# I need to make a makeshift thermocouple given limited resources in the third world

I'm living in a part of the world where scrap electronics are hard to come by. Recycling these are big money in the impoverished SEA region I'm living in.

I have a Fluke 289 with a type K thermocouple that is reasonably precise. I really don't want to cut this and re-weld until I have to as the cable is really not very long to begin with.

Basically, I have access to a lot of aluminium wire, steel wire, some copper wire, some lead/tin alloy solder wire, and some very thin unknown alloy of nichrome I could take from inside a small power resistor. It's so thin though, like 44 AWG maybe.

I'm in need of a thermocouple that can give at least 10 mV of change in potential with only 1-2° C change at the welded junction. I can use my Fluke to calibrate this hackjob of a thermocouple.

Basically, I'm just asking if I can get anywhere near enough potential change using the materials at hand in order to trip a comparator to turn on a circuit.

The application is low temperature; I'd like to maintain 37-39° inside an insulated box at all times.

Some rudimentary experiments have yielded very poor results using copper and aluminum. Like 5-10 of millivolts across a 20° temperature change. I could get a lot of nuisance tripping with stray RF and parasitics at those low levels.

I haven't attempted to spot weld a junction yet but I can do so with the materials at hand.

• Can you get some old discrete transistor (a 3904 type, for example) in a small enough package to thermally couple to whatever you're measuring? The b-e forward voltage at some fixed current will probably change enough to be useful to you. Commented Apr 28, 2023 at 16:17
• Type K thermocouples have an output much less than 10mV per 1-2C More like 1/100th that. Thermocouples are probably not what you want. Also, were you aware thermocouples measure temperature differences between each end of the thermocouple? Not absolute temperature? Every thermometer that uses a thermocouple is using another sensor to determine the temperature at one end of the thermocouple and the readout is just adding the temperature difference indicated by the far end of the thermocouple to that other sensor's reading. Commented Apr 28, 2023 at 16:21
• That means the real purpose of thermocouples is measure temperatures higher than the temperature that other sensor can withstand (often silicon or a thermistor). If you're measuring near room temperature only near room temperature it's not required. Wasteful even if you need to build a custom cold junction compensation circuit (that's the name of the absolute temperature sensor in a thermocouple circuit). It would be like carrying water to the river. Commented Apr 28, 2023 at 16:26
• @DKNguyen " thermocouples measure temperature differences between each end of the thermocouple" - that's a useful simplification in the case of this question, but keep in mind that it's not universally true. There is at least one case where the voltage is essentially independent (+/- 1.5uV) of the cold junction temperature over a reasonably wide range of cold junction temperatures (0 to 40°C, say) because of the nonlinearity. It's not even 100% monotonic with temperature. However it's only really useful for high temperatures, and the alloys are expensive, so kind of a moot point. Commented Apr 28, 2023 at 16:34
• If you are open to other ways of measuring temperature, the cheapest and easy way is by measuring the voltage across a biased PN junction. The circuit shown here electronics.stackexchange.com/questions/531190/… is a Vbe multiplier that will give you around 10mV variation for every degree Celsius. (There are better ways to measure temp, of course, but since you are limited in resources and your system should be accurate at about one-two degrees this might be sufficient). Once you calibrate the sensor you can add more circuitry... Commented Apr 28, 2023 at 21:57

A thermocouple is an extraordinarily poor choice for measuring and maintaining a temperature in that range unless there is something very odd you are not divulging.

A thermocouple front end consists of the thermocouple itself, a “cold-junction” temperature sensor which generally measures the temperature where thermocouple materials transition to copper (eg. terminal block or connector) and some circuitry to amplify, combine and/or digitize the two inputs. The effect of the CJC sensor on the measured temperature is very close to 1:1 for most thermocouple types under typical conditions.

In your case you might be able to just use the CJC sensor to measure the chamber temperature and avoid the thermocouple entirely. A diode-connected transistor makes a reasonable to good temperature sensor in that range, depending on how you use it (~ -2mV/K to -200uV/K), and even a cheap 1% precision thermistor will outperform most thermocouple circuits. An RTD is more stable again. Both the latter can be had via mail for a few dollars.

If you would like to play with thermocouples for educational purposes, you can multiply the sensitivity by creating a “thermopile” with many junctions connected electrically in series. Type K is around 40uV per K so you would need many junctions to approach mV/K. You also need many “cold” junctions, so many pieces of wire thermally in parallel transmitting heat from hot to cold junctions- that’s the inescapable thermodynamics of the situation. You can make the wires as thin as you like to try to decrease errors due to heat transfer and the voltage won’t change, however the source resistance will increase. I would expect the materials you mention to be similar to worse to type K- iron is one additional material you could try. The traditional laboratory approach to the CJC is to immerse the cold junctions in an ice-water slurry, thus maintaining them near 0 degrees C. Not very convenient, but it can be relatively accurate with crude resources.

• Always grateful for your highly educated responses. Thermocouple does not seem to be what I am looking for, indeed. Thank you for taking your time to provide such high quality education to passionate hobbyists such as myself. I have learned more than you can know from all your comments and answers across EESE. I am interested in playing with them for educational purposes, but will be approaching this problem with a PN junction approach from hereon. Thanks again. Commented Apr 29, 2023 at 6:46
• Mail to my address in Vietnam is notoriously difficult for shippers to find. I do not know why I didn't bring my full electronics supply kit with me as the additional weight would have amounted to maybe 2 Kg. Hindsight is 20/20. However, I think I can manufacture an RTD device quite simply using materials I have at hand. Commented Apr 29, 2023 at 7:08

Thermocouple sensitivity is 100% determined by the two metals the thermocouple is made of. http://www.mosaic-industries.com/embedded-systems/microcontroller-projects/temperature-measurement/thermocouple/types-wire-element shows that your sensitivity will top off at much less than what you're asking for.

If you need more than that, you need to amplify the signal, or shift to a thermistor.

• Thank you for the insightful link and the response. I know now that a thermocouple is not what I need for this application. Commented Apr 29, 2023 at 6:48