search

QuantAsylum / QA40x

Releases (no source) for the QA402 and QA403 Audio Analyzer.
https://quantasylum.com/products/qa402-audio-analyzer
15 stars 1 forks source link

Noise Spectral Density Measurement (Rt Hz) depends on window function #31

Closed NoiseFree closed 1 year ago

NoiseFree commented 1 year ago

Hi,

I encountered a bug with the noise spectral density measurement. I wanted to validate the measurement with a known DUT. So I prepared the following OpAmp circuit (Non-Inverting Amp with OPA828, metal film resistors, input shorted with 50 Ohms):

image

Simulation says, it has a noise spectral density of around 4.2 nV/rtHz (broadband noise between 1kHz to 10kHz):

image

The measurement with the QA403 depends strongly on the window used:

image

From a user point of view the spectral density shouldn't depend on either the FFT size nor the window function used. So far, Rt Hz mode compensates for FFT size only, like it is discribed in the GUI. But I think the scaling isn't done correctly yet. A correction factor for window functions should also be involved, like it is mentioned here:

https://www.ap.com/technical-library/fft-scaling-for-noise/

Looking forward to your answer. Regards

QuantAsylum commented 1 year ago

Hi @NoiseFree, can you verify if on your plot that the RMS is reported correctly regardless of window setting? Yes, I see the shift in trace as you do

NoiseFree commented 1 year ago

RMS value stays the same regardless of window setting

QuantAsylum commented 1 year ago

This has been fixed for release 1.174 out later today. In this release, you can verify for RECT or other windows (HANN, FlatTop, etc) the displayed Y value of measured noise remains the same when RTHZ is enabled via the DBV context menu. Please confirm at your convenience, and thanks very much for reporting this.

QuantAsylum commented 1 year ago

Link is here: https://github.com/QuantAsylum/QA40x/releases/tag/1.175.

Please re-open if you see an issue.

NoiseFree commented 1 year ago

Thanks for the fix! I can confirm that the measurement is not window dependent anymore.

But there is still an issue with the display. It seems that it's off by about 1 dB.

To proofe that I measured the spectral density of an 1MOhm resistor at room temperature. The circuit is the same non-inverting amp as shown above with the 50 Ohms removed.

image