Hans Camenzind's (free, downloadable) Book

Kevin Aylward wrote: [...snip garbage...]

You need to just stick to whatever simpleminded and non-discretionary work that dead-wood laden cesspool Texas Instruments gives you, and don't worry about anything else.

It is clear that you suffer from severe brain damage and arrested development from too many years of excessive substance abuse; and , yes, it is an established scientific fact that many of the hallucinogenic compounds do in fact arrest intellectual and emotional development so that you remain at age 16 or whenever it was you started- no matter how old and decrepit you become physically.

Sno-o-o-o-o-rt! ROTFLMAO!

...Jim Thompson

--
|  James E.Thompson, P.E.                           |    mens     |
|  Analog Innovations, Inc.                         |     et      |
|  Analog/Mixed-Signal ASIC's and Discrete Systems  |    manus    |
|  Phoenix, Arizona            Voice:(480)460-2350  |             |
|  E-mail Address at Website     Fax:(480)460-2142  |  Brass Rat  |
|       http://www.analog-innovations.com           |    1962     |
             
I love to cook with wine.      Sometimes I even put it in the food.

Oh...

Yep. Unable to refute any of my technical arguments (e.g. violation of probability theory) so its down to personal insults as your only avenue of recourse left. Good one. Very original method.

Kevin Aylward snipped-for-privacy@anasoft.co.uk

SuperSpice, a very affordable Mixed-Mode Windows Simulator with Schematic Capture, Waveform Display, FFT's and Filter Design.

I think whether electronics is on par with physics debate somewhat boils down to the question of whether Art is is on par with Science. Both Art and Science are Important and Successful. Whether we like it or not, there is no question that a lot of engineering judgements made as to why we selected a particular circuit or architecture boil down to intuition/guesswork/prior-experience etc. that cannot be nailed down to logical arguments. The creativity involved is akin to artistic creativity of humans esp. those who are handicapped/challenged. The situation seems to be getting even more difficult nowadays when analog designers(including those doing IC/VLSI) are expected to design circuits working in a very hostile digital environments of supply and substrate noise, thermal gradients, emi,rfi etc. In a way successful analog designers today are even greater than the "greats" of the past(names like Widlar etc.).

Nevertheless, I would agree that a lot of well known circuits with names appended to them: e.g. "Darlington pair", "Wilson" or "Widlar" CS appear to be trivial.(An analogy would be "Rolle's theorem" in contrast with say "L'Hospital's rule" in any calculus book :) ).

The interesting question from an analog designer's perspective is, is designing an 18-bit self-calibrated "embedded" A/D converter more difficult/challenging(and more deserving of his time and energy) than analog neural network that can do difficult computational tasks with very low power?

Regards, Subhajit (An analog designer)

Its not particularly meaningful to compare the two - why stop there, why not debate pottery vs euclidean geometry....

But I certainly disagree strongly with your suggestion re. the "art" aspect. Its precisely the fact that its NOT art, its just rigorously applied (often fairly basic, usually classical) physics that makes analogue electronics so much fun. The best analog designers I have met/corresponded with/read their books are all very good "applied" classical physicists (listen to the physicists complain about that). Read any decent text on analogue design, and you pretty soon discover that its all just maxwells equations (even thermal stuff :). Good analog designers know this, and they know all the approximations (lumped vs distributed elements etc) that are in common use, and therefore when and where to apply them (or not).

the not-so-good analog designers never really know what is going on, and so stuff like "RF" is this scary, magical so-called art. Good experience is learning how relevant performance characteristics actually wind up in the real world (eg hard-to-tune circuits that are highly sensitive to tolerance, low or no noise margins etc).

obviously the detailed semiconductor physics is more complex than say faradays law (oh yeah - try complex 3D geometry with nasty sharp corners :), but again (all bar bleeding-edge research) is just physics, so can be learned, understood and applied.

IMO digital design is straightforward for good analog designers, who immediately look at L,C, dI/dt, dV/dt, impedances, reflections, crosstalk......although designing complex analogue circuits to work in the presence of digital "mess" is of course complex (but still not "mysterious")

I dunno about you, but I will design pretty much whatever I get paid to design - cash is good.

I like power supplies :)

Cheers Terry

I don't see that as relevant to the issue I am addressing. The deal in electronics is that is is engineering. It is not dealing with

*fundamentally* new things. It is only applying existing knowledge to new situations. its a different thought process. One is trying to figure out, how do I use what I have. The other is how do I create a new have.

Note, I am not implying that the "worth" of physics is better than electronics. I also agree that there is a certain amount of arbitrariness in all of this.

Oh but there is a question.

Thats my point, they can be.

Design is not magic. Its essential method is easily explainable.

Oh dear...

Ah...I developed an dependant proof of L'Hospital's rule that I was quite chuffed at.

Kevin Aylward snipped-for-privacy@anasoft.co.uk

SuperSpice, a very affordable Mixed-Mode Windows Simulator with Schematic Capture, Waveform Display, FFT's and Filter Design.

I liked your analysis of HF injection in tape recording. I am continually amazed by:

a) how many things are amenable to simple analysis b) how few "engineers" can actually perform said analyses.... c) how many engineer-turned-managers seem incapable of understanding simple mathematical proofs.

cheers Terry

Thanks.

Having said that, making models of systems is still something that in principle is not trivial. I was blown away when I first discovered that, there was still debate prior to the Atlantic telephone line as to whether or not the transmission line equation used was correct. After the fact, the derivation of T Lines (RLGC) is pretty easy for anyone with say, 2nd year year EE degree knowledge, but the first time people were trying these techniques, it was not so obvious.

But I agree, that it is somewhat disconcerting that many with full 4 year EE degrees, are unable to do the most basic of small signal equivalent circuits.

Indeed. Managers just can't seem to understand that it is impossible to do tomorrows work, yesterday.

Kevin Aylward snipped-for-privacy@anasoft.co.uk

SuperSpice, a very affordable Mixed-Mode Windows Simulator with Schematic Capture, Waveform Display, FFT's and Filter Design.

I do not see a fair comparison between Analog Design vs. Physics and pottery-making vs. euclidean geometry. The former pair includes an interplay of classical and quantum physics(90 nm MOSFET gates offered by state of the art fabs conduct current) with classical itself being subdivided between domains of electro-statics/dynamics(Electrostatic discharge(ESD), rf design, emi,rfi etc.),"EE", thermal and mechanical engineering(bandgap voltages can change with strain in package) and even some difficult applied mathematics("tones" in quantization noise spectra of delta-sigma converters). Whereas the latter pair pretty much involve just classical geometry. I.e. there is a enormous,enormous difference in complexity between analog design(esp. semiconductor microelectronics variety) & p.m.. However, I would assert that both analog design and p.m. involve "artistic" creativity and temperament.

I certainly agree that analog(ue) design can be fun. However, haven't had too many opportunities to apply formulae of classical physics directly in a design except for stuff like calculating mosfet transconductances, coupling caps. , self/mutual inductance in IC's and the like. It has mostly been taken care of engineering models like lumped R,C and Spice models of transistors. Whenever the opportunities to apply Maxwell's equations arose e.g. in ESD in an I/O pad, the problem became so hopelessly difficult that it became the fun job of a specialist or a CAD/EDA tool designer and straight got out of my hand. This is not to even mention practical economic realities that my embedded PLL or A/D will be "re-used" in more than one layout/geometric configuration so that even if I could apply Maxwell's eqns.,I do not know the exact boundary conditions to use, if I wanted to compute the noise injected by the logic circuitry in the chip.

Would like to mention two good texts on analog IC design: (1) Gray & Meyer (2) Grebene. Don't see too many physics formulae in them.

I gave the example of an embedded PLL,A/D or I/O pad with ESD pad designed in Europe, done on TSMC fab(Taiwan) to be "re-used" by 2 different companies in US and Canada respectively.

I don't think I used the work "mysterious" and take it back if I did. Also disagree with the notion that digital design would be "straightforward" for analog designers. Its like saying swimming would be straightforward for gymnasts. Similarly, I don't think analog design would be "straigtforward" for physicists. Each of these professions require a fundamentally different temperament for one to be successful in them. I don't see how a Bob Widlar could become a successful Ted Hoff(inventor of first micropressor Intel 4004).

Good for you! But perhaps not always good for others :).

I like neural networks (tho' haven't had an opportunity to work on them).

I would like to conclude with some remarks made by Chris Mangelsdorf (Analog Devices designer) in ISSCC 2004 in which he said that there is something "Wagnerian" about analog design in the panel discussion on the relevance of analog design with the onslaught of digital. Check the eetimes report on this in the press.

I think fundamentally we agree that analog design involves mostly principles of physics, tho' my own take is that the practical application of these principles is so difficult, that only people with a certain kind of artistic creativity and imagination can be successful in it.

Best regards, Subhajit Sen

of

that

them

only

to do

not

thats because its not a terribly fair comparison. But it is a good example of art vs science. There is a great deal of science in pm, almost none of which is understood by the artists who make the really cool stuff. its also a smart-arse remark :)

Fabulous argument! Thanks for taking the time to state your case so well :)

The basis of my (albeit cheeky) "point" is that the science is a

*pre-requisite* - art alone doesnt work very well. Many so-called electronics designers have very little scientific knowledge - just read some of the astonishingly stupid questions people ask in this forum (note: stuff you know is easy. stuff you dont know is hard...). I have worked with many so-called engineers (with or without tertiary qualifications) who just dont understand the physics involved (although im not an expert in QM or semiconductor physics, by any stretch of the imagination)

Part of what we are discussing here is our old friend semantics. I dont like the term "artistic" - having been to many art galleries, I especially dont like the association with modern artists (but damn those surrealists could paint :). I think what we are talking about here are aspects like creativity, inspiration and MIOA (IMHO) LATERAL THINKING. slavishly cranking mathematical handles is very unlikely to lead to funky new designs, which invariably involve the aforementioned aspects.

but without the detailed understanding of the relevant physics involved, creativity etc. arent likely to get you very far - witness the reams of drivel produced by idiots who try and tap into "free space energy" by adding magnets to transformers - great for a laugh though.

a lot of what gets referred to as "art" (eg RF, EMC etc) is NOT art, merely physics that is not understood by those making the reference.

aspect.

(often

so

their

design,

use,

the "except for" bit is the fun bit :)

I work with power electr I use my knowledge of physics to complement modelling tools - I look at the physical geometry of a circuit, and calculate (or often simply estimate upper and lower bounds of) stray inductance and capacitance thats likely to occur, then go feed those back into my spice sims. I like to ask "what-if" questions - how much L (or C) can that node tolerate before performance is degraded etc.

OTOH those without a deep understanding (which doesnt have much to do with education) of the relevant physics tend to blunder on unawares, creating sims that work well and circuits that dont. which is how I make most of my money :). pretty much all books on EMC say the same thing: analyse the circuit you actually build, not the one you think you built.

Actually I think that might be what Kevin was getting at - while the individual details are often nasty to calculate (esp. analytically) its all very well known. mind you, things like cavity resonances will stump most so-called engineers (many of whom dont even realise that a lumped model is an approximation).

partly thats just statistics though - normal distributions abound, in any group of people most are clustered around average (ie mediocre) and small numbers are either very good, or very bad. true for engineers, doctors, drain-layers etc.

have the former, not the latter - whats the title/isbn, got a review? (I collect eng books)

G&M dont mention ohms law, either. they expect you already know this stuff...

so

the

tolerance,

:),

all fairly well understood though. cf quantum cryptography, which is only starting to make the transition from physics to engineering...

the

"mysterious").

you didnt use that word, I did (sorry :). "art" implies mystery though - artists tend to be unable to explain their actions. Engineers bloody well better be able to explain it, lest they get their ears singed off at a peer review - who cares about the inspiration, what physics enables it to WORK.

I do however think digital design is merely a subset of analogue design. except when it turns into software, which IMO is not designed at all, merely cobbled together at high speed by over-qualified typists. Now SW is *art* - and bloody unreliable as a result - but its not the discussion we are having so I wont rant any further....

digital design is interesting in that it is not (at a larger scale) based on physics principles (ignore edge-related dynamics, meta-stability etc) - consider state machines for example, or von Neumann architecture - no classical or quantum physics hiding in there. there are correspondingly large numbers of digital "designers" with no understanding at all of electronics physics. which is fine when writing VHDL code, but tends to fall down when the first signal leaves the chip.....

I do a lot more analogue than digital design, but have done a lot of DSP, and use my analogue skills to great effect - spend a lot of time figuring out what to do, rather than start typing.....then figure out how to *prove* it works, stress-test it (eg throw junk data at it, watch it squirm) etc. whilst I am not fluent in VHDL, I have yet to come across ANY element of digital design that i cant handle with the toolset analogue has given me. The converse is NOT generally true for those who live in the digital world - a problem exacerbated by the fact that it is very cheap for universities to churn out EEs who only ever simulate things and write code - low resource requirements = high student numbers = more $$$

I am a realist, and I like to think I have helped a number of companies who have been stung by hiring idiots - if they know nothing about the field, its not hard to convince them you are an expert.....Hell, I did some work a few years ago for a company that wanted to make power supplies. Not one person knew ANYTHING about electronics, yet they spent $millions of VC and designed this thing that didnt go. I made it work, but the job turned into a nightmare (they wanted me to then sort out their thermals, then mechanics, then production line, all without increasing the funding - no way). I ended up having a hush-hush meeting with the primary VC guy, and told him that he was wasting his money on these guys, they would NEVER build a single working unit, even after I gave them a full functioning prototype. He was very grateful, and ended up pulling the plug (after wasting about $2,000,000). Bad for the idiots, but they were in the process of crashing and burning anyway. And if I ever need VC, well I now have a friend...

estimate upper and lower bounds, then triage your problem: it is either

a) no problem at all b) clearly a problem c) in-between therefore examine it more closely/accurately

FEA is definitely the way to do complex things, the maths is just too hard (and therefore too easy to screw up).

I guess I do agree with you (apart from the "a" word :), but think the creativity etc has to be predicated on a solid theoretical background

Cheers from Aotearoa, New Zealand Terry

Thanks.

I wish I could do more to apply more fundamental knowledge i.e. physics. The reason for this is driven as much from economics as from from "scientific aesthetics". I do not wish to spend too much on $$$ expensive EDA/CAD tools except perhaps for a good version of Spice.

There seems to be an increasing amount of "power electronics" within IC's too esp. with the need to reduce/optimize power and many of the concepts are are borrowed from box/board level power electronics. I am sure there are opportunities for applying your knowledge and experience in these. One "cool" design I have seen recently is a regulator that extracts RF power from coils surrounding the chip and using it to power the on-chip circuitry. The RF signals also contain modulated data.

Sorry it is: "Analog Integrated Circuit Design" Alan B. Grebene

1972, Litton Educational Publishing ISBN: 0-442-22827-9

I also personally like Roubik Gregorian's "Intro. to CMOS Op-amps"(Wiley) ISBN 0-471-31778-0, for CMOS analog design.

Digital (IC) design is now the domain of computer scientists and there is very little "EE" content. Wonder why EE depts. put up parallel courses on digital design when computer science depts. could teach them better. Perhaps the reason is digital chip designers will still need analog skills of the highest order to do the physical design( implement the state of the art PLL's and I/O's circuits).

My last comment is more philosophical(and controversial). More than pianists, "sitar" players, "martial art" practitioners etc., analog designers need a "guru"(teacher, mentor etc.) who serve as a role-model and inspiration throughout his career.

Thanks for the opportunity for a nice discussion.

Regards, Subhajit initiation.

Hi Subhajit,

message

technology.

lots

to

The

first

why

example

of

also

:)

some

stuff

many

dont

like

dont

could

cranking

which

adding

merely

electronics

with/read

the

(even

the

common

I use simetrix, and I am very happy with it. It could be better, but the price is right, and it has a lot of handy features for parameter stepping and the like.

of

importance,

diffusion

on

good

concepts

now thats cool. I read an article the other day that pointed out in a few years moores law will ensure chips have power densities approaching the surface of the sun.....thermal engineering is becoming a show-stopper. All sorts of funky technologies like closed-cycle evaporating coolers etc.

models....I

didnt

spice

the

to

"what-if"

is

with

my

all

is

any

small

thanx, added to my list of books to keep an eye out for

and

experience is

in

say

corners

be

only

noise,

in

well

peer

WORK.

merely

*art* -

having

based on

fall

yeah, VHDL made life a lot easier. I picked power supplies to try and make my career a little less self-obsoleting. Power distribution infrastructure changes very slowly, so I am likely to have work for a long time, without needing to constantly study the latest man-made-mess (god, imagine being a windows programmer - not thanx!). Besides, far more programmers/digital "designers" are churned out of universities than EEs, so even more job security :)

DSP,

figuring

*prove*

etc.

me.

world -

to

resource

"Widlar"

more

to

who

its

few

person

designed

mechanics,

ended

he

working

$2,000,000).

hard

I was very fortunate that my first *real* job was designing huge (1kW - 1MW) AC motor controllers. I worked with a small team of extremely competent engineers, and learned more in 3 months than all 4 years of my degree. After I had been there for 6 months the "gurus" all quit, leaving me the 3rd-most senior engineer, and the only smps designer. I worked my ass off for about 2 years, learning a huge amount - in addition to new designs I also had to support the entire existing product range - whereupon I started to discover mistakes made by my "gurus" - very important that step was, it started me learning that EVERYONE makes mistakes, and a smart engineer plans for it :). I am thankful for all the screw-ups I have been involved with (more so the ones that werent my fault :) as they taught me so much.

cheers Terry

Hi Terry,

Just wanted to know what part of NZ

************************************************* Roy Hopkins :-)

Tauranga New Zealand

*************************************************

1MW)

After

3rd-most 2

discover

:).

--
Outgoing mail is certified Virus Free.
Checked by AVG anti-virus system (http://www.grisoft.com).
Version: 6.0.706 / Virus Database: 462 - Release Date: 14/06/2004

just over the hill :)

Terry Given Domes Engineering

109 Centennial Avenue Te Aroha New Zealand ph +64 7 8844 596 (voicemail) cell +64 21 422 400 (no voicemail) terry snipped-for-privacy@ieee.org

You could've deleted 90% of the crap that was above.

(1kW - 1MW)

When they upgraded the air conditioning in a bldg they installed those Danfoss variable speed controllers. Now when I go to the comm room, next door to the A/C room, all I hear is this high pitched whine coming from the motors and wiring. Like screeching! I dunno who did that design, but it sure is noisy!

yeah they can (and often do) scream like stuck pigs. We always used frequency "slushing" - pseudo-random variations of around 10% of switching frequency turn an annoying screech into a more pleasing hiss, by spreading the total acoustic energy over a wider band. Similar approaches are used to pass EMC tests, because of the averaging behaviour of the test instruments.

We once made up a demo for a trade show that performed "silent night" entirely with motor acoustic noise. I looked into direct paralleling inverters with a single motor, to then generate chords (we are talking 1MW drives here :) as the currents add linearly, but we never did it as we had actual work to do....I thought we should have made the motor say things like "help Im trapped inside this motor," but one of the programmers wired an analogue input up to the waveform generator, so we could feed an audio signal into our 250kW "speaker" - ah the fun we had :)

I also built a modbus driven "clonk" generator once, with a 250kW vector drive - I added serial comms to a very heavily loaded processor. A data burst meant the micro ran out of time, and interrupt priorities were such that the serial won, and the field orientation lost. Re-orientation was accompanied by a VERY loud "clonk" from the machine, the pattern of which depended on the data :). I had to change to a faster micro.....20MHz

80C196KC vs 16MHz, which solved my problem.

Cheers Terry