I have what I believe is a 2N 3904 638 transistor.

It measures 50MΩ at around 70°F.

  • How can I find a data sheet on this?
  • How would I go about finding a similar / replacement thermistor?

PS: I did try Google Searching for those numbers but didn't come up with anything definitive, so this is partially about learning what the right way to identify / look up components is. Also, if I need to graph the resistance over a temperature range I can do that as well.


  • 3
    \$\begingroup\$ That looks like a common transistor. 2N3904, the rest of the numbers are probably a batch code... I think you have a component under the heatshrink.... \$\endgroup\$
    – Spoon
    Jul 24, 2014 at 22:33
  • \$\begingroup\$ I was all "Wha? Thermistor? Someone need an edit?". \$\endgroup\$ Jul 24, 2014 at 23:20

1 Answer 1


It's an NPN transistor- a very common one. You can find the data sheet easily by searching. They're around a dime each in small quantities from any distributor.

They've tied the base and collector together so it's a diode-connected transistor.

Such a diode-connected transistor follows the ideal diode equation more closely than a real diode at lower currents.

In the simplest implementation you put a current (say 100uA~1mA) through the transistor and it has a forward voltage of around 600mV that changes at about -2mV/°C. However the forward voltage and the tempco vary from unit to unit.

By using two or three different currents you can cancel out the unit variations and also the resistance of wires going to the sensor by using a bit of math and choosing the currents carefully. It's possible to get interchangeability of sensors in the +/-1°C range typically without calibration or selection.

This exact method is used to monitor CPU die temperature in a PC- the diode is part of the CPU chip.

If you want to test if it is functional, many multimeters have a diode test function that should show a number such as something between 500 and 600 in one direction, and overrange (open) in the other direction. If it reads something like that, it's almost surely functional.

  • \$\begingroup\$ Diode connected transistors are my favorite temp. sensors. (from 77 to 400 K, beyond those limits, it gets weird/ different between devices... at least from my very limited observations.) But for the TIP31C made by fairchild, they all follow a single curve, with a temperature dependent linear offset.. more tomorrow. \$\endgroup\$ Jul 24, 2014 at 23:40
  • \$\begingroup\$ @GeorgeHerold thanks for the tip. The TIP31C has lower junction to case thermal resistance, what can give faster and more precise results. TO-92 package (like from a LM35DZ temperature sensor), has this "problem" for heat sink or surface temperature measuring. \$\endgroup\$ Jul 25, 2014 at 3:57
  • \$\begingroup\$ George's tip is that the TIP does not follow the same curve, though. So perhaps it's a foul tip, and they belong in the tip? \$\endgroup\$ Jul 25, 2014 at 4:03
  • \$\begingroup\$ A foul tip, and still caught out. :^) @DiegoCNascimento, I made a mistake I use the TIP32C (pnp) I can then stick the collector tab to what I measure and feed it from a positive current source. Here's a link to a data sheet. dropbox.com/s/qxxoamj2hcn2nly/Diode%20calibration%20table.doc I tested another batch of transistors from fairchild and they all fell on the same curve. (not just one date code.) I haven't tested other transistors, nor other manufacturers. But the single point calibration is nice! \$\endgroup\$ Jul 25, 2014 at 13:44

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.