Path: utzoo!attcan!uunet!husc6!uwvax!oddjob!uxc!uxc.cso.uiuc.edu!uxe.cso.uiuc.edu!mcdonald
From: mcdonald@uxe.cso.uiuc.edu
Newsgroups: comp.arch
Subject: Re: History of personal computing (LONG
Message-ID: <46500024@uxe.cso.uiuc.edu>
Date: 7 Aug 88 13:53:00 GMT
References: <5946@venera.isi.edu>
Lines: 35
Nf-ID: #R:venera.isi.edu:5946:uxe.cso.uiuc.edu:46500024:000:2013
Nf-From: uxe.cso.uiuc.edu!mcdonald    Aug  7 08:53:00 1988



>In fact the DG Nova was a 16-bit version of a 12-bit machine (the PDP-8)
>and architecturally is atrocious compared to the PDP-11 at least in my
>opinion. It has been a long time since I programmed a Nova, but it had
>only 4 registers and NO stack push or POP instructions.

>I suppose you could say it was interesting but you wouldn't want
>to program it in assembly language.

Actually, not very interesting to an old PDP-8 hand - but then not
really so bad to program. Can you say "RISC"? There are several
interesting anecdotes about this line. First, there was a thing
called a "Supernova", which had semiconductor memory with a 300 nsec
cycle time - this was 1971 or 1972. Second, DG had a really interesting
construction technique - this was a Nova 1200. It had no case per se.
There was a front panel and a power supply and a card cage. These
things existed a separate objects, and bolted together to make the 
computer. Remove one for servicing and the whole thing fell to pieces.
The power supply was one of the first switching mode supplies used
in computers, and it had a horrible bug. There was no large capacitor
on the output. It was supposed to depend on all those 0.01 microfarad
things accross every IC. Well, the -5volt supply didnt power many
IC's - it was just for the (analog) sense amps in the core memory. 
Problem was, it also powered the current loop in the ASR33 teletype,
which used mechanical relays to generate output. This produced
horrendous glitches on this line. If one occurred at the wrong point during
a read cycle, the data got read correctly, but got put back in core wrong
(for you newcomers, in core memories a read destroyed the contents of the cores,so every read had to be followed by a write). Unfortunately their
diagnostics were so constructed that they could never, ever, catch this
fault. The refused to admit it existed until I wrote a test which caught it
within two or three keystrokes on the ASR33. A 50 cent electrolytic
solved the problem.

Doug McDonald