Audiophile Series: E3 - DSD vs PCM
DSD Records (DSD) is a trademark used by Sony and Philips for their system of digitally recreating audible signals for the Super Audio CD (SACD). It has sampling rate from 2.8 to 5.6 MHz (64 times more than CD quality 44.1 kHz) and bit depth is only 1 bit.
How
ridiculous! From episode 1 we know, as our listening range is 20-20000 Hz, so according to Nyquist frequency theory, 44.1KHz is fair enough not to miss
anything, then why we need to push sampling rate to Mega Hz range? And why bit
depth is only one bit where people are now craving for 24 bits instead of 16? Well,
cause its totally different from PCM algorithm.
First
of all, you need to know about PCM algorithm. If you aren’t familiar with it,
check my article Audiophile Series: Episode 2.
DSD
(Direct Stream Digital) takes a different approach. DSD seemed an ideal system to build the
new format around, particularly as it has a great copy protection. The main
attraction of DSD is its simplicity and, with that, cheapness of
implementation. It needs less processing than PCM and can use simpler, far less
expensive DACs.
DSD Algorithm:
DSD uses a single bit of information, and all
this information tells us is whether the current sample of the analogue
waveform is higher or lower than in the previous one. Compared with the over 65,000 different values 16-bit PCM has, the
two values (0 if the new sample if the signal is lower or 1 if it’s higher) of DSD appear mighty limiting.
That resolution shortfall is made
up by the very high sampling rate of over 2.8 million times a second – that’s
64 times the speed of CD. Standard DSD is sometimes called DSD64 for this
reason, with double and quadruple speed versions called DSD128 and DSD256
respectively. There's even a DSD512 spec, though none ever heard any material
encoded in it.
Standard
DSD recordings are still relatively rare compared with the PCM alternatives,
and these higher speed versions are extremely limited to hardcore enthusiasts.
DSD fans claim the format is as close to analogue as digital gets.
Downsides
It’s not
very practical to manipulate a DSD recording. All the things that are required
post-recording usually involve the DSD stream being converted to PCM to do the
processing and then switched back.
Just
about every studio recording made with DSD goes through this process. The
processed PCM is called DXD (the name seems like DSD but it’s actually 24 bit
352 kHz PCM).
When
it comes to measurement, DSD also has issues with high levels of noise compared
with PCM. Clever processing techniques allow the engineers to shift the noise
above our hearing, and optimise the performance in the audible region. The
high-frequency noise is generally filtered out.
DSD vs PCM
There’s no exact way of striking equivalence between PCM
and DSD but it’s around the same as 24-bit/88.2kHz PCM. In terms of dynamic range, DSD is claimed to be around 120dB
in the audible frequency range. In comparison, CD measures around 96dB, and
24-bit/192kHz recordings have a theoretical maximum of around 144dB.
Compatibility
There are two kinds of DSD files, DSDIFF and DSF, but you
should always choose the DSF version if your player is compatible – it’s better
at holding a track's metadata information (artist/song name, album art etc).
Reference:
