Virtual Systems 1443tem. This is not the optimal way, however, to integrate
digital processing into a system.
All devices, analog or digital, impose quality limita-
tions on the performance of the system. In a properly
designed digital device, the major performance limita-
tions are due to the conversions from analog into digital
and back. Properly done DSP will not introduce signifi-
cant distortions or other artifacts into the signal. Early
analog to digital converters had significantly worse
distortions than early digital to analog converters. With
modern converters the pendulum has swung back the
other way with analog to digital converters often having
less distortion than digital to analog converters. In any
case, the majority of the distortions and other quality
degradations in a properly designed digital based audio
signal processor are due to the converters.
This suggests that we should consider carefully how
many A/D and D/A converters are used in our systems,
with an eye to minimize the number of converters in
any given signal path.
This requires a change in how we treat digital audio
devices in our systems. No longer can we consider them
to be interchangeable with traditional analog compo-
nents. Instead we must use system design practices that
will allow a reduction in the number of converters our
audio must go through.
One powerful technique is to group all the digital
devices together in only one part of the signal flow of
our systems, and use digital interconnects between the
devices instead of analog. While not all of our digital
processors are available with digital interconnections,
many of them are.
The most popular two channel consumer digital
interconnect standard is known as SPDIF, while the
most popular two channel professional interconnect
standard is AES3. The European Broadcasting Union
(EBU) adopted the AES3 standard with only one signif-
icant change. The EBU required transformer coupling,
which was optional under AES3. As a result, this inter-
connect standard is often called AES/EBU. Many prod-
ucts are made with these interconnects, and converters
are available to go from SPDIF to AES3 and from
AES3 to SPDIF.
There are many interfaces that carry more than two
channels. One that is popular in the home studio market
is the ADAT interface, which carries eight channels.
Most of the interfaces that originated in the home studio
market are limited in the distance they can be run.
To address the need for greater distances and larger
numbers of channels in professional applications,
CobraNet was developed. It also differs from the other
digital interfaces in that it allows point to multipoint
connections instead of only point to point. This is due to
it running on Ethernet, which is an industry standard
computer networking protocol. Today there are several
different digital interface systems sold that use some or
all of the Ethernet Standard to transmit digital audio.
By grouping as many of our digital devices in one
portion of the system as possible, and making all the
interconnections between them in the digital domain,
we have minimized the number of conversions our
signal has gone through, and maximized the potential
performance.38.2.3.5 SynchronizationDigital audio consists of a series of consecutive numeric
samples of the audio, each of which must be received in
sequence. If the samples are not received in proper
sequence, or if samples are lost or repeated, then the
audio will be distorted.
In order for digital audio devices to interconnect
digitally, both ends of each connection must run at the
same sampling rate. If the source is running at even a
very slightly faster rate than the receiver, sooner or later
the source will output a sample that the receiver is not
ready to receive yet. This will result in the sample being
lost. Similarly, if the source is running at even a very
slightly slower rate than the receiver, eventually the
receiver will be looking for a sample before the source
is ready to send it. This will result in a new false sample
being inserted into the data stream.
In a simple chain of interconnected digital audio
devices, it is possible for each device to look at the
sampling rate of the incoming digital audio, and lock
itself to that incoming rate. One problem with this
system is that the sampling rate as recovered from the
incoming digital audio is less than a perfect steady rate.
It will have slight variations in its rate known as jitter.
While there are techniques available to reduce this jitter,
they add cost, and are never perfect. Each consecutive
device in the chain will tend to increase this jitter. As a
result, it is not recommended to cascade very many
digital audio devices in this manner.
If a single digital audio device such as a mixer will
be receiving digital audio from more than one source,
then this simple scheme for synchronizing to the
incoming digital audio signal breaks down, since there
is more than one source. There are two ways to solve
this problem.
One way is to use a sample rate converter (SRC) on
each input to convert the incoming sample rate to the
internal sample rate of the processor. Such a SRC will
add cost to the input, and will in some subtle ways