Hello folks,
I’m using the QA403 generator to test an ADC. I’m noticing that the measured output noise (using a different device) changes between the IDLE and RUND conditions.
It’s worse when the generator is in IDLE mode, by about 10 dB. Is this right? or is it particular to my unit?
Any insights on why this might be?
Best
/rgc
The output attenuator was disabled recently due to glitches when changing levels.
Download 1.193 to compare: Releases · QuantAsylum/QA40x · GitHub
Keep a copy of 1.193 if you need the lowest distortion idle generator, but bear in mind the level glitches.
Disabling the output attenuator was hopefully just a temporary band-aid solution to the glitches, as it severely restricts the usefulness of the QA’s generator for what most of us need the device for- low level noise and distortion testing using small stimulus signals.
Here’s a comparison of the sine at 5mV on the latest 1.197 software:
Here’s version 1.194, at the same 5mV, which uses the hardware attenuators we paid for. So the D/A is generating a high level, clean sine and the attenuators are attenuating down to 5mV and taking the noise down too. The way it was designed.
And here’s what a 0.0068uF cap does to the noise. So we have tons of noise AND low level D/A distortion as we are at the bottom end of the D/A’s resolution.
The whole reason the QA was such a useful product was the A/D and D/A could be kept in the optimal ranges for the lowest noise and distortion. Now for bench mode operation with idle (95% of my work), the latest software is unusable for me. Yes, the auto operations control the attenuator, but they are less than 5% of my time.
I’ve gone back to an old bench generator I built 35 years ago until this is fixed properly. The trace above (old bench gen) has an inherent 0.09% THD+N and yet it is way cleaner on my scope.
Here’s a 500uV sine, Gen1, on version 1.193. Perfect.
Here’s the idle gen 1 switched off. But note, the relay attenuator is still active (it remembers the last setting of output), so your noise completely disappears and the true output is close to zero, as it should be.
Here’s the idle gen switched off on version 1.197! Note the approx 460uV of noise, almost as much as the 500uV I want.
Here’s the idle gen switched on at 500uV. Note it adds to the noise and the figure is completely wrong. It’s RMS added to the excess noise at almost a 50:50 ratio.
The decision made with the output attenuator for idle mode in version 1.194 (Jan 2024) was poorly thought out and needs to be revised. Many of us and much of the industry work at low levels, mV and uV for MC/MM and line levels of 150mV or so. The generator must be precise, noise free and not outputting anything when turned off.
Note: You need to completely uninstall the current QA40x application and then install 1.193 if you want to run with a properly working idle generator. Hopefully this issue I have highlighted will become a priority as I don’t think we should be marooned on an old version while other bugs and features feed through, like some of the really useful additions of the last two updates.
Thank you for the insight, much appreciated. I confirm that the noise goes away with 1.193.
Hopefully it gets fixed soon.
Have a great day!
/rgc
Hi.Perhaps it would be appropriate, if it were possible, to have a checkbox in the “Edit” “Settings” menu that would allow the user to choose the attenuator management mode: that of QA40x version 1.193 or that of the current version (1.197). Each person could then choose the attenuator management version that best suits what he or she needs to do.
@Claudio Can you conceive a good reason why you would ever want to choose an inferior idle generator performance option in an audio analyzer?
The idle generator output glitches (in 1.193) when adjusting levels across the -2.5dBV to -2.45dBV (749-750mV) thresholds and the -7.45dBV to -7.5dBV thresholds, perhaps due to the D/A being sent its value before the gain/attenuator/relays have been set, giving a momentary full output burst.
This what it looks like for the 749mV to 750mV point.
Surely a mute for those range change points is a better idea than hobbling the idle generator with poor low level performance,distortion and noise, as most times you are adjusting levels on an idle/bench generator, you are not performing measurements until after you set the new level. A 200mS mute would easily cover those few points.
@restorer-john- I appreciate your discussion on this and all the work you did. I was not aware of this problem with the newer software. I did have a problem when switching between Idle and run where the software would hang up, but that problem went away after a later rev. I am sure we will get the most out of the Generator before too long…
Hi @restorer-john. My proposal was made only to quickly solve both problems, pending a final solution to the problem. Certainly what you propose is a better solution, but I don’t know what reasons prompted Matt not to adopt it. Perhaps there are valid reasons why Matt did not adopt your proposal. What I propose is a compromise to try to still get a result, but again, your solution, if workable, is certainly better.
Yes, maybe the better solution is @Claudio’s checkbox for “glitchless” DAC idle. In that mode you’d have zero dropout and the output relay stages would always be set to 0 dB, but quality at lower levels would suffer. And if you uncheck “glitchless” then you’d have a 3-5mS glitch, but much better quality. And important change would be that the glitch would be a temporary mute while waiting for the relays to settle, while previously the output could inadvertently go too large while awaiting the relays to settle. A new release is close that also has the direct entry for Gen2 and some other fixes. Let me check.
You do that Matt, and I’ll buy you a six-pack of your favourite cold beer!
I’m running 1.193 on my main lab machine in the meantime.
The actual transition points are the only points where you’d need the temporary mute. That means only 2 or 4? I’m not sure how the relay attenuator/resistor attenuator and opamp gain stages are set up, as I haven’t opened the QA. It’s really nice not having to dismantle something for once. All I know is the combinations when I built (kit) my PC based analyzer back in 1995 were mind numbingly complex for the input scaling. Same for outputs I’m sure.
Don’t laugh, it still works nearly 30 years later. I’ve had to keep a few Pentium machines alive during those years running win95 just for that old girl. The software is DOS so it’s easy to kill it (in a win95 DOS window) when it crashes- which is often.
The QA is like an S Class Mercedes by comparison.
