Here are brief descriptions of the electronic relay systems I've built:

"System/270":
This is a MIDI-based system using 8051-based microcontroller
boards and shift register I/O.  The processor board is 6"x6" and is
single-sided (to make it easy to produce the prototypes).  It contains
an 8031, 2 ROM/RAM sockets, 12 bits of parallel I/O, and MIDI IN,OUT,
and THRU.  There are also 2 LEDs to make debugging easier (ha!).

The input boards consist of CD4021 shift registers arranged 24 bits to
a card.  These connect in series for as many input bits as necessary
and the first board connnects to a processor board.  The input lines are
active when taken to ground (through a 1N4002 diode for protection).
Some boards may have 8-position DIP switches for setting MIDI channel
number, starting note number, and so on.  The program scans the boards
and generates MIDI note-on/off sequences as appropriate; it also acts
as a MIDI-merge so that several input processors can be daisy-chained.

The output board also contains 24 bits using CD4094 shift register
latches and ULN2803 sink drivers.  Again, enough boards are
connected in series to cover the entire rank and the first one connects
to a processor board.  The program accepts MIDI IN data and implements
a 16-manual 1-rank unit organ.  The rank's identifier number and number
of pipes is burned into the program EPROM.  MIDI system exclusive
messages are used to specify which rank is to play on which keyboard at
which pitch offset or which keyboard is to be coupled to which other
keyboard.  The program calculates the current state of the pipes and
shifts the resulting bit pattern out to the drivers.

This system seems to work rather well.  Tom Dimock did the board layout
and construction for this system and is installing on in his home
instrument.  He used Adobe Illustrator to design real art deco circuit
boards. I did the logic design and the programming (8051 assembler).
Advantages: lots of flexibility ("flexibility through programming": a
great slogan!); compatibility with MIDI devices; plenty of horsepower,
even for large instruments (one processor per rank).
Disadvantages: 12Mhz clocking and possible RF interference; longevity of
EPROMs (rated for 10 years); output drivers are not short-circuit protected
(but other chips could be used that are protected).

"C1":
This was a prototype of a simple relay in which all of the switching
logic and magnet driving for a single note was on a single card.  Twelve
of those cards would make up the entire relay.  I didn't complete this
unit because it was not easy to make up different configurations.  It
might be a simple scheme for stock model organs or for the normal
couplers on a classical organ.  It would not have been practical for
highly-unified organs, however.

"C2":
This is a non-processor, non-clocked, relay system that I designed and
built for Cullie Mowers for use in the rebuild of a small Estey.  That
organ had just unison and octave couplers, and so a complex relay
wasn't necessary.  However, Cullie used new OSI slider chests with
electric pull-down action, and so the relay had to handle about
500ma for each note.  I built this using UDN2987As, 8 of them per board
to make a 61-pole stop switch.  The outputs were ganged via ribbon
cable and drove 1-octave boards using the same chips, this time with
2 outputs ganged for each magnet.  This is a simple design and might be
useful for adding a few unit stops to an existing electro-pneumatic relay.
Its inputs and its outputs can be wired directly in parallel with such
a unit.  The main problem with it was a (temporary, I hope) unavailability
of the chips; Allegro (was Sprague) is the only maker.  These chips are
also short-circuit protected, a real feature.  This relay operates on
the standard rectifier and therefore does not need a special +5 supply.
Advantages: protected driver chips; no clocks or other complex logic;
"organ-like" (each board is a 61-position switch); fairly easy wiring via
ribbon cable and IDC connectors; runs on standard rectifier.
Disadvantages: bulky for large unit organs; somewhat costly due to IDC
connectors; UDN2987A is the ONLY chip I could find that would work, and
it is single-sourced.

"C3":
This is another relay for Cullie.  This time, the organ was a 2/9
Tellers unit organ with normal couplers; each rank played at only one
pitch per keyboard, so the unification wasn't all that massive.  The
whole thing needed about 29 gang switches.  Because of delivery problems
with the ULN2987A chips, I changed to a multiplexed system in which
the coupler and unification logic first processes "C", then "C sharp",
and so on.  The master clock board, input boards, and switch boards
are all about 4x5 and use 64-position DIN connectors onto passive
backplanes containing the key signals for 3 keyboards and the necessary
clocking and latching signals.  The switch boards implement 8' and 4'
switches from each of the 3 input keyboards and merge the output into
a single wire for each octave.  By shifting the ribbon cables in the
connectors, you can have 16' or 2' or 1'.  Wires run to output boards,
each of which has 2 8-bit shift register and driver chips.
These drivers, Micrel 58P42, aren't cheap but they are short-circuit
protected.  The master clock board has a single-step mode and LEDs to
show the status of the input lines as an aid to debugging.  I have
designs for several other types of switch boards and output drivers.
At an FCC-legal 8.4 Khz clock rate, this system scans the keyboards
about 700 times per second. The chips are mostly HCmos, and so
it requires a regulated +5 supply.  The inputs are active high
and can easily be +15, even when the contact also drives a magnet.
Advantages: more compact than "C2" because only 1/12th of the switching
logic is needed; easier wiring than "C2" because the switch cards'
inputs are on a backplane; various switch card can be mixed and matched;
built-in debugging via LEDs; all chips are standard or at least multiple
sourced (well, the protected versions of the UDN5842 are just from
Micrel at present); FCC-safe clocking speed with good input
scanning rate.
Disadvantages: needs +5 power supply; less flexibility than with a
microprocessor.

Larry Chace ([log in to unmask])