these RTL or what? - test-drb@ccmp.vtda.org

6 Oct 2007

...

Subject: Re: these RTL or what?
   From: shoppa_classiccmp at trailing-edge.com (Tim Shoppa)
   Date: Sat, 06 Oct 2007 08:40:15 -0400
     To: cctech at classiccmp.org, cctalk at classiccmp.org

Allison <ajp166 at bellatlantic.net> wrote:
  In the end ECL was a way to speed but always at
such a high system cost 
 and complexity it was often behind the curve for integration and delivery. 
It depends on what you're doing.

ECL was perfect for custom-built lab and military hardware. Follow
a few simple rules and even a bozo like me could reliably lay out the
PCB's. Contrast that with 74F technology where you couldn't even figure
out if ground at the center of the board was the same as the ground at
the edge of the board :-). 
Having used most logic from transistors to ECL100k and some oddballs 
inbetween I'd agree. ECL was excellent for mixed signal and fast 
front end stuff.  My favorite uses were programable /n for PLLs and
frequency counters.  ECL was far nicer without ground noise and some 
of the transmission line difficulties that the faster/fastest TTL 
(and CMOS) were really nasty driving. Ringing and reflections on a
board, bus or interconnect could really ruin your day. That made ECL 
nice for fast intersystem connects were the cables were for reasons 
coaxial cables or other shielded schemes.

...
 Above the onesies-twosies level things weren't so
clear. There's a big
leap between a back-projector or array-processor made at the onesies-twosies
level and the world where VLSI becomes economical. Gate arrays helped
span this gap but that gap had been pinched to nonexistence by the mid-80's. 
The magical thing that really impacted logic design indirectly be 
it discrete transistors or the fastest of the fast was simply size.
The faster the logic was the closer all the sourrounding bits had
to be to capitalize on it.  Otherwise the rule of thumb of 1nS/ft
took over never minding load capacitances.  Witness the Cray round 
machine (YMP?). There is a long history of systems compaction, 
cooling and speed interactions in computers.

...
 Minicomputer makers like DEC who also had their own
fabs were in an odd boat...
the process-leading CPU chips couldn't utilize the fabs built to
deal with them because DEC didn't sell enough CPU's. In the end
a vast army of interface and peripheral chips seemed to keep things
churning well enough that they kept their fabs for many many years past
where I was convinced they couldn't be economically viable. 
Bingo there.  While a few chips were economically successful it was
only with the help of silicon foundries like WD, SMC and AMD to get
needed volumes. In the end the greatest value of silicon hill was
in its sale with a few licenses kicked in. 

Allison