Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.1+some 2/3/84; site dual.UUCP
Path: utzoo!linus!philabs!cmcl2!seismo!ut-sally!mordor!dual!paul
From: paul@dual.UUCP (Baker)
Newsgroups: net.arch
Subject: Re: "When to go to CMOS", or, "Is Schottky dead"
Message-ID: <836@dual.UUCP>
Date: Thu, 11-Oct-84 12:13:05 EDT
Article-I.D.: dual.836
Posted: Thu Oct 11 12:13:05 1984
Date-Received: Sat, 13-Oct-84 01:11:59 EDT
References: <537@turtlevax.UUCP>
Distribution: net
Organization: Dual Systems, Berkeley, CA
Lines: 57


> Myriad new high-speed CMOS IC's are coming out from lots of new
> startups as well as some of the old fogies.  AMD, Analog Devices,
> Cypress, Integrated Device Technology (IDT), International Microcrcuits
> (IMI), Lattice, Logic Devices, TRW all have LSI devices (memories,
> processing units), pending or available.  Other companies, such as
> National, TI, Zytrex, RCA, have got high-speed CMOS MSI chips
> (registers, gates, decoders, etc.).

Most of these high-speed CMOS circuits appearing seem to be about half
the speed of LS parts.  It is is a bit hard to tell as many of the CMOS
manufacturers only publish "typical" specifications.  These are quite
useless for design purposes as you are not guaranteed to only get
"typical" parts.  They do appear to be quite useful for slow speed
parts of circuits which are not connected to external buses, where it
is desirable to conserve power.  In looking at retrofitting some of my
existing designs, it is surprising just how few parts are candidates for
this.

Zytrex did have a lot of double page advertisements headed "Who killed
Schottky dead?", with, as usual, no indication of the parts they produced.
After many further enquiries, it turns out there are supposed to be two
speed selections, one about as fast as LS the other about as fast as S.
The second are completely unavailable and I'm still waiting for samples
of some the first types.

> It almost seems like it is near the time when a design may consist
> entirely of CMOS components.

Not for a few years yet I think.  Where CMOS will do very well is in the
area of gate arrays and other large scale circuits.  It appears that most
of the slowness of the current high-speed CMOS is in the output buffers.
Also the range of parts available in high-speed CMOS is very much smaller
than for LS or S or even F series of TTL.  This is particularly noticeable
if attempts are made to get parts.  At the moment many 20 pin, 8 bit parts
are available (HCT373, HCT244, etc).  I believe HCT00s and simple parts
are also available, but you do not reduce power consumption much with these
as the simple LS parts do not consume much anyway.  I am still having
trouble getting samples of HCT174s and HCT393s - not exactly high-tech
but not exactly available either.

> How close is high-speed CMOS to price parity with S or LS?  Some of the
> components (MSI) are dominated by packaging costs, so should be no more
> than 5% more expensive than their bipolar counterparts.  How about LSI
> components?

Nearly all parts costs are dominated by "what the market will bear" rather
than by any real costs of manufacture.  At the the moment high-speed, or as I
call them High-Cost, family of CMOS parts seem to cost about three times the
equivalent LS parts, sometimes more.  For example a 74LS670 has a cost of
about 70 cents.  I was quoted about $5.00 for a CMOS TTL compatible version.
Another thing to consider is that many of the less common parts are only
made by one manufacturer, sometimes a rather obscure one.  Purchasing 
departments tend to give engineers a hard time about specifying parts like
this.

Paul Wilcox-Baker.