Mr Ian Primus wrote:
<snip>
That actually would work pretty well, since the
mechanism already intends to have a continuous loop of
tape. You would just have to shorten it somewhat.
Waiting 12 minutes for that bit to come back around
would be a pain.
If you shortened the tape, and a somewhat shorter foil
splice and had both the record and playback heads
egaged at the same time, then you could use the foil
sensing splice/contacts and have that generate an
index pulse to start timing.
Of course, you're still limited to 2 bits wide, since
eight track tape recorders still only have a stereo
head, with a solenoid to move the head.
Now, if you could find a quadraphonic 8 track recorder
(rare as heck)...
-Ian
The other problem you have with tape even with acceptable delay between
the bits appearing is that you want to do a write after read. Say you
shift in a value from the read head, process it and need to write it
back over the previously read location.
with a drum you can have a write head and a read head and have them
synced to either a timing track, or a tach disk or other device and have
a pretty precise pulse when you read and when you need to write.
but with tape, you'd have to be able to get the read and write heads
precisely set so that if you wanted to write a value, you'd put it down
on the tape over the previous position.
I guess the other obvious solution to that problem would be to read
everything into a register, and then write it for later retrieval,
whether it was modified or not.
writing all the time like that doesn't appeal to me but it would work.
main thing that would kill you would be the latency of how long till the
bits reappeared.
I know that one of the attributes of the drum on the lgp30 that was
taken advantage of was the fact that there was skew between tracks that
one could write modified values to on other tracks, and you'd juggle
between several tracks at different radial positions knowing you'd not
wait a whole rotation to dispose of the data you needed to save, and
could calculate at faster than the drum's rotation latencey. With only
a few tracks and a very long time between appearances of data that would
be pretty useless.
Jim