AD2 Measured noise frequency dependent on sampling frequency

[Joh]

I am trying to measure the noise in an analog circuit using the AD2 with BNC Adapters (CH 1, 10x Attenuation, DC Mode, Ground-Spring). I enabled the FFT-View and activated "Linear RMS Averaging" over 10 Samples and measured in "Repeated Mode" without trigger. The measured circuit can produce a voltage between 0-3.2V but should output approximately a stable 1.55V during test conditions.

I expected to see one or more frequencies in the FFT that produce most of the noise. However I was taking three measurements with three different sampling rates and got three very different looking FFTs.

1. Measuring with Samplingrate 20kHz: FFT shows a peak at ~8kHz

2. Measuring with Samplingrate 50kHz: FFT shows a peak at ~18kHz

3. Measuring with Samplingrate 100kHz: FFT shows a peak at ~32kHz

In my opinion the location of the peaks in the FFT should not depend on the samplingrate. Now I am unsure whether the FFT peaks are simply measurement artifacts of if I am missing something.

I attached screenshots of the three measurements.

Thank you for any help in advance!

 

Edited June 1, 2023 by Joh

[attila]

Hi @Joh

What you are seeing is probably the alias of the 32kHz (or higher frequency signal).
https://en.wikipedia.org/wiki/Aliasing
In the Scope by default average samples are stored so the magnitude of aliases is highly reduced but not eliminated.

32kHz @ 25kHz -> 25kHz - (32kHz - 25kHz) -> 18kHz
32kHz @ 10kHz -> 2x10kHz + 10kHz - (32kHz - 10kHz) -> 8kHz