When playing certain .DSF tracks, for example “Hotel California” in DSD64 format, the song plays fine for the first ~50sec. After that, it just keeps getting quick “pauses”. It is like the track is stopped and resumed almost instantly, like within 0.1s. I am not sure but this could be what others describe as stuttering? Anyway, this problem only happens when I have PEQ filter enabled. That being said, some other tracks in different albums just play perfectly even with exact the same setting (i.e. PEQ enabled).
What I have tested:
1). Direct DSD + PEQ disabled = perfect playback
2). Direct DSD + PEQ enabled = stuttering
3). Default SDM modulator and filter + Upsampling to DSD256 + PEQ disabled = perfect playback
4). Default SDM modulator and filter + Upsampling to DSD256 + PEQ enabled = stuttering
5). Playing the PCM version of “Hotel California” in SDM mode + Upsampling to DSD256 + PEQ enabled = perfect playback
Any help would be appreciated. I am going to pull my hair out soon…
Thanks for help.
Unfortunately the problem remains the same. I tried all of the suggestions above.
I also set multicore=1. Still the same.
Sorry for being a Linux laymen here, which command should I use to check the cpu usage on HqplayerOS? Mpstat?
Yea htop works better!
Just checked, all cores run pretty low, all below 40% usage throughout the whole track.
That makes me wonder my Amanero combo384 maybe the source of problem?
Is there a verified version of Amanero firmware for HQplayer?
Okay so I was dealing with 2 problems. First was the dropouts like above mentioned, Second was Pop when changing DSD to PCM (or reverse I don’t remember).
I can confirm that by flashing the firmware of Amanero to 1099c solve the poping issue. Disregard the suggestion on the official page about using 2006be for native DSD on Linux. That pop gave me PTSD honestly.
Now onto the dropouts, I have run some tests and still cannot identify the component(s) that causes the issue. Not to mention fixing it.
Here are what I tried:
Not only HQPlayer OS, but also Desktop version on Windows10 on a different computer (a gaming laptop). Same problem.
Delete all content in every pipieline, but left the matrix “Enabled”. So that no PEQ nor alteration done to the sound basically. However just by clicking the “enable” button on the matrix tab, even empty content, the problem repeats.
Monitoring the CPU usage and frequency. Both were fine. So it seemed the dropouts were not related to any of those. But dropouts were so consistent for a track. Always occur at specific time stamps.
Changing integrator and rate converter. Tried all possible combinations. FIR2 and XFI shorten the dropouts, like from 0.5s to 0.2s. But still detectable.
This is all I have now. Although I am unsure if I can solve this on my own, I will continue.
Next step…idk maybe changing the Amanero to XMOS.
I have the same dropout problem when playing native DSD above DSD64 while my HAF convolution filters are enabled. My roon core is antipodes s40 +s60 psu. I disable the filters or disable native DSD playback and problem goes away.
Antipodes doesn’t say…only that it is at least as powerful as Nucleus+. It should be, considering it can function as a core and player…of HGPlayer, Squeeze, Sonos. I use it as a roon core only for the time being and my Lumin X1 as the Streamer/DAC/PreAmp. I think to get the power required for convolution filters enabled and smooth DSD128, 256, 512 play will require a server of the Taiko Audio Extreme level
Your problem is simply insufficient computing power for such tasks. Applying filters to DSD (no matter IR .wav or text PEQ) is very challenging. When handling 5.1 DSD256 sources, my i9-12900k with 3080Ti can only use lower order integrator like IIR plus non-EC modulator for 11-channel IR filters.
As you can see my screenshot, although CPU loading was not that high but nearly all P-cores run at full clock rate and also occupied nearly 8GB of GPU memory. You need a very powerful PC for DSPing DSD.
Thanks so much! This could be the answer to my struggle. One of my computers is a NUC, no CUDA for sure. The other one has a RTX2060 but I always doubted CUDA wasn’t truly enabled. This may also explain why perfect playbacks using convolution filers because it is lighter than SDM pipeline.