SACD - gapless playback [Archive]

Ghitulescu

13th January 2015, 09:20

The first thing to do is to copy the Sacd with a PS3 with Sacd Extract which makes an ISO.
http://code.google.com/p/sacd-ripper/downloads/list
Instructions ;
http://code.google.com/p/sacd-ripper/source/browse/trunk/readme
Then you extract the dff files from the ISO with scarletbook.exe.
http://dl.dropbox.com/u/1370915/Audio/scarletbook.rar
Then you can convert Dsd to PCM (2.0 and 5.1) with several programs like Weiss Saracon, KORG AudioGate, Philips ProTECH Audio Format Converter, Pyramix Virtual Studio, maybe DiscWelder and maybe Foobar with foo_input_sacd.dll ;
http://sourceforge.net/projects/sacddecoder/files/
More informations there ;
http://www.stereo.net.au/forums/showthread.php/36026-SACD-DFF-file-extraction-backups

I wonder whether there is a guide for gapless discs (live, electronic music etc.), sort of BIN+CUE....

SeeMoreDigital

13th January 2015, 10:17

I wonder whether there is a guide for gapless discs (live, electronic music etc.), sort of BIN+CUE....It's actually not too difficult to create your own .cue files ;)

pandy

13th January 2015, 10:21

Btw - conversion from DSD to PCM is always lossy...

Ghitulescu

13th January 2015, 10:28

It's actually not too difficult to create your own .cue files ;)

The point was different.
Gapless discs, if compressed, will always be problematic to "rejoin", because of the buffer size differences (read padding) on both coding and decoding steps.

Yes, I know that DSD to PCM are lossy, however I am confident that the losses are not that great, in particular with titles that are available only as SACD (and are, according to Sony, processed in PCM then converted to DSD). On the other hand, HiRes FLACs or WAVs can be played on most of my players.

SeeMoreDigital

13th January 2015, 10:36

So you want to take individual tracks from different SACD/DSD sources and join them all together into one long track. And create a cue file that allows you to navigate to each individual track?

pandy

13th January 2015, 10:58

The point was different.
Gapless discs, if compressed, will always be problematic to "rejoin", because of the buffer size differences (read padding) on both coding and decoding steps.

Well - real GAPless playback is Player functionality and it is not related to file format whatsoever...

Yes, I know that DSD to PCM are lossy, however I am confident that the losses are not that great, in particular with titles that are available only as SACD (and are, according to Sony, processed in PCM then converted to DSD). On the other hand, HiRes FLACs or WAVs can be played on most of my players.

I would say that decimation filter reconstructing DSD (necessary step from DSD i.e. Pulse (1 bit) Density Modulation to PCM (multi bit) ) is directly responsible for sound characteristic (and as such it can be FIR or IIR and can have various unique flavors). Not sure if those filters (i mean software implementation) are "analog" or "digital" but i assume that from subjective perspective they can be perceived as "worse" or "better" (i.e. lossy or not).

feisty2

13th January 2015, 12:59

DSD...? never heard of it before, it's a commercial name for PDM according to Google
why would any one store audio in this format?

SeeMoreDigital

13th January 2015, 16:06

DSD...? never heard of it before, it's a commercial name for PDM according to Google
why would any one store audio in this format?In this case, DSD stands for Direct Stream Digital (http://en.wikipedia.org/wiki/Direct_Stream_Digital). The audio stream format has been used on Super Audio Compact Disc's (SACD) since 1999 and is still being used...

Indeed, I 'pre-ordered' some new SACD disc releases yesterday ;)

Music Fan

13th January 2015, 19:22

The point was different.
Gapless discs, if compressed, will always be problematic to "rejoin", because of the buffer size differences (read padding) on both coding and decoding steps.
Interesting but succint, could you give more explanations about this ?
Are you sure gapless discs are problematic with the method I gave ?

Ghitulescu

13th January 2015, 20:17

So you want to take individual tracks from different SACD/DSD sources and join them all together into one long track. And create a cue file that allows you to navigate to each individual track?
Yes, exactly.
But not exactly :) I want to have a huge file that should have cue points or indexes. I don't want to join files for the reason I already gave and is repeated below.
Well - real GAPless playback is Player functionality and it is not related to file format whatsoever...
Well, even if a gapless-aware player can play the tracks with pause, if the tracks themselves are padded with NULL samples (to round the last encoding frame to its full size) the playback is still pausing (NULL samples are silence in audio). In short, like a file of 513 Bytes that has to be stored on two 512B sectors (while the rest of the 511 Bytes are padded with NULLs).
So while not technically an issue of the format, it still is related to the compression algorithm, and since a compression algorithm is related to a format (like MPEG to .MPG, MPEG1 layer 3 to .MP3 etc etc).

I might be worring too much, since I don't know the frame sizes for FLAC nor the indexing system of the SACD.

I hope it's clear now. If it's not possible, then it's Ok.

Thanks for input.

SeeMoreDigital

13th January 2015, 21:06

Ghitulescu,

As far as I'm aware Foobar offers a few DSD tools that might be able to help you. Not that I use it myself...

Sufficed to say, if you're project is related to any other audio formats types, then it's outside the scope of this topic.

Cheers

pandy

13th January 2015, 21:23

Yes, exactly.
But not exactly :) I want to have a huge file that should have cue points or indexes. I don't want to join files for the reason I already gave and is repeated below.

Well, even if a gapless-aware player can play the tracks with no pause, if the tracks themselves are padded with NULL samples (to round the last encoding frame to its full size) the playback is still pausing (NULL samples are silence in audio). In short, like a file of 513 Bytes that has to be stored on two 512B sectors (while the rest of the 511 Bytes are padded with NULLs).
So while not technically an issue of the format, it still is related to the compression algorithm, and since a compression algorithm is related to a format (like MPEG to .MPG, MPEG1 layer 3 to .MP3 etc etc).

I might be worring too much, since I don't know the frame sizes for FLAC nor the indexing system of the SACD.

I hope it's clear now. If it's not possible, then it's Ok.

Thanks for input.

Well assume that payer need to be capable to decode 2 data streams at the same time and provide gapless fading between 2 data streams (trough fading/mixing) - so DSP need to be at least 2x times faster.

But perhaps my impression of gapless playback is different than Yours.

Sample padding/aligning seem to be less important.

pandy

13th January 2015, 21:27

DSD...? never heard of it before, it's a commercial name for PDM according to Google
why would any one store audio in this format?

Marketing... hype... audiophilism...

Concept of 1 bit delta sigma stream with strong noiseshaper sounds odd but lot of people believe that this format is superior to DVD Audio...

Ghitulescu

13th January 2015, 22:08

Well assume that payer need to be capable to decode 2 data streams at the same time and provide gapless fading between 2 data streams (trough fading/mixing) - so DSP need to be at least 2x times faster.

But perhaps my impression of gapless playback is different than Yours.

Sample padding/aligning seem to be less important.

Actually not.
The player should work like Nero Burning Rom(e), using a buffer, and concatenate the files seamlessly.

Padding is very important.

pandy

14th January 2015, 10:11

Actually not.
The player should work like Nero Burning Rom(e), using a buffer, and concatenate the files seamlessly.

Padding is very important.

In random mode you need to have capability to decode 2 streams and to perform fading between them - padding may help but real gapless playback assume that concatenation points (splices) overlap (at least partially). Buffering may work but can't provide random access and latency is unavoidable...

Music Fan

14th January 2015, 12:06

I'm not sure to understand what you mean ; is there a way (something to do at the authoring step or when burning the Sacd) to avoid the "pop" (or "glitch") sound that we hear between some tracks on some Sacd players ?
I have 3 Sacd players and the only one that does not have this problem is the Oppo. I found this explanation on Oppo's website concerning a firmware ;
-Some users have reported hearing pops or crackles between SACD tracks.
This issue happens to certain discs that have data frames that are not aligned with track boundaries. This version removes such pops or crackles.

What does exactly "track boundaries" mean ? Can we change it without conversion in pcm and re-make the Sacd (with Super Author for example) to avoid these "pop" sounds on all players and not only on Oppo (and maybe others) that solved the problem with a firmware ?

And I'm astonished you talk about Nero Burning Rom because I believed it was not Sacd compatible, Imgburn or UltraIso are generally used.

Ghitulescu

14th January 2015, 14:39

What does exactly "track boundaries" mean ? Can we change it without conversion in pcm and re-make the Sacd (with Super Author for example) to avoid these "pop" sounds on all players and not only on Oppo (and maybe others) that solved the problem with a firmware ?
That's exactly what I meant.
Since DSD is a stream of bits, and SACD/DSD does not allot a fixed number of bits/bytes (it depends on the music) ie VBR, but the SACD sector is 2048 bytes, I am not sure how the track boundaries are set (they should be at sector boundaries - but this does not correspond to a fixed/foreseable place in the stream).

On the other hand, PCM has a fixed bitrate (CBR) and it's easier to "scrub"/cue to a position, knowing any of size and time.

And I'm astonished you talk about Nero Burning Rom because I believed it was not Sacd compatible, Imgburn or UltraIso are generally used.

I was not talking about SACD in relationship with Nero, just mentioning Nero for its ability to create gapless tracks from separate files. I never used ImgBurn for audio.

Ghitulescu

14th January 2015, 14:46

In random mode you need to have capability to decode 2 streams and to perform fading between them - padding may help but real gapless playback assume that concatenation points (splices) overlap (at least partially). Buffering may work but can't provide random access and latency is unavoidable...

I realised I was misunderstood.

I don't want padding.

I only explained why I don't want padding.

Gapless playback does not require two streams. It simply requires that once a stream ends, the next one should be in line to be processed. No player I know avoids a buffer - therefore indeed the content of two separate streams coexist temporarily, but the files themselves should have been processed one at a time.

This is how all hardware players (should) work. CD players work this way, if pure audio players (not MP3 players with computer drives)
Now, if the tracks/files are gapless, but the player pauses each time it loads the next one, this is IMHO due to two main causes
1. track boundaries (VBR) do not match sector boundaries - this way the player disregards the padded real bits and starts with the next track (which should have included the discharged bits)
2. track boundaries match sector boundaries but the player is programmed to read the tracks sequentially as tracks, and not to process the stream as a stream (ie ignoring indexes and stuff - but updating the information on display like TRACK 03 TIME 15:45.73)

I think the first case is the most common case.

Ghitulescu

14th January 2015, 15:07

In random mode you need to have ...

Gapless playback does not require two streams. It simply requires that once a stream ends, the next one should be in line to be processed. No player I know avoids a buffer - therefore indeed the content of two separate streams may and will coexist temporarily, but the files themselves should have been processed one at a time.

Live albums (or electronic music like JM Jarre, M Oldfield etc) are not made to be played in random mode, as they tell usually a story.
Yes, I know two (yes, only two) movies whose reels are designed to be played in a random order and still tell a coherent story. But these movies are the exception.

Music Fan

15th January 2015, 12:48

1. track boundaries (VBR) do not match sector boundaries - this way the player disregards the padded real bits and starts with the next track (which should have included the discharged bits)
Ok, and do you believe there is anything feasible to make match track boundaries with sector boundaries, keeping Dsd as is ?
How to know if they match before exporting, burning, and listening Sacd ?
I'd like to avoid conversion in PCM, otherwise remake a Sacd would be a nonsense (after conversion, I would let PCM in PCM and make a dvd-audio).

And I wonder how did Oppo to remove these "pop" sounds, because I guess that if it can be done in real time by a standalone player, it can be done with some audio tools on a pc, but I may be wrong.

Ghitulescu

15th January 2015, 13:37

Well, the way I knew a player should play discs (not files) has been listed a few replies above.
Files are always played differently, because they are differently organised.

The (original) question has been asked, because all the methods involve the step of "breaking" the ISO into tracks (id est DSD files) and then these are converted, individually, to PCM (Would it be possible to cvoncatenate these files like COPY /B X+Y, or they have a header and stuff?).

I was specifically asked if there is any method that would extract the whole music into a single file with indexes (or any other marks), and this to be converted to PCM.

For CD-DA there are at least two methods (BIN+CUE and Extract from sector X to sector Y).

I would not go the conversion if the music would have been available natively in PCM or if my player would have known SACD (actually the player knows but sends the SACD as bitstream and the AVR has no DSD codec; the analogue way is not possible).

Music Fan

15th January 2015, 13:57

the way I knew a player should play discs (not files) has been listed a few replies above.
What post exactly ? Because I don't see anything about the way Oppo players handle these pop sounds.

all the methods involve the step of "breaking" the ISO into tracks (id est DSD files) and then these are converted, individually, to PCM.
Ok, but this does not answer my questions. How to do it keeping Dsd as is, if possible ?
I recall that these pop sounds are not present on all Sacd's, and on some sacd's, this can be heard only between some tracks but not between all tracks.
I also noticed that some tracks make pop sounds on my Pioneer player and not on my Sony, and the reverse for other tracks with the same Sacd ! There is no rule, very strange.

Ghitulescu

17th January 2015, 22:59

My explanation was generic. Because I am not the developer of Oppo, nor of any other manufacturer. But this is how I would do it as a programmer.

And for the second part, I don't xxx need DSD, DFF, DSFF or whatever files. I don't xxx care about losses, or anything.

What I want is to convert my SACDs into PCM. The original amplifier was scraped and the new one has not 6-analogue and no DSD-decoder.

Now, could we please return to the original question - how to rip SACD seamlessly into a gapless format? Please notice the title was wrongly set by a moderator.

Music Fan

18th January 2015, 00:22

My question is the same than yours, except about the fact that I don't want to convert it in PCM.
Can the padding with NULL samples be done on DSD or only on PCM ?

These topics may help you ;
http://code.google.com/p/sacd-ripper/issues/detail?id=6#c30
http://www.computeraudiophile.com/f11-software/sacd-ripping-using-your-ps3-part-2-a-7495/index19.html#post114279
http://yabb.jriver.com/interact/index.php?PHPSESSID=4pv5ejv5m7udhl874n2683d8o1&topic=77883.msg531650#msg531650
http://www.computeraudiophile.com/f11-software/sacd_extract-dsf-output-problem-18114/

For my question, I wonder if a simple timecode change in the cue (created by Sacd Extract and used by Super Author) may help to make match track boundaries with sector boundaries to avoid padding (if not possible in Dsd or too complicated).

Ghitulescu

18th January 2015, 08:50

Ghitulescu

18th January 2015, 12:14

What post exactly ? Because I don't see anything about the way Oppo players handle these pop sounds.

Well, what I am after is to find a method to save the whole content in a single file and maybe some pointers to the track beginnings.

About OPPO I haev none of their items, I considered them too expensive (in Europe), but there is no problem in making the playback gapless (one only has to have some willingness), simply remove the padding NULLs from the last 2048 bytes of each file. I don't know whether Oppo does it this way, but this is the way I would do it myself.
-----------------------------
That was a post I forgot to send after the browser crashed. I thought I have posted it and did not checked up.

Music Fan

18th January 2015, 17:50

Ok, but you said that it was not possible in Dsd, thus I guess I can't remove the padding NULLs from the last 2048 bytes of each file if I don't make Dsd => Pcm conversion.

For the timecode change I suggested (to actually change the place of index), it may work if the precision of the Sacd's timecodes (frames ?) is more accurate than SACD's sectors (2048 bytes), to chose exactly the debut of a sector to put index, and I'm afraid it's not the case.
I believe that for cd's, the precision corresponds to a 75th of second, I don't know for Sacd's.

For your problem, it seems that some software dsd to pcm converters don't make noise (pop sound) between tracks, I read it in one of the topics I listed.
I also read that the pop sounds could come from a misinterpreted digital silence, because there are 2 ways to code silence and some converters understand only one the two, but I don't know if it only concerns PCM. If yes, this is probably not the good explanation because some players that don't convert Dsd into Pcm also make these pop sounds.
When Dsd is directly converted into analog, no pop sound should be heard, while it happens.

With Sacd Extract, you have the choice to extract each song separately (from the ISO) or the whole album in one file. When done in one file, you get the timecodes in a cue file.
You may need to extract twice : one time for the whole album to get timecodes in a cue file, the other time to get tracks separately to remove padding of each track (if pop sounds are still there after dsd to pcm conversion of the whole album). Then you append all pcm files and you use the whole album with the cue file (the songs will be shorter but probably too slightly to have to change timecodes).

Ghitulescu

19th January 2015, 10:33

Ok, but you said that it was not possible in Dsd, thus I guess I can't remove the padding NULLs from the last 2048 bytes of each file if I don't make Dsd => Pcm conversion.

True, this is what I said, assuming DSD files are bitstreams, although I've reckon that DSD (or the DFF or the DST or the others) might be compressed (like MLP and similar), and thus easier to detect the NULLs.
With Sacd Extract, you have the choice to extract each song separately (from the ISO) or the whole album in one file. When done in one file, you get the timecodes in a cue file.

I'll check this software to see what's able to do.

I also read that the pop sounds could come from a misinterpreted digital silence, because there are 2 ways to code silence and some converters understand only one the two, but I don't know if it only concerns PCM. If yes, this is probably not the good explanation because some players that don't convert Dsd into Pcm also make these pop sounds.
That is exactly what I want to avoid. Such clicks/pops/noises/even silences are extremely annoying in a live album.

If SACD Extractor can convert the whole album in a PCM format, and it doesn't creates pops/clicks/silence of its own, the problem is solved.

Music Fan

19th January 2015, 13:09

I don't believe it can make conversion, unless a new version added this option. You will have to use a software Dsd to Pcm converter.

DSD (or the DFF or the DST or the others) might be compressed (like MLP and similar), and thus easier to detect the NULLs.
The compression is named DST ;
http://en.wikipedia.org/wiki/Super_Audio_CD#DST
Mandatory for multichannel (5.0 or 5.1, both are possible but not together), optional for 2.0.
Dsd is stored in Dsf or Dff files (whatever compressed or not in DST).

But why are NULLs easier to detect when compressed ? And how to detect it ?

pandy

19th January 2015, 14:21

IMHO conversion DSD to PCM is unavoidable (DSD is recorded DAC stream so at some point it is analog signal - SACD/DSD DAC is nothing than pair of high precision switches constantly switching between to reference voltages for example +-5V to simplify comparison), DSD stream have constant speed and as such can be packetized in regular chunks of data ("sectors"), for reducing transitions frequency some lossless transformation may be applied ( like bit-flipping http://www.eecs.qmul.ac.uk/~josh/documents/ReissSandler-2001-AES111-sdmcompression.pdf ).

Music Fan

19th January 2015, 14:42

IMHO conversion DSD to PCM is unavoidable
Unavoidable for what ? To make some changes or in all cases ?
Some Dsd DAC make dsd to analog directly, without PCM step. That's why Dsd was invented.

http://media.soundonsound.com/sos/mar07/images/korgdstopology_l.jpg

http://www.ps3sacd.com/images/PCM_vs_DSD_480.gif

http://www.canadapromedia.com/Articles/sacd/pcm%20and%20dsd.gif

Ghitulescu

19th January 2015, 15:15

But why are NULLs easier to detect when compressed ? And how to detect it ?

Because the original uncompressed stream of bits has to be fit into an integer number of sectors, each 2048 bytes. To fit a VERY POSSIBLE non-2048 multiple stream into a 2048-multiple file, padding is added, voluntarily or not, usually NULLs. Although statistic noise may be used, the safest procedure would be to voluntarily add the padding NULLS otherwise risking that the statistic noise to contain the file marker (thus leading the decoder into thinking the "garbage" is sound). A track can only start at sector boundaries, like eg in FAT systems. And like in FAT, the rest of the sector, in particular where the sector was long :) (floppies) could contain garbage (IIRC one ancient copy prevention method used that "garbage" zone as a marker).

pandy

19th January 2015, 15:21

Unavoidable for what ? To make some changes or in all cases ?
Some Dsd DAC make dsd to analog directly, without PCM step. That's why Dsd was invented.

http://media.soundonsound.com/sos/mar07/images/korgdstopology_l.jpg

http://www.ps3sacd.com/images/PCM_vs_DSD_480.gif

http://www.canadapromedia.com/Articles/sacd/pcm%20and%20dsd.gif

SACD/DSD is not digital in terms of processing - this idealized analog representation - only just before conversion to voltage i.e. analog audio - you can't process DSD for example mixing or fading - all you can is cut - to perform fade such signal need to be converted to something more editable like PCM.

Pictures are misleading as normal processing need to be done anyway in PCM domain (and this is not showed on those pictures) - they are OK only if analog source is directly sampled by DSD modulator and such bitstream is used to create SACD by duplication - but i assume not many SACD is AAD - most of them use anyway high sampling/bitdepth PCM and DSD modulator is implemented as final stage of processing and as such DSD modulator implementation may vary (noise shaper characteristic) - at the end DSD bitstream need to be converter by mentioned pair of switches that switch between reference voltages (+-V) - analog signal quality directly depends from accuracy of those switches (so error introduced by theoretical and practical implementation is unavoidable).
But yes, DSD is recorded bitstream that need to be directly converted by switches to voltage and lowpass filtered to suppress noise.

(In other way marketing scam that can be overpass by 8 bit DAC with correct noise shaping and four times oversampling at a half of bitrate - i.e. DSD is waste of resources - but i understand that some recordings are available as high res only trough SACD).

Sorry for being completely OT.

Because the original uncompressed stream of bits has to be fit into an integer number of sectors, each 2048 bytes. To fit a VERY POSSIBLE non-2048 multiple stream into a 2048-multiple file, padding is added, voluntarily or not, usually NULLs. Although statistic noise may be used, the safest procedure would be to voluntarily add the padding NULLS otherwise risking that the statistic noise to contain the file marker (thus leading the decoder into thinking the "garbage" is sound). A track can only start at sector boundaries, like eg in FAT systems. And like in FAT, the rest of the sector, in particular where the sector was long :) (floppies) could contain garbage (IIRC one ancient copy prevention method used that "garbage" zone as a marker).

Well "Zero" in perfect life should look like 010101010101010101 - idle tone (but in real life this probably never occurs as small noise is usually injected to improve delta sigma stability).

Not sure if this explain "Zero" and click problems related to DSD (so maybe important from gapless playback):
https://code.google.com/p/dsf2flac/wiki/ClicksAtStartAndEnd - alternatively to proposed one - small noise i.e. dither (or rather digital representation of dither after DSD conversion) may be applied to digital signal (so instead of null, feed FIR with low level for example -120dBFS noise).

Music Fan

20th January 2015, 12:25

Because the original uncompressed stream of bits has to be fit into an integer number of sectors, each 2048 bytes. To fit a VERY POSSIBLE non-2048 multiple stream into a 2048-multiple file, padding is added, voluntarily or not, usually NULLs. Although statistic noise may be used, the safest procedure would be to voluntarily add the padding NULLS otherwise risking that the statistic noise to contain the file marker (thus leading the decoder into thinking the "garbage" is sound). A track can only start at sector boundaries, like eg in FAT systems. And like in FAT, the rest of the sector, in particular where the sector was long :) (floppies) could contain garbage (IIRC one ancient copy prevention method used that "garbage" zone as a marker).
Ok, but I believe that this is as true for compressed than for uncompressed Dsd.
Thus I still don't understand why it's easier to detect when compressed and how to detect the Nulls in a Dsd stream on pc.

Music Fan

20th January 2015, 12:57

you can't process DSD for example mixing or fading
I know, but that's not what I was talking about, I just said that Dsd could be converted in analog without PCM step between Dsd and analog.

In other way marketing scam that can be overpass by 8 bit DAC with correct noise shaping and four times oversampling at a half of bitrate - i.e. DSD is waste of resources -
I guess it's a joke :)
Aside the marketing reasons, Dsd has been invented to be closer to the original analog signal than PCM and to simplify the work of the DAC.
Here is what I understood from some readings about another aspect ;
PCM has a limitation that Dsd does not have : the number of samples by frequency. When you increase the sample frequency in PCM, you only gain more high frequencies (unuseful for most people), but you won't have more details or dynamic between 20 hz and 20 khz (but you can increase dynamic with 24 bit instead of 16 bit at the same sample frequency).
In Dsd, when you increase sample frequency (which is not possible on Sacd, fixed on Dsd64), you get more samples and details on all frequencies (on the whole bandwidth).
But this is off topic.

foxyshadis

20th January 2015, 13:43

PCM has a limitation that Dsd does not have : the number of samples by frequency. When you increase the sample frequency in PCM, you only gain more high frequencies (unuseful for most people), but you won't have more details or dynamic between 20 hz and 20 khz (but you can increase dynamic with 24 bit instead of 16 bit at the same sample frequency).
In Dsd, when you increase sample frequency (which is not possible on Sacd, fixed on Dsd64), you get more samples and details on all frequencies (on the whole bandwidth).
But this is off topic.

Which is rather pointless, because unless a DAC is completely broken, it'll perfectly reconstruct any frequencies below the Nyquist limit; that's the whole idea behind the Nyquist–Shannon sampling theorem and why even budget DACs sound so good. Adding more samples, whether PCM or DSD, just wastes space at that point.

Music Fan

20th January 2015, 13:53

Which is rather pointless, because unless a DAC is completely broken, it'll perfectly reconstruct any frequencies below the Nyquist limit; that's the whole idea behind the Nyquist–Shannon sampling theorem and why even budget DACs sound so good. Adding more samples, whether PCM or DSD, just wastes space at that point.
In this case, what's the difference between Dsd64 and Dsd128 ?

foxyshadis

20th January 2015, 14:44

In this case, what's the difference between Dsd64 and Dsd128 ?

I'd call it an effective way to part businesses and audiophiles from their money, but I'm rather the cynical sort.

Music Fan

20th January 2015, 15:48

Mmmh ... that's your right, but I can't content myself with this, I need a technical explanation.

Ghitulescu

20th January 2015, 18:38

Mmmh ... that's your right, but I can't content myself with this, I need a technical explanation.

IIRC there is no difference apart from the higher bitrate. Like there is no difference between PCM 16 and PCM 24 except for the bit depth.

IIRC higher bitrates are used in studio, and the bitrate of SACD was chosen to allow at least one hour of music on the DVD5. As foxyshadis correctly pointed out, the difference is only for the customers money - AFAIK there is no software to process music in bitstream (DSD), only in PCM (like ProTools). This is confirmed inter alia by Sony. That means that everything one would say about how good SACD would sound (over DVD-A) is purely bull-caca. Pure DSD streams, recorded with no processing (like Korg or Tascam) and played back on the same gear might be a different story.

I don't know any standalone that would play DSD at higher than SACD bitrates. But thsi may change with the 4k SACD :) :) :)

pandy

21st January 2015, 14:40

I know, but that's not what I was talking about, I just said that Dsd could be converted in analog without PCM step between Dsd and analog.

Well... i can imagine processing in DSD but this is mathematically complex and very challenging task...

I guess it's a joke :)
Aside the marketing reasons, Dsd has been invented to be closer to the original analog signal than PCM and to simplify the work of the DAC.
Here is what I understood from some readings about another aspect ;
PCM has a limitation that Dsd does not have : the number of samples by frequency. When you increase the sample frequency in PCM, you only gain more high frequencies (unuseful for most people), but you won't have more details or dynamic between 20 hz and 20 khz (but you can increase dynamic with 24 bit instead of 16 bit at the same sample frequency).
In Dsd, when you increase sample frequency (which is not possible on Sacd, fixed on Dsd64), you get more samples and details on all frequencies (on the whole bandwidth).
But this is off topic.

Sorry for OT

Nope it is no joke - 8 bit PCM, with 4 times oversampling (i.e. 176.4/192kHz) and correct noiseshaping will provide better objective audio parameters than DSD with half of required bitrate - this is scientific fact...

Not sure what are You talking but in CD Audio (44.1kHz 16bit) you may have 144dB dynamics with correct noise shaping in reference to human auditory system but frequency range is limited to approx 22kHz and some people complain on this.
That's why higher sampling rates systems was created even is majority of population don't perceive difference between 44.1 and 48kHz.
Difference between DSD and PCM is simple - DSD use one bit and very aggressive noiseshaping and 64 times oversampling, PCM use between 16 and 24 bits and between NOS - 4 time oversampling.

Noiseshaping is optional and in case of 192kHz 24b it can probably provide more than 250dB SNR from human auditory perspective.

Noiseshaping is key to understand DSD and PCM and it is applicable to both systems.

As i can imagine simple linear processing natively in DSD domain then anything more complex in unavoidable way need to be done in PCM domain and reality of DSD recording is that most of them are created from PCM sources.

Not sure what is more close to analog signal series of 1 and 0 or values in range +-32767 - however i can't imagine any natural sound produced as series (even fast repeated) of 0 and 1 (or rather -1, +1).

Absolute silence is coded in PCM as 0, there is no such thing as absolute silence in DSD and if there will be something like silence in DSD then it will look like 0101010101010101010101010101010101010101 or rather -1+1-1+1-1+1-1+1-1+1-1+1-1+1-1+1-1+1-1+1-1+1 tell me which one of methods is more close to analog sound.

Use SoX @sox --multi-threaded --buffer 131072 -S -D -V %1 %1-output-p16d.wav rate -v -s -I 44100.0 gain -n -3.0103 dither -f improved-e-weighted -p 16
if You doubt that noiseshaping is not applicable to CD...