interpreting data from thermocouple without ADC

First, I'm not an engineer, electrical or otherwise. And basic circuitry in undergrad physics years ago is the extent of my formal training in this area. Please be kind :)

I want to experiment with using a Type K thermocouple connected directly to GPIO pin(s) of a Raspberry Pi, using an RC(resistor-capacitor) charging circuit as outlined in this article: https://learn.adafruit.com/basic-resistor-sensor-reading-on-raspberry-pi?view=all

Instead of a photoresistor, I would use a thermocouple, of course. And I realize, too, as the article points out, that this is not as accurate as using an intermediary ADC (like MCP3008). But I'm curious to see just how (in)accurate such a simplified circuit would be.

I haven't put the circuit together yet, but I don't anticipate any problems with data acquisition. But the RC time equation (obviously) involves the relationship of resistance, capacitance and time. Not voltage. And the reference charts I've found for Type K thermocouple temperature conversion compare the temp to Volts, not Ohms.

And I'm not convinced the old standby V = IR would help here, because I don't think that I (current) is constant thru this circuit. If I'm wrong, I'd be happy to hear it - because then, whatever the current is, voltage and resistance would be proportional to one another. Then I could use an ice water bath for a 0 Celsius / 0 voltage reference point, and I could plot any other points based on the proportional rise. At worst, I'd probably need to take another reference reading (e.g. inside a pre-heated oven). But I'd like a simpler mathematical solution (i.e. convert voltage to resistance) which would allow me to use the existing charts, if I could.

Any ideas?

• What temperatures do you want to measure ? Do you need to log and graph them? Sep 28, 2018 at 14:39

A thermocouple has a very low voltage output, and a constant resistance. It will not work with the Adafruit RC circuit.

As it says on that page you've linked to

It cannot be used with sensors that have a pure analog output
like IR distance sensors or analog accelerometers.


You need to use a thermistor, a temperature sensitive resistor, to work with a simple RC circuit. These will work well.

As your comment indicates that for your temperature range, thermistors are infeasible (my favourite supplier has a gazillion to 125C, and only two going to 300C) and thermocouples your only solution, we need to consider how to read a thermocouple.

A thermocouple has very low resistance, and a type K outputs around 41uV per degree C. It copes with a bit over 1000C, where the output would be about 41mV.

If you are hell-bent on measuring (a) a voltage and (b) this very tiny voltage, with a capacitor charge timing type of solution, then it's possible. The voltage output would have to be converted to a current with which you charge your capacitor, instead of the variable current passing through a resistive sensor from the +ve rail.

However, once you've built a voltage to current converter, added a capacitor and used it with the Pi timer, you have built a single-slope ADC. You're only a component away from having built a charge-balancing ADC, which works better, and another few components from a delta-sigma converter, which works yet better. So the question has morphed to 'what sort of ADC do you want to build?'

I have years of experience of having made bad build/buy decisions, both at work and for hobbies, that have killed projects. Decide what your project is. If it is temperature measurement, then buy an ADC. If building the ADC is the project, then go right ahead and build one of those. Don't try to do both.

I've just bought an ADS1115 2/3/4ch 16bit ADC module from fleaBay for a few pounds, which on the most sensitive range will resolve a type K to 0.2C directly, and read three of them at the same time, for example two measurement and one reference. Hooks up with a 2 wire I2C interface, works off 3.3v, libraries available. Just sayin'.

• To Neil_UK: thanks for making that distinction (thermistor vs thermocouple), I didn't know that! The application I had in mind was high-temp (up to 2000F). I found an advert/data-sheet for an NTC thermistor from a Japanese firm claiming a temp range as high as 1000C, though the wording implied this represented a sort of breakthrough, and I found no other references to its price/availability. Most other available thermistors I found seemed to max out at 300-500C. Is there some more specific type of thermistor I should be looking for, or is this the max temp tolerance I can expect to find? Jan 29, 2017 at 0:30
• Thermocouples are made from relatively refractory metals, thermistors from a blob of nasty ceramic, so yes, I'd agree that 1000C for a thermistor would indeed be a breakthrough, I'd be surprised if they were easily obtainable above 'semiconductor' temperatures. Jan 29, 2017 at 5:09
• "However, once you've built a voltage to current converter, added a capacitor and used it with the Pi timer, you have built a single-slope ADC." So - crudely - if I put a resistor in series with the thermocouple as a V-to-I converter, then add a capacitor (also in series), I'd get this "single-slope" ADC? Jan 30, 2017 at 7:49