First steps, Measuring THD(N) in loopback, L+R channel THD difference

A hello to the round,
I am hans and since yesterday owner of a qa403.
i first followed the steps in the “getting started” guide.
i noticed that the THD values between the L+R channels are about 10dB apart.
shouldn’t the values be approximately the same? Also a change of the bnc-cables and crossing of the inputs brought no change.
i would be very happy about help.

best regards hans

Hi @Hans, ideally there’d be no harmonics. But the low-level harmonics and hash (non-harmonic tones) will be present in varying degrees, and will change as you change your full scale input and input levels. Take a look at the plots from the product brief (page 14 and 15). The QA403 THD Loopback you should be able to recreate using the AMP THD versus Frequency plugin located on the Automated Test menu.

Let’s run an faster version of the plot on page 15 of the PDF linked below.

Please try this:

  1. Verify your reported USB voltage (status bar at bottom of app) is 4.75V or higher
  2. Select File->New Settings
  3. Increase FFT size to 32K
  4. Select HANN windowing
  5. Short Inputs L- and R-
  6. Connect L+ OUT to L+ IN, and R+ OUT to R+ IN
  7. Selected Automated Tests->AMP THD vs Frequency

Fill that out as follows:

Let it run. Share your result. The plot for the unit on my desk is shown below.

1 Like

Hi @matt, thanks for the quick reply. i followed your instructions. the qa403 warmed up for 30 minutes, then the measurements were repeated several times, using two sets of cables (30cm/100cm length). the results are reproducible for each set of cables.

30cm cables

100cm cables

Hi @hasi, thanks for sharing the plots. I suspect the second plot you shared (with 100 cm cables) are being degraded due to power line being picked up by the cables. These were 64k fft, and with hann window a 50 Hz power line component has a significant contribution to THD even though it’s not at 40 or 60 Hz (2nd or 3rd harmonic). You could confirm that by going to a 256K FFT and running the test again with the 30cm cables.

Below is the unit on my desk with Gen1 set at 20 Hz -10 dBV, and Gen2 set at 50 Hz at -20 dBV. With FFT size of 64K, the THD is -16 dB, but with FFT size of 256, the power line is ignored. But my guess is that on the 100 cm plot, your 50 Hz leakage through the longer cable is around becoming significant.

Looking at your 30cm plot there are two things that jump out at me. The first is the red trace with THD around -100 at 6 kHz. In the QA403 brochure linked HERE you can see a similar outlier but not quite as much. Additionally, your plots show very good THD performance around 2.5 kHz at -20 dBV for the left channel that isn’t present in the brochure plot. the plot I shared above shows similar THD performance but at 4 kHz.

In the end, the limits are tied to the converter limits. Can you please email support at quantasylum and provide your serial number there (Help->About, then look for the line that starts with “SN:”) and we’ll pull the manufacturing test data on that unit. I’d like to see what that was.

1 Like

Hi @matt, thank you for your time and the detailed explanations.
I have sent you the serial number.
I have, at your suggestion, connected a 2m long usb-cable to the qa403 regarding the interferences and placed it in the middle of my lab, away from the supposed sources of interference, on a chair and repeated the measurements.

30cm cable

100cm cable

your prediction was of course correct, as you can see. now the cable length almost doesn’t matter anymore. i didn’t expect such a strong influence. now i’ll test some scenarios to enable some interference-free measurements and ergonomic working. the chair method is not the last word :wink:
regarding the outlier i would be pleased about a message.

best regards hans

1 Like

Hi @hasi, thanks for sending the SN. I pulled the test data on your unit and nothing is unexpected and of course everything passed before going out the door.

But I also wrote a small app to look at the statistical differences across all the QA403 units that have been made thus far. On these larger sample sizes, there is roughly a 5 dB difference between THD mean on left and right channel, with the left channel being better (note: The QA403 swaps the left and right channel for layout reasons). The standard deviation on the THD measurement shows the right channel has about 20% higher standard deviation. So, these numbers all come together and at the end of the day, you can probably find particular DACs and ADCs with wide ranges of performance between left and right channels if you look at enough units. It also suggests that grading a future audio analyzer might make sense, and charge a premium for really good performance (aka tightly matched).

When a QA403 gets tested, the focus is primarily testing at the hardware relay boundaries, because we want to make sure the difference in output (level, noise, THD, etc) at -12 and -13 dBV is very similar. This is because as you transition from -13 to -12 dBV, you’ll hear a relay click and we’re trying to detect a missing part, wrong part or a missed solder joint. Additionally testing is done at very high output levels (+17 and +18 dBV) because this helps shake out problems in the supplies (for example, a backwards diode in the isolated DCDC converter allows it kind of work, but the voltage is diminished and the LDO that follows the DCDC is thus in dropout and not doing it’s job. So, you test at +18 dBV and make sure the THD is as expected to detect this. Outside of the relay input and output boundaries, we rely on the ADC manufacturers test limit. In other words, we don’t check performance at -30 dBV, -40 dBV, etc.

I’ve put two plots next to each other. The plot on the left is from the QA403 product brochure, and the plot on the right is your data. Note in the product brochure there is an outlier trace on the left channel at -30 dBV. Yours looks to have an outlier trace at -30 dBV on the right channel.

In the product brochure, note how the performance tightens as you get to 6 kHz, while both my plot and your plot showed a spread at 6 kHz that was closer to the spread shown at 1 kHz.

Finally, note that your unit is showing THD at -123 in the left channel at lower input levels and higher frequencies. My unit showed even better performance than that at 6 kHz.

In summary, your unit is functioning withing in the limits of the ADC manufacturer’s limits and our limits. Your left channel performance is quite good–better than what was shown in the product brochure. We’ll look into potentially sorting units in the future and charging a baseline price for nominal performance and a premium for units that are more tightly matched.

Hi @matt,thanks for your work.
I have repeated the automatic measurement by hand, because I can not distinguish the colors monotonously and I want to get to know my qa403 better.
it is not the left channel at -20dBV that is better, but the right channel at -5dBV. according to my series of measurements, the left channel is constant in its performance/values, the right is except for a small range (-5dBV drive and 3.5-4.0kHz) teutally worse and more unsteady.
i am not sure if this is a normal deviation.
a few screenshots for illustration.

best regards hans

Hi @hasi, these are within the range of normal deviations.

PS. You can try to “Dark on Light” UI settings to see if it helps with viewability. Edit->Settings and then tick the “Dark on Light” display options.

Hello @matt, thanks again for your explanations and your patience.
with the colors I meant the Graphtool, but I have already come to the setting options, am just still in the training :wink: .
you also made a mistake in the assignment, where you attributed the positive outlier to the left channel, which happens in the right channel.
to the conclusion still the reason of my last questions, in none of the THD graphs to find to the QA403 (yours, elektor test (a particularly good QA403), diyaudio, audiosciencereview) are to be seen in a channel such jumps as with my QA403 in the right channel. all graphs, which I saw in the net are more or less linear in the course, with the tollerances mentioned by you.
from this arose my worries of a hardware error. perhaps you can reassure me in this regard, with a brief explanation. after that, I will not bother you in this regard :wink:
by the way, i am very happy with the QA403 and thankful to you for its development!

best regards hans

Hi @hasi, the performance variability is unfortunate, but it’s normal. For example, your -30 dBV THD loopback at 3 kHz on right channel is less than median, but your right channel performance at -5 dBV THD loopback at 2.5 kHz is more than median. I think the solution here would lots more testing on our end, and find the units that are delivering very tightly matched performance across frequency and levels, and charge a premium for those at some point. And for the units that don’t pass the additional screening, we rely on ADC and DAC manufacturer limits and sell those at the normal price. But I’m not sure where the outliers might exist. For example, your outlier is at -30 dBV. I don’t know if some units would have outliers at -40 or -50, or at -43.2 at 4.3 kHz. At some point, we must always rely on ADC and THD limits as tested by the ADC/DAC manufacturer.

Hello @matt,
thank you for your reply.

best regards hans