Direct Stream Digital and the Super Audio CD: A Revolution in Digital Audio

1
Direct Stream Digital and the Super Audio CD A
Revolution in Digital Audio
2
Whats wrong with my CD?

• The standard Compact Disc currently in use has a
  frequency response of 0 to 20 kHz and dynamic
  range of 96 dB.
• While humans can only detect tones up to 20 kHz,
  higher frequencies have been shown to affect
  timbre of perceived sound.
• Human hearing has a dynamic range of about 140 dB.

3
What is our current situation?

• Nearly all digital audio in the past has used a
  scheme called Pulse Code Modulation or PCM.
• PCM represents an audio waveform as points on a
  grid.
• Sampling rate determines the interval between
  data points on the time axis.
• Bit resolution determines the interval between
  data points on the amplitude axis.

4
PCM represents audio as points on a grid.
5
What is Direct Stream Digital?

• Direct Stream Digital (DSD) uses a completely
  different scheme.
• Audio waveform is represented in a pulse train.
• The density of pulses determines the height of
  the waveform.

6
DSD represents audio as a pulse train.
7
How does it work?
8
So whats that Delta-Sigma thing?
9
Another way to look at itMatlab Code

• function modSigdeltaSigma(sig)
• Input is any signal 'sig'
• Function outputs delta sigma modulated version
  of input
• Normalize the input signal so that its maximum
  amplitude is 1
• (only scales down, not up)
• if (max(sig) gt 1)
• normSig(sig./(max(sig)))
• else
• normSigsig
• end
• Set the initial value for the modulator
• if (normSig(1)gt0)
• mod1
• else
• mod-1
• end
• Set the initial value for the sum
• sigma0
• Create an empty array for the output signal

for i1length(normSig) Delta
deltanormSig(i)-mod Sigma
sigmasigma delta Modulator if
(sigma gt 0) mod1 elseif (sigma lt
0) mod -1 end Add 'mod' to
end of output stream modSig modSig,
mod end Convert 'modSig' to binary (all '-1'
become 0) modSig (modSig 1) modSig
modSig./2
10
Lets give it a try
11
Another example
12
And one more
13
Sample outputs
14
(No Transcript)
15
(No Transcript)
16
Getting around.

• PCM can be converted to DSD by means of a
  delta-sigma modulator. This time the input to
  the modulator is digital instead of analog.
• DSD can be converted to PCM by means of an FIR
  filter. This is a non-recursive filter that
  basically averages out the pulse train and then
  quantizes to the desired number of bits.

17
Is DSD, in fact, better?

• Lets look at the specs

18
But wait, thats not all

• PCM cant deliver its promised frequency
  response because it must use low pass filters to
  prevent aliasing, a type of distortion.
• Not only that, but the steep filters used to
  prevent aliasing often cause time alignment
  problems in the high frequencies.
• While the 144 dB dynamic range of 24 bit PCM is
  quite impressive, it is probably unnecessary.

19
Further more

• DSD can only deliver its promised 120 dB of
  dynamic range in the 0 20 kHz range.
• Since it uses noise shaping to achieve such a
  high dynamic range, it sacrifices dynamic range
  in the upper, inaudible, frequencies for improved
  dynamic range in the lower, audible, frequencies.

20
A quick look at noise shaping
21
Things get hairy. (Even more complications to
the issue)

• Low cost PCM A/D converters are already using
  delta-sigma modulators, because they are
  inexpensive and easy to implement.
• High quality DSD converters use multi-bit
  delta-sigma modulators because they are free from
  certain types of distortion. The signal is later
  converted to the 1-bit stream.

22

• All current digital signal processing operations
  require a multi-bit signal. DSD workstations must
  convert to multi-bit for processing, then back to
  1-bit for final delivery.
• Many audio professionals wonder whether a high
  sampling rate, 4 or 5 bit, system using
  delta-sigma modulation might be a better solution.

23

• The Super Bit Mapping Direct technology from Sony
  allows a high precision down-conversion from DSD
  to CD quality PCM. The result is better sounding
  CDs.

24
DSD on Disc, the Super Audio CD

• Single sided, double layer disc.
• One layer contains the standard CD version of the
  disc, on the other is the SACD version.
• Due to loss-less data compression, the SACD layer
  can hold both a Stereo and 6- channel version on
  the same disc!

25
More goodies

• The SACD layer can also hold text, graphics, and
  video.
• Completely backwards compatible with standard
  CDs.
• 74 minute playing time. (Same as CD)

26
Copy Protection

• On a standard CD, digital information is
  represented as small pits of varying length.
• On a Super Audio CD, the widths of pits are
  varied as well in order to implement a digital
  watermark.
• Requires special optical sensor, wont play on
  existing DVD-ROM drives.
• Discs without the special watermark wont play.
• Very difficult to pirate.

27
The Contender

• DVD Audio (DVD-A) can hold 2 to 6 channels of PCM
  audio.
• Quality can range from 44.1 kHz/16 bit to 192
  kHz/24bit.
• Playing time varies depending on quality and
  channels.
• More difficult to author than the SACD

28
Politics

• DSD/SACD is backed by Sony and Philips, two huge
  companies with lots of marketing clout.
• These are the same companies that developed the
  original CD.
• The DVD-A is backed by a conglomeration of
  companies who often do not get along.

29
Conclusions

• 99 of people will probably not be able to tell
  the difference between high resolution PCM and
  DSD.
• DSD/SACD will probably win the marketing war.
• DSD/SACD is the most effective means available to
  deliver digital audio to the consumer.
• DSD is not the most effective solution for
  professional recording and processing, although
  it is a large step above other schemes in use.

30

• DSD/SACD will be much more beneficial to very
  dynamic and acoustic music (such as Classical and
  Jazz) than to pop and electronic music.
• Only some people will have good enough equipment
  and ears to hear the difference between a
  standard CD and a SACD.
• It will take the SACD a long time to reach the
  mainstream market if ever.