THD measurement lower than 127 dB

1

Maybe it is ghost chasing, but I tried to figure out the best “floor” readings I can get out from my QA403.

Connecting a QA403 to a tower PC running Debian Linux 13, Trixie, I was also able to run the QA40x software.

while connected via a straight USB-B to A cable, a loopback to itself in single-ended on left channel allow THD measurement to cross the -126 dB level and SNR is very close to 110 dB. All 3 other inputs have shorting plugs.

When using an external 5V wall adaptor to drive the QA403 via a modified USB cable, I am able to get a THD measurement under -127.5 dB and SNR almost hit 111 dB.

A direct USB cable show the voltage displayed by the QA403 to be 4.75VDC or a bit above.

The use of an external power supply and a modified USB cable show 4.97 VDC or a bit above.

The QA403 I have is able to go so low only if certain conditions are met.

The GEN1 output must be set to -0,23 dB / 1kHz.

Right channel is muted

GEN2 output is set to -150 dB / 23 kHz. (so as not to interfere)

FFT using flat-top window and set to 1024K.

Sample rate 48K

Full Scale Input is +6 dB

Average is 100

When 100 readings are completed, I read -127.51 THD or lower and 110.94 dB SNR or higher.

Changing the orientation of the unit, I was able to get the 60 Hz at lower than -142.7 dB.

The right channel does not have numbers as good. (2 or 3 dB difference on both SNR and THD)

Having the left channel set to -165 dB / 23 kHz and all 4 inputs shorted, the highest peak I get is for 60 Hz and it is blow - 155 dB. The rest of the 20-20,000 bandwidth is below -160 dB.

1 Like
2

If you want to get better, use a low distortion (lower than the QA403’s DAC) 1kHz source. The QA403 DAC is the limiting factor, not the ADC.

1 Like
3

I had read about QA403 DAC being less capable than the ADC.

However, finding a cleaner 1 kHz source seems harder than I thought, especially when cost is a limitation. Exploring the internet has brought some possibilities to the surface. Some seems too good to be true. Others are not available anymore.

I will keep digging.

A pearl may cross my path.

4

Yves, this is one of the lowest distortion Oscillators that I have come across. Maybe there are other options that I have not learned about.

Mark

1 Like
5

Thank you Mark.

You are right.

I saw this one through my research and found nothing else getting close to these specs.

I have sent an email to answer questions.

Do they ship to Canada? What would be the total cost for a unit?

I would still have to build or get a 35V power supply and proper case, but that is not a problem.

1 Like
6

65 Euros international shipping. It’s on the page.

Mark

1 Like
7

Its hard to reduce several harmonics simultaneously (you can cascade twin-T passive notch filters to do this but you lose a lot of amplitude and risk high noise levels). A simple passive low-pass won’t buy much harmonic attenuation for the low harmonics.

Perhaps you can measure harmonics separately using a different filter on the DAC output for each measurement (twin-T’s for 2nd and 3rd, low pass for the rest?). Of course you’ve got to be sure distortion level of the passives used in these filters.

Characterizing the DAC can be done by knocking out the fundamental with a notch - standard approach really, then you can measure the residual for the DAC (it will depend on amplitude and loading no doubt, so watch out for these). Given the known residual of the DAC you can in theory subtract it from other residual measurements you take.

1 Like
8

@MarkT And here you show me that there is plenty that I really do not understand. I know the filter shapes that you are talking about. And the general flow of the filtering. How to do this in DSP is not something that I have tried to do.

9

It cannot be done in DSP, you have to manipulate the analog signal to improve the headroom between harmonic(s) you are interested in and the overall signal, using ultra-low distortion components.

This is commonly done when measuring an external DUT by using a notch on the fundamental (getting an extra 40dB dynamic range is routinely possible) - here we want to do the opposite, suppress the harmonics and not the fundamental, which is intrinsically harder in practice.