[PROBLEM] White noise playing DSD audio

[MikeyFresh]

I have opened a discussion in MPD forum: https://github.com/MusicPlayerDaemon/MPD...sions/2066

Out of curiosity maybe you can re-rip some SACD MCH tracks using the DST decompression option. While the resulting file sizes will be larger, storage is relatively cheap these days, and this test would reveal whether or not this white noise you hear is an artifact of on-the-fly DST decompression.

Stereo 2-ch playback of .dsf at 96kHz is OK?

It is a bit strange it is due to the decompression, because the same file that does not work for me in my Moodeaudio + AVR, works for Tim in his. Also, the same file works for me in other platforms (bluray in the same AVR and computer with a Sound Blaster XFi HD USB).

Understood but it might be an interesting test to isolate the exact issue, and I would suspect the RPi might have trouble with DST decompression if it is one of the lesser spec'd CPUs, I don't see where you might have posted which RPi model you are using here. 

I do know that the RPi 3B was pushed to thermal overload and auto-shutdown while actually ripping a MCH SACD with the DST decompression setting. I can't remember if I ever tried that with the 3B+, but suffice it to say the 3B or lower spec'd CPU Pi models do find the task of DST decompression during a rip challenging, maybe the same is true during playback?

[jcucurull]

Yes it is a RPi 3B+. However if the issue were the decompression, why adding and additional step of resampling to 96Khz with SoX fixes the issue? The RPi is more overloaded and the decompression still takes place.

By the way, same thing happens with Stereo files.

[MikeyFresh]

Yes it is a RPi 3B+. However if the issue were the decompression, why adding and additional step of resampling to 96Khz with SoX fixes the issue? The RPi is more overloaded and the decompression still takes place.

By the way, same thing happens with Stereo files.

OK, I understand this better now. I did ask you about what happens with stereo files in my post #18, and yes I agree that the 96kHz resampling only places an additional CPU burden on the Pi. 

I am certainly not arguing with you and have no similar unsolved issue on my end, I was merely suggesting a somewhat easy test involving a single rip with the DST removed from the resulting tracks. That said, since we now know that the issue is plaguing stereo tracks as well, then it can't be DST related as only the MCH tracks have DST.

[jcucurull]

Ah you are right, stereo files do not have the compression. So then it is not this the issue...

My guess is that the resampling of ALSA is not working correctly, but I don't understand why 192Khz FLAC files do not present the issue. One reason might be 96 is multiple of 192, while the output of the DSD converted to PCM is likeky to be multiple of 88. But still, this should not work with the setup of Tim.

[Tim Curtis]

Ah you are right, stereo files do not have the compression. So then it is not this the issue...

My guess is that the resampling of ALSA is not working correctly, but I don't understand why 192Khz FLAC files do not present the issue. One reason might be 96 is multiple of 192, while the output of the DSD converted to PCM is likeky to be multiple of 88. But still, this should not work with the setup of Tim.

I forwarded the download link to @TheOldPresbyope and @Nutul.

[Nutul]

while the output of the DSD converted to PCM is likeky to be multiple of 88.

By standard definition, DSD sampling rates are multiple of 44.1..., that is, DSD64 is sampled at 64 x 44.1 kHz.

ETA
I got the file, unfortunately I won't be able to test it on my moOde setup until Monday afternoon sometime...

[TheOldPresbyope]

Ah you are right, stereo files do not have the compression. So then it is not this the issue...

My guess is that the resampling of ALSA is not working correctly, but I don't understand why 192Khz FLAC files do not present the issue. One reason might be 96 is multiple of 192, while the output of the DSD converted to PCM is likeky to be multiple of 88. But still, this should not work with the setup of Tim.

I forwarded the download link to @TheOldPresbyope and @Nutul.

The streams info I get from ffprobe for this file shows no unexpected features; it's the same as I get for DSD files I've obtained from commercial sites.

The file plays flawlessly on moOde 9.0.4 installed on a Pi 3A+ with Khadas Tone1 USB DAC (formats: . S32_LE, SPECIAL, DSD_U32_BE).

I don't have any HDMI-driven audio gear but I'm wondering if that's a red herring.

The symptom "white noise playing DSD audio" suggests to me that the DSD bitstream is being modified upstream of the conversion to PCM. Perhaps there's an unintended software volume modification.

As an aside, the new software volume feature "DSD format will output 0dB when using Software volume" is awesome!

Regards,
Kent

[Nutul]

No hint of any white-noise / hiss over of the music itself, and this is an album I do know, very well...
The audio chain (no DSP / EQ / whatsoever):

Code:

Source:        USB/Spare 4/Music/@/test/01 - So Far Away.dsf
Source format: DSD 1 bit 2.822 MHz, Stereo
Decoded to:    DSD 1 bit 2.822 MHz, Stereo, 5.645 Mbps
Output format: DSD 1 bit 2.822 MHz, Stereo, 5.645 Mbps
ALSA mode:     Direct (hw)
Audio chain:   MPD -> hw -> Device
(...)
Volume type:   Fixed (0dB output)
(...)
Device:        EVER SOLO DAC-Z8

[jcucurull]

The noise is very clear and loud when produced and it is constant. Believe me that if the problem were reproduced in your system it would be very noticeable. It is missing for me tot test a FLAC at 176Khz or 352Khz (multiple of 44.1 at a high sample rate) and see if it happens the same.