Results 1 to 21 of 21
  1. #1

    ES9018, DSD to PCM or NOT ?

    Hi all,
    I saw a video from ESS Sale Office(China) recently, ESS engineer Mr. Zhang said that the 9018 isn't 1bit DSD DAC chip, and DSD to PCM conversion will be done in 9108.

    I also got other opposite informations from computeraudiophile.com about this. Mr. Michal who is the designer of Mytek, here are his words:
    "Ok,the way DSD is implemented in our DAC:
    DSD 1 bit> digital DSD 3 pole filter @ selectable 50,60 or 70kHz>6 bit output > 6 bit DAC. The filter, standard in all sacd players (typically analog) helps to filter out of band noise inherent n DSD. Such filtering requires multiplication which produces result in more bits than the 1bit input. It's actually in away better than 1 bit because all info is preserved but 6 bit conversion is more accurate than 1 bit. It's better to keep it at 6 bits than round it to 1 bit, for example. We are still talking about 64xFS stream of 6 bit data , NOT any PCM."

    According to Mr. Michal's descriptions, in my understand the ES9018 does 6-bit SDM and the DSD SR is still 2.8M(or 5.6M), no any PCM.

    The answers added to my confusion. Who is right?

    Thanks!

  2. #2
    To my knowledge, it's the other way around. Since the ESS chip is a delta sigma DAC, it will convert PCM to delta sigma.

    As for two-level SDM (DSD) vs multi-level SDM, the signal will stay SDM.

  3. #3
    My god, an OWN GOAL made by ESS China themselves ?

  4. #4
    Sophomore Member
    Join Date
    May 2012
    Location
    USA Virginia
    Posts
    305
    Since ESS has incorporated an ASRC in their chips as jitter-reducing solution (Time Domain Jitter Eliminator), they need to DSD trough PCM in order to use it.

  5. #5
    Disruptive Audio_ELF's Avatar
    Join Date
    Mar 2009
    Location
    England
    Posts
    11,055
    Blog Entries
    1
    I believe everything (both DSD and PCM) is converted to 6bit SDM that the DAC supports... That isn't the same as converting from PCM to DSD (or vice versa).
    Eloise
    ---
    ...in my opinion / experience...
    While I agree "Everything may matter" working out what actually affects the sound is a trickier thing.
    And I agree "Trust your ears" but equally don't allow them to fool you - trust them with a bit of skepticism.
    keep your mind open... But mind your brain doesn't fall out.

  6. #6
    Quote Originally Posted by Audio_ELF View Post
    ... That isn't the same as converting from PCM to DSD (or vice versa).
    It's really about how one defines it. It's all PCM really, even 1-bit two level (DSD). PCM is much more than the 2's compliment integer based weighted consumer delivery format that we associate. All Delta-Sigma Modulators output a flavor of PCM, single bit or multi-bit, depending how they are configured.

    To me, the issue when saying something about "converting from PCM to DSD (or vice versa)" is whether the sampling rate has changed, and whether decimation and low-pass filtering was required. That's when the losses can occur, especially down-sampling.

  7. #7
    Senior Member barrows's Avatar
    Join Date
    Jul 2008
    Location
    9,300', above Boulder, CO
    Posts
    3,566
    Quote Originally Posted by tailspn View Post
    It's really about how one defines it. It's all PCM really, even 1-bit two level (DSD). PCM is much more than the 2's compliment integer based weighted consumer delivery format that we associate. All Delta-Sigma Modulators output a flavor of PCM, single bit or multi-bit, depending how they are configured.

    To me, the issue when saying something about "converting from PCM to DSD (or vice versa)" is whether the sampling rate has changed, and whether decimation and low-pass filtering was required. That's when the losses can occur, especially down-sampling.
    And the ESS 9018 runs at high enough rates that no down sampling is necessary with 64 or 128 DSD input. Additionally, to Sonic, while the ESS 9018 does have an internal ASRC and DPLL, these elements of the DSP section become inactive if one chooses to supply a synchronous masterclock to the ESS chip: this is the way Ayre, and Wavelength Audio use the chip.
    EVERYTHING MATTERS

    Sonore µRendu with Sonore Signature Power Supply--Buffalo IIIse/Legato 3/Sonore Async USB, or Buffalo II w CCHD-575 XO/NTD1 v.4/Engineered USB interface-DIY Ncore 400 Stereo-Focus Audio FS888 cap mod-DIY Parallel AC Filter-Cardas Clear power cables-Cardas Clear audio cables-Curious USB Link-Synergistic Black Fuses

    Design/Build Consultant with Simple Design/Sonore

  8. #8
    Sophomore Member
    Join Date
    May 2012
    Location
    USA Virginia
    Posts
    305
    Quote Originally Posted by tailspn View Post
    It's really about how one defines it. It's all PCM really, even 1-bit two level (DSD). PCM is much more than the 2's compliment integer based weighted consumer delivery format that we associate.
    Careful, when I said that that DSD is just Sony brand name of 1 bit PCM, the believers in DSD magic said that I don't know what I am talking about

  9. #9
    Quote Originally Posted by SoNic67 View Post
    Careful, when I said that that DSD is just Sony brand name of 1 bit PCM, the believers in DSD magic said that I don't know what I am talking about
    I don't believe in any DSD magic, but DSD most definitely is not the same as 1 bit PCM. Listen to a 1 bit PCM signal sometime. It has > 100% distortion, and zero dynamic range.

    Now I will agree that today most modern PCM of DSD converters are variations on SDM, and that the primary difference with DSD is the decimation filtering which it doesn't undergo. But DSD is a differential signal which needs to be integrated to look like analog, and PCM is a decimated stair-steps which need to be steeply low-pass filtered to look like analog.

    I suppose that this argument is really about semantics in many ways. But it is far more accurate to say that both DSD and PCM are subsets of SDM than that DSD is a subset of PCM. It may be a subset of the process which SDM creates PCM, but the end modulation pattern is a very different animal without the benefit of the decimation filtering.

  10. #10
    Quote Originally Posted by TarnishedEars View Post
    Now I will agree that today most modern PCM of DSD converters are variations on SDM, and that the primary difference with DSD is the decimation filtering which it doesn't undergo. But DSD is a differential signal which needs to be integrated to look like analog, and PCM is a decimated stair-steps which need to be steeply low-pass filtered to look like analog.

    I suppose that this argument is really about semantics in many ways. But it is far more accurate to say that both DSD and PCM are subsets of SDM than that DSD is a subset of PCM. It may be a subset of the process which SDM creates PCM, but the end modulation pattern is a very different animal without the benefit of the decimation filtering.
    For me, you hit the nail directly. It is about semantics, and term definition. And, there seems to be no industry consensus in this corner of the audio A/D-D/A converter biz.

    The general definition of PCM is "short for pulse code modulation, a sampling technique for digitizing analog signals, especially audio signals", without further clarification of specific formats. But PCM was and is used broadly in the telecom industry, long before the advent of the CD, to describe various transmission formats, some 1-bit two level.

    I personally feel uncomfortable using the terms Delta-Sigma Modulator or Sigma-Delta Modulator (DSM/SDM) to refer to data format families, as they refer to a thing, a processing element, not the format it produces. And depending how those modulators are configured, they can produce an infinite variety of 1-bit and multi-bit outputs, including the 2's compliment integer weighted multi-bit we think of as PCM. The only reason decimation is required is to lower the sampling rate frequency to make it deliverable.

    The term DSD is clearly a marketing term for 1-bit two level modulation, which also existed long before its being coined. So except for the consumer market its aimed at, it really isn't a useful descriptor.

    We agree Ears, I'm just looking for the appropriate terms.
    Last edited by tailspn; 07-02-2013 at 10:55 AM.

  11. #11
    Quote Originally Posted by tailspn View Post
    The term DSD is clearly a marketing term for 1-bit two level modulation, which also existed long before its being coined. So except for the consumer market its aimed at, it really isn't a useful descriptor.
    What's wrong with the current term Direct Stream Digital?

    In my opinion the marketing term is a very apt descriptor of the technology behind the format which captures direct digital stream produced by a delta sigma modulator.

  12. #12
    Quote Originally Posted by Hiro View Post
    What's wrong with the current term Direct Stream Digital?
    Nothing's wrong with it. My question is what's the correct term for the output formats that the various Delta-Sigma Modulator configurations output, which include 1-bit two level (DSD). I submit all of them are in the family generically called PCM.

  13. #13
    Quote Originally Posted by tailspn View Post
    My question is what's the correct term for the output formats that the various Delta-Sigma Modulator configurations output, which include 1-bit two level (DSD).
    If you're capturing a direct digital stream from a delta sigma modulator the correct term for such formats is Direct Stream Digital.

  14. #14
    Señor Member Jud's Avatar
    Join Date
    Mar 2010
    Location
    Born in Bethlehem (PA)
    Posts
    13,034
    Blog Entries
    7
    Quote Originally Posted by Hiro View Post
    If you're capturing a direct digital stream from a delta sigma modulator the correct term for such formats is Direct Stream Digital.
    Thanks Hiro, I think tailspn really needed some help from an expert on this topic.
    The fairest thing we can experience is the mysterious. It is the fundamental emotion which stands at the cradle of true art and true science. - Einstein
    One never knows, do one? - Fats Waller

    MacBook Pro, A+ -> 2 Mapleshade Clearlink USB Plus flanking Regen (powered by LPS-1) -> iFi micro-iDSD -> Spectral DMC-12 & DMA-150 -> Vandersteen 3A Signature.

  15. #15
    Quote Originally Posted by Jud View Post
    Thanks Hiro, I think tailspn really needed some help from an expert on this topic.
    No problem.

  16. #16
    Señor Member Jud's Avatar
    Join Date
    Mar 2010
    Location
    Born in Bethlehem (PA)
    Posts
    13,034
    Blog Entries
    7
    Quote Originally Posted by Hiro View Post
    No problem.
    OK, gotta give you that one. :-)
    The fairest thing we can experience is the mysterious. It is the fundamental emotion which stands at the cradle of true art and true science. - Einstein
    One never knows, do one? - Fats Waller

    MacBook Pro, A+ -> 2 Mapleshade Clearlink USB Plus flanking Regen (powered by LPS-1) -> iFi micro-iDSD -> Spectral DMC-12 & DMA-150 -> Vandersteen 3A Signature.

  17. #17
    Senior Member barrows's Avatar
    Join Date
    Jul 2008
    Location
    9,300', above Boulder, CO
    Posts
    3,566
    Quote Originally Posted by Jud View Post
    OK, gotta give you that one. :-)
    LOL, jud! Way to go… Time for a nice bike ride.
    EVERYTHING MATTERS

    Sonore µRendu with Sonore Signature Power Supply--Buffalo IIIse/Legato 3/Sonore Async USB, or Buffalo II w CCHD-575 XO/NTD1 v.4/Engineered USB interface-DIY Ncore 400 Stereo-Focus Audio FS888 cap mod-DIY Parallel AC Filter-Cardas Clear power cables-Cardas Clear audio cables-Curious USB Link-Synergistic Black Fuses

    Design/Build Consultant with Simple Design/Sonore

  18. #18
    Masters Level Member
    Join Date
    Apr 2010
    Location
    Finland
    Posts
    7,594
    Blog Entries
    13
    Quote Originally Posted by tailspn View Post
    And depending how those modulators are configured, they can produce an infinite variety of 1-bit and multi-bit outputs, including the 2's compliment integer weighted multi-bit we think of as PCM.
    Well, in most cases, delta-sigma converters specifically don't use two's complement binary coding, but they use variant of unary coding instead (commonly also called "thermometer code").

    So in SDM, bits don't have the 2n weighting, but each bit has equal weight instead. By controlling how the bits (converters) are used in both parallel and in time dimensions, you have enormous amount of flexibility compared to old fashioned PCM where each bit and sample have a strict pre-defined relationship.

    Nothing's wrong with it. My question is what's the correct term for the output formats that the various Delta-Sigma Modulator configurations output, which include 1-bit two level (DSD). I submit all of them are in the family generically called PCM.
    It is "PDM" = Pulse Density Modulation, or "PWM" = Pulse Width Modulation. These two are closely related. You can also create both without delta-sigma modulator in analog domain. So for clarity you can see "SDM" = Sigma-Delta Modulation or "DSM" used to emphasize variant that has been specifically produced by a delta-sigma modulator.

    Mola-Mola DAC talks about "PWM Noise Shaper" which practically means delta-sigma modulator...

    PCM is a different beast and separate from the two above, even if those two use more than two absolute states to represent a pulse.

    I still believe that DSD originally mean "Direct Sigma-Delta", but someone in marketing department decided that consumers wouldn't understand at all what "Sigma-Delta" means there, so instead they invented "Stream Digital" for the same letters... But that would need some Sony/Philips insider to tell the truth.
    Signalyst - http://www.signalyst.com
    Developer of HQPlayer

  19. #19
    Quote Originally Posted by Miska View Post
    Mola-Mola DAC talks about "PWM Noise Shaper" which practically means delta-sigma modulator...

    PCM is a different beast and separate from the two above, even if those two use more than two absolute states to represent a pulse.
    Well, that's what's so confusing to the unwashed like me. I had the privilege of spending several hours with Bruno a few weeks ago, and both he and Andreas, IIRC, spoke about all of these modulation data subsets falling under PCM. Like Coded Modulation was a superset of all the possible modulation schemes. There are many high level papers available detailing that author's point of view about modulation theories and practices, but not one I can find that even defines the terms so they make sense to me.

  20. #20
    Masters Level Member
    Join Date
    Apr 2010
    Location
    Finland
    Posts
    7,594
    Blog Entries
    13
    Quote Originally Posted by tailspn View Post
    Well, that's what's so confusing to the unwashed like me. I had the privilege of spending several hours with Bruno a few weeks ago, and both he and Andreas, IIRC, spoke about all of these modulation data subsets falling under PCM. Like Coded Modulation was a superset of all the possible modulation schemes. There are many high level papers available detailing that author's point of view about modulation theories and practices, but not one I can find that even defines the terms so they make sense to me.
    Well, if you generalize to a level that having any number means having PCM, then yes. Maybe they also consider rational and even irrational or complex numbers being PCM. But maybe mathematicians wouldn't agree about categorizing all numbers as "PCM".

    I wouldn't use such wild definition. Maybe something like "a system where discrete sample value X at discrete time point Y fully describes a discrete signal state". (sounds like lot of BS still...)

    I don't think the Wikipedia explanations are bad either:
    Pulse-code modulation - Wikipedia, the free encyclopedia
    vs
    Pulse-density modulation - Wikipedia, the free encyclopedia
    and since the adjacent equal states are usually held, it turns into
    Pulse-width modulation - Wikipedia, the free encyclopedia

    Now with 1-bit, your pulses would have equivalent levels of
    1 = +2V
    0 = -2V

    But nothing says that you need to stick to two levels, so instead
    [1 1] = +2V
    [0 1] or [1 0] = 0V
    [0 0] = -2V
    Now you have a three-level SDM already.
    Signalyst - http://www.signalyst.com
    Developer of HQPlayer

  21. #21
    You should write a book!