# How to Build a Thermocouple Amplifier

A Thermocouple is a terrific way to measure temperature. The effects of temperature change on dissimilar metals produces a measurable voltage. But to make that measurement you need an amplifier circuit designed for the thermocouple being used.

While researching “Zero Drift Amplifiers” as a follow-up to my video on Instrumentation Amplifiers I noticed the little schematic the front page of the LTC1049 datasheet which is shown here. I thought it was an ideal example of an analog application where some gain and some “gain helper” were needed to accomplish our useful little application of amplifying a thermocouple probe.

In the video I don’t really talk much about the thermocouples themselves other than the type I see most of the time which is type K. If you’re not already familiar with the construction of these probes you can find an informative write-up on thermocouples and the different types on the Wikipedia page and you might also want to check out the Analog Devices app note if you would like to know more. What I will cover is a reliable and precise way to read from these probes, seen in the video below and the remainder of the post after the break.

Different thermocouples sensors have a different temperature coefficients meaning that they will generate different amounts of voltage for the same change in temperature, usually specified in volts per degree of Celsius (v/◦C). Knowing the temperature coefficient of a sensor is only half the equation, we also need to nail down the zero point, meaning that we establish a calibrated reference point. Applying a known temperature such as immersing the sensor in ice water would be a simple if inconvenient way to establish a known reference temperature. Basically we could zero out and measure the change in volts per degree C from there. Below is a graph showing

Alternatively we could use a Cold Junction Compensator (CJC) such as the LT1025, a chip made to not only replicate the different temperature coefficients of the various thermocouples, but also give us a pretty reasonable calibration.

Behind the scenes the CJC acts as another thermocouple or thermometer and changes the voltage as seen by the thermocouple, which is at room temperature in our case, and also corrects for some other non-linearities. With the thermocouple being driven by the CJC the output of the thermocouple is then fairly linear and fairly calibrated.

For this quick demo I went one step further and used a chipset from Linear Technology called the LTK001 (PDF) which is comprised of an LT1025 CTC and a matched amplifier known separately as the LTKA0x. A quick glance at the specifications for the LTKA0x shows a lot about why it works in this application: it has a high open loop gain and very low input currents and input current errors. Analog Devices has a good write-up on open loop gain (PDF), we can talk more about bias current and errors in the future.

The schematic I used is here and was a composite of a couple of different sample circuits, if I were to do this circuit as a production run I think I would include a trim pot as well (and a calibration procedure.).

The PCB layers are shown below and I can make the Gerbers available if anyone is interested.The bottom is a ground plane fill which is why ground traces aren’t easy to spot.

Using this circuit you can realize a simple thermocouple amplifier which should give you enough gain to interface to your favorite controller providing it has an Analog to Digital Converter (ADC). As this reads out in millivolts per degree Celsius you will have to do the conversion to Fahrenheit in software which should be straightforward.

## 46 thoughts on “How to Build a Thermocouple Amplifier”

1. Thinkerer says:

Once upon a time thermocouples required special connectors since using ordinary screw terminals would make an additional pair of dissimilar-metal junctions. Can you compensate that out of these circuits?

1. SOI Sentinel says:

I deal with thermocouples industrially, and generally, the amplifier connection is to normal screw terminals these days. It’s a single point and they are compensated for, I believe. Now, if you have other junctions along the way, then yes, you must use the correct matched set of terminals/plugs to get an accurate reading.

1. Bil Herd says:

We used to account for things like the kovar leads on TO-5 cases having a small additive temperature coefficient depending on whether it was inside a feedback loop or not, etc. etc.

These days with software usually involved with a measurement, a good calibration procedure can usually take out a lot issues that creep in, with the results being better the closer to ideal the original signal is.

Also the Cold Junction Compensator uses an approximation for “bow” correction and appears to spread the approximation out over the various outputs, another source of error at some level.

2. Marvin says:

If all your other dissimilar junctions are at the same temperature the Seebeck EMF is equivalent to a single junction of the first and last metals.

3. If you want to measure in the 0.01’s of degrees, then no – the only option is to have a well defined hot and cold junction. I’m working in a company that does measure in that range using thermocouples and yes, there is quite some material controlling being done on the physical side of things

4. Don says:

Dealing with industrial signal converters, there is usually a temperature reference bonded to a leg of the terminal block to provide an accurate CJC.

2. netbeard says:

Unless you’re measuring insanely high temperatures, there’s no good reason not to use a pt100 or pt1000 RTD instead of a thermocouple. They’re more reproducible, no drift, and no need for cold junction compensation.

1. Yes, there is. You can make any amount of thermocouples cheaply, quickly, and easily. All you need is the thermocouple wire itself and a way to weld the tip. RTDs, not so much. From a maintenance perspective, it makes a lot more sense to keep a spool of thermocouple wire on hand, than needing a drawer full of spare RTDs. Also, when a thermocouple breaks, you can replace it in seconds with a pair of wire cutters and a propane torch.

1. chango says:

Or weld it with a big cap. Non precious metal thermocouple wire (T) are cheap and come spooled and jacketed in zip cord like insulation.

1. Matt says:

Or a rewound savaged transformer from a microwave oven… I’m currently using these when I need to weld small stuff (or melt something just for the sake of it)

2. tekkieneet says:

I used DIY carbon arc welder – alligator clip with mechanical pencil lead and a small adjustable bench supply (less than 10V and an amp or so). Worked pretty to weld the twisted end of the wires into a nice small round junction.

I tried using the plasma arc from my solid state Telsa coil adjusted all the way down to 10-ish watts, but it was too hot and harder to control.

3. christoph says:

How exactly is the big-cap-welding done? I have some Type K wire left and I’d like to give it a try.

3. Matt says:

I tend to use the good old OP07 and any cheap temperature sensor for cold junction compensation. Data processing is done in software. Works really well – with three-point calibration (ice water, boiling water, melting solder) the accuracy is 2-5K with respect to +/- 0.5K calibrated temp sensor.

1. chango says:

This is the way to go, especially where you’re measuring high temperatures. An OP07 and TMP35 each feeding into separate ADCs on a micro is usually good enough. Way cheaper than those fixed function thermocouple amp ICs.

1. Bil Herd says:

We used to use the OP01 and OP05 which suffered from some popcorn noise back in the 70’s, and then the OP07 when it came out, all from from PMI. They used to make some great stuff, looks like they were bought by Analog Devices. We used to add a matched transistor pair to the front end of an OP07 as our instr amp.

1. Bil Herd says:

Cool Frank.

2. joo says:

One pitfall with the MAX31855 you can have, is to rely on the internal fixed amplification. With the J-type version the sensitivity is set to 57.953 µV/°C which is fine if you measure 1000°C, at say 200°C according to a J-type datasheet you would need 53.895 µV/°C, which results in a whopping 15K error.

4. Artenz says:

I usually grab an AD7792. It comes with pre-amp, 50/60 Hz filtering, current source to detect broken sensor, and extra ADC channels for measuring cold junction with NTC. Only needs NTC, resistor, and a few caps as external components.

1. Bil Herd says:

Cool. I like being able to mux in a gain and zero reference and null out in software providing the mux is mostly neutral.

1. Bil Herd says:

Cool. My last vid used AD so I switched it up to Linear Tech.

5. I think an Attiny861 with differential gain could be used as a “simple stupid” alternative

1. Bil Herd says:

All depends on what you want to get out of it which is the cool part of designing. There is often a near-infinite number of ways to do the same thing.

6. Ken says:

This is kinda an old school analog approach?
Today, you’d use a thermocouple to digital converter like a Maxim MAX31855.
Okay that part sucks for RF susceptability and reading below freezing temperatures.
But calibrating and drift of thermocouple amps is a hassle.

1. Bil Herd says:

This circuit is mostly pre-calibrated, the amp is a zero-drift type which typically do their own drift correction which is what led to showing it.

7. Ren says:

Okay, my first impression on reading the title of this blog was “amplifying sound using thermocouples”…
The World Book Encyclopedia I had as a child had a “candle powered radio” build instruction which used a multitude of
dissimilar metal junctions to power a small amplifier for a crystal diode AM receiver.

1. Bil Herd says:

There are thermocouples that directly power a valve just from the voltage they produce which is kinda cool.

1. tmk says:

I added a safety valve on my vintage stove that just does that. A thermocouple is heated by the pilot light, which provides enough juice to keep a valve open.

1. Antti says:

You can make this without electricity by using closed tube which flame heats. As pressure grows inside tube it keeps valve open.

2. Marvin says:

It’s possible to evaporate thermocouples onto split mica or blown glass with so little thermal lag that they are microphonic – responding to the change in temperature caused by adiabatic compression of air of the sound waves.

Ok, it’s mostly off topic but someone mentioned sound and I may not get another chance to drop that into a conversation this lifetime. :P

1. Bil Herd says:

I figured there was a path between motion and voltage in there but assumed that it would be limited to low frequency. It is hard to work the word adiabatic into a casual conversation. :)

My question would then be would there be desirable qualities of a sound (pressure) made this way compared to other methods?

8. Eric says:

Bil,

Would it be possible to get the gerbers (or Eagle) files you offered?

Thanks!

1. Bil Herd says:

Hi Eric, I am working now on getting that up for you, I will follow-up when done.

1. Eric says:

Hi Bill,

Thanks! I’ll just buy one since you have them for sale – it saves me from making yet another Mouser order. Thanks for making them available.

Eric

9. Theo says:

I like LT’s explanation of why you need an IC for cold junction compensation.

In the LT1025 datasheet, they are talking about the cold junction traditionally being held in a slurry of ice water and say “To date, IC manufacturers efforts to make microminiature thermos bottles have not been totally successful. Therefore, an electronically simulated cold-junction is required for most applications.”

I love it when someone sneaks a joke into a datasheet.

10. manoj says:

I want to measure differential temperature. Two (identical) thermocouples are connected in back to back fashion so that I get signal proportional to difference in temperature only. This signal is very small, typically 1 or 2 mV. How to correctly amplify/measure it?

11. I need to know the calculated accuracy of this design! please tell Sir….

1. Bil Herd says:

Hi Rajinder, There are many things that can effect the accuracy such as the temperature range you want to use the circuity itself at and things like absolute accuracy vs relative. For best results you would actually test/calibrate by checking it at a low temperature and a high; freezing water and boiling water being two readily available sources. There are published specs for the components and there are various versions that cost a little more. You can email me at bil @ hackaday.com directly if you have other questions or an example use.

12. meriemasma says:

how can i interface through spi a thermocouple max6675 with pic 16f876 with proteus (micro ci)

13. Jonathan says:

Hi Bil,

This is a great write up and video! Do you have any of these thermocouple boards for sale? I’d like to purchase some.

1. Jonathan says:

Thanks Bil!!