Nov/Dec 2000 45

Thermistors are inexpensive, readily available components.

Better yet, they can greatly enhance the performance
of your projects. Learn how to put them to work for you.
By William E. Sabin, W0IYH
1400 Harold Dr SE
Cedar Rapids, IA 52403
[email protected]
Thermistors in
Homebrew Projects
T
hermistors are interesting
compnents that Amateurs can
use to enhance their projects.
Variations in circuit temperature that
affect gain, distortion and control func-
tions such as receiver AGC or
transmitter ALC can be compensated.
Dangers of self-destruction of over-
heated power transistors can be greatly
reduced. Oscillator drift can be greatly
reduced. This article discusses thermis-
tor properties and shows three exam-
ples of how they can improve project
performance. We will see first that
some easy mathematics improves the
understanding.
Mathematics of Thermistors
A thermistor is a small bit of intrin-
sic (undoped) semiconductor material
between two wire leads. As temper-
ature increases, the number of liber-
ated hole-electron pairs increases
exponentially, causing the resistance to
decrease exponentially. This expo-
nential nature is seen in the resistance
equation:
R T R T0 e
T0 T
( ) ( )

j
(
\
,

1 1
(Eq 1)
where T is some temperature in
Kelvins and T0 is a reference temper-
ature, usually 298 K (25C), at which
the manufacturer specifies R(T0). The
constant is experimentally deter-
mined by measuring resistance at
various temperatures and finding the
value of that best agrees with the
measurements. A simple way to get an
approximate value of (this is usually
all we need in ham-gear design) is to
make two measurements, at room
temperature, say T = 25 C (298 K) and
T0 100C (373 K) in boiling water.
Suppose the resistances are 10 k and
938 . Eq 1 is solved for :
= =
ln
ln
R T
R T0
T T0
( )
( )
,

,
]
]
]

j
(
\
,

1 1
938
10000
1
373
1
298
3507 (Eq 2)
Usually, the exact value of temper-
ature is not as important as the ability
to maintain that temperature. A better
estimate of , if needed, can by achieved
by a linear regression method using the
program THERMIST.BAS, download-
able from the ARRL QEX Web site.
1
This program takes the logarithm of
both sides of Eq 1, which provides a
linear relationship between log(R(T))
1
You can download this package from the
ARRL Web http://www.arrl.org/qexfiles/.
Look for THERMIST.ZIP.
46 Nov/Dec 2000
and (1/T0 1/T). Five equally spaced
data points are input to get the slope of
the line, which is .
The following examples illustrate a
few of the main ideas of thermistor-
circuit design that can be employed in
a number of similar situations.
MOSFET Power-Transistor
Protector
It is common practice to compensate
the temperature sensitivity of power
transistors. In bipolar (BJT) transis-
tors, thermal runaway occurs because
the dc current gain increases as the
transistor gets hotter. The runaway
condition is less likely to occur in
MOSFET transistors, but with excess-
ive drain dissipation or inadequate
cooling the junction temperature will
increase until its maximum allowable
value is exceeded. A common proce-
dure is to mount a diode or thermistor
on the heat sink close to the transis-
tors, so that the bias adjustment
tracks the flange temperature. Refer-
ences 1 and 2 give detailed discussions
of this.
A special problem occurs when a
sudden large increase in transistor
dissipation occurs. The flange temper-
ature changes rather slowly because of
the thermal capacitance (heat storage)
of the heat sink. However, the junction
temperature rises much more rapidly
and can rise above the maximum limit
before the correction circuit has a
chance to function. If a thermistor is
attached to the ceramic case with a
small drop of epoxy as shown in Fig 1 it
will respond much more rapidly (short-
er time constant) and may (not guar-
anteed) save the transistor. The circuit
(Reference 3) of Fig 2 detects a case
temperature of about 93C and com-
pletely turns off the FETs until the case
temperature drops about 0.3C, at
which time the FETs are turned on
again. A red LED on the front panel
warns of an over-temperature condition
that requires attention.
In Fig 2, a 4.7 V Zener (D1), R1 (metal
film) and Rth (thermistor) are a voltage
divider with an output of 0.6 V.
If this voltage decreases slightly
(because the resistance of Rth de-
creases slightly) Q1 starts to come out
of saturation, Q2 quickly goes into
saturation and the gate voltage of the
FET goes to a low value, turning it off.
At the same time, the 20-mA LED
(RS 276-307) lights. The voltage divi-
der equation is:
0 6 4 7 . .
+

j
(
\
,
R
R1 R
th
th
(Eq 3)
Fig 1The thermistor attached to power MOSFET with a drop of epoxy.
Fig 2Schematic of a
MOSFET
temperature-
protection circuit.
If we solve this for R1, we get:
R1 R 6 831 . th
be placed in a cool location, not
directly on the heat sink.
The RadioShack precision thermistor
(RS 271-110) is rated at 10 k 1% at
25 C. It comes with a calibration chart
from 50C to +110C that can be used
to get an approximate resistance at
some temperature. We are most inter-
ested in FET case temperatures in the
range of 70C to 100C. It is assumed
that very close temperature knowledge
is not needed, but in order to be sure
that the thermistor is okay, I measured
its resistance at 20C (68F) and in
boiling water (100C). The circuit of
(Eq 4)
If the value of Rth is known at some
temperature, then R1 is the value that
activates the circuit at that temper-
ature. An interesting feature is that
the voltage across the thermistor
never exceeds about 0.6 V, and this
greatly reduces the self-heating of the
thermistor, which could otherwise
cause a substantial error in the circuit
behavior. The temperature vari-
ations of Q1 and D1 are small sources
of error, so this simple circuit should
Nov/Dec 2000 47
Fig 3A VFO temperature-controller schematic. The eight 200- resistors labeled Heater serve to warm the VFO enclosure
Fig 4Schematic of a three-band VFO.
L122 t #24 AWG, wide spaced on a T68-7
core
L217 t #24 AWG, wide spaced on a T68-7
core
L322 t #24 AWG, wide spaced on a T68-7
core
L42 t #24 AWG link, close wound on L1
core
SW1Electroswitch D4C0312N
CtuneHammarlund RMC-50-S
Fig 2 is intended for 50C or greater
when the 10-k thermistor is used. At
lower temperatures, the circuit of Fig 3
would be better.
The following procedure was used to
get the desired temperature control:
1. The MRF150 MOSFET has a
maximum allowed junction tem-
perature of 200C. The thermal
resistance
JC
from junction to case
is 0.6C/W.
2. The maximum expected dissipation
of the FET in normal operation is
110 W.
3. I selected a case temperature of
93C. This makes the junction tem-
perature 93 + (0.6) (110) = 159C,
which is a safe 41C below max.
48 Nov/Dec 2000
4. The FET has a rating of 300 W
maximum dissipation at a case
temperature of 25C, derated at
1.71 W/C. At 93C case temper-
ature, the maximum allowed dissi-
pation is 300 1.71 (93 25) =
184 W. The safety margin at that
temperature is 184 110 = 74 W.
5. A very simple way to determine the
correct value of R1 is to put the
thermistor in 93C water (let it
stabilize) and adjust R1 so that the
circuit toggles. I found that 8450
was the nearest standard value for
a metal-film resistor. At 93C, the
measured value of the thermistor
was about 1230 .
Measurements of the circuit
sensitivity determine the temper-
ature values at which the circuit
toggles on and off. I replaced the
thermistor with a resistor decade box
and measured resistances of 1224
and 1234 . Solve Eq 1 for the T that
corresponds to each value of R:
T
R
R0 T0

j
(
\
,

ln +
(Eq 5)
which is easy to perform with a hand-
held calculator or math program such
as MathCAD. Using the two values of
R, I found a temperature range of about
0.3C.
Fig 5The three-band VFO with temperature control. Five of the eight 5-W resistors are
mounted on a circuit board to the right of and slightly below center. The other three 5-W
resistors are wired together slightly to the left and above center. The thermistor is not
visible in this photo.
Fig 6Temperature
compensation of IF-
amplifier and AGC
circuitry: (A) IF amplifier,
(B) AGC threshold, (C)
AGC gain. Each
compensation circuit uses
two resistors and a
thermistor (RthA, RthB or
RthC).
lined with
1
/4-inch plexiglass and the
inside surfaces are lined with
1
/4-inch
Styrofoam sheets. The tuning and
bandswitch shafts are thermally
insulated from the outside world,
using plastic shaft couplers. Plexi-
glass blocks attach the box to the front
panel. Electrical grounding is via an
RF choke (for dc) and several 0.01 F
capacitors (for RF).
The temperature at the thermistor
location is maintained within 0.1C,
as determined by thermistor resist-
ance measurements. Using the meth-
od of the previous example to get the
temperature range and knowing the
frequency drift versus temperature
coefficient of the VFO in parts per
million (PPM), the frequency change
can be found as follows:
A Temperature Controlled VFO
The circuit of Fig 3 is used to control
the temperature of the three-band VFO
shown in Fig 4, inside a thermally
insulated enclosure. The Wheatstone
bridge circuit with an LM339 compar-
ator as a null detector is more sensitive
and less temperature dependent than
Fig 2. The 339 works quite well at a
level of 0.5 V at each of its two inputs.
This circuit is preferable at lower
temperatures, such as 30 to 35C, where
the thermistor resistance is in the 8 k to
7 k range.
I use eight 200 , 5 W metal-oxide
resistors at the output of the LM317.
The total heat applied is 4 W to main-
tain 33C, and the resistors are placed
so that their heat is distributed
uniformly: Half are placed near the
bottom and half near the top. The
thermistor is mounted in the center of
the box (see Fig 5), close to the tuned
circuit and in physical contact with the
oscillator ground-plane surface, using
a small drop of epoxy.
The enclosure is homemade from
sheet aluminum and angle stock. It is
large enough that it has almost no
effect on frequency. The outside is
Nov/Dec 2000 49
F F PPM T
VFO
Plugging in some numbers encoun-
tered for a 4.0 to 6.5 MHz VFO with
PPM = 500/C and Delta T = 0.1,
before temperature compensation,
F

6, 000, 000 Hz
500
1 000 000
0 1 300
, ,
.
(Eq 7)
which indicates that some improve-
ment is needed, in particular the
temperature coefficient. If the enclo-
sure is massive or well insulated, the
rate of temperature change can be
slowed down. When a small temper-
ature range of 0.1C was maintained,
the VFO temperature coefficient was
improved by a factor of about seven
with a negative-temperature-coeffi-
cient capacitor, C1. Inexpensive poly-
styrene capacitors, for example
Mallory type SX, have a well-con-
trolled negative (120 PPM/C) temp-
erature coefficient that is intended to
offset the positive temperature coeffi-
cient of inductors. (Caution: To pre-
vent damage, use pliers on the leads
as a heat sink when soldering.) No
other temperature compensation was
needed (probably fortuitously).
Temperature-compensating capac-
itors are available from Surplus Sales
of Nebraska (www.surplussales
.com). Other suitable fixed ceramic
capacitors are Vishay 561 series, type
10TC NP0, which I have found are
excellent. No variable trimmer caps
are needed because each VFO band
has a 25 kHz margin at each end and
a calibrated analog dial is not used.
The two-turn feedback winding in the
J310 drain satisfies all three coils. The
three inductors use Carbonyl-TH,
T68-7 cores (white paint), which are
claimed by Amidon to be the most
temperature-stable mixture at normal
room temperatures.
The oscillator finally turned out to be
very slightly overcompensated. Over
the 0.1C range, the frequency varies
20 Hz or less, with a period of about
five minutes. Superimposed is a very
slow drift of average frequency that is
due to settling of component values,
including possibly that of the
RadioShack thermistor. These gradual
changes became negligible after a few
days of continuous operation.
One problem that is virtually elim-
inated by maintaining a constant
temperature is the retrace effect on
cores and capacitors. Because of
retrace, components subject to a subs-
tantial temperature transient of some
kind may take several hours to recover
their previous L and C values. The
thermistor may also show a retrace
effect.
Reference 4 shows ways to perform
the temperature compensation opera-
tion and gives further references. An
especially good method is to toggle the
value of Rx (Fig 3) slightly so that a
variation of 0.5C is created inside the
VFO enclosure, and then do the temper-
ature compensation. Because of the
small average power dissipation in the
VFO plus controller, it is economical to
let the VFO run continuously so that
initial warm-up drift (measured less
than 1 kHz) and retrace are avoided.
If the VFO is mixed with crystal
frequencies and then bandpass fil-
tered (the mix-master approach) the
final local oscillator (LO) can be quite
stable and very clean spectrally. This
is especially so if the crystals
(20 PPM/C maximum) are temper-
ature compensated, temperature con-
trolled or even phase-locked to a
reference (see Reference 5). An LO
frequency counter (see Reference 6)
offset by the IF is a very simple and
excellent way to read the actual RF
signal frequency to within 50 Hz, if
the reference crystal is of high
quality and perodically adjusted to
WWV. The frequency stability is quite
adequate for HF SSB/CW, which are
the primary applications for this
equipment. Listening tests confirm
that in SSB speech, slow frequency
changes of 50 Hz are hardly noticed.
The reason for the three bands of the
VFO (4.0-4.5, 5.0-5.5, 6.0-6.5 MHz) is
to minimize spurious mixer products
due to harmonic intermodulation that
can slip through the LO bandpass
filters.
Fig 6 sketches three examples. Fig 6A is
the final IF amplifier stage that is part
of an IF amplifier module. The
thermistor compensates the gain
variation of the module. Figs 6B and 6C
are part of the receiver AGC circuit. The
thermistors compensate for variations
in AGC slope (dB/V) and vary the AGC
threshold.
In all cases, use a resistor decade box
to determine the resistance-temper-
ature correlation and plot an RT curve.
Then look for a resistor-thermistor cir-
cuit that seems reasonable. A
MathCAD or Excel worksheet is then an
elegant way to get the component val-
ues experimentally by comparing the
thermistor-circuit temperature curve
with the desired RT curve. Usually, a
close approximation is good enough,
and perfection is not justified. (Cata-
logs offer a wide assortment of thermis-
tors for these projects.) The method
suggested in Fig 6, one thermistor and
two resistors, is a simple combination
that gives a good approximation to
the desired RT curve. In many cases,
one of the resistors can be deleted.
Reference 7 shows other useful and
interesting thermistor applications.
(Eq 6)
Gain and AGC control
Thermistors are used to stabilize, or
to vary in a controlled manner, the gain
of an amplifier or an on-off threshold.
References
1. N. Dye and H. Granberg, Radio Fre-
quency Transistors, Chapter 4, (Boston:
Butterworth-Heinemann, 1993).
2. H. Granberg, Wideband RF Power Ampli-
fier, RF Design, Feb 1988. Also Motorola
RF Applications Report AR313 (1993),
p 424.
3. W. Sabin, W0IYH, A 100-W MOSFET HF
Amplifier, QEX Nov/Dec 1999. See also
Letters to the Editor, QEX, Mar/Apr
2000, pp 60-61.
4. ARRL Handbook (Newington: ARRL)
1995 to 2000 editions, Chapter 14. Order
No. 1832, $32. ARRL publications are
available from your local ARRL dealer or
directly from the ARRL. Check out the full
ARRL publications line at http://www
.arrl.org/catalog/.
5. See Chapter 17 of the ARRL Handbook
for a description of the Ten-Tec Omni VI
Plus transceiver.
6. Radio Adventures Co, RR4 Box 240, Sum-
mit Dr, Franklin, PA 16323; tel 814-437-
5355, fax 814-437-5432; information@
radioadv.com; http://www.radioadv.com.
Model BK-172. For a description, see the
1999/2000 ARRL Handbook, Chapter 26.
7. P. Horowitz and W. Hill, Art of Electronics,
Second Edition (New York: Cambridge
University Press, 1989).