In 1976, I selected a metal-contact keyboard for a new
Hewlett-Packard terminal.  In the course of that investigation, I
uncovered several items which may be of interest to the
automatic-music subscribers.

A monograph by Kenneth E. Pitney titled "Ney Contact Manual -
Electrical Contacts for Low Energy Uses" was published in 1973 by
the J. M. Ney Company.  This ~200-page book contains a wealth of
information on the general theory of contacts; interactions of
alloys, atmospheres, surface oxides, contact voltages, erosive arcs,
and contact pressures.

One of the author's interesting observations concerns the production
of  insulating "friction polymers" where sliding contact systems
utilizing platinum-family materials catalyze the production of
polymers from the vapors of organic hydrocarbons. (Gold and silver
contacts do not produce friction polymers.)

Another observation concerns threshold voltages for puncturing oxides
and other contaminating layers.  In general, TTL logic levels are
marginal unless used with silver or gold contacts.  Direct operation
of high voltage, inductive loads (read:  "solenoids") caused
considerable erosion if sufficient thermal mass was not present in
the contacts.

I also came across a paper published by researchers at Bell Telephone
Labs which discussed the problems of designing coin-operated pay
phones.  Their challenge was to produce reliable contact closures
using the energy of a falling dime (a "thin" one, at that).  The
toughest environment was a remote phone booth situated on the New
Jersey Turnpike!  They achieved reliable operation by use of
multiple/paralleled palladium-gold alloy contacts.

The keyboard I selected (produced by the Hi-Tek corporation of Santa
Ana, CA) used quadfurcated Au-Pd contacts for high reliability.  The
contacts were similar to the tines of a fork, and designed such that
dust and dirt particles under one of the individual contacts would
not prevent the other three from closing.

These mechanical contacts suffered from contact bounce.  Our debounce
solution was implemented digitally in the keyboard-scanner integrated
circuit (as >100 resistor-capacitor networks were not practical for
space and cost reasons).  The debounce algorithm recognized the fact
that a typist cannot repeat a single key faster than 8-10 times per
second (pianists have some tricks for higher repetition rates,
however). Therefore, key down/up events with shorter than 100ms
duration had to originate from contact bounce.

We implemented a simple 3-bit counter clocked at 80 Hz.  The counter
was reset (and reported KEYDOWN) for any contact closure, and
reported KEYUP only after 8 successive non-keydown clock cycles.

This type of algorithm is easily implemented in today's digital
computers. The algorithm can be made symmetric such that upon change
of state -- after a predetermined stable period -- the new state is
immediately reported. Resistor-capacitor networks introduce a
propagation delay for one (or both) state changes.  A final advantage
of the computer algorithm is that its time constant is a program
variable; you don't have to get out the old soldering iron to twiddle
things!

John D. Rhodes - AA7HL  Vancouver, WA  [jrhodes@teleport.com]¶
"We used to be able to do that, but the technology's improved
 -- and it's no longer possible."