[sdiy] design experts

[Magnus Danielson]

From: "Czech Martin" <Martin.Czech at Micronas.com>
Subject: RE: [sdiy] design experts
Date: Fri, 15 Aug 2003 10:46:56 +0200

> Ok, I know some guys who have no "high level education",
> i.e. a Ph. D. or M.S., but are very experienced, so
> most of the time they can solve most problems.
> 
> So with a lot of experience you can get something without
> too much theory.

Especially if the experience is varied.

> But there are times when this is not enough.
> If you have never thought about high frequency (i.e.
> lumped elements do not hold, but traveling waves),
> Maxwells equations (sounds scary, but can often be cooked
> down to simple arangements where they are easy to solve,
> gives good insight in capacitance, inductance and radiation
> problems, coupling problems) and the like,
> some problems remain "black magic" for ever.
> 
> Without solid theoretical background you have extreme
> problems to understand and plan experiments, as well
> as interpreting the results in such situations.

Um, I beg to differ with you here... you can come a long way without a "solid
theoretical background", but there are a few things you really must learn.
... and I know alot of people which should have this "solid theoretical
background" that fails to have an intuitive understanding about how many
things work. They passed the exam, but that's about it.

Maxwells equations are marvelous when you understand them, but except for a few
they don't have much bearing on real work. A number of rules of thums on whats
happening and being able to reason around them as a basis knowing the existence
of the Maxwell equations is what kicks most of the times.

Few MSc are of much use when they come out of school. Often one has to teach
them how things are done and not done. They are not able to accept overall
designs based on rule-of-thums and some head-counting but must simulate the
hell out of the problem and deliver a 30 page report to come up with another
result - and naturally it doesn't work as adverticed when it comes mounted into
the lab. This is why I sometimes say: "If you can't solve a problem with pen
and paper, then it is not worth solving!" meaning essentially, many times you
can solve things easier by rought methods, but it takes experience in both
practice and theory to know what works and what doesn't both as a design and
as a short-cut in analysis.

> Example: some people do not know magnetic fields and inductance.
> (I have to admit that I did not really understand in 
> University. Now, after 12 years of more thinking I understand better).
> They ask me for specs for output load "30pF".
> So I keep telling them that:
> -you need to take bondwire and PCB inductance into account
> -you may need to model even a transmission line
> 
> but to no avail. They do not understand.
> 
> The outcome is that I/O drivers are overdimensioned,
> which gives neat curves/rise time for a pure capacitive
> load, but a "ringing" desaster for complex RLC impedances,
> not to mention strip lines etc.

Too quick risetimes is a hassle! It lowers signal-integrity and it increases
EMC problems, two supprises in one!

> This is also unfortunate if the people who design PCBs are
> only educated as draftsperson (?). So basically they do not
> know what kind of disaster they are doing, the result is
> very often a terrible S/N of the IC in system, where it is
> good in our own design evaluation boards.

I agree. PCB-layouters usually don't have the skills needed for the work they
really do. I have yeat to meet a PCB-layouter that "does the right thing" on a
PCB straigh away. On the other hand, it is really crazy that "Schematic design"
and "PCB layout" is such a parted field. I blame the totally incompetent and
crappy tools since you need to be expert in how to use each set of these tools.

As for your own design evaluation boards, that's dangerous... I've often found
datasheets and appnotes of IC-manufacturers lacking relevant information for a
good analysis and understanding of a chip which is needed for a good design.
It's sad but true.

Cheers,
Magnus