I have almost the opposite of this question: Multiple Parallel LED Drivers with Single Thermal Derating NTC Thermistor

I have an array of devices which depending on operating condidion may run hot or not. I will have an NTC sensor for each, but I'm not concerned about sensing them individually, whichever runs the hottest will be used as feedback for thermal throttling and fan speed. Most likley case say 15 out of 16 will be close to room temperature and one is running hot and I want that temperature to control my system.

Planed operating range is 0-60 degrees, nominal is 20 degrees, warning/fan increase at say 40, thermal throttling at 50 and trip the system at 60.

In the off chance that say 10 of 16 is running at 39 degrees but due to some summing effect the system sees 45 and runs the fan unessesarily high, that's ok. Accuracy is not that important, but it can't be a straight sum of temperatures either.

What methods are there to cheat on this? Cost is a factor so no 1-Wire digital sensors and MCUs. Bias each NTC individually and OR them with diodes and apply some compensation factor? Is there a standard solution for this problem I can't seem to find? My google-fu is usually good, but this week it's not working for me.


simulate this circuit – Schematic created using CircuitLab


whichever runs the hottest will be used as feedback for thermal throttling and fan speed.

If you want an analogue solution use a precision rectifier circuit for each channel and tie all the outputs together. The NTC sensor that produces the highest voltage output will win the battle to imprint its voltage on the common output line. All other sensors will lose the battle and not affect the common output line.

This should work as a two input channel device: -

enter image description here

Whichever input is most positive will superimpose itself at Vout.

  • \$\begingroup\$ Neat! Any lower tier solution without opamps? \$\endgroup\$ – winny Feb 5 at 13:01
  • \$\begingroup\$ @winny schottky diodes would not work very accurately but they might work a bit. It's down to how big the signal voltage is and how sloppy the output signal could be tolerated. \$\endgroup\$ – Andy aka Feb 5 at 13:05
  • \$\begingroup\$ Pretty sloppy I suppose. Higher Vcc would help, but this is already set at 5 V. Bias and nominal NTC resistance can be chosen. Shottky+silicon diode to even out the temperature dependence? Cheat once more and place the diodes in a "cold" area far away from the NTCs? \$\endgroup\$ – winny Feb 5 at 13:12
  • \$\begingroup\$ .... or just use op-amps LOL \$\endgroup\$ – Andy aka Feb 5 at 13:13

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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