There is a part of my circuit that uses resistance to output a voltage, in which is compared to a voltage threshold to output a digital signal.

However +Vcc can vary and any voltages too far from 5V will cause this comparator to function wrongly and toggle at the wrong resistances, since the output voltage is a percentage of +Vcc while the voltage threshold is determined by a zener diode, which won't change even if +Vcc is changed. Which makes calibrating the comparator useless since any change in +Vcc throws the entire module off.

I'm not sure which of the two solutions are the best to implement, or to do both?

  1. Use a linear regulator to maintain a 5V +Vcc signal, so that the output voltage will always be calibrated since +Vcc won't change at all.
  2. Make it such that instead of using an absolute-value threshold like a zener diode, use a voltage divider so, like the output voltage, is relative to +Vcc, therefore it will remain calibrated (the comparator will still toggle at the same resistance).

I'm kind of in for #2 because it feels like "good practice" to work relative to +Vcc, but it also introduces some problems like it is difficult to compare two voltages directly. #1 also seems nice because it allows us to work with numbers directly.

If I do both one of them may feel unnecessary. #2 may make it such that a wide range of voltages may be used (up to a certain power limit), but may be harder.

#1 may make it such that everything will be more accurate, especially when there are some unavoidable constants that won't vary with +Vcc such as the diode forward voltage.

So, #1 or #2, or both? Or something else entirely? :/

  • \$\begingroup\$ To answer the "something else entirely" part we probably should resolve this seeming XY Problem by knowing more about what you need that voltage exactly for, then maybe someone can suggest an alternative. \$\endgroup\$
    – PlasmaHH
    Jan 28, 2015 at 13:38
  • \$\begingroup\$ @PlasmaHH That output voltage is compared to a voltage threshold, which then outputs a digital signal which will turn on a darlington pair which turns on an electric fan. The digital signal part is not important in here; what I am worried about is that changes in +Vcc may throw the threshold calibration off. What this module does is that a thermistor varies in resistance, which changes the output voltage and the digital signal. The threshold is calibrated manually such that it will flip at right about exactly 22 degrees Celcius. \$\endgroup\$ Jan 28, 2015 at 13:51
  • \$\begingroup\$ For accuracy the voltage reference will be better. A voltage regulator is good to ~1-5% (depending) and also the power supply rail can droop some if there is some other part of your circuit drawing power. \$\endgroup\$ Jan 28, 2015 at 14:26
  • 2
    \$\begingroup\$ Show us the circuit. \$\endgroup\$
    – Andy aka
    Jan 28, 2015 at 14:27
  • \$\begingroup\$ @GeorgeHerold Thanks for your answer (comment). I agree, if I take special care in my design I can get a relative voltage reference that is accurate enough, while we should not depend on absolute values with a regulator since it isn't accurate enough, especially with the other parts of the circuit. \$\endgroup\$ Jan 28, 2015 at 14:43

1 Answer 1


What you should do depends on how you are driving the NTC thermistor.

If you are sensing temperature using a voltage divider formed by an NTC thermistor and a fixed resistor, then it will definitely work better to use a voltage divider for your switching threshold also. Both dividers should be connected to the same voltage. You can consider using capacitors in parallel with the resistors to improve the bandwidth of the voltage dividers if the voltage has a lot of ripple.

But if you are sensing temperature using an NTC thermistor driven by a constant current source, then you would be better off using a stable reference voltage for your switching threshold.

A hybrid option would be to create a low noise voltage rail (could be 3.3V from an LDO) to drive your NTC thermistor divider and your comparator divider. If the main voltage rail is very noisy, or if you need the utmost in precision this may provide more reliable operation.

  • \$\begingroup\$ Thank you for your answer. It had clarified a lot. Yes, I am using a voltage divider with an NTC thermistor as a resistor and a fixed in another, so I'll go for the relative threshold. The hybrid (as mentioned in the question) also sounds nice, but it would require a voltage drop in which it may not be able to drive the fan. I'll try to implement it though, to see if it improves the results. Does using 3.3V to drive a relay/transistor and then using the nonregulated 5V to drive the fan sound nice to you? \$\endgroup\$ Jan 28, 2015 at 23:18
  • \$\begingroup\$ Just a quick comment. If you are using a comparator to compare two voltages, the comparator could still be powered from the higher voltage, but use the lower, quieter and more tightly regulated rail for comparison purposes. Not trying to change your mind. Just pointing something out that you may not have thought of. \$\endgroup\$
    – user57037
    Jan 29, 2015 at 4:40
  • \$\begingroup\$ Oh, I get what you mean. It is such that the comparator will still flip at almost exactly the same point, since the voltage is regulated, but when it flips, a 5V output can still result. Nice idea! :) Never thought of that. \$\endgroup\$ Jan 30, 2015 at 2:58

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.