I need to sample an analog pressure sensor (TSCDRRN015PDUCV) with very high accuracy. The pressure sensor is currently connected to an instrumentation amplifier (AD623) and then sampled by my MCU's ADC. The sampling frequency is very low, something like 20-30 Hz.

I want to be sure to reject as much noise as I can. Does it make sense to add a first-order passive/active low-pass filter to get rid of HF noise between the sensor and the amplifier? If yes, what would you suggest as a cut-off frequency? Could it be something like 40 kHz, for example?

Is it also worth it adding a buffer between the instrumentation amplifier and the ADC?

  • \$\begingroup\$ Are you trying to capture peak or transient pr changing values, of just steady-state values after the pressure has stabilized after a change? \$\endgroup\$
    – AnalogKid
    Feb 1, 2023 at 17:28
  • \$\begingroup\$ High accuracy will require NIST traceable sources. High precision doesn't require that. What is the signal bandwidth of interest? (No, I don't care the least bit about your sample rate, just yet. Just what you consider to be signal and what you consider to be outside the signal band.) What exactly is the precision? And if you really mean accuracy, then what's the accuracy and how do you expect to be certain of it? \$\endgroup\$ Feb 1, 2023 at 17:29
  • \$\begingroup\$ @AnalogKid I need to measure the value of pressure change during air flow in a sealed chamber at intervals of about 50ms. \$\endgroup\$
    – mdir
    Feb 1, 2023 at 17:56

2 Answers 2


You may as well go all the way down to about 40% of your sample rate. Any frequency content much higher than that will just get aliased anyway. If you need to capture transients faster than that, you need to rethink your strategy.

It doesn't make too much of a diff whether the filter is before or after your amplifier.

The output impedance of your 623 is good enough -- you shouldn't need a buffer between it and your ADC

  • \$\begingroup\$ Thank you for your answer. So you mean I can go for a fc= 100Hz for example? \$\endgroup\$
    – mdir
    Feb 1, 2023 at 17:57
  • \$\begingroup\$ Wrong direction. 12Hz is more like it, for a 30Hz sample rate. \$\endgroup\$ Feb 1, 2023 at 18:15

If your filter is to reject harmonics of your signal from causing aliases, then indeed you want a cutoff that's less than 50% of your sample rate. So less than 10 Hz for a 30 Hz sample rate.

However, if the pressure is going to be varying slowly (so is also free of these harmonics), and if the sensor and amplifier noise is wideband, and if the ADC sampling is sufficiently wideband to down convert high frequency noise, and if this wideband noise is causing you a problem (that's a lot of ifs), then you will see an improvement from almost any corner frequency of filter. For instance, if your ADC aperture time is sufficiently small to be sampling noise from 100 kHz, and your amplifier is generating noise there, then a 100 Hz low pass filter may reduce your noise amplitude by a factor of 10 (sqrt(100k/100)), making some other assumptions about the spectrum of the noise.

Using a 10 Hz low pass filter will slow the response of the system to step changes. If you need to see steps with low latency, then use a Bessel filter, these settle the fastest of the major filter types. Although Bessel have the least aggressive transition band, this is not a problem if you are filtering high frequency noise.


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.