Topic: Adi-2 Pro FS: minor processing issue (1LSB asymmetric offset error)

Hi, 1st post here.

I'm the lucky owner of a Pro FS and it surely is the most versatile unit around, both for music playback and measurement tasks, with outstanding features and specs.

I'm currently doing some specialist low-level high-resolution measurements with heavy time-domain averaging of sample-synchronous periodic data blocks to dig deep into the analog noise floor down to the LSB of 24bit data. Checking signal intergrity for a basic loopback cable arrangement, it seems I have found a minor DAC data processing glitch somewhere in the device firmware or the DAC-chip itself.

The error shows up as positive (>=1) sample values of a 24-bit input data stream all being lowered by one (when measured with the ADC, running with an analog gain of 10 to reduce its error contribution). That means, for example a sample sequence of -1,0,1,2,3 is output as -1,0,0,1,2.

Of course, the error produced is way below analog noise floor and thus quite probably irrelevant, nonetheless I find this worth of further investigation. Also, this could all still be "user error" although I tried to take care as much as possible.

Please read the full story over at AudioScienceReview: https://www.audiosciencereview.com/foru … by-1.6865/
There, I present the effect of the error both on a low-level bit by bit staircase as well as a sine signal, and was able to model  (and later correct for) the error in an emulator which predicted the outcome to great precision.
If these observations are real and could even be backed up by other independant measurements it is easy to fix this (very) minor issue. I'll be glad to provide further information as needed.

2

Re: Adi-2 Pro FS: minor processing issue (1LSB asymmetric offset error)

Hello Klaus,

this error is already fixed. I am currently testing a new firmware (v188) to see if our re-implementation of the whole output paths did screw up something else (and the fix was NOT that easy).

BTW, you can use Dig. Out Source Main Out also in Multichannel mode and easily verify the error wasn't there (at the digital outputs) by using the 24 bit Bit Test.

BTW 2, this issue is not present in the ADI-2 DAC.

Regards
Matthias Carstens
RME

Re: Adi-2 Pro FS: minor processing issue (1LSB asymmetric offset error)

Hello Matthias,

great to hear this issue is already known and will be fixed in the next firmware. That's why we love RME!

MC wrote:

BTW, you can use Dig. Out Source Main Out also in Multichannel mode and easily verify the error wasn't there (at the digital outputs) by using the 24 bit Bit Test.

Ah, OK. You are entirely correct, it's stated in the text, but it is missing in the block diagram "USB Class Compliant Multi-channel mode" (pg.38), the selector isn't present compared to the previous diagram for stereo mode. That's why I overlooked it yesterday, just saw the diagrams and stopped further reading, d'oh :-(

Anyway, I now checked this digital cable loopback routing in multichannel mode**) and the effect of eg the volume control at -6dB is reflected in the output (dithering), and when set to 0dB I now get the expected correct bit-perfect output!

So it seems the output stream of the DSP rendering is correct (and it does auto ref-level "wrap arounds"at the transitions), but the actual feed to the AK4490 is not... I understand it's not so straightforward to debug/fix this.

Big thanks for taking care of the isseu, and I look forward to try the new firmware.


**) ran into a slight problem there, neither Foobar2000 nor Audition ASIO channel selection/mapping to AES output did work. With Foobar I couldn't fix it at all, and with Audition it only worked when I send to analog-1/2 I get a signal at AES and all other output (like it were still 2-channel mode)??

4

Re: Adi-2 Pro FS: minor processing issue (1LSB asymmetric offset error)

Yeah, some software can't cope with too much flexibility. Needs a reset of formerly stored properties, maybe.

BTW, if I have a new firmware would you mind checking it? Can it be Mac, or does it need to be Win?

Regards
Matthias Carstens
RME

5 (edited by KSTR 2019-02-26 10:07:24)

Re: Adi-2 Pro FS: minor processing issue (1LSB asymmetric offset error)

Yes, I'll be glad to beta-test a new FW. I'm on Win7/64 and would need the corresponding FW updater.

BTW, I have a few other topics for future convenience/usability updates, the most important one is an additional "true analog levels" meter screen where the actual voltages (and max I/O voltage ranges) at all physical input/output ports can be monitored, preferably in dBV (who needs/uses those dBu's anyway) and Vpk. This would be helpful, no need to calculate the true levels manually every time a setting is changed. Since the device knows the effective analog gains applied to the individual inputs/outputs at any given moment anyway this should be feasible... will open an extra thread for this...

6

Re: Adi-2 Pro FS: minor processing issue (1LSB asymmetric offset error)

The unit does not measure or show RMS, only peak.

Regards
Matthias Carstens
RME

Re: Adi-2 Pro FS: minor processing issue (1LSB asymmetric offset error)

I know, and that's OK, RMS is not needed anway. Peak levels is what would matter here.

8

Re: Adi-2 Pro FS: minor processing issue (1LSB asymmetric offset error)

Sorry that it took so long. Once we fixed this issue we found another one - and that one was very time consuming in testing and fixing. FPGA firmware 192 should fix both issues and provide perfectly looking LSB presentations. Note that this is not an official release, that one will come later, together with other goodies.

http://www.rme-audio.de/download/fut_wi … ro_192.zip

http://www.rme-audio.de/download/fut_ma … ro_192.zip

Regards
Matthias Carstens
RME

Re: Adi-2 Pro FS: minor processing issue (1LSB asymmetric offset error)

>> Once we fixed this issue we found another one

What bug is/was this one, so we know the difference for testing, while no release notes given.

10

Re: Adi-2 Pro FS: minor processing issue (1LSB asymmetric offset error)

We do not plan to disclose that one. It was only mentioned as explanation why it took so long.

Regards
Matthias Carstens
RME

11

Re: Adi-2 Pro FS: minor processing issue (1LSB asymmetric offset error)

I know it sounds weird but beta firmware make some minor changes with sonic sinature of adi-2
(i was made update during playing music)
so i decide to go back to official 185/90

12

Re: Adi-2 Pro FS: minor processing issue (1LSB asymmetric offset error)

That is both weird - your post and your logic in it. Let's just assume the atomic-small changes of audio data in this firmware are audible - isn't it logical then that the 'new' sound is the correct one, and you should not go back to the broken one?

Regards
Matthias Carstens
RME

13

Re: Adi-2 Pro FS: minor processing issue (1LSB asymmetric offset error)

"the correct one" where? on the LSB level only? 
I don’t know exactly which part of the signal processing was covered by the reprogramming of the FPGA. Perhaps this is an anti-placebo, but the sound of the beta firmware seemed more "sluggish." And this is despite the fact that before that I had updated the firmware and did not notice any degradation of the sound.

Re: Adi-2 Pro FS: minor processing issue (1LSB asymmetric offset error)

Without a scientific double-blind test of two different devices with different firmware, any comment on "sound quality" remains speculation, really...

Regards
Daniel Fuchs
RME

15

Re: Adi-2 Pro FS: minor processing issue (1LSB asymmetric offset error)

I know about this argument, but the method of blind test with a grueling repetition of one piece of music is completely different from the usual listening experience. With the help of a blind test, you can "prove" the absence of a difference between 24bit lossless and 320 kbs mp3 audio. But this experiment does not mean that there are no differences between them.

Re: Adi-2 Pro FS: minor processing issue (1LSB asymmetric offset error)

As of now, I don't seem to hear any sonic differences with the new FW.

And, @dir, you have to update your knowledge about blind testing. Blind test does *not* mean a short piece of music is used over and over, nor does it mean there is any other significant "test stress". One can integrate long-term blind testing seamlessly into the usual recreational  listening experience and once you do that you will get meaningful results with high resolution.

17 (edited by dir 2019-03-07 11:58:25)

Re: Adi-2 Pro FS: minor processing issue (1LSB asymmetric offset error)

Perhaps I am mistaken, but I do not insist on my opinion. This is just an 'first impression' (which is often true). In any case, the blind test involves the serial downloading of the new update, and in such a way that I was not aware of. It's easier for me to stay on the old firmware and wait for the official update.

Re: Adi-2 Pro FS: minor processing issue (1LSB asymmetric offset error)

OK, I finished some measurements with the new beta firmware and can confirm the problem has been fixed:
https://www.audiosciencereview.com/foru … ost-157421

Big thanks to Mathias and the whole RME team for the excellent and professional way this small issue has been handled!