Martin McCormick wrote:

> The old device is a Votrax Personal Speech System, a speech
> synthesizer that was popular in the early eighties.  The device has a
> DB25 mail connector on its parallel port, but the pinout is totally
> unlike anything else.

  Ahh yes, but that was Votrax.  Was it completely encapsulated in epoxy
resin to make it un-repairable, like the Type-n'-Talk?  I think I have
two of these, but at least one was faulty.  They used a serial interface
which also had handshake problems.

>  This makes me wonder why have both an ACK and a Busy?  If there is an
> ACK pin and the peripheral stops acknowledging characters, then the
> data stops, anyway.

  An interesting question.  Essentially, the two functions do exactly
what they say, that is "Ack" is pulsed to mean "I have seen your Strobe
pulse and read the data; you may release it now", while "Busy" means "I
an NOT ready for more data".  The correct sequence is to poll Busy
before attempting to send data, and to wait for Ack afterward.

  If you can be sure your device will immediately and always latch in
data on Strobe, you can use a simple monostable pair to generate Ack
from Strobe, otherwise what you should really do is to provide an extra
latch AND the monostable pair, with Busy operating as always.

  I think http://www.hut.fi/Misc/Electronics/pc/interface.html should
have links about this.

  The simplest handshake is a two-phase Strobe and Acknowledge where the
data is asserted, Strobe is asserted, Acknowledge is asserted to
indicate the data has been latched, Strobe is then cleared and the data
de-asserted (wild), while Acknowledge is cleared only when the data has
been digested and the receiver is ready again.

  Lap-link (parallel) uses, I believe, a counter-phase system where data
is asserted, Strobe is asserted, Acknowledge is asserted only when the
data has been fully processed, new data (nybble) is asserted, Strobe is
cleared, and Acknowledge cleared when this new data has been processed.
I do not claim that laplink's pin allocations are the conventional ones
though!

  The big problem in all this relates to what happens when a printer
goes off-line; i.e., Busy was true then goes false.  If a character was
sent and Strobed in the interim, whether or not it was Ack-ed, it may be
quite unclear whether it should be sent again or not.  Or if the printer
is switched off and on or connected and disconnected (i.e., switch-
boxes!)  This sort of uncertainty really fouls up network printing,
particularly of graphical data!

  Cheers,
        Paul B.