QuantAsylum QA400 and QA401

Hi, ASIO401 author here. Thank you for your interest :) I'm glad that the driver was already reported to work with at least 3 different ASIO Host Applications, that was more than I expected for code that has seen very little testing until now.

I see a few idiosyncrasies like left/right reversal, and polarity reversals using the FR2 mode.

Ah, I think I see what you mean. I think I remember accidentally observing some of these issues during development, but I didn't think much of them because I trying to fix much bigger problems at the time.

I have created entries in the ASIO401 issue tracker to act as my to-do list and so that people are aware of these issues: channel swap issue and polarity reversal issue. I'll try to take a look at them over the next few days.
 
Is there a possibility that the driver will also work with the QA400 (now or later)?

Honestly I have no idea. I don't have a QA400. Technically if the QA400 and QA401 happen to share the exact same USB protocol then it's conceivable ASIO401 could work out of the box with the QA400, but of course that seems unlikely.

If only small changes are required to make ASIO401 work with the QA400 then I'd be happy to make them, but I'd have to know what the USB protocol differences are in the first place.
 
Cool, there is a opening, I would be happy to try Arta (and maybe a few others).

The QA400 is a very different animal. It uses the old Cypress USB stack, which had some serious problems for us once MSFT moved to Windows 1803 (A Win10 release about a year ago). At that time, we moved the QA401 to the WinUSB stack, which has been rock solid and a great move even though the migration was painful at the time.

-Matt
 

mkc

Member
Joined 2002
Paid Member
Hi Matt,

Just out of curiosity. Would you consider making enough QA400 HW information available so a thirdparty would have a chance of making a new firmware? I'm only thinking pin connections and such.

I respect your decision to not do so. Just wanted to hear what you think.

Cheers,
Mogens
 
Account Closed
Joined 2001
Hi, ASIO401 author here. Thank you for your interest :) I'm glad that the driver was already reported to work with at least 3 different ASIO Host Applications, that was more than I expected for code that has seen very little testing until now.



Ah, I think I see what you mean. I think I remember accidentally observing some of these issues during development, but I didn't think much of them because I trying to fix much bigger problems at the time.

I have created entries in the ASIO401 issue tracker to act as my to-do list and so that people are aware of these issues: channel swap issue and polarity reversal issue. I'll try to take a look at them over the next few days.
Thank you. This is not a big deal usability-wise, since once aware a user can simply change ARTA settings to work-around. In the FR2 mode simply select "invert phase of input channel" and you're good to go.

Dave.
 
Can a totally fanless media-player-style PC like the Minix Z83-4 cope with the requirements?

Yes, there's a blog post below detailing the QA401 running on a low-spec $120 Win10 machine. If you've been following the developments on Tractor, you might know it's a program for driving automated testing and logging the data directly to a database. The small machines obviously won't be able to run a database, but the next release of Tractor will support logging your test data to a database in the cloud that QuantAsylum runs. You just click a button, and every test you run gets push to a web service. The cloud service will be free.

This lets you construct an audit trail on the devices you manufacture. You scan the serial number, run the tests, and if you ever want to know what the THD performance was on the unit with the serial number WU32X you can find it in the cloud. Or if you are wondering how amp gain this month compared to last month after you switched to a new thin-film resistor supplier, you can find it in the cloud.

Store as many tests as you like. If you are testing a lot in a certain part of the world and need the cloud database moved closer to you to help with response times, we can do that too.

So, very low-spec machine can easily function as the bay controller with full logging to a redundant database. There's a link below to Tractor running tests on a 300W class D amp.


Low-Cost Test Setups


YouTube

-Matt
 
Last edited:
Matt -

Great work on all the QA401 updates! Lots of good stuff in there. Just downloaded and installed V 1.707 to get caught up on everything since August. I can see that boatloads of work has gone into those on your end! Interesting you found/solved Jen's jitter discovery in the FPGA.

Now if the QA401 just used USB 3.0 (err, cough, USB 3.1 gen1 lol) for the power supply... :p :) (Sorry, couldn't resist, bought a Dell Precision 7730 a few months ago. 3.1 gen 1 and gen 2 ports all over the place, USB 2.0 is so 2017). USB type-C is good for 100W these days, and means never having to guess if your USB (Type A lol) plug is upside down on not.
 
Last edited:
Interesting you found/solved Jen's jitter discovery in the FPGA.

Yes, that bug report led to another side benefit too. There's an experimental mode on the QA400/QA401 that was released in 2013 or 2014 called "melt noise" based on an AKM paper Demian had sent long ago regarding using N converters to acquire a signal. Every time you double the number of converters used, your noise drops by 3 dB. In the paper they paralleled 8 converters.

The melt noise feature acquires the signal N times. And since the box has precise control of the samples going out and coming in, they are precisely lined up. And if you do 8 acquisitions, that's the same as having 8 converters. The N acquisitions are precise in time alignment, and then they are averaged in the time domain (squashing the noise) and then the FFT of that average is displayed.

And eventually, THD should equal THD+N because you've melted all the noise away. But it never quite got as close as expected and I couldn't understand why. But this was due to the jitter. And once the +1/-0 sample jitter issue got fixed, then it started to better match the simulations.

Below you see loopback in the experimental mode with "melt noise" turned on. MeltNoise2 needs several more minutes to run to match the THD (and it's already been running for several minutes). MeltNoise1 has already converged and you see the THD and THD+N are very close.

PS. To turn on the experimental menu, just place a file in the QA400 or QA401 data directory called "experimental" (no extension) and the menu will appear on next re-launch of the app. Melt only works on the left channel.

PPS. AKM published the technique to explain how to extend the dynamic range of a converter. It's primarily useful if you are looking to find harmonics in what you are measuring that might be below the noise floor of the converter. But as you can see, a lot of tones get uncovered that are likely sigma delta idle tones from either the DAC or ADC. Thus, the ultimate utility of the mode is limited IMO. Though still interesting. That's why it's buried in the experimental menu.
 

Attachments

  • MeltNoise1.jpg
    MeltNoise1.jpg
    254.2 KB · Views: 551
  • meltnoise2.jpg
    meltnoise2.jpg
    248.3 KB · Views: 558
Last edited:
Yes, that bug report led to another side benefit too.

Oops - I see a typo on my part - should be " Jens' ". Sorry Jens!

Ah very interesting about the experimental parallel conversion mode. I think that I can see that, in general. I know that paralleled op amps (with output current balancing resistors), used for higher headphone amplifier output current, increase S/N since the op-amp noise is uncorrelated.

In this case am I understanding correctly that essentially the effect would be paralleled uncorrelated converters? Here is a paper:

Correlation of Signals, Noise, and Harmonics in Parallel Analog-to-Digital Converter Arrays

Except instead of paralleled converters just the single converter taking multiple uncorrelated samples? Or maybe they would be considered correlated since the time intervals are known?

Cool that the jitter fixed made the "melt noise" math work! :) Impressive it converged within -0.8 dB of the THD. Love it when fixing one thing solves some other lingering issue in anything I'm working on.

Hey a question I've been meaning to ask. For measuring power supply noise (page 49 of the QA401 manual) I understand the resistive divider to keep the QA401 input DC below 5V (I'm measuring a 12Vdc switching wall wart in this case). I read more than 5V would turn on some protection diodes. Just wondering if you could share a bit more of what is going on there, since the inputs are capacitively coupled? The protection diodes must be between the inputs and coupling caps. Are the caps rated at 5Vdc? Would be especially nice if it were possible to hook higher DC voltages to the QA 401 input since >5Vdc is a real "accidental" possibility with headphone amps. The rails are typically +/- 12 Vdc or +/15 Vdc. if something internally goes wrong (DIY issues or part failure) the output can hit the positive or negative supply rail. There are 2 or 3 failure modes with the O2 headamp which can cause that to happen.
 
Last edited:
May have figured it out (QA401 5V DC input limit)

Matt ...thinking about it some more (I haven't looked at the QA401 circuitry) I think I may have figured it out (DC 5V input limit). There is probably a resistor to ground between the coupling cap and op-amp input(s) on the QA401 for op-amp input bias current ground sink/source. Since the coupling cap is initially a short at t0+ with charging current, the full applied front panel input voltage (AC and DC) would exist across that resistor for an instant, and hence at the op-amp input for an instant.

So the 5V clamp diode may have gone across that resistor to protect the op-amp input. Which works, but with repeated clamps, and/or high(er) voltage clamping, the diode might degrade?

Also the QA401 attenuator - the manual seems to indicate it won't help with the DC input issue, so must be an AC attenuator only?

What I really should do with the headphone amp testing, if I think there might be a DC output issue (sketchy unit), is use a 3/4 / 1/4 resistive divider on the QA 401 input (like the QA 401 manual page 49 mentions). So even if the headamp outputs +/-15Vdc at some point, the max going into the QA401 would be 15Vdc * 0.25 = 3.75. Although the resistive divider, the series resistor part anyway, will introduce a little bit of thermal noise.

Thanks for the help and information!
 
Last edited:
I have created entries in the ASIO401 issue tracker to act as my to-do list and so that people are aware of these issues: channel swap issue and polarity reversal issue. I'll try to take a look at them over the next few days.

These issues are fixed in ASIO401 1.1. Starting from that version, ASIO401 sticks with the convention that a positive sample value represents a positive voltage on the + terminal.
 
The QA400 is a very different animal. It uses the old Cypress USB stack, which had some serious problems for us once MSFT moved to Windows 1803 (A Win10 release about a year ago). At that time, we moved the QA401 to the WinUSB stack, which has been rock solid and a great move even though the migration was painful at the time.

-Matt

The USB issue was an oversight on my part :( sorry (what good is 40 years+ of IT experience?). So, actually, I'm better of switching to QA401.
 
Some experiments in replacing the input/output OPA1612 bipolar opamps in the QA401 with some new OPA2156 CMOS opamps. I'd not recommend doing it without a lot more study. And there are some very easy pre-amps you can make if you want to measure high Z things. But this is an important opamp in the evolution of opamps.

OPA2156: HiFi in CMOS

This is an awesome article! It makes me really proud of all the hard work that went into making the OPA2156. Thank you for writing it.
 
Some experiments in replacing the input/output OPA1612 bipolar opamps in the QA401 with some new OPA2156 CMOS opamps. I'd not recommend doing it without a lot more study. And there are some very easy pre-amps you can make if you want to measure high Z things. But this is an important opamp in the evolution of opamps.

OPA2156: HiFi in CMOS

Thanks, Matt.

I was unaware of this CMOS op amp part. Your article about it is a real eye-opener. I would never have believed that a CMOS op amp could get down to 4 nV/rt Hz.

Cheers,
Bob