Build 99 HQP stopping and shrieking

I just did a test using 3.13B2 and playing a DSD (1x) to PCM directly in HQP. I am using the Traditional Filter in the DSDIFF settings from suggestions by Jussi.

The DSD song did play down sampled to PCM

Slider bar acted strange while song played, jumped ahead a bit at beginning (like fast playing) then song started, it then stopped, then moved, etc.

Have to eat lunch and do some more snow shoveling here, then I will try using Roon and see if the same slider bar behavior occurs. I will also try to queue up a few songs and let it play.

I am afraid you are right and that it may be an MSB ASIO issue.

I am going to try my HUGO and see if I get the same issues. I am glad you told me about Windows 10 as that was going to be my next thing to try. I guess I will skip that

Regarding DSD playback stopping, I am seeing a very repeatable pattern .

  • While playing DSD64 (selecting the album and using Play Album -> Play Now), the time/waveform slider stops on Roon exactly after 12 minutes and 38 seconds.
  • But music continues to play, it finishes playing the current track and plays the next track too. Then stops.
  • If you ask Roon to skip forward, It skips two tracks.

Previously, I thought the music stops where the time slider stopped, but looking closer, I realized it is always after a fixed amount of time. Would be interesting to see if this time is specific to my setup or anyone else is seeing similar behavior. 12 minutes 38 seconds of stereo DSD64 is approximately half a gigabyte of uncompressed PCM, if that rings any bells to the devs re buffer size.

I was able to repeat the above every time. Both Roon and HQPlayer on Mac- >playing to NAA -> Hugo TT. Hqplayer (3.13.0b2, DSD->PCM 384KHz or direct DSD)

1 Like

oooā€¦fun KMan. Mine just stopped once again (Iā€™m back to running 3.12 incidentally). Iā€™m going to run it with a long track so its easy to keep an eye on the timing. Iā€™ll also run it with a DSD128 fileā€¦that should fail twice as fast if it is a buffer issue as you suggest.

As anticipated Roon froze at 6:20 on a DSD128 Opus3 DSF fileā€¦Kman QED! I had only loaded a track that was about 7 minutesā€¦it played out in HQplayerā€¦but roonā€™s progress bar was frozen right at 6:20. Then I loaded a longer DSD128 track (Reference Recordings DSF file) and once again Roon froze at exactly 6:20. As KMan reported playback continued all the way through the 13:54 track and went onto the next track. Noteā€¦Roon was still showing that it was frozen on the prior track at 6:20. The next DSD128 track was 8:31 and it played to its conclusion (which was a total of 22:26 in HQplayer). Playback then stopped in spite of another track in the Roon queue. Now onto DSD256 for more fun and faster tests.

HQPlayer on Mac Mini > playing to NAA > Auralic Vega. HQplayer (3.12) Roon build 99. Output in HQ player was DSD128 (5644800)

1 Like

Great debugging detective skills guysā€¦ I guess @brian will be around shortly to pick up the pieces.

KMan are you dropping this all to Jussi over at CA? Iā€™m busy running DSD256.

Copied these posts over to CA/HQplayer thread for Jussi

Zorntel, thanks for confirming and investigating it further. Just saw you posted on CA. We may be getting somewhere hereā€¦

As anticipatedā€¦againā€¦the Roon timeline froze at 3:10s on a DSD256 fileā€¦incidentally output to DSD128 in HQplayer. Roon and HQplayer had a lot of trouble negotiating the start of playbackā€¦starts and stopsā€¦but then by 10s or so playback was smooth. This was regardless of whether I was hardwired on gibabit ethernet (through a gibabit switch) or over a fast wifi connection. Once again the track finished playing and went onto the next one which played to its conclusion and then playback stopped. Roonā€™s timeline was frozen on 3:10s.

Thanks for the investigation guys. Iā€™m on it.

Just to make things interesting I took the same DSD256 file and sent it to my IFI micro iDSD which handles DSD256. Once again Roon froze at 3:10. Just to make things more interesting I then put HQplayer in pass through mode (SDMdirect)ā€¦its not doing any processingā€¦just handing the files right to the DAC (or in my case the NAA). Interestingly Roon and HQplayer had just as many handshake problems getting started. Then you guessed itā€¦freeze at 3:10s. This seems to suggest that the problem is not HQplayer backing up on the processing side.

And one last ā€œweirdā€ test that may mean something to Brian. I once again ran the SDMdirect DSD256 file and paused playback after about 10 seconds. This disengaged HQplayer. I then clicked at about 2:50 and started playbackā€¦ran right through 3:10. Thus when the buffer is reset the ā€œ3:10ā€ resets.

Hope all of this helpsā€¦going back to listening to PCM!

Something really suspicious about 6:20 at DSD128:

(6 * 60 + 20) * (44100 * 128) = 2145024000
2**31                         = 2147483648
(6 * 60 + 21) * (44100 * 128) = 2150668800

Clearly, weā€™re overflowing a signed 32bit integer while counting samples. Only really one place where that could be.

That explains the stuck time counter, and Roon ending playback after the next song, and the weird next button behavior.

Well I am glad I am not crazy.

Brian, I have tried every variable and it still stops and sticks on the counter and refuses to move to ā€œthe next songā€.

Look forward to the fix and glad you listen to all of us

This is good news. Priaptor, there is always hope!

Weā€™re putting a build into alpha testing tonight with a fix for this. Once itā€™s stable, weā€™ll release it.

Thanks for tracking down the pattern, @Kman and @zorntel. That made it much simpler to figure out the problem.

3 Likes

Would be happy to test the alpha if you so wish.

Ditto on testing the alphaā€¦Iā€™m real good at making those DSD files fail :-)ā€¦and I have both 32bit and 24bit DACs. All the credit to KManā€¦you are man of observation!..a good trait for a scientist!