Thanks for all the great ideas so far! In no particular order:
* Delay lines: I have a junk magnetostrictive DL on the way. I
don't know if it works.
* The idea behind using cassette heads is just to test out my logic
and R/W amplifiers at s-l-o-w speed. A baby step. The high-speed
stuff would come later. I do have a small lathe (a Taig), so I can
handle some of the machining, provided the drum is small (around 4"
dia x 9" long). If I can rig a tool-post grinder I might be able to
do all of it.
* As for the drum coating, it would either be plated with nickel or
coated with a mixture of ferric oxide and epoxy.
Now for some more questions:
I haven't settled on the machine architecture yet. I was thinking a
smaller word size would be better because the recirculating registers
would have lower latency; that is, a 16-bit word would halve the word
time vs. a 32-bit word. Does this make sense?
One of the big problems with drum machines is the need to ensure
instructions are optimally placed on the drum. To lighten the load a
bit I'm considering making it a zero-address stack machine. Then I
only have to worry about the occasional random access reads/writes.
I was thinking I would implement a data stack and a return stack,
each being made of short recirculating buffers to hold the top two or
three stack entries, with the older entries swapped to longer buffers
with corresponding longer access times. Allowing for the overall
insanity level of this project, is this seem a sane strategy?
-Bobby