Friden (was: IBM 604

From: John Allain <allain_at_panix.com>
Date: Tue Jul 3 11:32:54 2001

I actually dusted the thing off and tried it out for probably the
first time in 5 years. After figuring out how to get going (there
are still 5+ knobs that I still don't know the function of) and
cranking out some more ozone, here's the resullts.
Oh, its about 11" cubed and 40~50 lbs dense.
Needs more space for its carriage to move around.

> It certainly seemed MUCH longer.

One place division seems to take about 0.8 seconds
and there are ten places, so it may be possible
that No division will take more than 8~9 seconds, with a
properly tuned machine. If you want, you can set the division
place "pointer" just one digit to the left irregardless of the
numbers involved and it will "count up". THIS takes much
longer (EG 1 minute for a result of 400) since it has to count
all the way up the hill instead of counting up each "step" and
then stepping to the next digit. (example 345 is 29X that of
3+4+5 in terms of time involved). I could guess that if someone
didn't want to beat his machine up quite so much he might use
the slow method. Mine in its semi-maintained state will only give
100% correct answers with the slow method. Ouch.

> But I can't remember how "five states" were achieved.

My hunch is that duoquinary or whatever was set up to make
computer math more palateable for non mathematicians
(first C.S. degree = 1966? _at_UNC) and therefore more customers.
I would be not at all surprised if the implementation was just 6 bits
per digit (1,2,3,4,5,shift), it is only 80% added waste over pure binary
and required less effort for people to work with. IBM was probably
worried that its users were base 10 bigots and base 8 wouldn't be
as popular. Hunch.

I had some conversations with a 6xx user at G.E.20 years ago
in case anyone wonders why the interest.

John A.
Received on Tue Jul 03 2001 - 11:32:54 BST