QA402 - Frequency Response Measurement

Hi all

I just got my QA402 delivered and started to run some tests. My only doubts so far are related to frequency response meaurements. It seems that FR measurement is not perfectly working yet. I tested in loopback, single ended mode. I’m note sure if I did not get the entire concept of the QA402 or some features are currently missing. Here is what I have found:

  • Using an exponential chirp seems the only way to measure a frequency response. The WhiteNoise option is not working and a “normal” frequency sweep is not available (although in the data sheet it seems that some measurements were made with a sweep)

  • 48kHz is the only sampling rate which is working (while 96kHz produces a graph (very low resolution at the low end), 192kHz does not)

  • The measurement with the Chirp has very low resolution at the lower end

  • The frequency response is down roughly -0.25 dB at 20kHz (fs=48Hz). According to TI datasheets, the DAC and ADC should do better. Is this frequency response due to an analog LPF in front of ADC or at the output of DAC? If so, is the cutoff frequency switchable according to the sampling rate?

Thank you for any insight.

Hi @Avo

The white noise should be working at 48K (at higher sample rates, the right and left are swapped which is fixed in the release later this week). What you should be able to replicate at 48K sample rate with random noise and loopback is shown below. The FR generator is set to -12 dBV, and the RMS measurement is showing very near that (-12.8 dBV–it’s not exact given the nature of noise).

If I run it through a RC with a corner at 1.46 kHz, I get the following:

Note that this is random noise and the FFT yields equal bins widths, which means you’d expect to see a flat line in loopback. An RTA with equal octave bins will need pink noise to show a flat spectrum.

To increase resolution especially at the low end, you will need a longer FFT. A very short FFT will have very little energy at every frequency. And in some cases, there wont’ be enough energy at all to make a measurement. That will result in an annunciator “FFT too small”. This shows up at 48K with a 2K FFT, and a 192K sample rate shows this message when FFT is smaller than 32K

With suitable FFT size, here’s a sweep in loopback looking at low-end response. This is a 192K sample rate with a 512K sweep. 3 dB point is well below 10 Hz. But in general, you want to be careful about putting too much emphasis on chirped measurements at band edges.

If you are really interested in knowing the response at band edges, you can run a stepped tone sweep (See Automated Tests->AmpFrequencyResponse). That test will elt you pick your levels and frequencies to sweep.

Thank you Matt for the explanations. I managed to do a decent measurement of the frequency response using a chirp. Whitenoise and stepped tone measurement did not work yet but I’m sure that I will figure this out soon.
Loopback THD measurements show a significant (as long as -100dBV is significant) peak in the spectrum at 50Hz (mains frequency) when I measure single ended, this completely diaspears in balanced mode. Overall the QA402 is a really cool device and once the SW is more mature, it will for sure become my favorite audio measurement tool.
Thank you and best regards

Hi @AVO, yes, the sensitivity to mains is definitely there for some measurements. You might find a positional sensitivity too if you are near a transformer or machine. In general, lower impedance sources (single or balanced) that are well shielded should show very good performance at mains frequencies. And as you note, balanced can indeed further assist.

There’s another SW release out (0.993) today that hopefully addresses most of the issues you might have been seeing.