[Author Prev][Author Next][Thread Prev][Thread Next][Author Index][Thread Index]
Re: gEDA-user: C++ HDL
John Doty wrote:
> On May 11, 2009, at 3:04 PM, Joerg wrote:
>
>> al davis wrote:
>>> On Monday 11 May 2009, Joerg wrote:
>>>> I just hope universities don't rely on that route alone when
>>>> teaching RF design.
>>> Why not? It's no worse than some of the other choices they
>>> make, including some that you vigorously defend.
>>>
>> Like what?
>>
>>
>>> Considering how much proprietary stuff they rely on, and by
>>> doing so they reject real opportunities for learning, System-C
>>> as a sole way to teach a particular part of the curriculum seems
>>> rather benign.
>>>
>> It's ok as long as they don't neglect real HW design and testing, with
>> solder irons and all. I have found younger candidates to be seriously
>> lacking in that respect.
>
> I agree, but I find it even more shocking that a 21st century
> engineer would not know programming. Programming skills are as
> important as literacy to a scientist, mathematician, or engineer
> these days. They cannot be left to specialized programmers any more
> than writing can be left to specialized scribes.
>
It depends on what your main work is. For example, for a mechanical
engineer programming isn't terribly important. Other than what can be
done with a spreadsheet or MatLab and such, but that's probably not the
kind of programming you meant.
For me (analog guy) it's a little more important but not nearly as much
as for a signal processing guy. We do a lot of stuff with <gasp> Smith
charts. So yeah, I can piece together some uC code but it's not my daily
bread and at the end of the day it's finished by others. Afterwards I
take a real good look at POR/BOR handling because that's a very sore
spot with many engineers these days.
>> There comes a time when us older guys can't
>> help them anymore. The topper was a professor at my university who
>> said
>> that soon everything will be ICs, that transistors and most of that
>> discrete stuff would go away. I burst into laughter in the
>> auditorium, a
>> bit embarrassing ...
>
> Well, in case you haven't noticed, it's headed that way. My CCD
> measurement chains of 10-20 years ago were full of discrete
> transistors. IC's were not up to the job at the power levels required
> for a space instrument. Present day versions have no discretes, but
> use less power, and are faster and quieter. It's physics: the scaling
> laws tell you that in most cases, smaller transistors with shorter
> interconnections are better. You can only go so far down this road
> with discretes. Still need a big power transistor? Those are mostly
> IC's, too: millions of tiny transistors in parallel.
>
For low power, mass market or super high budget stuff, yes. Problem #1
is most of the students in that auditorium believed it would go this way
for everything because the professor said so. So they went on to become
digital HW engineers, SW engineers and <swallowing hard> sales engineers.
Problem #2 is: How do you get the HV power switcher that I am going to
design next into a PQFP package? Or the UHF power stage that probably
follows? Or the laser driver I did last year? As chips they'd melt down
in milliseconds and leave one hellacious hole. Ok, you, I and some
others are still on earth and young enough to do discrete designs for a
while but in 20-30 years new engineers have to follow or we as a society
will have a real problem. I am seeing the first signs of that already.
> What hasn't changed is the need to understand the physics of
> individual transistors: if anything in mixed signal VLSI that's
> *more* important than it was in the days of discretes.
>
Yes, it's crosstalk city on there. Then electromigration which, in
discrete design, is usually of no real concern. Had to suppress quite
some cussing during my first chip design.
[...]
--
Regards, Joerg
http://www.analogconsultants.com/
_______________________________________________
geda-user mailing list
geda-user@xxxxxxxxxxxxxx
http://www.seul.org/cgi-bin/mailman/listinfo/geda-user