Last week we talked in general about Thermocouples. This week I am going
to shift gears and format the discussion like an FAQ.
Also this week, I would like to acknowledge the help of Toby from
Labjack. bToby lent me some of his experience in
difficulties his customers have had with TC’s. The Labjack website
is worth a visit for anyone looking for inexpensive USB-based data
Q: Do I need to calibrate my thermocouple?
– The Merriam-Webster online dictionary (www.m-w.com) gives the
following two definitions for calibration (among others):
1: to standardize (as a measuring instrument) by determining the
deviation from a standard so as to ascertain the proper correction
2: to adjust precisely for a particular function
When I think of calibration, I think about taking a particular device
and adjusting it so that the output is what is expected when a standard
input is given. For instance, when exposed to a fresh-water-ice bath
at sea level, I would expect some electronic thermometer to measure 0 C
or 32 F. Similarly, I would expect that when exposed to a pot of
boiling fresh-water at sea level, the same device will read 100 C or
212 F. If the device is reading incorrectly, I expect to adjust it
(somehow) so that it can perform these tasks as expected.
When it comes to a Thermocouple, there is nothing to adjust. But that
does NOT mean that the thermocouple cannot be calibrated. (You win,
Jeff K.) Definition 3 above talks about measuring the deviation from a
standard. So, while we wouldn’t want to, we could quite happily use
a thermocouple that exhibited a LINEAR deviation from expected
behavior, as long as we understood and applied the previously
determined deviation when we reported a measured temperature.
From a practical standpoint, we don’t want to use thermocouples that
have a deviation from expected performance. Thermocouples or
thermocouple material usually comes with a certificate of compliance
that states that the material is certified to behave within the
performance laid down in some standard somewhere. In practice,
thermocouple material obtained from a reputable source that comes with
a certificate of compliance is usually good enough to justify putting
that material into service. A prudent precaution, however, might be to
subject that material to an ice-bath and a pot of boiling water just to
make sure that the thermocouple really does behave in an acceptable
manner. And, whenever thermocouples are left in service for an
extended period of time, they should be checked to make sure that their
performance has not degraded over time. Many factors can cause a
thermocouple to behave badly, and in general, they should be replaced
at regular intervals. Remember, TC material is cheap, a damaged test
article because of improperly measured temperature is not. If your TC
fails your hot/cold test, throw it out and make a new one. Check the
thermocouples in the described manner before every test until you
develop a feeling for the reliability of Thermocouples.
Q: How do I connect a thermocouple to my measurement system?
Bottom Line: You need a system that is set up to handle thermocouple
inputs. There are so many factors that are involved in connecting a
thermocouple to instrumentation that I could talk about it for 10
pages. At the very least, the temperature of the terminal block where
the thermocouple is electrically connected to the instrumentation MUST
be known. Using polynomials that predict the spurious voltages created
by the thermocouple-material / instrumentation terminal-block, these
junctions can be compensated for only when the temperature is known.
This usually involves the inclusion of some Integrated-circuit
temperature sensor into the terminal block of the instrumentation.
This is called Cold Junction Compensation or CJC for short.
In general, specialized signal conditioning equipment is needed in
order to accurately measure temperatures with thermocouples. If you do
not have such equipment, DON’T USE THERMOCOUPLES to make your
temperature measurements. You cannot hope to be even CLOSE to the
actual temperature if you do not use appropriate equipment.
www.omega.com is (again) an excellent resource for this type of
equipment, and National Instruments also makes dedicated signal
conditioning equipment for thermocouples.
Q: How do I convert the voltage to temperature?
As I mentioned last week, thermocouple voltage is related to
temperature by a high-order polynomial curve fit. So, there are two
1. Determine the thermocouple voltage.
(This involves Cold Junction Compensation as mentioned above). This
usually means that you determine the amount of voltage created by the
cold junctions and add that voltage from what is measured on the
For instance, if the thermocouple is a type T (copper/constantin), and
your instrumentation terminal block is made of copper, you would have
to calculate how much voltage is created by the SECOND
copper/constantin junction (formed where the constantin wire is
connected to the copper terminal block). In order to do this, you will
have to know the temperature of the terminal block. Once you
understand how much voltage is due to this ‘Cold’ Junction, add
that voltage from the voltage you are reading on the input where the
thermocouple is connected. This is your thermocouple voltage.
2. Plug the voltage into the high-order polynomial published by the
thermocouple material manufacturer to determine the temperature at the
In practice, this is handled automatically by the specialized equipment
I mentioned above. The best case scenario you can hope for is to
already have equipment that will perform the CJC based on the type of
TC you are using and output a scaled voltage signal that corresponds to
the temperature at the junction.
In the old days this had to be done in software. Nowadays, TC
equipment is so advanced that you rarely have to think about this
Q: My measurements are off by several degrees, outside of the
published accuracy of a TC, why?
This is an open ended problem with a million possible causes. The
majority of problems of this sort, however, can be rectified by
answering the following questions:
-Are you correctly connected to appropriate equipment?
– Does your Thermocouple pass the two-point test (ice-bath, boiling
water) mentioned above?
– Is your equipment calibrated?
– Is your signal free of noise?
– Is your Thermocouple where it needs to be?
Temperatures can vary substantially with location even in a
well-controlled environmental chamber.
More esoteric causes of this type of problem can be: loses due to
radiation, convection or conduction, or short circuits where your
thermocouple might be making an electrical connection (or two) with
some other conductor.
If you become totally stymied by a problem such as this, start from
basics. Do the ice-bath / boiling water test. If you cannot get your
equipment to measure those temperatures correctly, throw out your
equipment and start again.
Q: There are like 100 different types of thermocouples out there,
which one should I use?
That depends on your temperature range and application. Your best bet
is to find a good supplier and see if they have a selection guide.
They spend way more time on this kind of stuff than I ever could. Once
again, I am going to plug OMEGA Engineering. www.omega.com. I go to
their ‘Temperature Handbook’ EVERY time I need to make a
temperature measurement that I have never done before.
Q: Can I ‘share’ the output of my thermocouple with two different
data acquisition devices?
No. You cannot split the signal so that one thermocouple feeds two
different devices, for instance.
Q: How do I attach my Thermocouple to my UUT?
Any way you can. I’ve soldered TC’s to adhesive copper tape, used
polyimide tape, crazy glue and tie-wraps. Electrical insulation
between your TC and a UUT is desirable if a ground-loop could be
created in the TC. Whatever you use, remember that the thermal
characteristics of the attachment method become the thermal
characteristics of your thermocouple. The more thermal mass you create
with your connection vehicle, the poorer your measurement response to
temperature changes becomes.
The best method I have come across is to solder TC material directly to
adhesive-backed copper film and sticking that directly to my UUT.
Some mechanical attachment might be indicated f you are going to
temperatures above or beyond the capability of your adhesive. Remember
to keep the solder bead and tape small to avoid introducing extra
But wait, there’s more: This week’s bonus question:
Q: If I solder my thermocouple together to form a bead, will that
affect the accuracy?
No. In fact, you could splice a piece of standard wire between any two
points in the TC loop AS LONG AS THE TWO ENDS OF THE SPLICE ARE AT THE
SAME TEMPERATURE. It’s called the ‘Law of intermediate metals’.
So, as long as the solder bead is the same temperature at both
‘ends’, you have no problem.
TC’s are pretty easy to use. They are cheap to use in high
channel-count deployments and are very flexible in type, configuration
and temperature range, but they DO require great care in usage.
If you use TC’s in your system, take the time to make sure you are
doing it correctly!
Next week: I am going to change gears from the re-runs and talk about Thermo-electric-Coolers (TECs) or Peltier junction devices.
Thanks for reading!