discrete transistors

From: Allison J Parent <allisonp_at_world.std.com>
Date: Mon Oct 19 17:55:29 1998

< can not compete with a high-end monolithic microprocessor. The laws of
< physics conspire against it.

They do. There are ways around that. Tricks like massively parallel or
very wide words to name a few.

< Motorola claims to have started shipping ECL integrated circuits in 1962

In 1967 I got my first MECL 1000 parts to play with as they had dropped
to hobby prices, still have a few. They were difficult to work with and
fast by any standrd then. FYI: I was applying them as linear devices. I
do know for sure that in '64 those parts existed but they were somthing
like $20-30 each. Then again I remember seeing the infamous 709 opamp
the early 60s at $120 each! I may add that a friend gave me the
"engineering junkbox" from his company in 1968 and the parts in there
besides transistors included RTL (914, 923, MC7xx) and a few opamps
of the 709/741 series. I learned a lot working with those parts. I was
only lucky kid!

< Why were people still building computers using discrete transistors for
< years after that? Many of the high-end computers used discrete
< implementations of non-saturating logic that was very similar to ECL.

The bias points for ECL were hard to control and level shifts were a
source of errors. Also they were more expensive than RTL and the DTL
(pre ttl) parts. Also ECL requires transmission line style layout
and that would not be widely adopted until three or more layer PCBs
would be in use. There were also those that objected to the power
required by ECL and it's resulting heat (back to the bias shift problem).
The problem is once ICs made an appearance in the mid 60s they were being
used but only where cost justified them (they werent cheap! nor were they
fast!). Another item forgotten is the design cycles were measured
in years then, so by time the massive transistor machines started running
the SSI ICs were starting down the price curve. In practical terms the
first massive transistor machine was the TX2 an experimental one of a
kind. It would only about 10 years later when the first IC machines
started appearing in the late '60. By then the 1nS/ft propagation limit
was becomming obvious to designers (ask Cray!).


Allison
Received on Mon Oct 19 1998 - 17:55:29 BST