Connect with us

SACD / DVDA vs PCM

Discussion in 'Electronic Design' started by ted, Jun 9, 2004.

Scroll to continue with content
  1. ted

    ted Guest

    Does anybody here understand exactly how these formats work? (i.e.
    without marketing hype or hot air)

    My understanding is that both SACD and DVDA are both lossy methods of
    recording audio, mainly because the one bit serial stream cannot be
    recovered into its original analogue source using the simple low pass
    filter technology used in playback equipment.

    I appreciate that a lot of people claim they sound better, but surely
    from a technical point of view, these methods are worse than the
    original PCM?

    Or am I missing something???

    Anybody have a a www pointer where the formats are described in decent
    technical detail?? (again, without the hype)


    Thanks

    Ted
     
  2. Tim Wescott

    Tim Wescott Guest

    They're not lossy because of the 1-bit conversion back to analog,
    they're lossy because you cannot guarantee a compression ratio with
    lossless compression. This is because any chunk of data will be more or
    less self-correlated, which is what lossless compression algorithms look
    for.

    As soon as you yield to the real-time requirement of having a fixed data
    rate that's less than the original data rate you have locked yourself
    into using lossy compression.

    Does anybody claim that these are better than the original sampled but
    uncompressed data? Most real performance claims are going to be based
    on oversampling the data to the ADC, which works because it's much
    easier to build a good reliable interpolation filter in digital hardware
    than in analog, and sampling the ADC faster eases the requirements on
    the necessary analog filters.
     
  3. Ben Bradley

    Ben Bradley Guest

    I don't see/buy that argument.
    In modern A/D converters, PCM is derived from the 1-bit bitsream
    generated by sigma-delta (or is it technically delta-sigma?)
    converters, so the 1-bit bitstream is indeed "closer" to the original
    analog signal than is PCM.

    Here's a page describing it (though it seems to have some hot air):

    http://www.cdfreaks.com/article/95/6
    IIRC, DVDA has several audio formats available, all PCM at various
    bit rates and depths, but there's also lossless compression done to
    the PCM. Here's one article on it:

    http://www.meridian-audio.com/w_paper/mlp_jap_aes9_1.PDF
     
  4. Guy Macon

    Guy Macon Guest

    If the above statement was false you could run the data file through
    the compression again and again, eventually compressing it to a single
    bit that decompresses back to the original.
     
  5. ted

    ted Guest

    Does anybody here understand exactly how these formats work? (i.e.
    Well, look at it this way. SACD samples the audio faster, and
    generates 64 bits of samples 44100 times per second, whereas PCM
    generates 16 bits. So 64 encoded bits per sample look a lot better
    than 16. However, the pulse density decoding process for SACD is a
    simple low pass filter on the bit stream. This means that, just like
    an equivalent PWM decoder, you can only get 64 discrete analogue
    levels per sample (unless there is some hidden technique that can
    generate more levels than this), this corresponds to six bit
    precision! not much when compared to PCMs 65536 levels for the same
    sample interval.

    I appreciate that if you SACD encode a slowly moving waveform, the
    analogue output after the LP filter can follow it much better as more
    density bits are used in the LP pass process. However you are now not
    "sampling" at 44100, but at a much lower rate.

    To do some simple calculations, a SACD stream 65536 bits long would
    generate the same analogue accuracy as a 16 bit PCM signal. However
    the input audio voltage level has to be steady over this sample time,
    meaning it has to be 46Hz or less. What this means is that SACD can
    produce "better" bit accuracy than PCM, but only at audio frequencies
    below 46Hz.

    I also appreciate all the arguments of what the ear can hear etc, but
    this is purely a technical discussion.

    Have I got something wrong somewhere??

    Yes but surely, these A/D converters are also lossy in the sense that
    they do not produce the resolution they claim.

    The Crystal 24 bit CS5396 96Kbps A/D as a typical example. Although it
    claims 24 bit resolution, it is only capable of producing about 17 bit
    "real" per-sample resolution. If you put a steady (or slowly moving)
    voltage input into a
    CS5396, you will get 17 steady bits, the other 7 bits are just noise.
    I appreciate the noise is HF random and kind of cancels itself out
    when averaged over many samples. But at the claimed 96kbps rate, the
    discrete resolution is 17 bits only.

    Ted
     
  6. a bit OT but I got this from rec audio pro
    martin



    "Why should we subsidize intellectual curiosity?" -- Reagan, '80
     
Ask a Question
Want to reply to this thread or ask your own question?
You'll need to choose a username for the site, which only take a couple of moments (here). After that, you can post your question and our members will help you out.
Electronics Point Logo
Continue to site
Quote of the day

-