976 Chapter 25
up for those. A major positive is having everything
under one roof, dispelling the problems of multiple-part
interconnectivity.
The second approach very much follows the
signal-flow approach of a conventional analog mixing
console, using multiple DSPs as required, processing
being applied in-line as and where it is needed. On very
large consoles the mixing processing itself can take on
the look and feel of a tank-style array, but other than
that the layout of the signal paths has remarkable paral-
lels to analog.
25.23.2 A Practical Digital Mixer
As with the discussion of analog consoles, which
revolved around the description of a particular design,
so this section uses as its basis the architecture of a real
digital mixing console. It is shown in its basic form. As
the reader is well aware, pin-by-pin details of imple-
mentation, bells, whistles etc. can rapidly mushroom
and the weight of resulting detail tend to obscure; as it’s
not too difficult to figure out how most of this is done,
they have been omitted for clarity. It is important to
remember that in terms of lines carrying audio signals,
it is accurate, due to the use of the serial audio format
outlined below. Shown here in a mid-size 64-by-24
format, this particular design’s premises were simplicity
and scalability (it can be readily made bigger or
smaller) and has proven to be robust and reliable, using
no scary technology and with nothing running on the
edge. Over the years this basic architecture has grown
and evolved through generations of increasingly
powerful DSP and support devices with the odd effect
in this blighted world that it has actually become
progressively simpler to build with time. Also the
steady and welcome improvement in integrated
converters has resulted in the overall performance blos-
soming to the extent that this, along with other digital
mixer designs using comparable technology, owe
nothing to analog in performance whatsoever.
It is assumed, of course, that the control surface has
been undertaken as the separate design exercise that it
largely is; this discussion concerns the signal-processing
side of things.
25.23.2.1 Serial Audio Format
Nearly all converters and like peripherals such as
AES/EBU format transmitters and receivers use in
common a serial digital interface; this is usually set up
as to be two sets (left and right) of 32 data bits per
sample frame (64 total), meaning a data rate of
3.072 MHz (for a 48 kHz sample rate). This is a very
tame and robust rate and can be run around quite
happily without fear of corruption, and as such is used
as the nearly sole means moving audio data around in
this console. Adopting this serial format also minimizes
the amount of data format changes required.25.23.2.2 InputsInput signals are applied to whatever form of convertor
or interface is required: microphone amplifiers or
line-level inputs into A/D converters, AES/EBU into
AES receivers, and subsequent sample rate converters.
Sample-rate converters (SRCs) are necessary since it is
unlikely (unless a whole amount of trouble has been
gone through to synchronize the whole system of which
the console is a part) that other digital sources will be
and remain in word/data-rate synchronization with the
console. The recorder may well be, but typical
AES/EBU devices, such as outboard effects, or remote
sources, rarely will be. If it is considered necessary (on
the basis that anything that messes with data unneces-
sarily is a bad thing), the SRCs may be bypassed for
sync’ed sources, but frankly SRC’s today have artifact
levels so low as to be considered quite blameless.
At this point, all the data is in native format (the
convertor serial standard), travelling in pairs—mono
signals (microphones, say) in pairs and stereo sources as
left/right pairs per data line. For a 64 input console, this
means 32 data lines.
The channel signal processing is done four channels
(two pairs) per input DSP, Fig. 25-147. The DSPs used
here very conveniently have native format inputs and
outputs (being designed to work with normal
converters), making interfacing really simple. They are
also easily powerful enough to do four well-featured
channels worth of signal processing. Typically, this
would be high-and low-pass filters, a four-band para-
metric EQ and limiter/compressor/gate dynamics, and
delay (memory is attached to the external memory inter-
face of the device to support this if required). The
channel DSP has spare input and output capability,
which can be implemented if required as selectable
direct channel outputs, keying inputs to dynamics, etc.25.23.2.3 Mix StageThe 64 channel outputs are taken from the 16 channel
processing DSPs as 32 output lines and applied to the
mixer stage(s). Fig. 25-148.
The ominously large device labeled FPGA (field-
programmable gate array) into which all those lines