Amplifying a thermocouple (for learning)

[solderjs]

Newbie here, trying to learn.

I have multimeter that came with a thermocouple. I was curious and decided to measure it on the voltmeter setting and found that seems to change maybe 0.2mV per degree above room temperature.

Then I thought "wouldn't it be nice to be able to amplify this?" and so after watching Dave's video I went down to the engineering building of the local university and checked their vending machine for an op amp. I found an LF347N (similar to a LT074).

After much trial and error I was eventually able to get a capacitor with 1v charge to read 3.55v using a 5k1 as RF and a 2k as R1. This checks out (3.55 = 1 + (5.1/2)).

Then I tried replacing the 5k1 with a 10k and the 2k with a 1k expecting to be able to see something in the range of 2mv - 10mv when holding my finger to the thermocouple, but instead I saw a reading of 1.5v and it didn't seem to vary at all in relation to the temperature of the probe.

Then I smelled the magic smell and got that sinking feeling that I'm going to have to make a visit back to the vending machine.


A couple of questions:

Since the voltage of the battery is only in reference to its internal mechanism (i.e. neither the positive nor negative terminals have any voltage in reference to any other "ground"), and considering that the thermocouple likewise only really has voltage relative to itself, can I consider the negative end of the type K "connectable" to the "ground" of the battery? Should I be using two of the LF347s op-amps to amplify each individually (like this https://www.analog.com/-/media/analog/en/products/image/functional-block-diagrams/ad8494-8495-8496-8497-fbl.png?h=270&hash=D6B26C5A2AFAB2DCB6D79A3A465D0F9E3016ACB7)  so that it's relative to itself instead of the battery?

I thought that the thermocouple might be generating a low enough voltage that it could just be hooked up to the in+ and in- directly, but with between 25v/mv and 100v/mv gain I don't think it would, and my tests seemed so wildly different different that I wasn't able to draw any conclusion from them (which makes sense considering a single degree C would result in somewhere between 5v and 20v of change). Is the feedback loop approach the correct approach?

Why did the thing burn up? What did I do wrong?

[tpowell1830]

First, welcome to the forum. Second, as always, when trying to get help on the forum with a specific circuit, always post your schematic and a couple of pictures of your setup.

No one can guess what happened to your circuit if we don't know what it is.

Third, if you do a google search for "thermocouple op amp circuit", you will get a boatload of examples.

Hope this helps...

[kosine]

The first circuit on this page using a 358 (or equivalent) is about as simple as you'll get:

https://www.electroschematics.com/12610/how-to-play-with-thermocouples/

Not the best way to accurately measure the temp as explained later on the page, but it should get you started.

Note the Seebeck chart just above the circuit diagram shows that above 0C the thermocouple output voltage changes by just 40uV per degree. Not completely flat, but good enough for many simple applications.

[jysd]

Just want to add that for cold junction compensation you should not just add the (cold junction) temperature to the hot junction reading. This works reasonably well for K type as it’s reasonably linear, but for other types it can put your measurement way off.

Instead read the temperature at the cold junction and convert that to the equivalent Emf voltage for that specific temperature (lookup tables available). Add this voltage the voltage read at the hot end, then convert that to degrees. (If you’re amplifying the hot junction reading you could either first divide that reading by the op gain then add the CJC voltage, or, multiply the CJC voltage with the op gain, add it to hot reading and then divide the sum with the gain)

[soldar]

I have multimeter that came with a thermocouple.

I would suggest you keep that thermocouple safe and buy another one to experiment with. They are very inexpensive on ebay or other sites.

[solderjs]

@tpowell1830 Thanks for the warm welcome!

I'm excited to join in on the journey.

No one can guess what happened to your circuit if we don't know what it is.

True dat... see attached.

Third, if you do a google search for "thermocouple op amp circuit", you will get a boatload of examples.

In my googling I've mostly found things where a newbie is asking a question and the thread quickly goes dead - like this one, for example: https://forum.arduino.cc/index.php?topic=546682.0

I've also found things were people just say "it's too hard, pay $15 at Adafruit or $8 at DigiKey for an ASIC".

@kosine Thanks.

I've also found links like the one mentioned above ( https://www.electroschematics.com/12610/how-to-play-with-thermocouples/ ) that show a diagram that can ostensibly be used with any 50¢ op amp and explain that theoretically it'll work, but I haven't found anyone saying "yeah, I did this, and here's the problem I had, and here's how I realized what I was doing was wrong and here's the working end result".

Case and point (in the article linked above):

Because I had limited time and resources, I purchased a microcontroller-compatible MAX6675 thermocouple module and a K-type thermocouple from eBay.

People say you can do it, but I haven't had success and I haven't found a thread anywhere where the person posting the question comes back to say "thanks, that worked!" or "I figured it out, here's how:"

I'm not quite sure how, but so far I've burnt 90¢ worth of op amps. I'm okay burning another $15 worth of them if that's what it takes to learn. However, if I can skip ahead a little and only burn another $2 worth, that would be great too.

What would be most excellent would be if someone who is more experienced and also has some spare op amps, batteries, resistors, and thermocouples laying around happened to have the bug of curiosity bite them and come back and say "yeah, I got it working in a jiffy with link-to-schematic and an LXNNN" or, perhaps more likely, something like "I tried, but oddly I couldn't get a reading until 100˚C and it was only stable within 20˚C, which sucked, so I added an x value cap which kinda helped, but I think that the X, Y, or Z parameters aren't fit for this, I bet one that has improved Z or adding a Q in series would work great. Be careful not to touch the in+ to the output on your breadboard while in- is grounded otherwise you'll break it but still get random readings, which will lead you down a rabbit whole of sadness".

(meanwhile I'll also be trying it out, but I think I may order a different type of op amp and it may take a few days for me to resume my journey)

Here's my schematic (attached).

P.S.

@jysd Once I get a reading that's anywhere in the reproducibly measurable range at all, I'll keep that in mind. Thanks.

@soldar It's just a cheap-o multi-meter (BTMETER BT-770M). In fact I'm using a different thermocouple than what it came with because the plastic wasn't rated for the temps I wanted.

[Kleinstein]

The simple circuit works in principle with a suitable OP.  As shown it uses single supply and thus needs an OP that is single supply. In addition, as there in no negative supply, it can not output a negative voltage.  This may be OK if the thermocouple is used for higher temperatures only.
However it can be a problem due to the OPs offset: there is about a 50% chance for it to be negative, and with cheap OPs like LM358 or LF347 it could be quite large (e.g. 5 mV range). So the same circuit may sometimes work or fail - if the offset is negative.

There is a way around this, by using a virtual ground - e.g. lifting the signal ground level some 1 V from the negative side of the supply.

Just for learning I would not use an all in one thermocouple chip - it's more to learn from the old way.
For a more accurate measurement I would recommend a better DC precision OP. With a suitable +- supply this would be something like OP07 or with a single supply something like LT1013 or maybe LTC1050, MCP6V27 or another modern AZ OP. However not many AZ OPs in a dip package, if needed.

Having a vending machine for OPs is really cool  .

[nfmax]

Looking at the data sheet for the LF347, and comparing the numbers there with your circuit, shows up the problem. The maximum output voltage swing (when the OPAMP is run from ±15V supply) is typically ±13.5V, in other words, the output will go no closer to the negative supply than 1.5V. This is a restriction arising from the output circuit of the device. So in your circuit, where the OPAMP is running from a 9V battery, it will never give an output less than 1.5V, no matter what the feedback network or input voltage are.

Furthermore, the maximum input common-mode voltage range typically runs from -12V to +15V (again with the OPAMP running from a ±15V supply), so with the thermocouple input voltage of a few hundred microvolts connected directly to its input, the OPAMP will not respond as expected.

What you need is to run the OPAMP from two batteries, so that it has ±9V supplies. That will bring the input & output voltages within the working range.

You should then see the output voltage change in response to temperature changes.

However, the thermocouple voltage is very small (1mV @ 25˚C), and from the data sheet the input offset voltage of the OPAMP can be as much as 10mV. The offset voltage can be either sign, and appears in series with the actual input voltage from the thermocouple. You can get round this in a number of ways:

• Ignore it. Just measure the output voltage, change the input temperature by a known amount, and measure again. The difference in output voltages gives you the change in temperature
• Adjust it. Modify your feedback network to inject a small voltage to compensate for the OPAMP offset, which you can adjust with a potentiometer. You need to design this circuit so the potentiometer gives an adjustment range of around ±10mV at the OPAMP input
• Use a better OPAMP (one with a lower input offset voltage)

I suggest you try all 3 techniques in turn, and learn as you go. If you use the first technique, and you know one of the two temperatures (e.g. use an ice bath for one) you will also have no need for cold junction compensation, assuming the temperature of your circuit is otherwise stable. Otherwise, it is an additional problem, but one that can be left for now.