How to Losslessly Compress DSD without converting to FLAC and why you should do it.

QuadraphonicQuad

Help Support QuadraphonicQuad:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.
DSD -> 32:1 decimation transcode -> 32 bit floating point @ 88.2kHz -> normalize to -0.01 (or thereabouts) at 24 bit 88.2k.

This ends up nearly completely nulling with the source audio. You hear silence and at most see a meter wiggle at -100db.

Trying to convert to any other PCM sample rate ends up noticeably lossy. Similar to older sample rate conversion codecs that worked poorly. Not minding the + or - 6db factor and going straight to 24 bit fixed will lead to either a 23 bit capture 6db quieter or a glaringly clipped/distorted copy from the +6db overs.

If you're using an AVR to listen with, pay attention to what kind of DA converters you are listening to! It will not have both PCM and DSD and switch them. It's going to be converting one to the other and they will almost certainly be PCM based. Handle your digital audio accordingly.

This one requires some hands on to get right. I don't remember names but there ARE some paid conversion apps for this that do those in-between steps. Properly evaluating levels and such.
 
There is also some issue with the LFE level being non standard across various disks. I've never really understood it.

its discussed here

The LFE bug

I've never noticed the 6dB loss in level. I use replay gain which normalizes the average album level to a prescribed standard at the time of playback.
 
Last edited:
Any Lfe anomalies will either come from the original mix itself or a mastering faux pas. If it's a different mastering it will probably have other sonic things going on differently and not come even close to nulling. When you subtract one audio file from another (after precise level matching) and get a near null with only noise floor level differences it can't be coincidental. Those were the exact same file upstream. If someone did only corrupt the timing or level of the Lfe in one mastering vs another, you'd see only the Lfe channel having significant output after a null test with the other channels silent.

I guess my point is that I place a lot of weight on null testing in the scenario like I described. There isn't any kind of audible file alteration that can masquerade as two identical data sets. (Or near identical down to what would be considered noise floor level artifacts.)
 
Maybe I’m missing it, but other than perhaps saving some storage space, I don’t really see the point of going to the effort of converting to .wv files.

And my time is worth a lot more than the cost of extra storage.
 
I use Kodi for everyday playback, but I like the Musicbee interface a lot and do use it from time to time. My problem with Musicbee is there is no provision for video playback, its strictly audio. Unless they have updated it recently.
Kodi supports WV DSD. I should add it.
 
Maybe I’m missing it, but other than perhaps saving some storage space, I don’t really see the point of going to the effort of converting to .wv files.

And my time is worth a lot more than the cost of extra storage.
Well yeah, that's the benefit. Massive storage savings + better tagging. Time wise it shouldn't take long unless you own 300+ SACDs, you can just let it encode in the background overnight and speed tag in like 15-30 min (assuming you have 100-300 albums on SACD.)
 
Well yeah, that's the benefit. Massive storage savings + better tagging. Time wise it shouldn't take long unless you own 300+ SACDs, you can just let it encode in the background overnight and speed tag in like 15-30 min (assuming you have 100-300 albums on SACD.)
The only way I see this a being a worthwhile exercise is if you’re starting from scratch and ripping your SACDs for the first time.

But if you’ve already ripped discs?

I’m just spitballing my math here, but re-converting like 1,500 stereo SACDs to wv would save you maybe 2TB. 2TB of storage is what…$50 give or take?

Even at minimum wage, I hope time is worth far more than that.

And if you have less than 300 discs?

Unless I can convert them all in 5 min, I just don’t see the point of converting those either to save $10 worth of memory. Particularly to a format which not every device/DAC recognizes.

Maybe if I could put them on a mobile device, but I don’t think my portable DAC/amps can handle wv files.

It seems to me the biggest benefit is to reduce size for, ahem, file sharing.
 
It seems to me the biggest benefit is to reduce size for, ahem, file sharing.
Well that certainly wasn't my intention! That's a big no-no.
Maybe if I could put them on a mobile device, but I don’t think my portable DAC/amps can handle WV files.
I can confirm that there are popular apps that work with these WV DSD files and that it outputs fine to a DAC. Wavepack doesn't psychoacoustically compress DSD (unlike DST), so it just deflates and gets sent to the DAC or outputted as regular DSD. Like I said, it's a space and metadata benefit, nothing more. Some people (like me because I am stingy with HDD space, especially when I have active projects eating space like crazy) will find great value in this, others won't.
 
Last edited:
Well that certainly wasn't my intention! That's a big no-no.

I can confirm that there are popular apps that work with these WV DSD files and that it outputs fine to a DAC. Wavepack doesn't psychoacoustically compress DSD (unlike DST), so it just deflates and gets sent to the DAC or outputted as regular DSD. Like I said, it's a space and metadata benefit, nothing more. Some people (me because I am stingy with HDD space, especially when I have active projects eating space like crazy) will find great value in this, others won't.
Huh? I’m fully aware that WV is a lossless compression.

And what do apps have to do with playing back dsd directly from portable devices like a phone or ipad?

What tagging can you do with wv which you can’t do with dsf? I’m able to do pretty much any tagging I want (or maybe more accurately, need) with dsf files.

File sharing may not be your intention but that’s pretty much it’s biggest practical application as near as I can tell.

If folks want to go through the exercise, more power to them. But I think there’s a reason so few do.
 
File sharing may not be your intention but that’s pretty much it’s biggest practical application as near as I can tell.

If folks want to go through the exercise, more power to them. But I think there’s a reason so few do.

I want to do it to reduce the amount of space my ripped music takes on my NAS. That's the biggest practical application in my eyes.

FLAC exists for precisely the same reason.

Space reduction is good for everyone - not just pirates - if that were not the case everyone would be ripping to uncompressed WAV files.
 
I can confirm that there are popular apps that work with these WV DSD files and that it outputs fine to a DAC. Wavepack doesn't psychoacoustically compress DSD (unlike DST), so it just deflates and gets sent to the DAC or outputted as regular DSD. Like I said, it's a space and metadata benefit, nothing more. Some people (like me because I am stingy with HDD space, especially when I have active projects eating space like crazy) will find great value in this, others won't.

I don't have native DSD playback on my rig. PC does the mch flac decoding and send out PCM. So, i can't send DSD as it is.

Doing the DSD-inside-WV how much affect the processing power needed for on-the-fly decoding into PCM? IIRC decoding DSD to PCM without any resample still need a decent horsepower.
 
DSD -> 32:1 decimation transcode -> 32 bit floating point @ 88.2kHz -> normalize to -0.01 (or thereabouts) at 24 bit 88.2k.

This ends up nearly completely nulling with the source audio. You hear silence and at most see a meter wiggle at -100db.

Trying to convert to any other PCM sample rate ends up noticeably lossy. Similar to older sample rate conversion codecs that worked poorly. Not minding the + or - 6db factor and going straight to 24 bit fixed will lead to either a 23 bit capture 6db quieter or a glaringly clipped/distorted copy from the +6db overs.

If you're using an AVR to listen with, pay attention to what kind of DA converters you are listening to! It will not have both PCM and DSD and switch them. It's going to be converting one to the other and they will almost certainly be PCM based. Handle your digital audio accordingly.

This one requires some hands on to get right. I don't remember names but there ARE some paid conversion apps for this that do those in-between steps. Properly evaluating levels and such.
I have been using dbPowerAmp for converting SACD rips to FLAC. This software converts to 96/24 by default, but it can be changed if 88.2/24 is preferred.
I have many rips converted to 96/24 because of compatibility reasons. My sound card driver was unable to play multi-channel 88.2/24 until recently, so I am sticking with 96/24 and I am happy with it. Moreover, I don't normalise the volume of the FLAC files, and keep them at -6db.

I am not an expert of the technicalities, but it seems there is disagreement about the advantages of 88.2 vs 96 with modern converters. See this thread:
https://forum.dbpoweramp.com/showthread.php?39978-DSF-batch-conversion-to-96-or-88-2-kHz
My ears don't notice a difference, but I haven't done a null test to check how the two sample rates compare.

I could play DSD directly, and send it to the AVR, but FLAC is so much convenient. I just keep the SACD ISO image as a backup, should I ever need it.
 
I don't have native DSD playback on my rig. PC does the mch flac decoding and send out PCM. So, i can't send DSD as it is.

Doing the DSD-inside-WV how much affect the processing power needed for on-the-fly decoding into PCM? IIRC decoding DSD to PCM without any resample still need a decent horsepower.
I remember having that problem on my old, old computer (Celeron processor). I could play MP3 and wav files with no problem but flac and wma lossless (any lossless compressed audio file) would sputter, skip or pause during playback. It was time for a much faster machine!
 
I don't have native DSD playback on my rig. PC does the mch flac decoding and send out PCM. So, i can't send DSD as it is.

Doing the DSD-inside-WV how much affect the processing power needed for on-the-fly decoding into PCM? IIRC decoding DSD to PCM without any resample still need a decent horsepower.
As far as I have seen, no. Even on low-end devices it seems to be decoding fine without using too much processing power.
 
Great, I bought an Oppo 103 and just read this thread, it is making by brain damaged 55+ year old head spin, someone could make a business out of doing this for some of us technically challenged people
 
I'd love to be able to compress dsf files for my portable FIIO that will play them, but man do they take up the storage space. Doesn't seem like this would work though.
 
I just tried converting some DSF files to wavepack using the tools and guide outlined in the OP and although that part was successful, I had no luck in getting any of the wv files to play across a number of devices. I tried playing them on my OPPO 103D, my Bluesound Node 2i and my Marantz AVR, all via USB, but nada. None of these devices recognised the audio files. I know the OP only identified a number of programmes that would be compatible but I thought I'd give it a go with these devices just in case. It looks like I'll be converting DSD to FLAC after all.
 
Back
Top