Trying to understand how to design circuits

[PeteS]

There are, of course, a large number of exceedingly good replies (from
well known and exceedingly good engineers) amongst the noise of some
trolling.

I look at design as where art and technology meet. One must have an
understanding (indeed, an intimate knowledge) of the fundamental
theory, but there is a huge amount of art in any non-trivial design.
How do I decide just which method I am going to use? Large parts of
experience, and lots of bouncing ideas around in my head to figure out
just what is going to work for me *in this situation*. It's rather like
painting a picture - what am I trying to portray or achieve with this
exercise? Ultimately, that's what leads me to 'put a resistor there,
choose that diode here' .

As someone said, all designs are tradeoffs (and that's true whether
they are digital or analog) for various factors.

Even apparently simple designs may have been considered for quite a
while (one might argue that the simpler [more elegant] the design *for
a given task*, the more difficult it was to design, but ultimately more
satisfying to the designer).

That said, there is no 'method to design' that would work for everyone
- that's why I consider the design side an art, not a technology - we
*use* technology, we live design [well, I do ]. Of course, we also
design the original technologies on occasion On other occasions we
study it to understand it so we may modify and use it. (There must be a
recursive in there somewhere...).

Cheers

PeteS

 

[infinite]

I can't sit by and see such a grave injustice go answered. If the
regular designers here are guilty of *anything* it's only forgetting
how tough a subject this is to get to grips with adequately.

Thats a good point. But how many regulars here admit that.

There are
plenty of demonstrably genuine experts here. I smell a troll
hereabouts...

Well face facts ... the regulars who talk the most spend more time
trolling, insulting those who actually understand the subject, then
getting down to talk about real electronics. Most of their answers are
easily got with a Google search and most times you find better answers
using the same method.

Rubbish. It's one of the very best!

Check up some serious analog /design design books, none recommend the
Art of Electronics and neither do most engineers I know. But you have
a right to an opinion as I/all do.

infinite

 

[lemonjuice]

HohiHohiHohiHohiHohiHohiHohiHohiHohiHohiHohiHohiHohiHohiHohiHohiHohiHohiHohiHohiHohiHohiHohiHohiHohiHohiHohiHohiHohiHohiHohiHohiHohiHohiHohiHohiHohiHohiHohiHohiHohiHohiHohiHohiHohiHohiHohiHohiHohiHohiHohiHohiHohiHohiHohiHohiHohiHohi

Now who is laughing

 

[Winfield Hill]

John Larkin wrote...

Barrie Gilbert did a chapter in Jim Williams' first (1991)
Analog Circuit Design book [1] ...

[1] my copy of which, improbably, is autographed by Bob Pease.

Now you'll have to get Jim to autograph your copy of Bob's book.
I do wish Barrie Gilbert would write a book. His papers,
occasional book chapters, and patents are entirely too short.
And it's too bad Robert Widler didn't write an IC design-tricks
book. All we have are a few papers and NSC design notes.

 

[Jim Thompson]

On 28 Dec 2005 07:46:25 -0800, Winfield Hill

[snip]

And it's too bad Robert Widler didn't write an IC design-tricks
book. All we have are a few papers and NSC design notes.

Was he ever sober enough to write a book ?

...Jim Thompson

 

[John Larkin]

With all due respect John, I would have to see *proof* that such "new"
topologies actually were superior to the well trusted existing ones.

Well, they did things that no oscillators that I'm aware of can do.
One, for example, oscillates at 800 MHz, has a TC of a few PPM/K,
starts instantly - within 1 ns - of an external trigger, then outputs
clock edges regularly on schedule with picosecond accuracy. It can be
stopped and restarted in 5 ns or so. I know of no "well trusted
existing ones" that can do that.

Another is an LC oscillator, asynchronously triggerable in 2 ns and
stoppable within one cycle, that has the longterm accuracy and jitter
of the best crystal oscillators. I *know* nobody else has ever done
this commercially.

Around my place, we do stuff like this all the time, because we
believe it's possible.

The number of times I have examined "new" circuits that actually achieve
no net benefit from an existing circuit are too numerous to mention.
There are many convoluted designs that achieve precisely, nothing. Most
people fool themselves. It is very hard to be objective about ones own
work.

Maybe nobody shows you the really good ones.

Its very unlikely that anyone can come up, today, with a real useful
topology that hasn't already been looked at. Sure, it does happen, but
not very often. Too many people have looked at this stuff for way too
long.

Disagree. New parts allow new things to be tried. uPs, dense FPGAs,
screaming ADCs allow architectures to be put on a board that would
have been ludicrous 5 years ago. These architectures need circuits.

I don't. The opportunity for a really new circuit that actually has new
value is < 1:1000.

Well, I guess I'm doing all of them, and for every one I do, you poor
drones are stuck with copying and tweaking the other 999.

John

 

[John Larkin]

Hi all:

I am trying to understand the process by which the design of circuits
is carried out.
How from the white paper begins the design of a certain circuit?

In many other areas one understands that there is a certain
structure, a certain order, a process. For example:

In the case of a writing one understands that there is a thesis, main
ideas, ideas of support, a conclusion. All this must be articulated
to achieve a certain objective. This is understandable.

In the case of a car is understood that exists the motor, the
electric system, the chassis, the panel, all they complying a
certain function and thus in many other areas as the software, Civil
Engineering, etc.

But in the case of the circuits all seems very confused (at least for
the novice). One doesn`t know how someone decided to put a resistor
here, there a diode, or a capacitor over there. At times seems that
certain circuits were discovered by accident. Which is the center and
which the periphery.

So the question is:
Once one has certain know-how of electronics as the funcion of the
components, the basic theory, etc
¿How to proceed from the white paper to go building a certain
circuit?

How to decide where to put a resistor, a diode, a capacitor, etc?

Thanks in advance by any comment.

Barrie Gilbert did a chapter in Jim Williams' first (1991) Analog
Circuit Design book[1], where he talks about this very issue, "Where
do little circuits come from?" Highly recommended, his bit and the
whole book.

*ALL* "new" design is a randam variation from an existing design.

All of yours, maybe.

If the design were *all* new, it would have say, no diff pairs, no
cascods, no source followers no etc, that is, it could only be an
aimless connection of component terminals, and could not possible
achieve anything.

Am I allowed to use resistors? Wires? Or would that make it
derivative?

If the new design had no random component, it would, by definition, be
derivable from existing designs, in which case it couldn't be genuinely
new. Random generation is the only way to produce a non derivable
result. If it is random then we have no control over it, by definition.
Its random. The brain is a Darwinian machine, and that is how it
produces "new" designs. Copying, Selection and Randam variation is all
there is.

John, you seen to think that there is merit in coming up with something
new. Why?

Because it's fun? Because it's profitable? Because it's beautiful?

Because it amuses me.

The brain can only do this by a random process, so what's
superior about generating something by accident? For example, it way
harder to copy a complicated arpeggio and play it fast, then to generate
new music. Its piss easy to hit some random notes on a piano.

Damn, what a bleak opinion you have of yourself.

John

 

[Winfield Hill]

Jim Thompson wrote...

Was he ever sober enough to write a book?

Yes, of course. Hey, the book probably would have been better
when he wasn't. BTW, since we don't have Widler any more, we'd
be pleased to see you write a comprehensive IC-design how-to,
with ideas and tricks on every other page. Playing second best
to a non-existant Widler book isn't bad.

And, it'd be a far better use of your time than trashing me and
other anti-Bush folks here on s.e.d. After you pass on, your
book will be a grand legacy, whereas your trashing...

 

[Jim Thompson]

Jim Thompson wrote...

Yes, of course. Hey, the book probably would have been better
when he wasn't. BTW, since we don't have Widler any more, we'd
be pleased to see you write a comprehensive IC-design how-to,
with ideas and tricks on every other page. Playing second best
to a non-existant Widler book isn't bad.

And, it'd be a far better use of your time than trashing me and
other anti-Bush folks here on s.e.d. After you pass on, your
book will be a grand legacy, whereas your trashing...

I could even write about Widlar so drunk he ended up sleeping sprawled
out on Jim Estep's living room floor... much to Jim's wife's horror
when she got up the next morning.

Or the time Widlar and Joel Karp got picked up for DUI by the San Jose
police... but let go since the brother of one of the COP's worked as a
technician for Widlar.

...Jim Thompson

 

[Rich Grise]

On 28 Dec 2005 09:43:32 -0800, Winfield Hill

I could even write about Widlar so drunk he ended up sleeping sprawled
out on Jim Estep's living room floor... much to Jim's wife's horror
when she got up the next morning.

Or the time Widlar and Joel Karp got picked up for DUI by the San Jose
police... but let go since the brother of one of the COP's worked as a
technician for Widlar.

Or, you could set your political issues aside, and write a killer book
on analog ASICs, or whatever suits your fancy - it's clear you have
many years of experience at that stuff, why not share it?

Good Luck!
Rich

 

[Joel Kolstad]

Or, you could set your political issues aside, and write a killer book
on analog ASICs, or whatever suits your fancy - it's clear you have
many years of experience at that stuff, why not share it?

I'd vote for a chapter, "Reminisces on drunks, liberals,Attila the Hun, and
other things that tickle my fancy" so long as Jim wrote it _after_ finishing
the hard-core stuff.

For those who aren't aware of this, Hans Camenzind -- the designer of the
555 -- wrote a book on analog IC design that's available for free download at
http://www.arraydesign.com/downloads/index.html or can be purchased in print
form at Amazon. I vaguely recall that Hans hung out on sci.electronics long
ago, but it's been years now since I've seen a post from him.

---Joel

 

[Pooh Bear]

I don't know about that.

When I was a kid I DID have a lot of "hack around" time in my dad's
radio and tv repair shop.

However I WAS obsessed with mathematical analysis of circuits, and
probably remain somewhat that way after all these years.

Take a look at my analysis of the MC1530 OpAmp. I designed that
circuit through mathematical analysis, then built a breadboard to
prove it, then integrated it... I was 23 years old at the time.

...Jim Thompson

I reckon mathematical analysis is fine once you've mastered a level of
practical competence.

How many times have we heard here that it worked in Spice but won't in real ife
?

Graham

 

[Jim Thompson]

I reckon mathematical analysis is fine once you've mastered a level of
practical competence.

How many times have we heard here that it worked in Spice but won't in real ife
?

Graham

Yep. The practical competence allows you to judge the worth of the
simulation.

...Jim Thompson

 

[John Miles]

For those who aren't aware of this, Hans Camenzind -- the designer of the
555 -- wrote a book on analog IC design that's available for free download at
http://www.arraydesign.com/downloads/index.html or can be purchased in print
form at Amazon. I vaguely recall that Hans hung out on sci.electronics long
ago, but it's been years now since I've seen a post from him.

Thanks for that pointer. Ordered...

-- jm

 

[Mark Fergerson]

Hi all:

I am trying to understand the process by which the design of circuits
is carried out.

What makes you think there is "a" process? IME it's several
subprocesses running in parallel, but there's one thing I think we all
agree on:

How from the white paper begins the design of a certain circuit?

Real simple; you first decide where you want to get to, then decide
how to get there.

In many other areas one understands that there is a certain structure,
a certain order, a process. For example:

In the case of a writing one understands that there is a thesis, main
ideas, ideas of support, a conclusion. All this must be articulated to
achieve a certain objective. This is understandable.

Then you need to describe "where you want to get to" and "how to get
there" as concisely as needed; but "as concisely as needed" means
figuring out where you can and cannot compromise (identifying design
criteria). "Criteria" is from the same root as "critical" which here
means "what must I do/not do?".

In writing, must you use simple declarative prose, or poetry? Is
Standard English the "best" way to express something or would say Haiku
work better? Depends; who's your target audience, and what are you
trying to get across? Do you want them to "get it" right away or should
they think about it a while?

Where are you trying to get to, how many ways do you know of to get
there, and how many are you willing to use?

In the case of a car is understood that exists the motor, the electric
system, the chassis, the panel, all they complying a certain function
and thus in many other areas as the software, Civil Engineering, etc.

This kind of thinking (analyzing a problem by starting with known
solutions) is not designing.

You appear to be confusing design philosophy with pattern matching.
Pattern matching is part of designing but not the whole thing.

In the case of a car we start not with four wheels, an engine, and a
chassis etc. but with the premise that we want to make it possible to
get x number of people and y amount of stuff a given range of distances
in a given range of time. Then we explore some possibilities, one of
which is an independently-operated machine that carries a few people and
a little stuff over pre-existing roads with non-negotiable
characteristics. You know about I.C. engines and stuff like that, so you
may wind up with something very like a car.

But if "independently-operated" is considered compromisable (not all
that "critical" a criterion), then buses and so on enter consideration.
If you decide you don't _have_ to use the roadways then trains and
planes become design possibilities. If the terrain allows, maybe you
draw on your sailing experience and design a land yacht.

Target identification first, then criteria selection, then
possibilities. Then maybe some recursion over the criteria...

But in the case of the circuits all seems very confused (at least for
the novice). One doesn`t know how someone decided to put a resistor
here, there a diode, or a capacitor over there. At times seems that
certain circuits were discovered by accident. Which is the center and
which the periphery.

IME you start at the end ("where you want to get to") then explore as
many possible paths as you know about, then pick one that satisfies as
many of the design criteria as possible. When (not if) the boss tells
you "that won't do", you begin the compromise process. Figuring out
which criteria are really critical can be a pain. ;>)

So the question is:
Once one has certain know-how of electronics as the funcion of the
components, the basic theory, etc
¿How to proceed from the white paper to go building a certain circuit?

Sometimes it's a matter of personal preference, professional
experience, or what your design house is best set up to do. Sometimes
these things compete as in you occasionally have to discard "traditional
thinking" or be left behind.

How to decide where to put a resistor, a diode, a capacitor, etc?

Parts-level thinking is usually a matter of designing the blocks or
interfacing blocks (would putting a resistor _here_ speed this
oscillator up vs. what's the impedance of this transmission line at that
frequency?), but that's an experience thing.

OTOH YMMV. ;>)


Mark L. Fergerson

 

[Tim Williams]

PLONK!

Gee Jim, considering the difference in political opinion you two have, I'd
think you'd give puckerface (lemonjuice) a little credit on that ;-)

Tim

P.S. Was on vacation a few days. Took AoE with