Re: Xilinx legacy situation

Peter Alfke top-posted:
> 
> Tim, you have to get over the idea of still
> getting something from your old chip investment.
> Xilinx FPGAs have become 100 times (!) cheaper,
> have added functionality and better software
> support since the days when you bought the
> XC4013s.

It's not the chip investment that's the *big* hangup, but the
equipment investment.  The chips were chosen deliberately in
pin-grid-array package so we could replace as and when we wanted
- including when/if they got blown up by misuse.  Throw-away ICs
we can live with - even at the price of PGA 4013s.  Throw-away
experimental units is another ball game.

> ( Anybody who tries to hang on to a 10-year old
> computer faces a similar situation, albeit to
> a lesser extent). That's the price of progress.

Hmmm.  Doesn't explain why the majority of PC/104 processors are
486 clones rather than Pentium clones.  Doesn't explain why
8051s still sell by the truck-load.  Sure, we have to stay
current, but we've done some very useful work with 10-year-old
computers!  (Old DOS boxes make excellent targets for
introductory learning about embedded systems, provided you don't
mind the bench space they occupy.)

If it ain't broke, why fix it?  Some concepts can be taught
using kit that's a lot more than ten years old.  We have one
experiment that uses equipment which must be around 25 years
old, and part of the point is to show that important effects,
very relevant to designing the latest and greatest ICs and
systems, can be seen with basic testgear and almost rudimentary
test rigs.

If we can teach _currently relevant_ techniques of FPGA design
with ten-year-old kit (and the FPGA experimenter kits are only 3
years old), why should we have to throw the baby out with the
bathwater?  If the DESIGN software still supports XC4000E, why
has the downloader stopped supporting a download cable which was
still being sold only a couple of years ago?  If the library
files are still there, why can't the software be set up to
access them?  (After all, the guts of the software isn't the
Windows front end with the selection boxes - or have I
misunderstood all those command lines scrolling through, which
appear to show someone using 10-year-old DOS to do the hard work
rather than shiny state-of-the-art Windows?)

> Your biggest stumbling block is the 5-V
> compatibility, which stops you from using
> really modern (and cost-effective and
> sophisticated )

Look, these are STUDENTS.  Second year undergraduates.  Some of
them only got their hands on an oscilloscope for the first time
just over a year ago.  Some only SAW an oscilloscope for the
first time thirteen months ago.  They are designing simple state
machines and the like.  The prime exercise culminates in
controlling a three-storey model lift.  That could be done by a
Xilinx 1000 series device.  For the PURPOSE, we don't need
modern, we don't need sophisticated, and spending money when we
have something already is certainly not cost-effective.  We're
NOT designing for production - where we have project students
and researchers pushing boundaries, of course we use
state-of-the-art.  But at the moment we find 5V 74 series great
for teaching gates and discrete logic systems, so we have a
range of 5V I/O units which those systems can work with.  And,
to benefit both us and the students, we re-use those units. 
Students produce traffic light controllers using 74-series
logic, using a PLD, using a PIC.  They control a lift with an
FPGA and with a PC/104 system.  We're trying to turn out rounded
engineers, who understand there are options when confronted with
a requirement.  Options come with baggage.  Later in the course,
they learn of the baggage which comes with picking a 5V option. 
But up to second year, we're happy for them to swim in a
uniformly 5V digital environment.

> devices. Sooner or later you will curse the
> @#^%$*!  5-V standard.

5V has lasted longer than any other.  All subsequent standards
have been superseded - today's standard will be unusable by the
state-of-the-art ICs in three years.

If we'd gone 3.3V, would we not be cursing THAT?

> Why not do it now!

Because we teach a LOT more than just FPGA.  Believe me, we look
at the voltage issue every year.  So far, we've had insufficient
cause to decide that we'll throw out and re-cast the 21
exercises per student occupying 32 lab sessions over two years
of the course which use 5V circuitry.

-- 
Tim Forcer               [EMAIL PROTECTED]
The University of Southampton, UK

The University is not responsible for my opinions