| Direct
Stream Transfer |
 |
|
 |
|
 |
Super
Audio CD |
|
|
|
|
|
|
|
Direct Stream
Transfer (DST) (cont.)
|
|
Direct Stream Transfer relies on data framing,
prediction and entropy encoding
|
As
a summary, all lossless coding schemes have
basically the same structure, which is shown in the
above figure : framing to select an appropriate
selection of the input signal, prediction to remove
redundancy and entropy coding to efficiently code
the samples and side-information. The key issue is
to jointly optimise these blocks to get an optimal
coding gain for a class of signals and additionally
to minimise their complexity such that the algorithm
is practically applicable.
|
In
the case of DSD which is a 1-bit oversampled audio
signal, the blocks can be significantly simplified
(especially the prediction block), since the input
and output signals only have two representation
values : 0 and 1. The lossless encoding and decoding
schemes of DSD can then be represented like in the
following diagrams :
|
|
DSD
Lossless Encoder
|
|
DSD
Lossless Decoder
|
Besides,
it has been found that a constant frame length of
L=37632 bits, i.e. 75 frames per second, provides a
good compromise between performance and system
requirements for 64x44.1kHz oversampled DSD signals.
|
Finally,
Direct Stream Transfer is a high-gain lossless
coding method which varies on a frame basis, thus
resulting in a variable bit rate which may be a
drawback for some applications, particularly in case
of disc or tape based medias. It is indeed desirable
to keep the bit rate on these medias constant, since
this corresponds most of the time to a constant
linear disc or tape speed. A buffer control
mechanism can then be added to convert the variable
bit rate at the output of a lossless coder into a
constant bit rate, which corresponds to a constant
gain.
|
As
a conclusion, the very high bit rate of the DSD
audio signals (2 stereo and 6 multiple channels
sound) can be stored entirely on a single disc layer
thanks to the adaptive Direct Stream Transfer
method. DST presents indeed a significant gain in
"lossless" data reduction and can be
easily adapted to disc media thanks to buffer
control.
|
|
|
|
|
|
|
 |
