20 Jan
2004
20 Jan
'04
6:48 a.m.
On Tue, 2004-01-20 at 07:45, R. D. Davis wrote:
SOrry, I misspoke, by 'repair' I meant 'replace'. Though my father had
an 01 (not 01A) tube with I swear an RCA base and a Westinghouse
envelope (seal on top of tube) or converse. I lost it as a kid (ouch).
But with TUBE
computers you just fix the broken part.:)
Oh really? :-) When was the last time that you saw someone pop open a
tube and replace or repair an open filament, repair a short between
elements or restore the vacuum in a gassy tube, etc., and then put the
tube back into use? > I have a new design (yes you read right) for a
tube computer, probably
> 75 envelopes, on my website. 20 bits, serial arith, drum memory
Neat! Slow, but most interesting.
Not even capable of keeping a 60 wpm tty at full tilt -- but few small
machines of the first gen could (eg. the LGP-21 took 7 seconds (7000 mS)
to checksum 64 contiguous words of memory). Relative speed is umm
relative.
> Somehow, I can't find the money to make it.
But I'm serious enough,
> there's an assembler and simulator there (wps.com/projects).
What appears to be the most expensive aspect of the
project thus far?
Prototype development I'd say. The remaining paper circuit design could
be done in "spare time". I have no pretense of being a better tube
designer than the greybeards. Reliability and work-first-time is to me
more important than speed, which mattered most 'back then', so dropping
margins (80 KHz clock) and fancy hardware (eg. no mult/div) and using 50
years of hindsight (eg. IO design!) and the fact the silicon diodes cost
$5/100, not $50 each (equiv $) makes that part doable.
Drum memory was scary though, I have a design (not on www) for a mthod
of construction, but it's still scary to construct. Switched-cap is
easier, and using a PC board you could get 500 cells per PCB easy, likel
1024 (big 2-sided boards) and with 20 of those you'd have a Kword.
4Kwords would be best. That's just a first-order conservative guess as
we didn't talk details on it.