Math and electrical desgin

On Friday, March 27, 2020 at 3:29:36 PM UTC-4, snipped-for-privacy@columbus.rr.com wrote :

self an applied mathematician? I would say that I do. As an analog design er how much do you use various mathematical concepts. Basic electronic cir cuit design does not seem to require a lot of theoretical math, where as, signal processing requires a lot of math.

bra. I am on my 3rd pass ( of the first 10 lectures) of Gilbert Strangs on

-line MIT course. I think I am finally getting it. I really want to know linear algebra because I want to be good at matlab which is rooted in linea r algebra. However, is linear algebra used in circuit design other than an occasional solving of two or three simultaneous equations? Linear algebra, as in matrix manipulation? V=I*R is linear algebra. :^) Is the online course from a math prof.? I always enjoy the math more when it's partially applied. I had a course on 'advanced engineering math' taught by an engineer (evening classes) And it was great. I really like a concrete 'thing' to stick my equations to.

George H.

al processing stuff is definitely math intensive and required to be good at it.

per proficient at maxwells equations.

annot accomplish his task with out the math?

I agree except for the Williams. It used to be heavily used here but is now on the book shelf in the next office. Because the computer kind of took over filter design.

So now it's the bible, AoE, Unitrode IC Data Handbook, A Practical Guide to Linux, VLSI Design, a J&J first aid kit, then lots of my own binders. In that order.

Oh, anbd one of my binders contains the brwing recipes.

culate how good it _could_ be, (b) figure out how to get there, and (c) if that's a win, proceed, but if not, think up something better and goto(a).

a lot of lore and old fashioned crank-turning.

My favorite type of math is 'back of the envelope' calculations. Those are fun... but it is always better with one or two other people. (you can sorta check each other..)

George H.

It ranks surprisingly low on the pecking order.

Definitely not. My sister (has a mathematics degree) is razzing me about that all the time but that's just how I am. IMO engineering is mostly instinct. People who don't have them can't be good engineers. Instinct comes with practice, lots of practice.

I look them up if needed :-)

The rest is done by SPICE.

I think it requires a lot of gut feel, more so than math. The math comes later when fine-tuning stuff or when deriving a transfer function. The latter is sometimes tough for me. Start solving a lengthy equation, reach end of line, turn paper sideways, formula grows and grows, still not enough length, tape another paper to the right side, pencil lead breaks off again, cuss a little ...

My honest opinion after over four decades of analog design is that linear algebra is not super-important when designing circuits. Way more important is constant "radar vision", looking for potential pitfalls, noice sources, stability issues, SOAs, margins and, very important, have a cost calculator run inside the brain all the time.

Sorry, but I don't really agree and RF is my home turf. I learned a lot more from ham radio than in all the course at the university and not a lot of math was involved.

True ... but ... one must have a very good gut feel of how Maxwell applies to stuff. Otherwise the whole design could turn into a nightmare at the EMC lab.

It does happen but rarely. For example, when you need to decide whether to switch from frequency domain to time domain as I just had to do on a project. That required a boatload of math and solving. Can't say that I particularly enjoyed that part of the project but as John Wayne used to say, man's got to do what man's got to do.

Now, about the virus and the election ... :-)

te:

urself an applied mathematician? I would say that I do. As an analog desi gner how much do you use various mathematical concepts. Basic electronic c ircuit design does not seem to require a lot of theoretical math, where as , signal processing requires a lot of math.

gebra. I am on my 3rd pass ( of the first 10 lectures) of Gilbert Strangs on-line MIT course. I think I am finally getting it. I really want to kno w linear algebra because I want to be good at matlab which is rooted in lin ear algebra. However, is linear algebra used in circuit design other than an occasional solving of two or three simultaneous equations?

This online course is from Gilbert Strang, who I think I can say, is consid ered one of the world experts in fundamental Linear Algebra.

The thing that is tough about linear algebra is that there are not a lot of applications that continually reinforce your expertise on the subject. We use simple algebra all the time. It gets constantly reinforced. With lin ear algebra, you learn the various concepts , which are quite difficult - o r at least hard to sort it all out, and then it does not seem like there a lot of applications for the --get er done - type of engineer. I am convinc ed though, that mastery of the subject will allow new insights into differe nt problems.

gnal processing stuff is definitely math intensive and required to be good at it.

super proficient at maxwells equations.

cannot accomplish his task with out the math?

Abebooks is good for that type of stuff.

George

Notice how I said RF/radio. It is the radio part - modulation that require s the math stuff. About ten years ago I realized that I could not really c all myself a radio engineer if I did not have the basics of DSP down. Most of the radio - except power amplifiers/antennas and LNAs is getting done in DSP. It is a tough mountain to climb, but I just could not shake the idea that I could not call myself a radio engineer (in today's world---professi onally) if I did not understand DSP.

There is a lot more tools to understand this stuff than there was 15 years ago, and things like gnu radio are quite amazing.

And then the next problem, how do you be a player in radio - DSP - if you a re not good at software (I am not - but I do want to learn the gnu radio ca dence of programming).

te:

urself an applied mathematician? I would say that I do. As an analog desi gner how much do you use various mathematical concepts. Basic electronic c ircuit design does not seem to require a lot of theoretical math, where as , signal processing requires a lot of math.

gebra. I am on my 3rd pass ( of the first 10 lectures) of Gilbert Strangs on-line MIT course. I think I am finally getting it. I really want to kno w linear algebra because I want to be good at matlab which is rooted in lin ear algebra. However, is linear algebra used in circuit design other than an occasional solving of two or three simultaneous equations?

gnal processing stuff is definitely math intensive and required to be good at it.

super proficient at maxwells equations.

cannot accomplish his task with out the math?

c math and git 'er done!

l equations. I use very basic forms of abstract algebra which includes Pos t algebras which includes Boolean algebra. But more than all the others pu t together what I use is basic algebra. I can't say I've knowingly used li near algebra although some of the things I've learned may well be part of t hat.

I like to think I straddle circuits and systems. I am decent at both but c ertainly not an expert at analog. The problem in my company is that there is not enough work in circuit design to keep one always busy. I also think that most circuit problems are better solved through a system approach. I f you get the system concept wrong, then the circuit is going to be wrong. If you know what you want at the high level, it is easier to tell if your circuit design is adequate.

My "Building Scientific Apparatus" has the wire gauge table book marked. GH

I look at the attenuation and step response curves, to see what type and order of filter is, basically, emotionally appealing. I can't really quantify how pretty I want a pulse to look vs rise time vs cost. I take a normalized prototype from Williams, scale it, plug into Spice and fiddle for standard values and beauty. Some programs design vanilla active and passive filters around standard value parts, but not around the parts that I have in stock.

I have job binders that I use frequently, but they go down in the basement archives when a design is over.

I'll second AoE as most used. GH

We play a game, in a meeting, of doing the math in our heads. To slide-rule sort of accuracy, not many digits, but close enough to see what matters.

We've got pretty good at it, so we can really impress visitors.

calculate how good it _could_ be, (b) figure out how to get there, and (c) if that's a win, proceed, but if not, think up something better and goto(a) .

as a lot of lore and old fashioned crank-turning.

on"

Personally, it was vital. Not all the time, but when math is needed there i s no substitute. Try explaining complex impedance to someone w/o math; it j ust won't make any sense. Or how a capacitor can conduct current. Or how a transformer works. Or how RF signal transmission works. Had a discussion on ce with an FAA person on why radar can pick up non-metallic objects like bi rds - it pretty much ended when I mentioned "dielectric constant." It reall y becomes indispensable when you start designing control systems or advance d signal analysis.

Sure, most of us aren't in the business of pure science where you generate an RF signal just to be used in experiments and such. We generally design RF solutions that fulfill the purpose of messaging and control. So there is always going to be some kind of modulation and nowadays lots of digital communications.

Oh, you can. I know only little about DSP though I do prescribe coding strategies to DSP engineers. In fact, I might some day leave this earth without ever having programmed a DSP.

Well, I am player in radio and I really suck at software :-)

What I found, especially over the last 20 years, is that younger engineers no longer truly understand the fundamentals of RF. With truly I mean gut feel, instinct and all. How RF propagates, how it mixes when you don't want it to, how it leaks into places it shouldn't, and so on. Just to give you an example:

In my younger years I was often tasked with the design or re-design or medical Doppler ultrasound systems. The main reasons why I was called in were that prior solutions were too noisy, had inadequate dynamic range or simply were too expensive. Even in the late 80's this was already highly processor-driven. So you have DSP programmers. However, they knew nothing about anything that was in front of the ADCs. That was my turf. Pulse train generators, mixers, filters with gliding frequency response over time domain, quite complicated stuff. So we had an almost perfect split, I did the analog parts, they did the signal processing. Of course, we communicated a lot.

This hasn't changed much except that I have largely left the field of med-tech. Now it's more industrial, consumer, oil/gas and aerospace. I am an advocate of doing anything in the digital domain that can be done there if cost and power consumption remain reasonable. However, I do not consider it a problem to hand off all the coding to someone else, just like I do that with micro controller based projects.

The further on you get with electronics/software the more maths you tend to need. Ohms law is pretty basic maths. AC theory is a little more challenging. Kirchoffs theory relies on simultaneous equations.

In software I wrote a PCB design program which was mostly elementary geometry.

For a USB oscilloscope I designed I had to get into signal processing. I had to use a Fast Fourier Transform for displaying a frequency graph.

Amen!

Top-down is generally the only approach that really works. Now we'll have to explain that to the next generations. All the ones who have served in the military don't need to be told, they know this already.

There are very few times when a pure top down approach is used. Then even when it is used that's only because the bottom portions are very, very similar to something you've already completed.

It is seldom a design project is just pure design. There are almost always questions near the bottom that can only be resolved by bottom up design for at least that portion.

One thing that bottom up design does is to make testing easier. When people talk about top down design they usually really mean top down decomposition and bottom up design.

In medical and aerospace all the time. Has to be that way, else you can land in hot water with the Federales. Without a top-down approach it is hard to maintain a proper design history file set.

My projects typically start at the system spec level. That gets drilled into and sub-projects are parceled out. That way the documentation is always first and will be fleshed further out during the design. Not as an afterthought like it unfortunately is with many others.

calculate how good it _could_ be, (b) figure out how to get there, and (c) if that's a win, proceed, but if not, think up something better and goto(a) .

as a lot of lore and old fashioned crank-turning.

Right... except in my game you're allowed an envelope (or bar napkin) to scribble on... sometimes many bar napkins... depends on the paper quality... and how good your pen is. :^)

It's mostly the spit balling of ideas.

George H.

on"