Suggest clever temperature transducer

[trains are bad]

I develop photographic film and I want to use a thermoelectric element to maintain a tub of water at either 20C or 38C, (+-.5C) depending on my process.

I plan on using an AVR microcontroller to do the thermostat but there's the issue of thermometry. Sure thermocouples exist but you need a special other junction to use them--an ultra-cheap thermocouple setup with digital readout would be awesome, but there must be something more clever.

Since I don't really need linearity, only two points, I could really use anything that varies resistance with temperature; such as a coil of wire or a reversed-biased diode or anything like that. Any ideas as to a cheap temperature transducer would be appreciated. I have the ability to output PWM up to 5v average and I have a 10-bit voltage ADC.

[Spider-TW]

Your controller should have a cold junction built in to it. for the price of a thermocouple I would go ahead and try it. Moore Industries had free 'worm' RTDs, but I just found that they stopped Jan 31st. some of the controllers on the 'bay come with a sensor and are cheap as well. It's hard to beat them for the price if they fit your application.

[trains are bad]

Well I don't have a proper temperature controller, only a microcontroller. It's ok though, because I found out that they invented the perfect device for me, I guess they are calling it a "thermistor". Amazing the things you can get nowadays.

Now, the proper way to use a thermistor would be to run a constant, very low current through it, and scale the voltage dropped across it with an op-amp to match the range of my ADC. But since I don't care about actually measuring temperature, and only want to hold the water at one point, it doesn't really matter what function of temperature the resistance of the thermistor is or even if it joule-heats, so long as it is SOME function of temperature, and I can measure a voltage that falls somewhere in the range of my ADC.

[Spider-TW]

Well I don't have a proper temperature controller, only a microcontroller. It's ok though, because I found out that they invented the perfect device for me, I guess they are calling it a "thermistor". Amazing the things you can get nowadays.

Now, the proper way to use a thermistor would be to run a constant, very low current through it, and scale the voltage dropped across it with an op-amp to match the range of my ADC. But since I don't care about actually measuring temperature, and only want to hold the water at one point, it doesn't really matter what function of temperature the resistance of the thermistor is or even if it joule-heats, so long as it is SOME function of temperature, and I can measure a voltage that falls somewhere in the range of my ADC.

That should be good as long as you have a thermometer somewhere. You will have to have a sheath for your thermistor or mount it on the outside of your tub. The bath would short the bare leads, something you shouldn't have to worry about with a commercial probe. A good sealed straw or some sealed heat shrink would work.

You could probably set up a resistor bridge to match your adc and avoid the extra op amp.

I once made a general purpose setup adc and coupled it to an 8 bit UART so that the byte I sent the UART triggered the corresponding channel (of 8) to transmit back. It worked well but fell apart over the years (breadboarded). Nowadays I would rather buy a little PLC brick or controller.

some of the ebay temp controllers go for about $50 or less. By the time you count up costs for boards and components, they are a pretty good deal. But if you have the parts and the time, building your own is fun. I keep all of my wife's old hair driers thinking I'm going to build something with them.