Handbook for Sound Engineers

Consoles 945

than a single address acquisition by a typical

imbedded microcontroller.

A/D conversion can be done in a number of fashions
for this system.

25.16.7 Control A/D Conversion: Central or
Distributed?

Central conversion means that there is just one A/D
converter in the processor rack frame. All the multi-
plexed rider-pot voltages hit one bus, which is then A/D
converted centrally and the result of the conversion goes
directly to the processor data bus. The obvious advan-
tage is low cost—only one converter. Disadvantages are
speed (a successive-approximation converter takes
several processor clock cycles to perform a conversion;
this can be obviated by using a very high-speed compar-
ator-type flash converter) and bus slewing (caused
mostly by bus capacitance). Since each rider-pot source
is not zero impedance unless it’s at one end of its track
or the other and CMOS analog transmission gates have
a finite on impedance, there is a definite time constant
involved, with the bus capacitance needing a certain
amount of time to charge to the correct potential (as
determined by the rider pot). The previous bus potential
can, of course, be anywhere depending on the previ-
ously selected position of the pot. Even if this time
constant can be made short with respect to an acquisi-
tion cycle, it even then really can make a nonsense of
256 level, 8 bit resolution!

Buffering each rider pot to present a known zero
impedance to the multiplexer is a partial solution; buff-
ering the multiplexer output—a seemingly obvious
solution—creates more problems. First, the buffer
output needs to be gated away from the bus for the
times it’s not addressed, so there is a transmission gate
impedance there regardless. Second, it has to be a very

fast follower if it isn’t to create worse slewing than the

bus! Remember that the multiplexers are switching at

processor or sub-bus speed. Suitable amplifiers tend to

be as expensive as they are fast.

Distributed A/D essentially means having a

converter on each subassembly (channel) or for a small

number of pots. There are, for example, proprietary

converters that continually read and cycle through eight

inputs independently of the main processor, yet allow

free access to the collected data. Either a batch of these

or the equivalent built from individual converters and

multiplexers allows the processor to work unhampered

by conversion-related hangups, while also keeping all

system interconnections digital. Similarly, this method

avoids gross bus-slewing (there would no longer be a

long analog bus). As always, there are difficulties:

1. Although 8 bit successive approximation A/D chips
   are now cheap, the number has grown.


2. There are more bits on the channel subassembly.


3. The multiplexers feeding the on-board converter are
   still switching at high speed—slewing inaccuracies
   are still possible. Clever priming algorithms can
   increase conversion accuracies while maintaining a
   high overall acquisition rate, almost as high as for
   switch-closure bytes. These set in motion a conver-
   sion on one channel, allowing plenty of time for
   switcher settling and so on before the result is
   looked for on the bus. During the idle period the
   computer is dealing with other setups and results
   from other channels.

Both central and distributed systems are success-

fully used in circumstances where ultimate speed isn’t

that important. Remember that this is not intended as a

real-time application and the actual amount of data is

small. Accurate enough resolution is reasonably easily

achieved.

With the cost of such devices becoming pocket

change, it is not unrealistic to throw a microcontroller at

each channel simply to perform these tasks; a signifi-

cant reduction in parts can be afforded.

25.16.8 Recall Display

Figs. 25-114 through 25-116 describe how all the rele-

vant console control positions can be digitized into

processor-manageable form. Storage on a mass medium

such as hard disk or network is a fairly simple computer

file-management exercise, as is recalling it. What to do

with the recalled information is now the question.

It is assumed that this particular requirement is infor-

mational recall only, not hardware reset (i.e., setting up

Figure 25-115. Simplistic manual reset system.

Storage

system

Rider

pot.

Adjust to

match meters

Real Recalled

Store Previously

stored

value