Flac/APE vs DSD

Flac/APE vs DSD: Do You Hear Any Actual Difference?

  • Flac/APE is better than DSD

    Votes: 0 0.0%
  • No difference to my ears

    Votes: 11 68.8%
  • DSD is better than Flac/APE

    Votes: 5 31.3%

  • Total voters
    16
  • Poll closed .

travellersol

Active Member
[No theories, only actual hearing results wanted]

What I am looking for is to hear reports of actually hearing and comparing between either FLAC or APE (the two very common lossless formats) and DSD. I have read enough theories and debates, the more I read the more confused I become. So I would like to ask a favor from AK members who can play digital formats through a decent setup.

The reason I ask for this is because my system and room acoustics are not optimal so it would be unfair to test/compare these two formats here. If you are running a good enough stereo setup then I hope you can help to enlighten me. I don't know what standard parameters to measure but I think if you have a system that can let you hear good sound stage (or imaging) from recordings that are known to have such quality, then I think it'd pass.

I know mine is not up to par because I could hear very good sound stage from a local dealer's sound room but not a tiny bit of it in my oddly-shaped bedroom using the same speakers playing the same tracks. I would have a chance to move to a better house in two years and hopefully can have a better environment to set up my gears, but at the moment it's just impossible in my place to do this. Your input could help me to decide what kind of lossless tracks I should collect before moving to a better place.

I'm not an audiophile so please pardon me if this question sounds naive. Please educate me:

Is DSD definitely sound better than Flac/Ape?

A reminder again: No theories, just actual hearing reports.
 
You need to specify some sample rate and bit precision, otherwise results will be meaningless.

Then there's the DAC topology to consider; DSD and PCM are entirely different signal representations. Some 'PCM' DACs use 'DSD-like' techniques for reconstruction...

Apologies for not following your brief, but it's not a simple comparison.
 
Last edited:
You need to specify some sample rate and bit precision, otherwise results will be meaningless.

Then there's the DAC topology to consider; DSD and PCM are entirely different signal representations. Some 'PCM' DACs use 'DSD-like' techniques for reconstruction...

Apologies for not following your brief, but it's not a simple comparison.
I did think about that, but there are so many variations out there, I really don't know how to set the comparison standard. The only thing I know is: we must use the same song, the same amp, the same cables, and the same speakers. About sampling rate and bit precision, I think they are inherently different aren't they, as DSD is 1-bit but FLAC and APE are multi-bit? Maybe we should just use the file size: the file used for flac or ape should be about the same size as the DSD file? But then again, most DSD files available out there are very huge, at least double to quadruple the size of flac/ape files–what can we do now? Or, maybe we should just use the most commonly used resolutions for each format, even if that results in DSD files being many times bigger than FLAC/APE files, because these are their most natural form? It 's like: if we want to see whether a truck or a sedan runs faster on the freeway, then just let them run as they are, regardless of their sizes or horse powers?
 
but there are so many variations out there

That's why I asked. Without knowing those details, the comparison is meaningless.

You might limit it to SACD and redbook CD versions of DSD and FLAC, respectively, although that would be slightly unfair, as SACD is a higher bitrate.

Essentially, you are asking about a comparison between a delta-sigma representation (DSD) and a linear PCM representation (FLAC).

FLAC is just a bearer for lossless, linear PCM; it should sound the same as any equivalent format, such as AIFF, APE, ALAC, WAV, CDA, etc. If it doesn't, there is something wrong with the file/stream decoder...
 
Essentially, you are asking about a comparison between a delta-sigma representation (DSD) and a linear PCM representation (FLAC).
Jargons make me even more confused. I just want to know which format sounds better from actual listening, without having to know all these. Trust me, I have read many tech specs/writings on this, yet I still don't know which sounds better, because my system is not up to par to compare them.

FLAC is just a bearer for lossless, linear PCM; it should sound the same as any equivalent format, such as AIFF, APE, ALAC, WAV, CDA, etc. If it doesn't, there is something wrong with the file/stream decoder...
Let's proof this with actual listening tests.

If we really want to set some numbers, then I'd choose the following:

DSD: 2.8 MHz, 1-bit
FLAC: 192 kHz, 24-bits

They will generate files very similar in size (201 MB vs 195 MB), according to https://dsd-guide.com/size-comparison-chart-various-formats-dsd-wav-flac-mp3
 
Jargons make me even more confused.

Well, yes, I suppose it is jargon, in that it is the technical language used to describe the digital signal sampling methods.

To try to explain them simply, PCM is a direct number representing the signal magnitude at each sample point. Delta-sigma is a sampling method that measures the signal change between each sample.

Let's proof this with actual listening tests.

It doesn't need listening tests to prove that lossless, linear PCM sample stream of identical bit precision and sample rate are the same; the fact that they provide LOSSLESS (perfect copy) delivery of the original, pure PCM stream provides that. If the codecs do not reproduce a perfect copy of the PCM source to the input of the DAC, they are not lossless; something is wrong with the implementation of the codecs.
 
Can we just put aside theories in this topics here, there are tonnes of it and I (and many others) have read enough. Let's provide some actual listening experience.

It doesn't need listening tests to prove that lossless, linear PCM sample stream of identical bit precision and sample rate are the same; the fact that they provide LOSSLESS (perfect copy) delivery of the original, pure PCM stream provides that. If the codecs do not reproduce a perfect copy of the PCM source to the input of the DAC, they are not lossless; something is wrong with the implementation of the codecs.
Yes, you are entire right, "lossless linear PCM sample stream of identical bit precision and sample rate are the same". I don't want to argue nor debate with you, this is not my original intention.

Can we compare Delta-sigma (DSD) with PCM now, again, not in theory, but in actual listening experience?
 
DSD: 2.8 MHz, 1-bit
FLAC: 192 kHz, 24-bits

2.8MSa/s, 1-bit DSD gives 2.8Mb/s. Per channel, so 5.6Mb/s for stereo.

192kSa/s, 24-bit PCM gives 4.608Mb/s, so 9.2Mb/s for stereo. Assuming a 40% reduction by FLAC, this would give 5.53Mb/s, so, yes, they'd be comparable file sizes/stream rates.

My gut feel would be to compare the uncompressed PCM stream against the DSD stream, but it depends what you are trying to compare; PCM vs DSD representation as a concept, or the relative performance of the stored/streamed formats.

I've spent too many hours doing formal DRT/DAM tests whilst developing 16k CVSD, LPC-55 and GSM voice codecs to want to do more formal testing... But at least we've established a baseline to allow some vaguely meaningful results.

I'll stop being a technical pedant, and let the golden ears vote.
 
Short answer: I own numerous classical SACDs, and 24bit/192kHz (PCM) FLAC downloads, and Blu-ray, plus a few Pure Audio Blu-ray, and Ultra HD Blu-ray, that are based on modern high-quality recordings, and all of these “hi-res” technologies are capable of excellent audio quality.

Before I get into a lengthier discussion of audio quality, I’d like to point out that there are other factors affecting the selection of consumer deliverables:
  • After listening to numerous multi-channel classical SACDs, plus numerous multi-channel Blu-ray classical, opera, and ballet recordings, I have concluded that multi-channel can be useful for classical music in an installation where the main L & R speakers must be widely spaced. (OTOH, in a small room where the main speakers are spaced 5 feet apart, there is not much benefit IMO to a center channel and surround speakers.)

  • HD video is indispensable for ballet, very useful for opera, and an enjoyable enhancement for classical music recordings.

  • I think that Pure Audio Blu-ray discs are particularly well suited for very large remastered "box sets", such as the Maria Callas studio recordings that were released as 69 CDs. It would probably take quite a while to download all of the Callas recordings in 24/96 FLAC. I think that releasing it on 69 CDs (30+ year-old digital storage technology) was not ideal. A few Pure Audio Blu-ray disks would have been better IMO.

  • Ease-of-use is a matter of personal preference. Some people think that downloads are easier, whereas some like the simplicity (and reliability) of dropping a disc into the tray and hitting the “Play” button.

  • One disadvantage of SACDs is that the SACD layer cannot be played on a PC. (The CD layer of a hybrid SACD disk can be played on a PC.)

  • Often SACDs (e.g., from Amazon) are cheaper than hi-res downloads (e.g., HDTracks).
Regarding audio quality of hi-res DSD vs. FLAC, following are my quick, preliminary thoughts regarding a complicated topic.

As a classical music lover, my criterion for assessing the quality of a recording is fairly simple: Does the recording sound like what I remember hearing in the symphony hall or opera house (where no electronics are used – no sound reinforcement – the sound is 100% natural). I recognize that memory is fallible – nonetheless this is my criterion. And I want any deviations from a live performance to sound pleasant vs. unpleasant.

I’m not interested in a comparison test procedure that uses PC software. I’d want to know the quality of the original recording, and mastering. (Garbage-in/garbage-out.) I’d want to know whether or not the test files were manipulated by software, and if so what artifacts of the software might affect the sound quality. And I’d want to know whether or not these files have any relevance to consumer deliverables that I might buy (e.g., CD, SACD, Blu-ray, high-res FLAC or DSD download). (I’m not going to use PC software to play music – so I don’t care what PC software sounds like.) I am only willing to listen to music that I like (which eliminates some tests I’ve read about). The good news is that I listen to music performed with no electronics involved, and I know what classical music sounds like performed live in its intended venue. (I have full season tickets to the symphony and opera, plus I attend several small-scale chamber performances every year. That’s how my ears are “calibrated”.)

A complicating factor in comparing different deliverables of the same recording is whether “bass management” or a “.1” channel is used to send low frequencies to the subwoofer. (IME these different methodologies affect the sound quality.)

Sound pressure level matching between different recordings is important when performing qualitative comparisons, and is difficult to guarantee. (I don’t own a sound level meter.)

Based on the equipment that I own, the ideal methodology would probably be to use two Oppo UDP-205 universal players, with each connected to different line-level inputs on a modern integrated amp that can switch inputs via the remote control. An SACD loaded into one UPD-205, and a 24/192 FLAC playing on the other UDP-205. However – at least today – I’m not willing to uninstall one of my UDP-205 and lug it into the room with the other UDP-205. The next best arrangement is probably to have an SACD in the disc tray, and have a 24/192 FLAC file of the same recording on a USB drive, and switch via the Oppo’s “Home” menu.

The good news for me is that recent performances of classical music are routinely captured and offered in hi-res audio. I’ll have to peruse my collection to see which recordings I have in SACD and 24/192 FLAC. I can think of only one: Reference Recordings RR-136 Saint-Saëns: Symphony No. 3 "Organ" (Kansas City Symphony). I just acquired “The Asia Tour” featuring Sir Simon Rattle and the Berliner Philharmoniker that came with SACDs and Blu-ray video (recorded in 24bit/96kHz PCM), but I’m having difficulty downloading the FLAC 24/96 files. (I could compare the audio quality of the SACD with the Blu-ray.)

My approach to listening-test-methodology would be to make informal listening impressions, where I control switching back and forth. I understand the concepts of ABX, and double-blind testing, but I’m not interested in spending time attempting to implement these methodologies.

Bottom line, making valid comparisons between DSD (e.g., SACD) and 24bit/192kHz PCM (e.g., delivered via FLAC download) is not a straightforward task.

FWIW, here’s my perspective. For the music that I like, there are many modern performances that were captured in hi-res audio (DSD or PCM) – and in a growing number of cases - Blu-ray (or Ultra HD Blu-ray) audio/video. Almost all modern performances were recorded in hi-res. My perspective is this: Why not buy the consumer deliverable that is in the same format that the recording was mastered in – vs. buy a version that was downgraded (and possibly transcoded) into 30+ year-old CD storage technology?

Perhaps over the next few days I’ll do some informal listening comparisons. For now, I can say that both SACD and 24/192 (and 24/96) FLAC are all capable of excellent audio quality. My favorite way to consume music is Blu-ray audio/video featuring hi-res surround-sound and HD video, followed by surround-sound SACD. (Ultra HD Blu-ray recordings are slowly becoming available for classical music, opera, and ballet.)
 
Last edited:
2.8MSa/s, 1-bit DSD gives 2.8Mb/s. Per channel, so 5.6Mb/s for stereo.

192kSa/s, 24-bit PCM gives 4.608Mb/s, so 9.2Mb/s for stereo. Assuming a 40% reduction by FLAC, this would give 5.53Mb/s, so, yes, they'd be comparable file sizes/stream rates.

My gut feel would be to compare the uncompressed PCM stream against the DSD stream, but it depends what you are trying to compare; PCM vs DSD representation as a concept, or the relative performance of the stored/streamed formats.

I've spent too many hours doing formal DRT/DAM tests whilst developing 16k CVSD, LPC-55 and GSM voice codecs to want to do more formal testing... But at least we've established a baseline to allow some vaguely meaningful results.

I'll stop being a technical pedant, and let the golden ears vote.
Thank you cpt, you did help to narrow down the spectrum of variables to a more manageable or meaningful range. Appreciated :)
 
Wow Robert, thanks for your input and hope to hear your report. No hurry, I can wait a coupe of weeks if that's what you need.

Can you also share with me some of those classical recording titles you cited above?
 
Can you also share with me some of those classical recording titles you cited above?

Here’s a few Bu-ray audio/video classical music box sets that I recommend:



41%2BEYqMSRUL.jpg




71Xp1l2S4oL._SX522_.jpg




51Qnrm9bKEL._SX342_.jpg




51a2QA8JC-L._SX342_.jpg





61Lk0bYLR7L._SY445_.jpg


81cFPd8JWTL._SX385_.jpg


61LD9vqdP3L._SX342_.jpg
 
First of all, you're not really comparing the same thing. DSD is a digital format, with FLAC is a wrapper of the PCM format.

So, is DSD better than PCM? It depends upon the specifics.

Understand that most DSD / SACD releases are first captured and edited using DXD - which is 352/24 PCM. :)

"DXD was initially developed for the Merging Pyramix workstation and introduced together with their Sphynx 2, AD/DA converter in 2004. This combination meant that it was possible to record and edit directly in DXD, and that the sample only converts to DSD once before publishing to SACD."
 
Some 'PCM' DACs use 'DSD-like' techniques for reconstruction...

The only DAC's that aren't using a delta-sigma process to physically create an analog signal are R2R DACS. 99% of what's on the market uses "oversampling" or delta-sigma internally. In fact, a lot of work and attention was paid to this initial conversion process. Everyone had a term for how they did the same basic thing, MASH for example was just a name for "speical noise-shaping" applied to the signal during the conversion.

First of all, you're not really comparing the same thing. DSD is a digital format, with FLAC is a wrapper of the PCM format.

Incorrect. WAV is a wrapper/container format for PCM. FLAC itself is a compressed format that's designed for high-ratio lossless compression of PCM. You get WAV out of FLAC, you make FLAC from WAV; but FLAC is not a wrapper. A wrapper implies the actual bitstream is physically in the file. FLAC bitstream is not a PCM bitstream.

One disadvantage of SACDs is that the SACD layer cannot be played on a PC. (The CD layer of a hybrid SACD disk can be played on a PC.)

I'll give you half-credit for this, but that's also wrong. You can extract the SACD layer off a disc with a PS3 or a few of the modified players. You can play back this data in DSD with the right software and a DSD capable DAC. I mean, you're not supposed to be able to do this...it's not something anyone can just up and do. But there's a difference between "cannot" and "can, with difficulty". Proof is in this morning's playlist.

Understand that most DSD / SACD releases are first captured and edited using DXD - which is 352/24 PCM.

The sad fact is you're half-true on this. A lot of SACD's, especially early releases, went through an intermediate PCM stage. But most content that was processed in DXD was converted to that from a DSD capture. Some SACD's we suspect were made with even lower rate PCM....like there are claims that one of the Santana SACDs was made from 24/96 PCM. I believe one of the Queen SACD's I have even states it was made from the PCM sources. That's part of what makes it a crapshoot. You can usually tell these releases because they've been dynamically compressed and approach 0dbSACD quite often. A release that's straight DSD from the tape usually has a much lower average volume.

The main issue is any comparisons between PCM and DSD/Delta-Sigma are like comparing apples and oranges. They're both fruit, they both grow on trees; but that's about where the similarities end. SACD was a solution to a "problem" while attempting to be an upgrade. It was designed to just skip PCM entirely and play-back a high-rate delta-sigma stream directly; avoiding the extra PCM conversion process. The issue is that by the time SACD came out...the general market was going the other way with MP3s and iTunes. They couldn't give a single care about the quality...they just wanted music and they didn't care how it sounded. Very few people actually sit and listen to music to listen to music...it's usually background. I'll admit, I don't dedicate time to just music...I'm usually on the PC while listening. Let's face it....guys like us...that actually care...we're in the minority.

Anyway...comparing data rates is useless for the same reason. When you get down to it; the data rate (just the rate of bits) for 2-channel DSD matches uses the same amount of bits as a 176.4khz/16-bit PCM file; so it's going to be smaller than a 192khz/24-bit. Both PCM and DSD have advantages, both have disadvantages.

192kSa/s, 24-bit PCM gives 4.608Mb/s, so 9.2Mb/s for stereo. Assuming a 40% reduction by FLAC, this would give 5.53Mb/s, so, yes, they'd be comparable file sizes/stream rates.

If you want to talk after FLAC compression...that's a different game. 192khz FLAC compresses pretty well since there's really not much of anything in the ultrasonics. The compression ratio of FLAC depends on the complexity. 40% is about right for 192/24. I did comparisons once with Hotel California ripped from the original DVD-Audio. Most of the files came in between 40 to 50% reduction in FLAC...it's total size was a gig and a half. I then converted it to DSD. I believe it did come out a bit more than the FLAC; but DSD does have lossless compression. Using DST compression knocked the entire thing down to around 600MB. WavePack has added a DSD compression option that achieves very similar rates.

DSD compresses down a WHOLE lot better than PCM.

Ultimately the quality of either is going to be what went in to it; bad mastering is bad mastering. A really good transfer of a master at 192khz will probably sound no different than DSD. The rest of it depends on stuff like the qualify of your DAC and the process in which your DAC converts PCM to delta-sigma internally. One of the things that really threw a wrench in to the whole idea of SACD taking over is they started using multi-bit delta-sigma for internal DAC operations as opposed to the original 1-bit; which from what I've been told would make the conversion process "easier".

It still boils down to the same variables...the sources...your equipment...it's almost as if the actual format itself doesn't make that much of a difference.

I had a discussion with a fellow ham-op about this. Digital signal sampling is pretty much his game...I don't know what he did other than it involved image acquisition over RF for alphabet agencies and the DoD....but he did admit that the "1 dimensional world of audio sampling" wasn't his his field of expertise and was often more difficult than the 2D image world. I asked his opinion of delta-sigma; he summed it up by saying "in the world of sampling...sometimes what makes the least sense winds up working well."

The only thing you can do is get some equipment and try them out. See which one YOU like. We can sit around all day and say which is better...but at the end it's just a subjective opinion based on our setups and how our brains are processing the sound. What's good for the goose might not be good for the gander.

I care more about which release has better mastering moreso than it's format or resolution...but then you wind up like me with 6 different pressings/copies of Aja to compare.
 
Last edited:
Incorrect. WAV is a wrapper/container format for PCM. FLAC itself is a compressed format that's designed for high-ratio lossless compression of PCM. You get WAV out of FLAC, you make FLAC from WAV; but FLAC is not a wrapper. A wrapper implies the actual bitstream is physically in the file. FLAC bitstream is not a PCM bitstream.
Regardless, my point is that DSD and FLAC are not parallel concepts per thread title. Forget the semantics.

That would be DSD and PCM. Or .flac and .dsf or .dff


A lot of SACD's, especially early releases, went through an intermediate PCM stage. But most content that was processed in DXD was converted to that from a DSD capture.
And then necessarily converted back. PCM was always in the signal chain.
 
Last edited:
I'll give you half-credit for this, but that's also wrong. You can extract the SACD layer off a disc with a PS3 or a few of the modified players. You can play back this data in DSD with the right software and a DSD capable DAC. I mean, you're not supposed to be able to do this...it's not something anyone can just up and do. But there's a difference between "cannot" and "can, with difficulty". Proof is in this morning's playlist.

I’ve read descriptions of how SACDs can be copied via “a PS3 or a few of the modified players”. Based on what I’ve read, some of the “hacking” procedures for copying SACDs are not straightforward, and my assessment is that these procedures would mostly appeal to computer hobbyists. This is different from dropping an SACD into the PC disc tray and directly playing the SACD layer (which is what I meant when I said you can’t play the SACD layer on a PC).

Similarly, I’ve read that it’s possible – but not easy – to copy Blu-ray and Ultra HD Blu-ray discs.

Apparently, some people enjoy performing these copying “hacks” – i.e., it’s a hobby. To each their own. However, I question whether this saves time, when you consider the time to learn the procedure, install and configure the software, and copy all of the discs, and edit metadata for each recording (which is reportedly required for classical recordings in order to sort out composer, composition, conductor, orchestra, soloists, etc.).

Others (like me) would rather drop a disc into the tray of a universal player (e.g., Oppo UDP-205) and hit the Play button. (Or use folders to organize downloaded hi-res files (e.g., from HDTracks), and use the Oppo’s remote to navigate the folders.)

A problem with using an external DAC is finding one that will handle all digital music formats (CD, SACD, Blu-ray, Ultra HD Blu-ray, hi-res FLAC download, and hi-res DSD download) – and that someone who is not a computer hobbyist can easily install and use – without investing significant effort in transcoding their music library.

The music genre is a major factor in deciding what works best for each individual. For classical music, I listen to one symphony at a time, or one opera at a time, etc. I don’t play one 3-minute-long song from one album, followed by a different 3-minute song from a different album, etc. The concepts of playlists or shuffle play are not relevant for me. I see no benefit to selecting music from a tablet or smartphone, vs. getting out of the chair once an hour to retrieve a disc.

For people who want to access music from a tablet or smartphone, it seems to me the question is which makes more sense: ripping discs that they own, or subscribing to a premium streaming service (e.g., Spotify Premium to Tidal HiFi) that provides millions of recordings.

OTOH, for people who prioritize state-of-the-art-quality digitized music, hi-res discs (SACD, Blu-ray, Pure Audio Blu-ray, Ultra HD Blu-ray) represent an easy-to-use option.

Bottom line, different people enjoy different music, and enjoy the hobby of hi-fi differently.
 
Last edited:
Back
Top Bottom