Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
The idea of a thermistor, or thermally sensitive resistor, has been around for over 150
years. Although one of his lesser-known discoveries, the first documented use of an NTC,
section), thermistor came from Michael Faraday in 1833. 1 After the initial discovery, it was
quickly realized that thermistors could be separated entirely into two different categories: NTC
and PTC thermistors. Interestingly, the classification didn’t solely depend on metallurgical
properties due to the fact that at certain temperatures some types can actually switch
categories! Silicon is one such example that exhibits NTC properties until 250K, where a positive
temperature coefficient sets in. 2 All thermistors are made using semi-conducting metallic
compound oxides such as manganese, copper, cobalt, and nickel, as well as single-crystal
semiconductors silicon and germanium.3 Many different types of thermistors exist for different
uses. The coated lens type, while not utilized in this experiment, is one example as seen in Fig. 1
on page 2. While Faraday was first to discover the thermistor properties of semiconductors,
Samuel Ruben was quickest to perfect it and seal it under a U.S. patent. Almost a century after
Faraday’s breakthrough, Ruben released his “Electrical Pyrometer Resistance” findings in which
cuprous oxide. After being cleansed in hydrochloric and nitric acid, a thin film of this oxide
remained that gave his thermistor a negative temperature coefficient without the drawbacks of
adding heat to the device, its resistance dropped noticeably and reproducibly. As a final notable
mention, Rueben explains similar phenomena occurred when mixing the cuprous oxide with
P a g e |1
cuprous sulphide, or melting antimony sulphide with
to name a few.
Reginald A. Fessenden write in 1893 in the Physical Review about the linear relationship
between increased temperature and resistance in a sample of copper. In their testing between
the ranges of -69⁰C and 123⁰C, the same range we have worked in, they explain how copper’s
temperature coefficient is a positive 4.18% per degree Celsius.5 Thus, they seem to have proven
very early on that, unlike semi-conductors, well-conducting metals do not exhibit strong, or
any, fluctuations in R vs. T linearity. We will prove in later sections that this century-old value
Although not explored in this lab, it was only a few decades later when physicists made
perhaps the most astonishing discovery relating to electrical conductivity. On April 8, 1911,
Kamerlingh Onnes and his cohorts experimented with vapor pressures of liquid helium to drop
denote this phenomenon as superconductivity and it involves similar quantum effects explored
in the theory section of this report on thermistors. Thus, historical experimenting has proven to
P a g e |2
us that the relationship between resistance and temperature can take wildly different turns
Theoretical Basis
In order to better understand the results presented here, the underlying physical
workings of the materials must be grasped. The idea of a negative or positive temperature
coefficient is simple. It is the deciding factor as to whether resistance increases with increasing
coefficient thermistor and attempt to grasp the type of curve drawn out over a 10⁰C-90⁰C
𝑅(𝑇) = 𝑘 ∗ 𝑇 (eqn. 1)
resistance”. This k value can, at face value, be either negative or positive depending on whether
we are using an NTC or PTC. A significant problem with this simple equation is the fact that
most, if not all, thermistors do not wield a simple linear, quadratic, power, or other pre-
determined relationship between its resistance and temperature imposed. Thus, naturally,
equation 1 above breaks down when large temperature ranges are used. To resolve this, two
geophysicists John S. Steinhart and Stanly R. Hart developed a 3rd power logarithmic equation
with four coefficients to solve for the inverse of temperature in all semiconductors. The
equation is as follows: 7
1
= 𝑎 + 𝑏 ∗ ln(𝑅) + 𝑐 ∗ 𝑙𝑛2 (𝑅) + 𝑑 ∗ 𝑙𝑛3 (eqn. 2)
𝑇
P a g e |3
The lowercase letters are individual Steinhart-Hart temperature coefficients. While in most
professional cases, a, b, c, and d can be solved using four data points to solve four simultaneous
equations, we can exploit the fact that our thermistor has a negative temperature coefficient
and simplify the coefficients. Both c and d will practically fade into insignificance, and we can
1 1 1
set 𝑎 = − 𝛽 ln(𝑅𝑜 ) and 𝑏 = to arrive at the 𝛽 parameter equation:
𝑇𝑜 𝛽
1 1 1 1
= − 𝛽 ln(𝑅𝑜 ) + 𝛽 ln(𝑅)
𝑇 𝑇𝑜
1 1 1 𝑅
= + ln ( ) (eqn. 3)
𝑇 𝑇𝑜 𝛽 𝑅 𝑜
where β is the new temperature coefficient of resistance. Before moving on, the final equation
1 1 1 𝑅 𝑅
= + 𝛽 ln (𝑅 ) → 𝑇𝑜 𝛽 = 𝑇𝛽 + 𝑇𝑜 𝑇 ∗ ln (𝑅 ) →
𝑇 𝑇𝑜 𝑜 𝑜
𝑅 1 1 𝑅
𝛽 ∗ (𝑇𝑜 − 𝑇) = 𝑇𝑜 𝑇 ∗ ln ( ) → 𝛽 ∗ ( − ) = ln ( )
𝑅𝑜 𝑇 𝑇𝑜 𝑅𝑜
Place both sides as the exponent of e and multiply by R0 to get resistance as a function of
temperature:
1 1
𝛽( − )
𝑅(𝑇) = 𝑅𝑜 𝑒 𝑇 𝑇𝑜 (eqn. 4)
P a g e |4
Equation 2 can be deceiving. As I previously mentioned, we have taken the Steinhart-Hart
equation and modified it to work succinctly for negative temperature coefficient thermistors.
Thus, in our case, beta will actually come out to be positive. Now, this experiment enabled us
to compare a thermistor to an actual conducting wire made from copper. The electro-chemical
properties of conducting metals such as copper have been fairly well understood for many
years now. Excluding the strange nature of semi-conductors, most metals exhibit a linear
increase in resistance with increasing temperature. To fully appreciate this fact, we have to
understand copper at the atomic level. Electric current is simply the passage of electrons
through a wire over time somewhat analogously to the flow rate of water passing through a
pipe. When we introduce the concept of resistance, you can picture driving thick nails into the
pipe to impede the flow of water through it. At the atomic level, these “nails” are actually
nuclei vibrating around their mean position. The electrons passing through the wire are either
smashed into these nuclei, impeding their progress, or deliberately thrown off course due to
the strong electric forces present between the positively charged protons and the electrons’
own negative charge. It is fairly simple to visualize and conceive the notion that if these nuclei
were to vibrate back and forth quicker and quicker, and with larger ranges, the electrons’
chances of hitting nuclei increases dramatically. This is precisely the phenomenon that occurs
when the temperature of the wire is increased. Thus, we can appreciate why conducting
metals increase resistance with temperature, and why NTC semiconductors can be labeled as
“strange”. The math behind regular conducting metals is a bit simpler relative to thermistors.
We expect the temperature coefficient of these metals to be as close to zero as possible, thus
we define it as such:
P a g e |5
1 𝑑𝑅
𝛼= ∗ (𝑑𝑇 ) (eqn. 5)
𝑅𝑜
We can easily form a function 𝑅(𝑇) by separating variables and integrating demonstrated in the
following steps:
1 𝑇 1 𝑅 1
𝛼 ∗ 𝑑𝑇 = ∗ 𝑑𝑅 → 𝛼 ∫0 𝑑𝑇 = ∫ 𝑑𝑅 → 𝛼𝑇 = (𝑅 − 𝑅0 )
𝑅𝑜 𝑅𝑜 𝑅0 𝑅0
𝑅(𝑇) = 𝑅0 𝛼𝑇 + 𝑅0 →
The physics behind the strange phenomena exhibited in this experiment on thermistors is
somewhat convoluted. Semiconductors exhibit conduction band energy states at the very high
energy end. Most electrons in semi-conductors are stuck in the low energy band bound by the
valence band. Hence, very few charge carriers are available at low temperatures. However,
when temperature is increased and heat energy is bestowed to the metal, many of these lower
energy electrons begin making the jump to the conducting band and free holes in the valence
band. This occurrence overrides the fact that atomic vibrations are increasing simultaneously,
P a g e |6
Apparatus
This experiment was carried out utilizing a relatively simple set of tools and apparatuses.
The central piece of this experiment was a tripod-supported, insulated canister used to house
the copper coil and thermistor. This canister was electrically supplied by a two-prong wall plug
to operate its internal resistance source. This source involved a simple resistance heater with a
coefficient of performance = 1. It was hand operated and spring loaded to return to the off
state in order to avoid a boiling water or fire hazard. Next, the copper coil and thermistor units
were both coupled to a lid engineered to seal off the canister and minimize heat loss. These
lids came equipped with a sealed access point for a thermometer, and a stirring apparatus to
enable uniform heat dissipation throughout the canister. Lastly, the lids were electrically
coupled to the thermistor and copper wounds and were subsequently equipped with positive
and negative terminals on the top. We utilized these terminals through alligator clips and color-
coated wires (Red: positive, Black: negative) that were fed into a Keithley 197 Amp Multimeter.
This multimeter allowed us to simultaneously feed current into the copper or thermistor coils
and determine the resistance through one device. This multimeter was stated to have an
accuracy of 1mΩ.9 Lastly, we had five medium to large beakers at our disposal in order to
achieve water just above the freezing point to cool down the canister. As an effort to cut down
P a g e |7
Procedure
We began this experiment by acquiring roughly 1/2 pound of ice (1 large beaker) and
mixing it with tap water. The instruments necessary for the procedure were conveniently laid
out beforehand, and thus we were able to begin immediately. Before connecting the canister
to the wall outlet, we brought the metal inside down to roughly 5⁰C by flushing it with ice-cold
water from the large beaker. Another pre-experiment task was to determine the lead
resistance on the wires we had for use. This was determined by inserting them into the
multimeter, ensuring the multimeter was on and reading resistance on a 2Ω scale, and rubbing
the leads together to remove a bit of corrosion. While the actual lead resistance was rather
“jumpy” and hard to measure, we averaged the value over a roughly ten-second interval. Once
satisfied, we left the water inside the canister, plugged in the resistance heater, and opted to
begin with the copper coil apparatus. We placed it snugly onto the canister, inserted the
thermometer, connected the alligator clips to the terminals, and observed the multimeter to
ensure a proper connection had been made. The temperature was brought up to 10⁰C for the
first data point, and the data collection was underway. We divided the remaining tasks as such:
Joseph Oxenham operated the resistance heater and ensured the temperature was brought up
slowly to the next data point. I maintained the homogeneity of the heat in the liquid via the
stirring apparatus and inserted the data points into Microsoft Excel. As previously mentioned,
the tables, formulas, and graphs were created prior to the experiment. Thus, we were able to
observe the linear relationship between Resistance and Temperature on the fly during the
copper experiment to ensure all equipment and procedures were optimal. We attempted, and
mostly succeeded, in taking data points for every 10⁰C increase in temperature. Therefore, a
P a g e |8
total of nine data points were taken ending with a 90⁰C data point. Once this portion of the lab
had concluded, the copper lid was removed and ice-cold water was again flushed in and out of
the canister until the water maintained a temperature of 5⁰C. The thermistor device was
housed by the same type of lid as the copper device, ergo the same procedures applied. Again,
we acquired data points for every 10⁰C increase in temperature beginning with 10⁰C and
culminating with 90⁰C. We were able to observe the relationship between (1/T-1/Td) and
ln(R/Rd) in real-time to ensure the experiment was correct. As the lab concluded, we dumped
out the water, removed all the components, turned off/unplugged electronics, and cleaned
Results
From our experimentation, we gathered data in the form Ohms (Ω), Temperature (⁰C
and ⁰K), and unit-less values of (1/T-1/Td) and ln(R/Rd) for the thermistor.
As far as resistance is concerned, we noted both the raw measured resistance and the actual
resistance with the lead resistance taken into account. All of these values can be viewed in the
appendix section of this report. We found the lead resistance to be 0.258Ω, thus the actual
value of resistance was determined via 𝑅𝑚𝑒𝑎𝑠𝑢𝑟𝑒𝑑 − 𝑅𝑙𝑒𝑎𝑑 = 𝑅𝑎𝑐𝑡𝑢𝑎𝑙 . With the Ractual and T
values in hand, we were able to make a graph demonstrating the linear relationship between
P a g e |9
the two in copper wire demonstrated below.
4
Resistance (Ω)
y = 0.0161x + 3.3978
R² = 0.996
3
0
0 10 20 30 40 50 60 70 80 90 100
Temperature (⁰C)
Now, as we derived earlier, eqn. 6 represents the change of resistance vs. temperature for
copper.
(𝑅 − 𝑅0 )
1 𝑅 ⁄(𝑇 − 0) 𝑆𝑙𝑜𝑝𝑒
𝑅(𝑇) = 𝑅0 (𝑎𝑇 + 1) → 𝛼 = ∗ ( − 1) = =
𝑇 𝑅𝑜 𝑅0 𝐼𝑛𝑡𝑒𝑟𝑐𝑒𝑝𝑡
𝑆𝐿𝑂𝑃𝐸(𝐶6:𝐶14,𝐴6:𝐴14) 1
.10 This automatically gives us the value 𝛼 = 0.00473112 . This is
𝐼𝑁𝑇𝐸𝑅𝐶𝐸𝑃𝑇(𝐶6:𝐶14,𝐴6:𝐴14) ℃
1 11
acceptable, as it closely matches the value given in the lab handout: 𝛼 = 0.00433 .
℃
However, we wish to verify that excel is indeed performing the calculation correctly. We take
P a g e | 10
the first two data points (10℃ and 20℃ respectively) and use them simultaneously to solve for
𝛺
𝑆𝑢𝑏𝑡𝑟𝑎𝑐𝑡 𝐸𝑞𝑢𝑎𝑡𝑖𝑜𝑛𝑠 → 0.113𝛺 = 𝑚 ∗ (20℃ − 10℃) → 𝑚 = 0.0113
℃
𝛺
𝑃𝑙𝑢𝑔 𝑖𝑛 𝑚 → 3.577𝛺 = 0.0113 ∗ 10 ℃ + 𝑏 →
℃
𝛺
𝑏 = 3.577𝛺 − (0.0113 ∗ 10 ℃) = 3.465𝛺
℃
𝛺
𝑆𝑙𝑜𝑝𝑒 0.0113 1
℃
𝛼= = = 0.00329 ℃.
𝐼𝑛𝑡𝑒𝑟𝑐𝑒𝑝𝑡 3.465𝛺
This value of α is fairly close, so we are comfortable with our calculation. As shown in the graph
above, our coefficient of determination is practically 100%, so we can again be sure our
Next we deal with our calculations pertaining to the thermistor apparatus. We were
informed from our handout that, in this case, R vs. T would not be linear. Instead, what we got
P a g e | 11
Resistance Vs. Temperature of a Thermistor
60
50
40
Resistance (Ω)
30
20
10
0
270 280 290 300 310 320 330 340 350 360 370
Temperature (⁰K)
The important concept to take away from this graph is the notion that resistance did indeed
decline with increased temperature. Temperature was displayed in Kelvins for the reason that
an absolute scale was necessary for generating the logarithmic resistance ratio graph to solve
for β. To achieve this, all Celsius data points were converted using this formula:
𝑇 0 𝐾 = 𝑇℃ + 273.15
culminating with eqn. 4. Our goal is to solve for β, (the negative temperature coefficient), and
𝑅
𝑅 1 1 ln( )
𝑅0
ln ( ) = 𝛽 ( − ) → 𝛽= 1 1
𝑅 0 𝑇 𝑇 𝑜 (𝑇−𝑇 )
𝑜
P a g e | 12
𝑅 1 1
Therefore, we need a graph demonstrating the relationship between ln (𝑅 ) and (𝑇 − ) to
0 𝑇𝑜
0
-0.0008 -0.0007 -0.0006 -0.0005 -0.0004 -0.0003 -0.0002 -0.0001 0
-0.5
y = 3422x + 0.044 -1
ln(R/Rd)
R² = 0.999
-1.5
-2
-2.5
-3
(1/T-1/Td)
Armed with the knowledge that β is a constant, we expect this graph to be perfectly linear (or
close to it). Excel tells us we have a coefficient of determination of practically 100%. Thus, our
data can be viewed as highly precise. In this case, our slope is our β, and it is rather close to the
lab handout’s expected range of 3530⁰𝐾 ± 80⁰𝐾. It was acquired using the formula 𝛽 =
𝑆𝐿𝑂𝑃𝐸(𝑀6: 𝑀14, 𝐿6: 𝐿14).10 We perform a sample calculation using the first and ninth data
4.132𝛺
ln(52.242𝛺)
𝛽= ⁄ 1 1 = 3,368.27⁰K
(363.150 𝐾 − )
285.150 𝐾
P a g e | 13
This value is very close to our experimental value, so we can ensure the calculation is correct.
Discussion
Overall we are very satisfied with the results of our experimentation. Reviewing our
coefficient of determination value (0.996) acquired for the copper coil (0.996), we had a very
precise set of data points. Knowing the accepted value for copper from the lab handout, we
can calculate the percent approximation error using the following formula:
1 1
|𝛼𝑎𝑐𝑐𝑒𝑝𝑡𝑒𝑑 − 𝛼𝑒𝑥𝑝𝑖𝑟𝑖𝑚𝑒𝑛𝑡𝑎𝑙 | | 0.00433 ℃ − 0.00473112 ℃ |
𝛿= ∗ 100% = ∗ 100% = 9.26%
𝛼𝑎𝑐𝑐𝑒𝑝𝑡𝑒𝑑 1
0.00433
℃
While small, an almost ten-percent error has to be accounted for. This is most likely due to the
inexperience in working the resistance heater. While it was stressed in class that a gradual
increase in heat over the few degrees preceding the data point would garner the best results,
we may have acted overly ambitious on this first run. Consequentially, we were not allowing
for uniform heat dispersion throughout the entire canister. This, in turn, caused resistances to
vary throughout the copper coil and subsequently cause our readings to vary. Another problem
we ran into on this first run is variance in the multimeter itself. When the time came to take
note of a resistance, the multimeter would frequently jump over a range of roughly 0.3Ω. Thus,
we were forced to quickly average this value in order to gather a data point. In hindsight, this
too may have been caused by the hastily controlled resistance source and non-uniformly
resisting copper wire. Returning to the work of Kennelly and Fessenden, we did, however,
P a g e | 14
demonstrate similar values to their experiments dating over a century ago. Their value being
1
0.00418℃ , we experienced a 13.2% approximation error relative to this α.
We had a bit more luck with in determining the temperature coefficient of the thermistor
apparatus. The lab handout states that the manufacturer of this thermistor already accepts an
absolute error of ±80 ⁰𝐾. Thus we can begin by computing the percent error of the
manufacturer’s value, and compare that to our percent approximation error of the
experimental value.
𝑒𝑟𝑟𝑜𝑟 800 𝐾
% 𝑅𝑒𝑙𝑎𝑡𝑖𝑣𝑒 𝐸𝑟𝑟𝑜𝑟 (𝑀𝑎𝑛𝑢𝑓𝑎𝑐𝑡𝑢𝑟𝑒𝑟) = ∗ 100% = ∗ 100% = 2.27%
𝑚𝑒𝑎𝑠𝑢𝑟𝑒𝑚𝑒𝑛𝑡 35300 𝐾
|35300 𝐾 − 34220 𝐾|
𝛿= ∗ 100% = 3.06% 𝑒𝑟𝑟𝑜𝑟
3530⁰𝐾
Thus, our experimental error is less than one percent greater than manufacturing error, so we
can be very comfortable with this value acquired. For the error that we did happen to obtain,
this, again, is likely due to ambitious resistance heating and other minute issues such as lead
corrosion, terminal corrosion, etc. We were, however, intrigued by the fact that our β value
came out to be only two ⁰K off from the attempt from the writers of the lab handout. While
they were forced to utilize a slide-wire Wheatstone bridge and we were blessed with a digital
multimeter, the 0.053% error from their attempt gave us something to further ground our own
attempt.
While most of the theories behind the behavior of conducting and semi-conducting metals have
been well understood for, at the very least, decades, this experiment does its job at solidifying
certain concepts and implications of these technologies. First, copper’s positive temperature
P a g e | 15
coefficient demonstrates why it makes such a useful tool for widespread use among our
layman’s terms, copper and other conductors will “sense” that extra or unruly current is passing
through via heat generation and will automatically increase the resistance to this current. While
this is not applicable to, say, replacing a fuse, it does function optimally for the uses for which it
is designed (e.g. local telephone or power cables). On the other hand, NTC thermistors such as
the one experimented on in this lab are vulnerable to such a runaway. However, these units are
not exploited for their load carrying capacities so much as they are for their sensitivity. In this
lab we observed a change in almost fifty Ohms in our thermistor versus roughly 1.2 Ohms in the
copper coil. Our conclusion on this device is that it would be very useful in application settings
and computer chip, this devices change in resistance could immediately be relayed to the
computer chip and the temperature could be changed according to the algorithm set in place
by the software engineer. To no surprise, the thermistor is very heavily used in all areas of
temperature control such as thermostats, ovens, refrigerators, A/C units, fire alarms, fever
thermometers, coffee makers, and more.5 In summary, both conducting wires and semi-
conducting thermistors have very practical uses for their R vs. T properties that will likely leave
P a g e | 16
APPENDIX – RAW DATA
Copper Coil
α = 0.00473112 (1/⁰C) )
Lead Resistance = 0.258Ω
Thermistor Apparatus
β= 3421.988495 (⁰K)
Lead Resistance = 0.258Ω
P a g e | 17
APPENDIX – DISC FILES
Root Directory
o Lab1_Temperature Coefficient_9-8-10.xls
experiment, all three graphs presented in this report associated with the
P a g e | 18
ENDNOTES
1
Cornerstone Sensors. (n.d.). A Brief History of NTC Thermistors. Retrieved September 18, 2010, from
Cornerstone Sensors:
http://www.cornerstonesensors.com/?LinkIn=http://www.cornerstonesensors.com/About.asp
?PageCode=Brief
2
Radio-Electronics. (n.d.). Thermistor. Retrieved September 18, 2010, from Radio-Electronics.com:
http://www.radio-electronics.com/info/data/resistor/thermistor/thermistor.php
3
Arnott, P. (n.d.). Thermistors: Thermal Resistors. Retrieved September 19, 2010, from PatArnott.com:
patarnott.com/atms360/pptATMS360/CircuitLabThermistors.ppt
4
Ruben, S. (1930, March 19). Electrical Pyrometer Resistance. Retrieved September 19, 2010, from
FreePatentsOnline: http://www.freepatentsonline.com/2021491.pdf
5
A. E. Kennelly and Reginald A. Fessenden. (1893). Some Measurements of the Temperature Variation
in the Electrical Resistance of a Sample of Copper. The Physical Review, 260-273.
6
Kes, D.v. (2010). The Discovery of Superconductivity. Physics Today, 38-43.
7
John S. Steinhart, Stanley R. Hart, Calibration curves for thermistors, Deep Sea Research and
Oceanographic Abstracts, Volume 15, Issue 4, August 1968, Pages 497-503, ISSN 0011-7471, DOI:
10.1016/0011-7471(68)90057-0.
8
Nave, R. (n.d.). Band Theory for Solids. Retrieved September 18, 2010, from HyperPhysics:
http://hyperphysics.phy-astr.gsu.edu/hbase/solids/band.html#c5
9
ValueTronics. (n.d.). Keithley 197A Autoranging Digital Multimeter. Retrieved September 19, 2010,
from ValueTronics International Inc.: http://www.valuetronics.com/Used_Keithley_197A.aspx
10
See Lab1_Temperature Coeffieicient_9-8-10.xls on accompanying CD-ROM
11
Instructions for the Use of NO. 2836. Thermistor Temperature-Coefficient Apparatus. (n.d.).
P a g e | 19