Date: Fri, 14 Dec 2007 14:52:29 -0800
From: "Rick Bensene" <rickb at bensene.com>
Some of the smaller drum-based computers from the late
'50's and earlu
'60's (transistorized) had drums that didn't rotate nearly as fast as
6000 RPM.
I don't know--the drum on the IBM 650 spun at 12,500 RPM, was 4
inches in diameter and about a foot long.
This is not a project I'd launch into lightly. In the 70's and 80's
Meshna would occasionally offer a mil-surplus drum unit for sale
usually from some some airborne installation (i.e. used in radar).
Lots of heads and no guarantee.
The precision necessary in the mechanicals is pretty daunting. The
heads fly at a few micrometers above the surface of the drum, which
pretty much mandates some sort of clean-room environment. The
bearings would need close to zero runout and getting the surface of
the drum coated smoothly enough, either by plating or with a magnetic
emulsion would also be an obstacle.
If you have a good precision machine shop and mechanical engineering
skills at your disposal, you could try it.
Personally, I'd rather work something up with a magnetostrictive
delay line memory--easier to build and quite robust--and just as
"vintage".
I recall that even in the 70's, drums had their problems. On the CDC
STAR, every so often, a station would go down because of a "drum
error"--more frequently than the disk drives encountered errors. You
could make the heads on a Univac Fastrand II drum "ping" just by
jumping up and down on the floor near the monster. I also recall
that the ADL folks at CDC were working on a "super drum"--a very high
speed unit that spun in a vacuum(IIRC) to be used as a paging store
for the STAR. I recall the Neil Lincoln mentioned that the time
between the observation window being clear and coated with drum
surface was typically a few minutes at best.
There's a nice story about the Manchester Mark I drum here:
http://www.computer50.org/mark1/gethomas/manchester_drums.html
About the only item of the correct size and tolerances that's readily
available today, it seems to me, would be the drum unit out of a
laser printer or copier. I don't know if the OPC layer could be
easily stripped off and replaced with oxide, but it might be worth a
try.
If you're really serious abou this, I'd be willing to donate a head
from a CDC 808 disk unit--4 bit parallel recording.
Cheers,
Chuck