I am in a measurement systems class at school, and we have a project to design a measurement system. I was reading my text book when a diagram caught my fancy, and I thought I would try to build it. I have a few questions regarding it.
Bipolar Transistor PTAT Cell:
http://i.imgur.com/Slkftk4.png
The governing relationship is that the difference between the two emitter voltages is equal to the thermal voltage times the log of the proportion of the currents
Vptat = Vt*ln(I/Is)
The first question I have is how to do a constant current source to bias the BJT's. I was thinking along the lines of using a Widlar Current source like so. http://www.ami.ac.uk/courses/ami4409_amsicd/u01/images/amsicd_btccsd_img08.gif
Secondly, the temperature that this setup measures is the absolute temperature of the transistor because that is what effects the voltage to current relationship. So, if I wanted to make this cell measure ambient temperature, could I just stick a heat sink over the top of the transistors so the sink would help keep the transistor close to ambient temp?
Thanks,
Jon
Best Answer
If you keep the current levels reasonable so that self-heating is not excessive, the temperature should track ambient fairly well, with some offset. A heatsink would reduce the offset. You can predict the offset from the thermal resistance figures found in transistor datasheets. Hint: You probably want to be more like 10uA/100uA than 1A/10A.
You could use the Widlar current source, but it will have a temperature dependency. You could consider just using resistors in a 10:1 ratio connected to a relatively high voltage (think about first-order correction schemes for Vbe). Or just keep the current sources at a constant temperature.
Note also that you could use the same transistor for the two voltages, just time-multiplex between two currents and look at the difference in voltages.