An array of polarizing pulse durations, τ P , can be declared in the following
dialog (Figure 4.9) by setting the polarizing parameters: “Minimum polarizing time”, “Polarizing step
size” and “Number of steps”. Zero is also included in the final data set, since zero pulse length
produces zero magnetization. For best results the maximum polarizing pulse should be several T1 time
periods. The T1 for a water sample is about 2 s. Therefore a reasonable range of polarization time
periods is 500 ms to 5 s, at 500 ms intervals.
Figure 4.8. The T1 in Bp pulse sequence diagram
From equation 4-1 the signal amplitude can be written:
S = S 0 [1 − exp( −τ P / T1 ) ] .
The output of the T1 macro is a plot of the signal amplitude as a function of polarization pulse duration.
A line of best fit is found and plotted. From this, the spin-lattice relaxation time and uncertainty are
calculated and displayed as shown below.
Figure 4.9. The T1 in the polarizing field dialog
T1 = (2280 ± 50) ms
Attenuation (E/E0)
0.8
Figure 4.10. T1 Analysis. This figure is
updated as data for each new polarizing 0.6
time is collected. The x axis is the
polarization time 0.4
0.2
1000 2000 3000 4000 5000

t (ms)
4.3.2 T1 in the Earth’s field
The longitudinal or spin-lattice relaxation time, T1e, in the Earth’s field can be measured by performing
an array of pulse and collect experiments with variable delay periods between the polarizing pulse and
the 90° pulse. The pulse sequence diagram for this experiment is depicted in Figure 4.11.
Figure 4.11. The T1 in BE pulse sequence diagram.

The relative signal amplitude (or equivalently the area under the spectral peak) is then plotted as a
function of this time. An array of delay values, t, can be declared in the following dialog (Figure 4.12),
by setting the pre-90 delay parameters: minimum pre-90 delay, pre-90 delay step size and number of
steps. For best results the maximum delay should be several T1 time periods. The T1 for a water sample
is about 2 s. Therefore a reasonable range of pre-90 delays is 500 ms to 5 s, at 500 ms intervals.
Figure 4.12. The T1 in BE dialog
The signal amplitude can be written: S = S 0 exp(− t T1 ) because the equilibrium magnetization in the
Earth’s field is essentially zero. The output of the T1 in BE macro is a plot of the signal amplitude as a
function of pre-90 delay. A line of best fit is found and plotted.
T1 = (2320 ± 10) ms
1.0
Attenuation (E/E0)
0.8
0.6
0.4
0.2
0 1000 2000 3000 4000

t (ms)
Figure 4.13. T1 Analysis
Note that even with a zero pre-90 minimum delay, as set in this dialog, there is a hidden pre-pulse
delay of 50 ms to allow the magnetization to adiabatically return to the Earth’s field direction and to
allow the switching relays to settle into their final state. In addition there is the 25 ms delay after the
90º pulse. This means that it will be difficult to measure T1 values below 100 ms.
4.4 The Pulsed Gradient Spin Echo (PGSE) Experiment

Pulsed gradient spin echo (PGSE) is an versatile experiment and can be used to measure molecular
self-diffusion coefficients and fluid flow.
The pulse sequence is shown below (Figure 4.14) and is based around the spin-echo 90° - 180°
sequence used for measuring T2 relaxation. However the difference here is that we also apply two
narrow magnetic field gradient pulses shortly after each B1 pulse.
90o 180o
x y B1 Pulses
τE τE
∆
Gradient Pulses
δ
Sample from here

Y component of
magnetization
time
Figure 4.14. Pulse sequence for measuring diffusion (polarizing pulse not shown)
During the first gradient pulse, the magnetization vectors in different parts
of the sample will precess at very different rates due to the (almost) linearly
varying field and so phase coherence is rapidly lost. If no molecular motion
occurs before the second gradient pulse is applied, then these vectors will
reverse their motion, due to the action of the 180° pulse, and so form an
echo as in the simpler T2 sequence. However if motion has occurred in the
direction of the applied gradient, either due to self-diffusion or flow - then
the re-phasing will be incomplete since the spins will experience a different
local field during the second gradient pulse. The result will be a modified
echo amplitude. If self-diffusion has occurred, the amplitude will be
reduced (since re-phasing will be incomplete), while for pure flow there will
be a phase shift. In this experiment we will measure the self-diffusion
coefficient of water. To reduce the effects of convection (a problem in large
samples) you can use a porous material soaked in water (such as a sponge or
small glass beads) as a sample. As long as the pore sizes are not so small as
to restrict the motion of the water during the experiment we will see a
diffusion coefficient equal to that of free water.
Figure 4.15. The PGSE dialog

Set up the experiment as for T2 relaxation. The B1 pulse parameters are chosen as before. The delay
between the pulses will depend on the T2 of the sample, but should be long enough to accommodate the
gradient pulses and the ring-down they produce (20 ms before and typically 100 ms after the gradient
pulse switches off). If you are performing this experiment in an area of high magnetic field
homogeneity you may find it necessary to spoil the homogeneity of the field by deshimming the probe.
This ensures that excitation transients caused by the switching gradients do not interfere as much with
the experiment. (However this is less of an issue with this experiment than T2 because of the dephasing
action of the PGSE gradients.)
The gradient parameters determine the number of pulse durations used in the experiment. In the above
example there are 10 steps. The first has a gradient pulse 5 ms long , the second 10 ms and so on up to
a maximum of 50 ms. All pulse will have an amplitude of 2 A. The zero length step is included if the
“Include first point” checkbox is selected. The “Gradient current” text box in the dialog sets the
gradient current in the device and informs the macro of the chosen value so that the correct calculation
can be performed. Finally, as with the relaxation experiments, you can choose to average data for each
separate set of parameters to improve the signal-to-noise ratio.
The output of this experiment, like the T1 and T2 experiments, is a changing echo amplitude, decreasing
in this case. The change in amplitude can be understood and analysed by employing a simple model for
diffusion and assuming that the molecules only move between and not during the gradient pulses.
Using this narrow gradient pulse approximation it can be shown that a spin at position r during the
application of the first gradient pulse acquires a phase shift γgδ ⋅ r. If it then moves to position r′ it will
acquire a second phase shift of −γgδ ⋅ r′ during the second gradient pulse. The net phase shift will
therefore be γgδ ⋅ ( r′ - r) . The total signal acquired in the PGSE experiment will be the ensemble
average of all the phase shifts exp [ i γgδ ⋅ ( r′ - r) ] weighted by the probability for a spin to start at
position r and finish at position r′.
The probability can be represented by an averaged propagator Ps(R,∆) where R = r′ - r . For self-
diffusion with coefficient D this propagator will be a Gaussian function
Ps ( Z , ∆ ) = (4πD∆ ) exp − Z
− 12
( 2
4 D∆
) [4-3]
and the echo attenuation function, assuming gradients along the z-axis direction, will be given by
∞
−1
E 0 ( g ) = ∫ (4πD∆) 2
exp(− Z 2 4 D∆) exp( iγδgZ )dZ [4-4]
−∞
which simplifies to
E 0 ( g ) = exp(−γ 2δ 2 g 2 D∆) [4-5]
Taking into account the finite duration of the gradient pulses leads to the exact expression determined
by Stejskal and Tanner1 for the echo amplitude, E, as a function of gradient field strength, g, namely
= exp[− γ 2 g 2δ 2 D(∆ − δ 3 )] ,
E(g)
[4-6]
E ( 0)
where γ is the gyromagnetic ratio of the observed nucleus, g is the gradient field strength of the
gradient pulse, δ is the length of the gradient pulse, D is the diffusion coefficient and ∆ is the delay
between the two gradient pulses. This equation can be linearized and used to calculate the diffusion
coefficient.
In Figure 4.16 the echo amplitudes are plotted and fitted to the Stejska-Tanner1 equation to obtain a
value for the self-diffusion coefficient of tap water.
1
Stejskal, E.O. and Tanner, J.E. (1965). J. Chem. Phys. 42, 288
The example in Figure 4.16 does not include the first data point. A common error in this experiment is
that this first data point is not on the line. This is often due to convection effects in large liquid samples.
It is sometimes difficult to produce Diffusion coefficient = (3.07 ± 0.07) × 10 m s

-9 2 -1
a straight line in this graph. Make 10

0
9
sure that the range of pulse lengths 8
7
is not too large relative to the echo 6
Attenuation (E/E0 )
5
time and compensate by using a 4
large current to ensure that there is 3
sufficient attenuation (2 A is the

2
maximum). We recommend a echo
time of at least 200 ms and a
minimum delay between gradient
0.1 0.2 0.3 0.4 0.5 0.6 0.7
pulse and 180 pulse of 100 ms. γ2 G 2 δ2 (∆-δ/3) (109 sm-2 )
Figure 4.16. PGSE Analysis
Attenuations greater than a factor of 10 are difficult to achieve with good signal-to-noise so this should
also be considering when choosing parameters. In magnitude mode it is typical to see the data curving
up at the end of the plot if the attenuation is too great. A line which curves down for larger attenuations
is usually caused by insufficient delay between gradient pulse and 180 pulse. In this case increase the
echo time or decrease the pulse step size.
4.5 The Carr-Purcell-Meiboom-Gill (CPMG) Experiment

4.5.1 The CPMG Pulse Sequence
In a spin-echo NMR experiment (section 2.4.4) a 180° B1 pulse is used to re-focus transverse
magnetization de-phased due to local magnetic field inhomogeneities. This pulse sequence consists of
two B1 pulses; a 90° pulse followed, after an echo time period (τE), by a 180° pulse. The echo forms
during the echo time period following the 180° pulse, such that the centre of the echo occurs at a time
τE after the centre of the 180° pulse. This methodology re-focuses any NMR signal which was de-
phased due to local magnetic field in-homogeneities; however, de-phasing due to spin-spin relaxation is
irreversible and so the echo amplitude is weighted by the spin-spin relaxation decay term, characterized
by the spin-spin relaxation time constant T2.
In the Carr-Purcell (CP) sequence, multiple re-focusing pulses are used to generate a train of spin-echo
signals. The CP pulse sequence diagram is shown in Figure 4.17. The first 90° excitation pulse excites
the signal. After a time period τE, a 180° re-focusing pulse is employed to generate the first echo.
Subsequent 180° re-focusing pulses are applied at intervals of 2τE to generate the additional echoes. An
echo forms at a time τE following each 180° pulse. As mentioned above, the amplitude of each echo
will be weighted by the T2 decay. Therefore a plot of the echo amplitudes as a function of time (where t
= 0 is the centre of the first 90° excitation pulse) will provide a record of the T2 decay of the observed
nucleus.
In the CP sequence each echo is sampled within an acquisition window centred about the centre of the
echo. Each acquired echo can be Fourier transformed to yield an NMR frequency spectrum. The echo
amplitude can be calculated from the spectrum as the integral under the spectral peak corresponding to
the sample. Alternatively, the echo amplitudes can be determined directly from the time domain data as
the intensity of the single point at the centre of the echo. The advantage of integrating the spectrum
over the time domain method is that the integral rejects all noise outside of the integration range.
The CP sequence is very sensitive to non-ideal tip angles. If the 180° pulses are not ideal (i.e. if they
vary across the sample or are slightly larger or smaller than 180°), then the re-focusing of
magnetization de-phased due to local magnetic field inhomogeneities will be incomplete. Thus the
amplitude of each echo will be weighted by the imperfections in the refocusing pulses as well as by the
T2 relaxation of the nuclei. Pulse errors in the CP experiment are cumulative, i.e. the errors in the
second echo are the result of not only the non-ideal nature of the second 180° pulse but also the first.
Figure 4.17. The Carr-Purcell (CP) multiple-echo pulse sequence
The Carr-Purcell-Meiboom-Gill (CPMG) pulse sequence is an adaptation of the CP sequence that

compensates for errors due to non-ideal 180° pulses. In this sequence, the 180° pulses used to re-focus
the magnetization have a phase of π/2 (90°) with respect to the initial 90° pulse. For example, a typical
CMPG sequence will employ a 90x excitation pulse followed by a train of 180y re-focusing pulses. In
this sequence, errors in the re-focusing pulses are corrected for every second echo and so cumulative
pulse errors are avoided; however, because only every second echo is corrected, an oscillation in signal
amplitudes between corrected and non-corrected echoes is often observed. The non-ideal tip angle
compensation can also be achieved by alternating the phase of the 180° pulses by π (180°). In this pulse
sequence the pulse train would is: 90x – 180x – 180-x – 180x – 180-x.
4.5.2 The Terranova EFNMR CPMG Experiment
The CPMG experiment dialog is shown in Figure 4.18 with some typical parameters. Most of the
CPMG pulse parameters are familiar from previous EFNMR experiments; however there are some new
parameters to consider.
The common pulse parameters are the same as those used in the core Terranova EFNMR experiments.
The only additions are the phase of the 90° and 180° pulses. These parameters take values in degrees
between –360° and +360°. Refer to Table 4-1 for the pulse phases required for tip angle compensation
using the ‘alternating 180° pulse phase’ and ‘constant 180° pulse phase’ sequences. The echo train
parameters, “Number of Echoes” and “Echo time (ms)”, determine the number and time spacing of the
re-focusing (180°) pulses. The total number of acquired data points cannot exceed 32768. Therefore the
number of acquired data points per echo multiplied by the number of echoes must be less than or equal
to this number. The echo time is constrained by the acquisition parameters and the ring-down of the
coil as in the single-echo spin echo experiments. This will be discussed further in the section on the
acquisition parameters.
Figure 4.18. The CPMG experiment dialog

The CMPG dialog provides the choice of two pulse sequences. In the first option: “Constant 180 pulse
phase” all of the re-focusing (180°) pulses have the same phase. In the second option: “Alternating 180
pulse phase” the phase of the re-focusing (180°) pulses alternates by π (180°). Refer to Table 4-1 for
the correct B1 pulse phases required for tip angle compensation with each pulse sequence.
Table 4-1. Chart of the relationship between B1 pulse phase and tip angle compensation
for the multi-echo pulse sequences.
Relative Phase Relative Phase
Tip Angle
Sequence 90°-180° 180° pulses
Compensation
- rad (deg) - - rad (deg) -
Alternating 180 phase 0 π (180°) Yes
π/2 (90°) π (180°) No
Constant 180 phase 0 0 No
π/2 (90°) 0 Yes
The acquisition parameters for the CPMG experiment are slightly different to the core Terranova
EFNMR experiments. Instead of choosing an acquisition time, the user chooses a dwell time, ∆t. Dwell
time is defined as the time between successively acquired data points in the time domain. ∆t takes
values that are multiples of 10 µs. The acquisition time is equal to the dwell time multiplied by the
number of points acquired per echo. Note that the “Number of points” parameter refers to the number
of points acquired per echo. The “Number of scans”, “Display range” and “Average” parameters are
the same as those used in the core Terranova EFNMR experiments.
The acquisition parameters must be chosen carefully in a multi-echo experiment. The echo time must
be long enough to accommodate both the ring-down of the B1 coil and the signal acquisition window.
Therefore the echo time must be greater than half of the acquisition time (number of points × dwell
time) plus the coil ring down (typically about 25 ms).
The processing parameters provide the user with several post-processing options in addition to defining
the spectral integration range. These options include zero-filling and filtering. A zero-filling factor of 1
equates to no zero-filling. A filter choice of “(none)” equates to no filtering. In multiple echo
experiments, in order to employ short echo times, acquisition times are typically short. In addition to
poor spectral resolution, this often means that the full echo is not sampled and so there is a sharp fall to
zero at the edges of each echo acquisition window. This sharp decay to zero introduces sinc-like
oscillations within the spectrum which can inhibit peak integration measurements. The problem of low
spectral resolution can be addressed by zero-filling at the edges of the echo. The oscillations in the
spectrum can smoothed by the application of a sine-bell-squared time domain filter which smooths the
edges of the echo to zero. It is important to note that such a filter will cause broadening of the spectral
peak and so the integration range must be increased accordingly.
Note: it is possible to obtain the echo amplitudes directly from the time domain data rather than from
an integral of the spectral peak. However, the latter method is preferred as it rejects all noise at
frequencies outside of the integration range.
4.5.3 Sample CPMG results
The following data sets were acquired using the parameters shown in the CPMG experiment dialog
(Figure 4.18). The time domain data was apodized with a sine-bell-squared filter and then zero-filled
by a factor of 4. The three data sets presented illustrate the two methods of compensating for non-ideal
re-focusing (180°) pulses.
The first example, Figure 4.19, is the original CP experiment. The re-focusing (180°) pulses have a
constant phase and there is no phase shift between the excitation (90°) and the re-focusing (180°)
pulses. Tip angle errors have a dramatic effect on the observed echo amplitudes. The apparent T2 is
much shorter than the anticipated value of a couple seconds.
FID data Spectrum (Echo: 1) T2 = (360 ± 40) ms

3
×10
100 20
0.6
0.5
Attenuation (E/E0)
Signal (µV) 50 15
0.4
0 10 0.3
0.2
-50 5
0.1
0 0.0
-100
1000 2000 3000 4000 5000 2160 2180 2200 2220 2240 1000 2000 3000 4000 5000
time (s) frequency (Hz) t (ms)
Figure 4.19. The original CP experiment employing constant phase re-focusing pulses. There is no (0°) phase
shift between the excitation (90°) and re-focusing (180°) pulses. Non-ideal tip angles have a dramatic effect on the
observed echo amplitudes.
Figure 4.20 shows a single-shot T2 measurement acquired using the CPMG sequence. The phase of the
re-focusing (180°) pulses is constant and there is a phase shift by π/2 (90°) between the excitation (90°)
and re-focusing (180°) pulses. The echo amplitudes yield a T2 value much closer to the expected value.
Deviations of this observed T2 from the actual T2 may arise from diffusive attenuation of the echo if
background gradients are present.

3
×10
15
0.8
50
Attenuation (E/E0)
10
Signal (µV)
0.6
0
5 0.4
-50 0.2
0
0.0
1000 2000 3000 4000 5000 2160 2180 2200 2220 2240 1000 2000 3000 4000 5000
Figure 4.20. A sample single-shot T2 measurement of water using the CPMG pulse sequence. The relative phase
between the excitation (90°) and the re-focusing (180°) pulses is π/2 (90°).
Figure 4.21 presents the results of a multiple-echo experiment where the phase of the re-focusing
pulses alternate by π (180°). There is no phase shift between the excitation (90°) and re-focusing (180°)
pulses. This method compensates for non-ideal re-focusing pulses in the same manner as the CPMG
pulse sequence.

3
×10
100
20
0.8
Attenuation (E/E0)
50 15
Signal (µV)
0.6
0 10
0.4
-50 5
0.2
0
-100
0.0
1000 2000 3000 4000 5000 2160 2180 2200 2220 2240 1000 2000 3000 4000 5000
Figure 4.21. A sample single-shot T2 experiment using the CPMG experiment with re-focusing (180°) pulses with
phases alternating by π (180°). The relative phase between the excitation (90°) and re-focusing (180°) pulses is 0.
Note that it often advantageous to deshim the field when performing the CPMG experiment to
minimise the effects of imprecise 180° pulses.
Earth’s Field MRI Experiments 5-1
5 Earth’s Field MRI Experiments

Once the basic NMR experiments have been mastered you can move onto the imaging or MRI
experiments. These come in 3 flavours which vary depending on how they sample k-space (defined
below).
1. Gradient echo imaging

2. Spin-echo imaging
3. Filtered back-projection imaging
The first two experiments sample k-space on a Cartesian raster, while the third uses radial sampling.
5.1 k-space
5.1.1 Introduction
Consider a small volume element (or voxel) in our

sample at position ~ r . This has dimensions
dV S = dx dy dz . The peak amplitude of the NMR
signal from this voxel during a normal pulse and
collect experiment will be given by equation 4-1.
dE (~
r ) = ρ (~
r )dV S
= γB E (B1 I )M 0 (~
r)
× [1 − exp(−τ P / T1 )]
[5-1]
× exp(−t / T2* ) sin(θ tip )dV S
ρ (r~ ) is called the spin-density function (because

M 0 is proportional to the density of NMR active
nuclear spins at position ~
r ). Figure 5.1. Small voxel within sample
In MRI we apply magnetic field gradients during (and in most cases before), the data collection
process, to give each voxel a unique NMR tag. In the most general case we apply a 90º pulse, apply a
~ ~
gradient G phase of duration tphase and then start sampling while applying a different gradient Gread of
duration tread. The effect of these gradients is to provide a time variation to the voxel amplitude having
the form:
dE (t phase , t read , ~ ( ~
r ) = ρ (r~ ) exp iγG phase .~
~
r t phase + iγG read .~ )
r t read dV s [5-2]
In equation 5-2 the first term in the exponential function applies a phase shift to the signal (remember
~
that θ = ωt = γBt = γG.r~t ). The second term is the phase shift which is acquired during the acquisition,
or data read out process. This equation becomes easier to understandable if we identity the equality,
~ ~
k = γGt , where k is called the angular spatial frequency. In this case we can write
~ ~
dE (k phase , k read , ~
r ) = ρ (~ (~ ~
r ) exp ik phase .r~ + ik read .~ )
r dV S [5-3]
If we take the special case where the read and phase gradients are along the (x, y) and z axes we have:
dE (k x , k y , k z , x, y, z ) = ρ ( x, y, z ) exp(i (k x x + k y y + k z z ) ) dV S [5-4]
The signal from the total sample may then be found by integrating over all voxels:
E (k x , k y , k z ) = ∫∫∫ ρ ( x, y, z ) exp(i(k x
x + k y y + k z z ) ) dx dy dz [5-5]
5-2 Earth’s Field MRI Experiments
This relationship states that the observed signal E is the 3-dimension Fourier transform of the spin-
density function ρ. In other words when we sample NMR data in the presence of gradient we are in
actuality sampling Fourier or reciprocal space. In NMR this is conventionally termed sampling k-space.
Clearly it is not necessary to apply all 3 orthogonal gradients to get spatial information. A single
gradient applied along 1 axis will produce a so-called “1D image” after transformation. e.g. if the
gradient is applied along the x axis alone we have:
E (k x ) = ∫∫∫ ρ ( x, y, z) exp(i(k x) )dx dy dz

x
or
E (k x ) = ∫ (∫∫ ρ ( x, y, z) dy dz ) exp(i(k x) )dx

x
[5-6]
The inner integral is a projection of the spin-density function onto the x axis and so the Fourier
transform of the signal E is equal to this projection.
Likewise applying two gradients can be used to obtain a 2D projection:
E (k x , k y ) = ∫ (∫∫ ρ ( x, y, z) dz ) exp(i(k x + k y) )dx dy

x y
[5-7]
In this case the projection is onto the x,y plane.
All three imaging possibilities are available using the Terranova MRI system.
5.1.2 A graphical representation of k-space
To avoid working with integrals all the time, a graphical representation of the MRI data sampling has
~ ~
been introduced. Because of the relationship, k = γGt , different points in k-space are “reached” at
different times. Therefore we can imagine k-space being “scanned” by the imaging experiment. In the
case of the 1D imaging experiment we start in the centre of k-space and then move along one axis
reading data as we go:
ky
kx
Figure 5.2. 1D k-space sampling
In the 2D experiment the phase gradient moves us along one k-space axis and then we read out data
while moving along the orthogonal axis. By performing a series of such experiments, each of which
changes the phase gradient amplitude, it is possible to completely sample 2D k-space.
ky
kx
Figure 5.3. 2D k-space sampling (only 1 quadrant shown for clarity)

3D k-space sampling works in much the same way except that there are now two phase gradients (say y
and z) and one read-out gradient (say x).
kz
kx
ky
Figure 5.4. 3D k-space sampling (only 1 octant shown for clarity)
5.1.3 Other ways of moving about k-space
In addition to a positive movement along a k-space axis it is clearly possible to go in the reverse
direction by reversing the sign of the gradient. This method is used in the gradient echo imaging
sequence.
Less obvious is the use of a 180° pulse to jump from one side of k-space to the other. If a point ( k x , k y )
has been reached in k-space then the accumulated signal phase shift will be: φ = k x x + k y y . A 180°
pulse will then reverse the sign of the accumulated phase to: φ = −k x x − k y y , effectively moving the
sampling point to the opposite side of k-space i.e. ( − k x ,−k y ). This trick is used in the spin-echo
imaging pulse sequence. (Also called spin-warp imaging).
Finally we don’t need to move though k-space along the Cartesian axes. By combining two or more of
the orthogonal x, y and z gradients together, it is possible to move in any direction in k-space, from
radial lines away from the origin (as used in filtered back-projection, the original imaging technique),
to spiral schemes designed to minimise the amount of time taken to sample k-space. Although these
methods have the advantage of rapid scanning they do require more complex processing schemes,
because of the non-Cartesian way of sampling k-space.
5.2 Preliminary experiments

To use the imaging macros you will first need to have run the AnalyseCoil and B1Duration macros to
determine the correct tuning parameters and 90°/180° pulse characteristics. You should also run
AutoShim with the sample to be imaged loaded into the probe to ensure the line-width is as narrow as
possible. (Narrow line-widths mean more resolution and better signal-to-noise ratios). Note that these
parameters are not visible in the imaging macros but they are still used (see section 5.2.1 below).
Finally run the PulseAndCollect macro to determine the B1 frequency. (If this is not set correctly the
image may not be visible and in the case of FBP it will be smeared). To choose the other pulse
sequence parameters, first run the T1-Bp and T2 experiments to determine appropriate polarizing
durations and echo times. In the following plots this has been done for a red pepper sample.
T1 = (900 ± 40) ms T2 = (550 ± 40) ms
0.8
0.6
Attenuation (E/E0)
Attenuation (E/E0)
0.6
0.4
0.4
0.2
0.2
0.0
0.0
500 1000 1500 500 1000 1500 2000
t (ms) t (ms)
Figure 5.5. Relaxation measurements on a red-pepper sample, made before starting an imaging experiment.
From this we determine that T1 is about 0.9 s and T2 0.6 s. A polarization duration (tBp) of 2 seconds is
chosen as a compromise between a short imaging time and best signal-to-noise. The echo time (for
spin-echo imaging and FBP) should be chosen so that it is significantly less than the sample T2,
otherwise there will be loss of signal – 200 ms is chosen here. Note that an echo-time of less than 100
ms is not recommended because image distortions will be introduced. For gradient echo and spin-echo
imaging the phase gradient duration (tgrad) should be as short as possible. Typically this will be about
50 ms as it is limited by the maximum available gradient current. (Although this depends on the chosen
field-of-view).
5.2.1 Viewing common parameters
To prevent clutter in the MRI user interface, some of the common parameters are not visible in the
imaging macros. However you can still check them, and if necessary modify them, by running the
macro CommonParameters in the MRI menu:
Figure 5.6. The common parameters interface.
Pressing the save button to copy the parameters to the user preferences file (see Section 3.2).
5.3 Gradient Echo Imaging

This form of imaging uses the concept of the gradient echo to traverse k-space. The pulse sequence for
the experiment is shown below:
BP tPolz
tPolz-delay
B1 90°pulse
1
tEcho
Read gradient
Read shim Gro
Gr
2
tGrad
Phase gradients Phase shims 3
Gp 4 5
tAcqu
tAcq-delay
Signal 6
Figure 5.7. The gradient echo imaging pulse sequence

• The experiment starts by setting up the shim gradients to the correct values. These should have
been determined by running the AutoShim experiment beforehand (see section 2.4.5)
• The polarizing pulse (Bp) is then switched on to magnetize the sample. The duration of this
pulse should be at least twice the sample’s T1 to get a reasonable signal level.
• Following the polarizing pulse there is a short delay (tPolz-delay = 50 ms) to allow the current to
switch off and the magnetization to rotate adiabatically until it lies along the Earth’s field
direction. This delay limits sample T1s to more than 100 ms.
• The 90° B1 pulse then excites the sample and following this the read, Gr, and phase gradients,
Gp, are switched on for a time tGrad to move the initial sampling point out into k-space. These
gradients result in a rapid dephasing of the magnetization.
• After the phase and read gradient pulses, the read-out gradient, Gro, is switched on and the
receiver starts digitising the signal induced by the precessing magnetization. The readout
gradient has the reverse sign to the shorter read gradient so that the spins rephase producing an
echo. Note that there is a short delay (tAcq-delay = 20 ms) between the end of the phase gradient
pulse and the start of data acquisition to allow for coil ring-down.
A k-space representation of the experiment is shown in the next figure. The numbers refer to different
points in the pulse sequence.
3 4 5 6
Phase axis
2
1
Read axis
Figure 5.8. Traversing k-space using the gradient echo pulse sequence
For a 1D experiment the phase gradients are zero and so k-space is 1 dimensional, for 2D images there
is a single phase gradient producing a 2D k-space data set, while for 3D imaging there are 2 phase
gradients producing a 3D k-space data set. All three experiments use a single read gradient. The
relationship between k-space data and final image is a n-dimensional Fourier transform. This FT is
applied automatically as the data is collected (although in the case of 3D data the 3D-FT is not applied
until all data has been collected).
Note that because of the acquisition delay (tacq-delay) we are unable to sample all of k space. Even if this
delay were zero, sampling a whole row of k-space requires a long echo time which may result in
significant signal reduction if the sample T2 is short. To minimise distortions which may result from
incomplete sampling of k-space, ensure that the echo time is long enough so that data acquisition
begins well to the left of the echo peak.
The macro interface used to control the gradient echo pulse sequence is shown below (for details of
each window parameter see the section 5.3.5):
Figure 5.9. The gradient echo imaging macro interface
Note that the parameters: Polarizing current, transmit and receive gain, tuning capacitance and 90
degree pulse duration are all used by this macro but are not adjustable here. It is assumed that these
have been correctly set using the standard NMR experiments before running this macro. See section
5.2.1 for details on how to view these parameters.
5.3.1 The confirmation window
Before an experiment is started a window will appear which summarizes the imaging parameters you
have chosen. These include the imaging direction, the gradient/shim strengths and currents, and the
estimated time taken to collect the image data. It also reports the folders in which the data will be
stored.
If everything looks correct then press the OK button to proceed.
Figure 5.10. The gradient echo confirmation window
While the experiment is running the status bar at the bottom of the macro window displays the data
collection progress. It shows (from left to right), the number of scans collected, how many of the phase
gradient steps have been completed and also the elapsed time and remaining time.
Figure 5.11. The imaging status bar showing experiment progress

5.3.2 1D imaging
Once the above parameters have been selected, you should start by performing a 1D imaging
experiment. First, choose a suitable experiment name and working directory. All the data collected by
the experiment will be stored here. Next choose the 1D option in the top left of the macro interface, an
image orientation of Y with 32 data points and a field of view sufficient to cover the sample in the Y
(read) dimension. Here we started with an image bandwidth of 32 Hz and a repetition time of 4
seconds. The 32 Hz means that with a 1 Hz linewidth (typical) we will have (ideally) 32 horizontally
resolved points in our 32 by 32 pixel image. The 4 second delay is longer than the sum of the
acquisition time ( t acq = N / bandwidth = 32/32 = 1 s ) and tpolz, and includes time for data transfer and
processing.
Decimated data (Scan: 4) Magnitude image (Scan: 4)
×103 ×10 3
1
8
0
-1
6
flt*m
-2 4
-3
2
-4
-5 0
0 0.2 0.4 0.6 0.8 -15 -10 -5 0 5 10 15
Figure 5.12. y axis projection for the red-pepper sample – 16 seconds
Initially run the experiment with the filter on and a few scans (see Figure 5.12). The image appearing in
the right side of the 1D window will be the projection of the sample magnetization onto the Y axis. On
the left side is the complex k-space data with the applied filter shown in red. Note that this data set has
been digitally shifted so that the central frequency has moved from 2189 Hz (the local f0) to 0 Hz.
Occasionally a noise peak will appear at the edge of the image – sometimes this can be removed by
decreasing the bandwidth until it is no longer visible (but note that this will reduce the image
resolution). However if the noise peak appears within the image you will need to reposition the probe
so that the noise goes away or the resonant frequency shifts to a more favourable region.
Once you are satisfied with the projection along the Y-axis, repeat the experiment for the Z and X axes.
Usually the latter, (longitudinal), axis is the most difficult and it is important that this axis is orientated
east-west with respect to the Earth’s field before you attempt to do this. (It may be difficult to minimise
noise at this angle and so a compromise might be necessary.)

×103 ×10 3
4
15
3
2 10
flt*m
0 5
-1
-2 0
0 0.2 0.4 0.6 0.8 -15 -10 -5 0 5 10 15

Figure 5.13. x axis projection for the red-pepper sample – 16 seconds
With all 1D projections looking good, you can proceed to the 2D experiment.
5.3.3 2D imaging
Again choose a new file name for your data and then select the 2D radio-button option. Select the plane
to image (these are the characters that appear in the image orientation menu), the field of view, and a
lower resolution (say 16 by 16 points) and then start the experiment.
The macro is written such that the most intense low frequency rows of k-space are scanned before the
high frequency rows, which typically contain less signal. In this way you can very quickly determine
whether the image will be a good one or not. At any time you can terminate the experiment by pressing
the stop button, see section 3.10 for details.
If the low resolution image looks good, then you can repeat it at higher resolution (32 by 32) although
addition accumulations may be required if the signal-to-noise ratio is poor.
Figure 5.14. 2D YZ red pepper image (4 scans) – 8 minutes
Figure 5.15. 2D XY red pepper image (4 scans) – 8 minutes
5.3.4 3D imaging
If 2D images in both the XY and YZ planes look good, then a 3D image can also be obtained. However
be aware that the total imaging time can be quite long if high resolution and good signal-to-noise are
desired. In the following example, a 1.2 hour 16 by 16 by 16 image of the red-pepper has been
collected using 4 scans per line of k-space. At the end of the experiment the 3D k-space data has been
automatically transformed and stored in the matrix pepper3d. Planes from this data set can be displayed
using the plot3Dslices macro (in the 3D menu), or more directly using the 3D surface plot option in the
3D plot window’s contextual menu.
Figure 5.16. yz pepper slices from a 16 by 16 by 16 experiment and a 3D surface rendering.

The left plot is generated using the macro multiplotDisplay in the 3D menu and the right plot using “display 3D
surface plot” in the 3D plot contextual menu.
5.3.5 Description of each gradient echo macro parameter
Image parameters
Image dimension ............. Choose what dimension the final image will have. Either 1D, 2D or 3D.
Image orientation ........... One/two or three letters which signify the read and phase dimensions. The
first letter is the read dimension the other two the phase dimensions. Only n
letters are visible for an n-dimensional image.
Matrix size ....................... The number of data points in each dimension of the final image data set.
Field of view .................... The number of mm covered by each dimension.
Pulse sequence parameters
Polarizing duration .......... Length of polarizing pulse.

B1 frequency ................... Resonant frequency as determined by pulse and collect experiment.
Phase gradient duration ... Length of phase and initial read gradient pulses.
Echo time ........................ Time between centre of 90° B1 pulse and centre of echo.
Bandwidth ……............... Frequency range of collected data in the read direction.
Repetition time ................ Time between successive scans of k-space.
Number of scans ............. Number of accumulations per line of k-space.
Output location
Working directory .......... Directory where the data folder will be stored.
Experiment name ........... Name used for folder holding data.
Buttons
Run .................................. Starts the experiment.

Stop ................................. Stops the experiment when current scan is finished.
Load ................................ Load a previous set of parameters (as stored in the data folder).
Shims .............................. Show and adjust the current shim settings.
Help ................................ Display this help file.
Close ............................... Close the macro window.
5.3.6 Parameter limits
Following are the limiting and recommended values for the parameters used in the gradient echo
imaging macro window. Limits are the range of values which the macro will accept, while
recommended are those values which will produce a reasonable image in a reasonable length of time.
Table 5-1 Gradient echo imaging parameter range

Parameter Units Limits Recommended
Matrix size - 1 to 512 8-64
Field of View mm 0 to 500 10-200
Polarizing duration ms 0 to 5000 500-4000
B1 frequency Hz 1000-5000 Measure it
Gradient pulse duration ms 0 to 500 50-100
Echo time ms 10-1000 100-500
Bandwidth Hz 1-200 16/32/64
Repetition time s 0.5 to 60 2-10
Number of Scans - 1 to 128 1-32
5.3.7 Example Images
All the following images have been filtered with a sinebell squared filter (shifted in the read
dimension). Double filtered means this filtering has been applied twice to improve the signal-to-noise
ratio. The images have also been zero filled from 32 to 64 pixels in each dimension.
Figure 5.17. 3 tube YZ image (32 by 32 pixels, 4 accumulation, 90 by 90 mm FOV, total imaging time 5 minutes)
Figure 5.18. 2 cylinder XY image (32 by 32 pixels, 4 accumulations, 2 step phase cycling, 160 by 160 mm FOV,
total time 5 minutes – note that the central cavity was empty when making this image)
3
Figure 5.19. Mandarin image (32 by 32 by 32, 4 accumulations, double filtered, FOV (110 mm) , total
imaging time 4 hours)
Figure 5.20. Two tube image (32 by 32 by 32, 4 accumulations, Gaussian filtered, FOV 100 ×100 × 200 mm total
imaging time 2.5 hours (using doped water). Tubes are 120 mm long and 20 mm in diameter and 32 mm apart.
5.4 Spin Echo Imaging
This form of imaging uses gradients in conjunction with a 90-180° pulse to traverse k-space. The pulse
sequence for the experiment is shown below:
BP
tPolz
tEcho tEcho
1 2 3
B1
Read gradient
tGrad
Phase gradient 1
Phase gradient 2
4 5 6
tAcqu
tDelay
Signal
Figure 5.21. Spin- echo imaging pulse sequence
• The experiment starts by setting up the shim gradients to the correct values. These should have
been determined by running the AutoShim experiment beforehand.
• The polarizing pulse (Bp) is then switched on to magnetize the sample. The duration of this
pulse should be at least twice the sample’s T1 to get a reasonable signal level.
• Following the polarizing pulse there is a short delay (tPolz-delay = 50 ms) to allow the
magnetization to rotate adiabatically until it lies along the Earth’s field direction. It also
provides time for the receiver to recover from the ring-down induced in the B1 coil following
the rapid switching of the polarizing field. This delay limits sample T1s to more than 100 ms.
• The 90° B1 pulse then excites the sample and following this the phase gradients are switched
on for a time tGrad to move the initial sampling point out into k-space. The read gradient is
switched on before the polarizing pulse, but has no effect until the 90° pulse is activated.
These gradients result in a rapid dephasing of the magnetization.
• After a time techo, a 180° pulse is applied which causes the magnetization to start refocusing,
producing an echo an additional techo seconds after the 180° pulse. During this period the read-
out gradient has a constant value, causing k-space to be traversed in the read gradient
direction. The echo position corresponds to the centre of k-space along the read gradient axis.
• Data acquisition is started tacq-delay seconds after the 180° pulse to allow for probe ring down
(typically 25 ms). Acquisition continues for tacq seconds.
A k-space representation of the experiment is shown in the next figure. The numbers refer to different
points in the pulse sequence.
3 4 5 6
Phase axis
1
Read axis
Figure 5.22. Traversing 2D k-space using the spin- echo pulse sequence
For a 1D experiment the phase gradients are zero and so k-space is 1 dimensional, for 2D images there
is a single phase gradient producing a 2D k-space data set, while for 3D imaging there are 2 phase
gradients producing a 3D k-space data set. All three experiments use a single read gradient. The
relationship between k-space data and final image is a Cartesian n-dimensional Fourier transform. This
FT is applied automatically as the data is collected (although in the case of 3D data the 3D-FT is not
applied until all data has been collected).
Note that because of the acquisition delay (tacq-delay) we are unable to sample all of k space. Even if this
delay were zero, sampling a whole row of k-space requires a long echo time which may result in
The macro interface used to control the spin-echo pulse sequence is shown below (for details of each
window parameter see the section 5.4.5):
Figure 5.23. The spin echo imaging macro interface
The only major difference between this macro interface and the gradient echo macro is the addition of
the phase cycle option. Select a 2 step phase cycle and a multiple of 2 scans if you are having problems
with image artefacts, (typically streaks running horizontally though the centre of the image). Also note
that the time between the 90 and echo peak is twice the displayed “Echo time” parameter. This means
that the spin-echo imaging sequence will produce less signal than the gradient echo method for short T2
samples if both have the same echo time setting.
5.4.1 1D imaging
Once the above parameters have been selected, you should

start by performing a 1D imaging experiment. First, choose a
suitable experiment name and working directory. All the data
collected by the experiment will be stored here. Next choose
the 1D option in the top left of the macro interface, an image
orientation of Z with 32 data points, and a field of view
sufficient to cover the sample in the Z (read) dimension. Here
we started with an image bandwidth of 32 Hz (suitable for a
Figure 5.24. The 6 tube phantom used in
1Hz line-width and 32 data points), and a repetition time of
the following experiments. The tubes are
2.5 seconds (the latter time can be found by entering a small 13 mm in diameter and 50 mm long. Each
number and running the experiment – a popup dialog will tube contains water doped with copper
suggest a sensible number). sulphate so that the T1 is about 500 ms.
Before an experiment is started a window will appear which summarizes the imaging parameters you
have chosen. These include the imaging direction, the gradient/shim strengths and currents, the
estimated time taken to collect the image data and also indicates where the data will be stored.
Figure 5.25. The spin-echo confirmation window
As data is collect it is first down-converted from the resonant frequency (in this case 2325 Hz) to 0 Hz
and filtered, removing all unwanted frequencies outside the desired bandwidth. In this way only a few
data points need be processed. The collected data is displayed on the left side of the 1D plot window
while the processed version of this – obtained by applying a magnitude 1D Fourier transform - is
displayed on the right.

×10 3
800
3
600
400
2
flt*m
200
0 1
-200
-400 0
0 0.2 0.4 0.6 0.8 -15 -10 -5 0 5 10 15

Figure 5.26. z axis projection of the 6 tube phantom: 8 scans in 30 seconds (90 mm FOV)
While the experiment is running the status bar at the bottom of the macro window displays the data
collection progress. It shows (from left to right), the number of scans collected, how many of the phase
gradient steps have been completed and also the elapsed time and remaining time (the following
example is for a 3D image).
Figure 5.27. The imaging status bar showing experiment progress
Unless you have an excellent signal-to-noise ratio, perform the experiment with the filter on and
accumulate several scans. The image appearing in the right side of the 1D window will be the
projection of the sample magnetization onto the Z axis. In this case the 3 peaks represent the projection
of the 6 tubes onto the Z axis. On the left side is the complex k-space data with the applied filter shown
in red. To obtain the best image resolution when performing this experiment you should ensure that the
frequency range across the image (the bandwidth) is greater than or equal to the line-width multiplied
by the number of image points. Notice that decreasing the bandwidth increases the signal-to-noise ratio
at the expense of resolution and vice-versa.
Occasionally a noise peak will appear at the edge of the image – sometimes this can be removed by
decreasing the bandwidth until it is no longer visible. However if the noise peak appears within the
image you will need to reposition the probe so that the noise goes away or the resonant frequency shifts
to a more favourable region.
Once you are satisfied with the projection along the Z-axis, repeat the experiment for the Y and X axes.
Usually the latter, (longitudinal), axis is the most problematic since it is important that this axis is
orientated east-west with respect to the Earth’s field to minimise image distortion. (It may be difficult
to minimise noise at this angle and so a compromise might be necessary.)

×102 ×10 3
5
4
3
flt*m
-5
1
0 0.2 0.4 0.6 0.8 -15 -10 -5 0 5 10 15

Figure 5.28. y axis projection of the 6 tube phantom: 8 scans in 30 seconds (90 mm FOV)

×10 3
200
0 3
-200
-400 2
flt*m
-600
-800 1
-1000
-1200 0
0 0.2 0.4 0.6 0.8 -15 -10 -5 0 5 10 15

Figure 5.29. x projection of the 6 tube phantom: 8 scans in 30 seconds (120 mm FOV)
With all 1D projections looking as expected, you can proceed to the 2D experiment.
5.4.3 2D imaging
In this case the first phase gradient is varied from experiment to experiment so that each row in k-space
is scanned. As with the 1D experiment, first choose a new file name for your data and then select the
2D radio-button option. Select the plane to image (these are the two characters that appear in the image
orientation menu) and the field of view, and a lower resolution and then start the experiment. (Lower
resolution helps because the signal which was in N data points in the 1D experiment is now spread over
N2 – the extra N experiments does help a bit, but remember is introduces extra noise so you are still N1/2
worse off than in the 1D case if you keep the resolution the same).
The macro is written such that the most intense low frequency rows of k-space are scanned before the
high frequency rows, which typically contain less signal. In this way you can very quickly determine
whether the image will be a good one or not. At any time you can terminate the experiment by pressing
the stop button, see section 3.10 for details.
If the low resolution image looks good then you can repeat it at higher resolution (32 by 32), although
addition accumulations may be required if the signal-to-noise ratio is poor.
Figure 5.30. 2D YZ image of the 6 tube phantom (8 scans) –16 minutes
Figure 5.31. 2D XZ image of the 6 tube phantom (8 scans) - 16 minutes
Note the reduction in signal-to-noise ratio in the second image as the data is spread out over a larger
area. This image would have benefited from more accumulations.
5.4.4 3D imaging
If the 2D images in both the XY and YZ planes look good, then a 3D image can also be obtained if
desired. However be aware that the total imaging time can be quite long if “high” resolution and a good
signal-to-noise ratio are desired. In the following example a 1.75 hour 32 by 32 by 16 image of the 6
tube phantom has been collected using 4 scans per line of k-space and 2 step phase cycling. At the end
of the experiment the 3D k-space data is automatically transformed and stored in the matrix 6tubes.
Planes from this data set can be displayed using the plot3Dslices macro or more directly using the 3D
surface plot option in the 3D plot window.
Figure 5.32. The 3D spin-echo imaging dialog
Note that it is possible using a second instance of Prospa to view the progress of the collected data set
while the pulse program is still running. Just load the 3D data set “data.3d” stored in the data folder
using the macro importData in the NMRI menu and then transform it using the macro
nDFourierTransform in same menu. Following are screen shots of these two macros as used to
transform the 6tubes data:
Figure 5.33. The importData macro found in the NMRI menu.

(Used to load the raw data into the matrix mat3d)
Figure 5.34. The nDFourierTransform macro found in the NMRI menu

(used to produce the 3D image from the k-space data)
Note that when re-processing the original data set you will need to use a filter which mimics the shape
of the k-space data. In the case of the read dimension this means a function which peaks to the left of
the centre. In this case I have used a shifted Gaussian designed with the filter macro in the 1D menu.
This applied more severe smoothing than a shifted-sinebell squared filter
Shifted Gaussian
1.0
0.8
0.6
y
0.4
0.2
0.0
0 20 40 60 80 100
x
Figure 5.35. A shifted Gaussian which peaks at 1/10 of the horizontal axis width
Following are 3D surface renderings of the 6tubes image after 4 and 16 planes of k-space have been
selected. These images where made by running the 3D surface plot macro – found in the 3D plot
contextual menu.
Figure 5.36
Left: 3D transform after 4

planes of k-space
obtained
Right: 3D transform after

16 planes of k-space
obtained, showing
increased resolution
along the x axis.
Figure 5.37. The 3D surface plot macro
5.4.5 A description of each spin-echo macro parameter
Image parameters
Image dimension ............. Choose what dimension the final image will have. Either 1D, 2D or 3D.
Image orientation ............ Three letters which signify the read and phase dimensions. The first letter
is the read dimension, the other two the phase dimensions. Only n letters
are used for an n-dimensional image.
Phase cycle
None ……………...……. Run each scan with the same 180 pulse phase.
2 step ………………...… Alternate the phase of the 180 pulse relative to the 90 between scans to
minimise the effects of incorrect pulse tip angles.
Polarizing duration .......... The length of polarizing pulse (should be 2-3 times T1).
B1 frequency ................... Resonant frequency as determined by the pulse and collect experiment.
Phase gradient duration ... Length of phase and initial read gradient pulses.
Echo time ........................ Time between centre of 180° B1 pulse and centre of echo.
Bandwidth ……............... Range of frequencies covered by the read dimension.
Output location
Buttons

Following are the limiting and recommended values for the parameters used in the gradient echo
imaging macro window. Limits are the range of values which the macro will accept while
recommended are those values which will produce a reasonable image in a reasonable length of time.
Table 5-2. Gradient echo imaging parameter range

Gradient pulse duration ms 0 to 500 50-100
Echo time ms 10-1000 100-500
Bandwidth Hz 1-200 4-100
5.4.7 Example images
Following are a series of experimental Figure 5.38. 3 tube phantom,

results demonstrating the sort of images dimensions:
you can expect to get with the spin-echo
sequence. Note that the phantom images Overall diameter: 73 mm
were obtained using water doping with Overall length: 105 mm
sufficient CuSO4 to reduce the T1 to Inner tube diameter: 20 mm
about 500 ms. The noise level was 3.5
µV rms. All images have been zero
filled to twice their original dimension.
Figure 5.39. 32 by 32 pixels 90 mm FOV (4 acc. 5 min) Figure 5.40. 64 by 64 pixels 100 mm FOV (16 acc
1 hr)
Figure 5.41. 3 cylinder phantom XY image (32 by 32 pixels, 4 accumulations, 2 step phase cycling, 150 by 150
mm FOV, total time 5 minutes – note that the central cylinder was empty when making this image)
Figure 5.42. 6 tube phantom. YZ image (32 by 32 pixels, 4/16 (left/right) accumulations, 2 step phase cycling) 110
mm FOV, total imaging time 15 minutes/1 hour. Note this is doped/undoped water)
5.5 Filtered back-projection imaging

This method could also be called radial imaging since it samples k-space in a radial manner. However
the title “filtered back-projection” refers to the processing method.
BP
tPolz
tEcho tEcho
1
B1
2 3
4 5 6
Gradient 1
Gradient 2
Gradient 3
tAcqu
tDelay
Signal
Figure 5.43. The filtered back-projection pulse sequence
In this experiment (which only supports 2-dimensions) gradients 1 and 2 (which could be any of x, y or
z) are applied simultaneously to move in a radial direction in k-space. (Gradient 3 is only used to apply
shimming). Because of the dead time after the 90° pulse it is not possible to start sampling from the
centre of k-space (the preferred method using this technique) and so gradients are applied after the 90°
pulse to move away from the centre of k-space. A 180° pulse then flips the sampling point to the
reverse side of k-space so that during the read-out part of the sequence the central part of k-space
sampled.
G2
2 6
1
θ G1
5
4
3
Figure 5.44. Sampling k-space with filtered back-projection for angle θ
The experiment proceeds by acquiring k-space scans for a range of evenly spaced angles, θ, between 0
and 360 degrees. These are then filtered and back-projected (an implementation of the polar Fourier
transform) to produce the image. Note that k-space will be sampled unevenly (more densely closer to
the centre) and that there are regions outside the circle in Figure 5.44 where k-space will not be
sampled at all. However since the most important information is generally close to the centre of k-
space this is not usually a great loss.
Also note that because of the acquisition delay (tacq-delay) we are unable to sample all of k-space. Even if
this delay were zero, sampling a whole row of k-space requires a long echo time which may result in
The macro interface used to control the spin-echo pulse sequence is shown below (for details of each
window parameter see the section 5.5.4):
Figure 5.45. The filtered-back projection macro interface.
Before an experiment is started, a window will appear which summarizes the imaging parameters you
have chosen. These include the imaging direction, the gradient/shim strengths and currents, the
estimated time taken to collect the image data and indicates where the data will be stored.
Figure 5.46. The filtered-backprojection confirmation window

As data is collect it is first down-converted from the resonant frequency (in this case 2300 Hz) to 0Hz
and filtered to removing all unwanted frequencies outside the desired bandwidth. In this way only a
few data points need be processed. The collected k-space data is displayed on the left side of the 1D
plot window while the Fourier transform of this is in the centre (i.e. the 1D projection onto the gradient
axis). The filtered version of this projection is displayed on the right. (This last plot is obtained by
multiplying the k-space data by a ramp and smoothing filter and then Fourier transforming.)
Decimated data (Scan: 2) Magnitude image (Scan: 2) Filtered k-space data

×103 ×10 4 ×104
15 4 2
10 1
3
amplitude
0
signal
5
2
0 -1
1
-5 -2
-10 0 -3
0 0.2 0.4 0.6 0.8 -15 -10 -5 0 5 10 15 -15 -10 -5 0 5 10 15
time (s) frequency (Hz) proj: 0 (mm)
Figure 5.47. FBP 1-D data set from a bottle of water. Left k-space, middle 1D projection, right – filtered
projection.
5.5.2 2D imaging
First choose a new file name for your data, then select the plane to image (these are the characters that
appear in the image orientation menu), the field of view, a resolution (say 32 by 32 points) and then
start the experiment.
The macro is written such that after each scan of

k-space the image to date will be displayed i.e. a
2D polar Fourier transform is applied after each
scan is performed. In this way you will see the
image build up from the beginning and very
quickly determine whether the image will be a
good one or not. At any time you can terminate
the experiment by pressing the stop button (see
section 3.10).
If artefacts appear in the image (typically streaks

through the centre) try running the experiment
with phase-cycling enabled. This should reduce
the effects of an imprecise 180° pulse. If image
quality is good you can just sample half of k-
space. Select the 180° option in the “Max Angle”
box to do this. Figure 5.48. Bottle of water - 2D YZ image (16
projections)
5.5.3 Reprocessing FBP data
The FPB macro stores the radial k-space scans to a file called data.2d and a plot called plot.pt2, both in
the experiment directory, and also to a global matrix with the same name as the experiment directory
(“fbp” in this case). This data set can be loaded, displayed and then reprocessed.
Because you need a special Fourier transform to process FBP k-space data, an additional macro has
been provided. This is call ProcessFBP and the user interface is shown below:
Figure 5.49. The FBP processing macro interface
This works on the current 2D plot which should have been previously loaded and displayed either from
the above mentioned plot.pt2 file or the data set data.2d. Make sure you select the maximum angle
correctly before processing the data.
The macro has several options. Firstly you can

choose several smoothing filters. By default the
main FBP macro uses the Gaussian filter – as
defined in the 1D filters macro. However here
you are free to choose other filters. (Press the
Filter button to edit these filters).
In addition you can zero-fill the data set to

provide more data points in the image.
Finally you can correct for small errors in the B1

frequency. A 1 pixel shift corresponds to a
frequency shift of bandwidth/matrix size. In the
above example this would be 1 Hz. Typical
values for this parameter will be 0, ±1, ±2 etc.
Fractional values will not work.
Press the “Apply” button to perform the filter

back-projection calculation and “Undo” to revert Figure 5.50. 3-tube phantom, YZ image (32 projections
zero filled to 64 by 64)
to the original k-space data set should you wish
to try a different set of processing parameters.
5.5.4 A description of each FBP macro parameter
Image parameters
Image orientation ............ Two letters which signify the horizontal and vertical directions in the final
image.
Phase cycle
None ……………...……. Run each scan with the same 180 pulse phase.
2 step ………………...… Alternate the phase of the 180 pulse relative to the 90 between scans to
minimise the effects of incorrect pulse tip angles.
Max angle
180 …………………….. Only scan ½ of k-space

360 …………………….. Scan k space completely.
Occasionally artefacts which appear with a maximum angle of 180° will be removed with the more
complete scan. This is because each scan of k-space is only partial.
Polarizing duration .......... The length of the polarizing pulse (should be 2-3 times T1)
B1 frequency ................... The resonant frequency as determined by the pulse and collect experiment
(this should be as precise as possible).
Echo time ........................ The time between the centre of the 180° B1 pulse and the centre of the echo.
Number of projections … The number of scans of k-space made. These will be evenly spaced between
0 and 360 degrees (or 180° if the max angle is 180°).
Bandwidth ……............... Range of frequencies covered by the read dimension.
Output location
Buttons

Following are the limiting and recommended values for the parameters used in the filtered
backprojection imaging macro window. Limits are the range of values which the macro will accept
while recommended are those values which will produce a reasonable image in a reasonable length of
time.
Table 5-3. FBP imaging parameter range

Echo time ms 10-10000 50-500
No. of projections - 1-1000 Matrix size
Bandwidth Hz 1-200 4-100
Appendix 1 A-1
A.1 Signal-to-noise calculation macro
# Fundamental constants
mu0 = 4*pi*1e-7
#gamma = 2.675e+08 predefined constant
hbar = 6.626e-34/(2*pi)
kb = 1.38e-23
Ts = 273+15 # Sample temperature
Nh = 6.7e28 # Number of protons in water per m^3
rho = 1.7e-8 # Resistivity of copper
# B1 coil parameters
d1 = 0.084 # Diameter
l1 = 0.130 # Length
f0 = 2240
w0 = 2*pi*f0 # Resonant frequency
DF = 400 # Noise bandwidth
wd1 = 0.25e-3 # Wire diameter (no enamel)
N1 = 3700
R1 = 0.351*pi*d1*N1 # Coil resistance
L1 = mu0*N1^2*pi*d1^2/4/l1 # Coil inductance (long solenoid)
Q1 = w0*L1/R1 # Quality factor (theory)
Q1 = 20 # Quality factor (measured)
# Bp coil parameters
dp = 0.171 # Diameter
lp = 0.17 # Length
Np = 600 # Number of turns
Ip = 6 # Current
Rp = 0.00858*pi*dp*Np # Coil resistance
wdp = 1.60e-3 # Wire diameter
Lp = mu0*Np^2*pi*dp^2/4/lp # Coil inductance (long solenoid)
Qp = w0*Lp/Rp # Quality factor (theory)
# Sample parameters
Vs = 5.67e-4 # Volume
# Calculations
B1 = mu0*N1/(sqrt(d1^2+l1^2))
Bp = mu0*Np/(sqrt(dp^2+lp^2))
pr "\n B1 = $B1$ mT/A\n"

pr " Bp = $Bp$ mT/A\n"
M0 = (gamma)^2/(4*kb*Ts)*hbar*Nh*hbar*Bp*Ip
pr " M0 = $M0$ A/m\n"
S0 = Vs*B1*M0*w0*Q1
N0 = sqrt(4*kb*Ts*DF*R1)*Q1
pr " S0 = $S0$ V\n"

pr " N0 = $N0$ V\n"
pr " S/N = $S0/N0$\n"
pr " L1 = $L1$ H\n"

pr " Lp = $Lp$ H\n"
pr " R1 = $R1$ Ohms\n"

pr " Rp = $Rp$ Ohms\n"
pr " Q1 = $Q1$\n"
pr " Qp = $Qp$\n"
A-2 Appendix 1
Output:
B1 = 0.0300403 mT/A
Bp = 0.00312694 mT/A
M0 = 6.29317e-005 A/m
S0 = 0.000301727 V
N0 = 9.33673e-007 V
S/N = 323.162
L1 = 0.733362 H
Lp = 0.0611147 H
R1 = 342.719 Ohms
Rp = 2.76557 Ohms
Q1 = 20 (measured)
Qp = 311.02
Appendix 2 A-3
A.2 A macro for calculating the signal producing efficiency of

the B1 and Bp coils.
# Analytical solution for Bp coil *************************

R = 0.085; # Coil radius in metres
maxz = 0.085; # maximum z value for calculation
minz = -0.085; # minimum z value for calculation
stepz = 0.001; # step size along the z axis
len = 0.170; # length of coil in metres
N = 600; # Number of turns
I = 5.67; # Current
Bp = [0:1:(maxz-minz)/stepz];
cnt = 0
for (z = minz to maxz step stepz) # Step along z axis
Bp[cnt] = -((z-len/2)/sqrt(R^2+(z-len/2)^2));
Bp[cnt] = Bp[cnt]+((z+len/2)/sqrt(R^2+(z+len/2)^2));
cnt = cnt + 1;
next(z)
x = [minz:stepz:maxz];
Bp = I*4*Bp*pi*1e-7*N/(len*2);
# Calculate effciency of Bp coil to generate magnetization

effBp = Bp/max(Bp)
drawplot("false")
multiplot("1d",2,2)
curplot("1d",1,1)
plot(x,effBp)
xlabel("text","position (m)","size",10)
ylabel("text","efficiency","size",10)
title("text","Magnetization efficiency of B_(p)","size",10)
axes("fontsize",8)
# Analytical solution field along the axis of the B1 coil

R = 0.041 # Coil radius in metres
maxz = 0.085 # maximum z value for calculation
minz = -0.085 # minimum z value for calculation
stepz = 0.001 # step size along the z axis
len = 0.130 # length of coil in metres
N = 4160 # Number of turns
I = 1 # Current
B1 = [0:1:(maxz-minz)/stepz];
cnt = 0
for (z = minz to maxz step stepz) # Step along z axis
B1[cnt] = -((z-len/2)/sqrt(R^2+(z-len/2)^2))
B1[cnt] = B1[cnt]+((z+len/2)/sqrt(R^2+(z+len/2)^2))
cnt = cnt + 1
next(z)
# Calculate B1 as function of position

x = [minz:stepz:maxz]
B1 = I*4*B1*pi*1e-7*N/(len*2)
# Calculate tip angle as a function of position

effTip = sin(pi/2*B1/max(B1))
# Calculate efficiency of B1 coil to detect magnetization
effB1 = B1/max(B1)
A-4 Appendix 2
curplot("1d",2,1)
plot(x,effTip)
title("text","Tip angle efficiency of B_(1)","size",10)
axes("fontsize",8)
curplot("1d",1,2)
plot(x,effB1)
title("text","Detection efficiency of B_(1)","size",10)
axes("fontsize",8)
# Display combined efficiency of both coils

effTot = effB1.*effTip.*effBp
curplot("1d",2,2)
plot(x,effTot)
title("text","Relative signal along coil axis","size",10)
ylabel("text","S/S_(0)","size",10)
axes("fontsize",8)
drawplot("true")
# Print out fraction of signal which will be observed
e = submatrix(effTot,15,155) # Restrict range to sample length

eff = sum(e)/size(e)
pr("Fraction of signal observed = $eff$\n")
Output:
Fraction of signal observed = 0.7437

EFMRI User Guide V1.0 PDF

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

EFMRI User Guide V1.0 PDF

Caricato da

Copyright:

Formati disponibili

i

COPYRIGHT © 2005 magritek Limited

Caution and Safety Considerations

Identify System Components

Terranova-MRI EFNMR User’s Manual

Table of Contents ................................................................................................................................... v

4.5 The Carr-Purcell-Meiboom-Gill (CPMG) Experiment............................................... 4-10

1.2 EFNMR Apparatus Description and Operation

Table 1-1. Typical coil parameters

Coil Parameter Polarizing Coil Gradient B1 Coil

Figure 1.2. A view of the three EFNMR/MRI coils

1.2.2 The PGSE Gradient Coil

The PGSE gradient coil is designed to provide a linear

1.2.3 The shim and imaging gradient coils

Covering the PGSE gradient coil are a set of 3 gradient

1.2.4 Excitation and Detection

Polarizing pulse 90o B1 pulse Applied pulses

1.3 The Spectrometer

Figure 1.6. The entire EFNMR system: Spectrometer and Probe

Figure 1.7. Front Panel of the EFNMR Spectrometer

2.2 Software Installation

2.2.2 To install the Terranova-MRI software

2.2.3 To install the USB device driver

2.3 Running Prospa

• The “GUI” menu allows the design of new macro interfaces.

2.4 Preliminary experiments

2.4.1 Coarsely tuning the probe

Figure 2.11. The AnalyseCoil macro window

Figure 2.12. The select folder window

Figure 2.14. Sample output from the AnalyseCoil macro.

2.4.2 Select an initial location for the EFNMR apparatus

The observed signal at a time t is given by:

S (t ) = S 0 exp(t T2* )cos(ω 0 t + θ ) , [2-2]

Proximity to an NMR lab or other labs employing strong magnetic

Fortunately it is not always necessary to find a region of very high

The orientation, as well as the location, of the probe is important for

Figure 2.17. The monitor noise macro

RMS noise = 3.2 µV Noise magnitude spectrum

2.4.3 Select initial parameters

Figure 2.19. Pulse and Collect dialog.

Table 2-1. Suggested Initial Parameter Values

FID data (Acc: 1) Magnitude spectrum (Acc: 1)

0 0.2 0.4 0.6 0.8 1 2100 2200 2300 2400

2.4.4 Using a spin-echo to find a signal

1.0 Sample from here

Figure 2.22. The spin-echo experiment interface

FID data (Acc: 1) Magnitude spectrum (Acc: 1)

FID data (Acc: 1) Magnitude spectrum (Acc: 1)

0 0.2 0.4 0.6 0.8 1 2160 2180 2200 2220 2240

Figure 2.25. The Autoshim dialog

FID data (Acc: 1) Magnitude spectrum (Acc: 1)

FID data (Acc: 1) Magnitude spectrum (Acc: 1) x = 2.86 y = -13.6 z = 12.89

0 0.5 1 1.5 2 2160 2180 2200 2220 2240 0 20 40 60 80

2.5 Optimising the FID

2.5.1 Fine tuning the Probe

2000 2100 2200

2000 2100 2200 2300 2400

2.5.2 Setting the Resonance Frequency

2.5.3 Setting the B1 pulse duration

Figure 2.32. The B1 Duration dialog

FID amplitude vs B1 length (Acc: 1)

0.5 1 1.5 2 2.5 3

Figure 2.33. The output of the B1 Duration experiment.