This is the first editorial article for this site.
Such editorial reviews will issue periodically (I hope at least
once a week). The aim is to carry pleasant for reading and easy
for comprehending information on a free subject (that is, to my
taste). Undoubtedly, the topic has to relate to the site's subject-matter.
At least, to some degree.
There I will consider a technology so far untouched
in the computer fields which still can influence habitual conceptions
strongly. And that is digital representation of everything out there,
and different methods of its processing.
The SACD standard appeared due to a simple commercial
reason. The date of payments under a licensing agreement to CD-Audio
technology developers was approaching, and the income received by
them was incredible. According to rumors, it was comparable to that
received by the developers from sale of CD equipment of their own
production. A way-out is obvious: Sony and Phillips were to create
a new format capable to force out CDs. And mainly to attract comparable
license assignments. That was a prehistory to SACD creation (Super
Audio CD). Obvious it's impossible to attract customers and companies
by a simple increase in capacity - the whole commercial sound recording
sticks to 74 minutes and capacity increase isn't worth realization.
And there lies only one way - to gather additional money for a 2-CD
set of classical music. But who is to blame that at that time there
wasn't 74-minute limitation. Size is out of discussion, so it's
necessary to develop completely new quality. And it shouldn't be
just an intensive jump up (we have got already DVD with 192 KHz
24 bit), but a qualitative new approach.
Many modern, decent AD and DA converters are build
on a so called 1 bit scheme. If I had had my way I'd have called
it the "digital PDM scheme", but I wasn't afforded such
opportunity =). An analog signal is represented in the digital form
with the help of PDM (Pulse Density Modulation) with a variable
quantization step. Pulse duration defines the level of an analog
signal. To be exact, it is defined by a ratio of the pulse duration
to the duration of the following pit (see the graph).
That is the energy transmitted in one period of
a decoded PDM signal. It gets clear that it's not difficult to make
a DA converter for such signal. Just take a capacitor and let it
integrate for some definite time period. No doubt that a real model
is more complicated, but it's based on the same principle. Besides,
the ADC has a perfect ratio price/quality. Conversion is fantastic
and the most problems concerning parallel and other DAC/ADCs disappear.
For example, this scheme is adaptive: if a signal has a small amplitude
the readings go more often and detailing increases. This, in its
turn, completely corresponds to well known nuances of sound perception
by a human being.
In modern Hi-Fi facilities such convertors are
used widely. But, unfortunately, information is kept in linear code
with a fixed number of bits per reading. It means that after such
ADC it's necessary to convert 1-bit PDM into a standard one, say
16 bit, with a fixed period of reading selection. To save possibly
compressed (with losses or without), to read, decode (unpack), reconvert
into successive PDM, transfer to DAC. Difficult. And the main thing
is that intermediate conversions bring in their own distortions.
Good (with low distortion level) intermediate converters cost expensive,
nullifying all advantages of 1-bit converters. The way-out should
conclude in getting rid of intermediate conversion. To transfer,
compress, store and process a successive PDM stream i.e. in the
form of a bit stream.
Of course, SACD developers followed this way. Recording
density increased several times as compared with CD. Judging from
today's point of view, it's ordinary. The bandwidth went up as well
from 44'100*16=705'600 bit/s to 705'600*4=2'822'400 (for each channel).
PDM stream was divided into bits with sampling frequency around
2.8 MHz (the same 2822400 bit/s). It's called DSD (Direct Stream
Digital). And then recorded on a CD. That's all. The others are
just details. For example, a bit wider aperture for an optical system
provides better scratch prevention (though just a little). And the
most important that it provides an opportunity to create two-layer
discs which are seen by a SACD drive as SACD, and by A CD drive
as usual CDs.
Further, if there is some information, one feels
like compressing it and converting into prevalent formats. The latter
can be executed simply (quantization step of the PDM stream is divisible
by standard 44'100), and the former requires new compression methods
(with losses and without) oriented to digital bit streams. Believe
me, they have been "living" for a long time already and quite successfully.
For example, on the base of associative or "entropic" coding. Some
of them are realized on hardware level without any problems.
Why don't we see similar technologies on PC? Who
prevents installing 1-bit convertors on sound cards? There we can
save on board layout - data on codecs are transferred successfully...
But no - there tells upon a parallel orientation of modern architectures.
It's inconvenient (for architectures, and therefore programmers)
to work with successive bit streams. Although some are doing it.
There arise nondestructive archivers and channels of successive
data transfer. I can only hope that in the near future music will
be kept and processed (right up to usage of different filters and
effects) in successive DSD-like formats.
Write a comment below. No registration needed!