I’m failing to reproduce this issue. Do you have some other DAC to test with? Preferably with different type of USB interface.
Does it happen if you use fixed output rate? Sometimes DACs have trouble changing clocks quickly.
Yes mentioned above, reproduced with another HQP recommended DAC (Pro-Ject Pre Box S2).
Are you hitting next track after 2-3 seconds? Through that entire Qobuz Disco playlist?
Yes always fixed at DSD256x44.1k due to problem with NAA UpBoard UAC2 input with changing sample rates (white noise blast, reported months ago).
Fixed output sorted that issue for non-amd version HQPe but there’s this issue now with amd version (not specific to UAC2 input).
This single channel issue with amd version I reported back 3 weeks ago with 4.28.0 but thought it was just UAC2 input issue but now have done more extensive testing than before
Only other way I can simplify system is to disable pipeline matrix for my headphones use.
I’ve already tried HQP OS, Client app, direct USB ALSA connection, simple PCM output with simple filters, fixed output sample rate. This is the ultimate in simplified setup.
I can’t simplify system any further apart from disabling pipeline matrix for headphones EQ.
I’m using string text for headphones EQ. It isn’t an issue with x64gen version.
Maybe that’s part of the problem, I’ll test later.
Yes, to the last track and then picking random tracks. And it just keeps working just as it should.
My hunch was correct @jussi_laako ! I assume pipeline matrix was disabled for you earlier or you had no EQ?
It was the last thing possible for me to ‘simplify’ the system.
With pipeline matrix disabled, no problems for me.
As soon as I change profile to my headphones EQ, problem returns!
Here is my EQ:
iir:type=peak;f=80;g=-2.3;q=1.4,iir:type=lshelf;f=105;g=5.5;s=0.71,iir:type=peak;f=165;g=2.0;q=1.4,iir:type=peak;f=1850;g=-4.7;q=1.6,iir:type=hshelf;f=2000;g=8.0;s=0.71,iir:type=peak;f=2120;g=3.0;q=3.0,iir:type=peak;f=2875;g=-2.3;q=3.5,iir:type=peak;f=5750;g=-6.5;q=4.0,iir:type=peak;f=7150;g=-1.7;q=6.0,iir:type=hshelf;f=13000;g=-8.0;s=0.71
Note: my matrix profiles are not a problem at all with non-amd version HQP Embedded or HQP OS ! Only an issue with amd version
And I don’t do any post processing
One question, “plot” function isn’t working for my headphones EQ when I now try .txt file for the first time.
Can you try plot please on HQP Embedded? On Desktop I get it after I press plot multiple times.
Preamp: -7.1 dB
Filter 1: ON PK Fc 83.00 Hz Gain -2.10 dB Q 1.000
Filter 2: ON LS Fc 105.0 Hz Gain 5.50 dB S 0.71
Filter 3: ON PK Fc 160.0 Hz Gain 2.50 dB Q 2.000
Filter 4: ON PK Fc 1750 Hz Gain -3.00 dB Q 2.500
Filter 5: ON HS Fc 2000 Hz Gain 8.00 dB S 0.71
Filter 6: ON PK Fc 2950 Hz Gain -2.40 dB Q 2.500
Filter 7: ON PK Fc 5620 Hz Gain -6.40 dB Q 4.200
Filter 8: ON HS Fc 10000 Hz Gain -6.00 dB S 0.71
Filter 9: ON PK Fc 11200 Hz Gain -3.00 dB Q 6.000
Is that amd version of Embedded?
The virtual cores are doing nothing there?
Yes…
Yes, those are not used much in this case.
I thought the single channel issue disappeared. It’s back with build 105 (amd)
Oh well, until someone else or you can reproduce I understand nothing can be done.
Build 104 works great so I’ll stick with that
Can you try pipeline matrix with headphones EQ and see if you can reproduce it? Again skipping tracks after few seconds and hitting next track, alternating sample rates
Here is my 11900K doing 44.1 to DSD256x44.1 with same settings as you (gauss-xla, 7ECv2) with build 105 (amd). More virtual cores action than your screenshot
Yes, that’s because HQPlayer sees it has more cores and auto-enables pipelining.
1 Like
Interesting, is this pipelining related to me seeing one channel issue with pipeline matrix headphones EQ?
Can you try that when you get a chance.
Repeating skipping tracks after few seconds and hitting next track, alternating sample rates (especially fast, consecutive 44.1k to 96k jumps)
@jussi_laako I feel like I’ve zoomed in on the issue now.
I’m running USB bootable HQP OS amd version 4.28.2 just to rule out any issue with my Ubuntu install.
While Daft Punk playing “Get Lucky” I switched back and forth between 2 different EQ profiles , multiple times switching and this tripped up something in HQPlayer and resulted in 1 channel audio again
So not really a change in sample rate issue - this was while 1 song was still playing.
As mentioned earlier, when my matrix profile is set to “No DSP” or matrix is disabled - this issue doesn’t exist.
It’s definitely pipeline matrix related.
Can you try to reproduce. Ideally try my same EQ (add more gain) - you don’t need to listen loud at all, it’s fair obviously when sound comes out only one side.
Pipeline matrix behaves differently between amd and non-amd version
No, very unlikely. It just allows spreading the load to more cores.
Now I know much better where to look at!
1 Like
Thanks and don’t forget to test/fix plot feature please! My text file for you to test :
Preamp: -7.1 dB
Filter 1: ON PK Fc 83.00 Hz Gain -2.10 dB Q 1.000
Filter 2: ON LS Fc 105.0 Hz Gain 5.50 dB S 0.71
Filter 3: ON PK Fc 160.0 Hz Gain 2.50 dB Q 2.000
Filter 4: ON PK Fc 1750 Hz Gain -3.00 dB Q 2.500
Filter 5: ON HS Fc 2000 Hz Gain 8.00 dB S 0.71
Filter 6: ON PK Fc 2950 Hz Gain -2.40 dB Q 2.500
Filter 7: ON PK Fc 5620 Hz Gain -6.40 dB Q 4.200
Filter 8: ON HS Fc 10000 Hz Gain -6.00 dB S 0.71
Filter 9: ON PK Fc 11200 Hz Gain -3.00 dB Q 6.000
More strangeness regarding pipeline matrix.
So if I directly import AutoEQ (oratory1990) .txt file seems to work ok. I haven’t been able to make HQP trip to 1 channel audio (yet). But I note oratory1990’s EQ in the AutoEQ txt file is PEAKS (PK) only…
My EQ that has been stumbling (with amd only) is based on oratory1990’s EQ presets below which also include LSHELF and HSHELF.
I use this:
iir:type=peak;f=83;g=-2.1;q=1.0,iir:type=lshelf;f=105;g=5.5;s=0.71,iir:type=peak;f=160;g=2.5;q=2.0,iir:type=peak;f=1750;g=-3.0;q=2.5,iir:type=hshelf;f=2000;g=8.0;s=0.71,iir:type=peak;f=2950;g=-2.4;q=2.5,iir:type=peak;f=5620;g=-6.4;q=4.2,iir:type=hshelf;f=10000;g=-6.0;s=0.71,iir:type=peak;f=11200;g=-3.0;q=6.0
And this below and both trip up HQPlayer.
Preamp: -7.1 dB
Filter 1: ON PK Fc 83.00 Hz Gain -2.10 dB Q 1.000
Filter 2: ON LS Fc 105.0 Hz Gain 5.50 dB S 0.71
Filter 3: ON PK Fc 160.0 Hz Gain 2.50 dB Q 2.000
Filter 4: ON PK Fc 1750 Hz Gain -3.00 dB Q 2.500
Filter 5: ON HS Fc 2000 Hz Gain 8.00 dB S 0.71
Filter 6: ON PK Fc 2950 Hz Gain -2.40 dB Q 2.500
Filter 7: ON PK Fc 5620 Hz Gain -6.40 dB Q 4.200
Filter 8: ON HS Fc 10000 Hz Gain -6.00 dB S 0.71
Filter 9: ON PK Fc 11200 Hz Gain -3.00 dB Q 6.000
1 Like
Reason to use ortatory1990’s actual EQ (on his reddit page here) is because AutoEQ is just an automated approximation
Left is AutoEQ approximation , right is what it should actually look like:
On a positive note with amd version with my 11900K
PCM96k to DSD256x44.1k with gauss-xla, 7ECv2
And no audio dropouts (using AutoEQ txt file)
Incredible optimisations here @jussi_laako ! Let’s just fix pipeline matrix tripping and it’s perfect ! 
In a way 10900K may actually be even better than 11900K with 2 more cores, similar speeds. And obviously 12900K king of all Intels.
Actually nothing special about the AutoEQ txt files - it fails too.
I was saving AutoEQ as my default - load and then press “apply” at the bottom.
But it turns out it was not being applied.
My default had remained as ‘No DSP’.
So each time I thought AutoEQ was working as default, it was actually my ‘No DSP’ profile.
I found this out because while music was playing I switched to AutoEQ profile expecting no change in sound quality, since it was supposed to be default. And volume dropped (preamp gain -7.1dB) and after skipping to next tracks, single channel audio came again.
So nothing special about any particular txt files - pipeline matrix itself is buggy with amd version.
Another thing to check is why default isn’t being saved like it should. Like after a reboot or restart hqplayerd, the default should be what has been ‘applied’.
Found the culprit. Let’s see if I can find a fix too. It is not related to the AMD build as such. In suitable situation, starting point in the data combined with the EQ causes limiter to trip. If you nudge volume one step to either direction and back it gets reset and your fine again.
Then there’s a bug in web interface that on first upload the .txt works fine, but on subsequent modification of the matrix, path to the file gets lost and it fails to load. This has been fixed now.
1 Like