Electronic and thermal feedback loops

On 9/26/18 9:12 PM, John Larkin wrote:> On Wed, 26 Sep 2018 19:05:43

-0400, Phil Hobbs > wrote: > >> On 9/26/18 5:23 PM, John Larkin wrote: >>> On Wed, 26 Sep 2018 14:21:36 -0400, Phil Hobbs >>> wrote: >>> >>>> On 9/26/18 9:41 AM, George Herold wrote: >>>>> On Tuesday, September 25, 2018 at 8:56:16 PM UTC-4, bitrex wrote: >>>>>> Say you have for simplicity two TEC modules, something like: >>>>>> >>>>>> >>>>> Is that the TEC and 'infinite' heatsink? Does it have forced >>>>> air cooling? (that will help) >>>>>> >>>>>> each mounted on one side to a large aluminum plane of heatsink, large >>>>>> enough to be considered more-or-less infinite in extent from the >>>>>> perspective of the module. On the other you have a block cooler with >>>>>> cooling fluid piped through the two coolers in series, and a temperature >>>>>> sensor to sense the cooling block's temperature, and the cooling blocks >>>>>> are connected in series for fluid flow. >>>>> Your first TEC thing? I think the biggest mistake with first TEC's is >>>>> to make the TEC too big, or the hot side heat sink too small... >>>>> (same problem.) >>>> >>>> Or the cold plate too large. Thermal diffusion is what sets the maximum >>>> loop bandwidth, so if you make the cold plate smaller, it helps a great >>>> deal. PID is no use once diffusion delays start to be important, >>>> because the phase shift increases without bound with frequency, so >>>> adding a derivative contribution makes things worse rather than better. >>> >>> Real-world PIDs usually don't use the derivative term at all. And if >>> they do, it's just a soft phase bump, not a real derivative. In >>> addition to the diffusion issue, there's noise to worry about. >> >> Some systems, e.g. the voice coil drive in a hard disk, really are very >> nearly pure second-order, i.e. displacement is proportional to I*V**2, >> so that the phase shift is close to 180 degrees out to pretty high >> frequency. Lightly-loaded, current-driven DC motors are also (nearly) >> like that. >> >> In that sort of case, you need the controller to have a (band-limited) >> derivative term in order to have a stable position for the 0-dB cross point. >> >> Otherwise I agree that the D term isn't as useful as it could be. >> >>> >>> The first PID that I designed, sophomore year at Tulane, was the >>> throttle control for a 32,000 horsepower steam turbine. It did use a >>> bit of D. >> >> Probably for the above reason. > > It was a cool system. Steam spins up a gigantic moment of inertia > (mainly the HP turbine geared way, way up) that spins the prop. That's > practically an integrator. Then the difference between hull speed and > prop speed loads the shaft with a squared term. Prop torque > accelerates the hull. The hull is loaded by the water proportional to > speed squared. So horsepower goes as speed cubed. > > What was amazing is that it behaved just as predicted. Beginner's > luck. > > A crusty old Chief looked at my simulation curves (printed sideways on > a Teletype and hand-colored) and said "That's just the way an > experienced hand would work the throttle valve" so we got the job. > > I just did dumb rectangular integration of one chunk of the system at > a time. The profs and control students told me that I had to do some > fancy runga-kuta interpolations or it wouldn't sim right. Runge-Kutta, predictor-corrector, and Stoer-Bulirsch methods work well only with systems whose coefficients are continuous functions. Runge-Kutta methods are self-starting, whereas the others require you to use some other method to get the first few points. The classical Runge-Kutta method uses four points to get fourth order accuracy, but doesn't supply an error estimate and (iirc) runs about half as fast as the others. Higher-order Runge-Kuttas and the others mentioned supply an internal error estimate, so you don't have to re-run the sim with a finer mesh to check it. (For the other methods, the error estimate allows you to write an adaptive solver, which changes the mesh size to keep the error in some desired range.)

For really smooth systems, Stoer-Bulirsch uses Richardson extrapolation to higher and higher order.

Rectangular integration isn't a bad choice at all for discontinuous systems, but on smooth ones it takes a lot more computation to get the same accuracy.

Cheers

Phil Hobbs

Or else scale the problem by a factor of a million or so.

Cheers

Phil Hobbs

Oh, duh... GH

They aren't the same "level" of nonsense though. a "thermal op amp" _flagrantly_ violates the second law of thermodynamics, like an op amp that could run off its own power. That's very nonsensical

The Griffith's text on quantum mechanics I liked a lot but I don't remember much of it at this point it's been a number of years. there's a course offered by IIT on YouTube I'm thinking about watching in sequence to refresh memory.

It's pretty light stuff IMO compared to books like the Jackson classical EM text. Classical physics actually seems more difficult math-wise.

I have a book on my shelf called something like "Theory and Design of Charged Particle Beams." it's about eight hundred pages long, it makes a great Valium substitute when I get insomnia from time to time.

One of the challenges of experimenting with a fluid cooling system is you can't really drink beverages at all for four or five hours before you start fiddling with the pumps and tubing and stuff.

It sounds a bit like one of those Zen-waterfall rock gardens they sell in Brookstone etc. which can be very relaxing but also sometimes makes one really gotta go. I'm pushing 40 my prostate probably isn't exactly what it used to be either, maybe I should get that checked more regularly.

No worse than a thermal inductor. The heat equation is first order in time, i.e. it has no oscillatory solutions. In a thermal SPICE model, all the capacitors have to go to ground, as well.

I can see how you might think that, but it's very much the other way round. You don't get very far into quantum in a one-semester course for EEs, followed by one semester of solid state for EEs.

Real graduate quantum textbook such as Itzykson & Zuber's _Quantum Field Theory_ or Fetter & Walecka's "Quantum Theory of Many-Particle Systems" (to cite two on the shelf opposite my desk) is at least as mathematical as Jackson and deals with much harder concepts. There's a lot more to quantum than to E&M.

Cheers

Phil Hobbs

A good Chemical Engineering "Transport" text would help you with that mass.

Disclaimer, I work in a Chemical Engineering Department at a major university doing Electronics and Physics, but I have borrowed the Transport Text for a Peltier project in the past.

The simplified charts in the back worked wonders for designing a circular chilled plate attached to a cubical mass, attached to a well characterized Lord Corp Peltier Module with built in fans and heatsink.

Steve

Do you have a fave transport book? Could be pretty useful sometimes.

Cheers

Phil Hobbs

Before I decide to become a quantum field theorist in my spare time I should probably read Misner, Thorne, and Wheeler that one's only 1300 pages. Woo Hoo!

I went to art school btw. They did have a physics requirement though, QM was an elective which I took and did poorly in at the time. I was 19 though so whatever, I did a lot better when I took the intro class as an adult student ~10 years later.

You don't need to be a field theorist. I'd settle for being a certified quantum mechanic :)

te:

rced

the

irst

ts

ortant,

so

r in

del,

se for

A good physics read are the Feynman lectures, which you can now get online from Caltech. The only problem with the text, is that there are no problems, at the end of chapters.

Quantum is way harder than classical, Jackson is near the top of E&M theory, yet not that hard... well you have to work at anything. But for classical, it's easier to make models in your mind. QM has all sorts of weird mixing, that I read papers about... and do measurements, but don't have a good model for.

I wish I had Feynman to explain the mixing of states and the driving E-M field.

George H.

e to

Phil, I'll ask the transport prof for you on Monday.

Steve

Is that the TEC and 'infinite' heatsink? Does it have forced

All the more so, then.

Some engineering E&M texts are in the same class as Jackson, e.g. Collin's "Field Theory of Guided Waves", Ramo (yes, the TRW Ramo), Whinnery, and Van Duzer's "Electromagnetics", and two of my faves, both by Roger Harrington: "Field Computation by Moment Methods" and "Time-Harmonic Electromagnetic Fields". I also bought Julian Schwinger's "Classical Electrodynamics" but haven't read it yet. Lots of good stuff out there. One engineering E&M book that I've used a lot is Taflove & Hagness, "The Finite-Difference Time-Domain Method". It helped me a lot with writing by big-iron optimizing simulator. Great book--just the right mix of tutorial and principles.

Cheers

Phil Hobbs

Feynman lectures didn't impress me much as a learning tool for newbies, looks like there are a lot of interesting nuggets in there for advanced students but sort of like Bob Pease's books they're good for people who are already fairly familiar with the material but you wouldn't want the guy as a ground-up kind of teacher. I've watched some of the Feynman lectures he doesn't impress me as a lecturer either his showoffy-style is distracting.

These "type" of guys can't at all suffer having to explain e.g. what an inductor or capacitor is to someone who doesn't know already. They come full equipped with high levels of cerebral narcissism and grandiosity, that stuff is beneath them and they don't make much effort to disguise it.

I've had instructors like that in regular-college as well, their classes are sink or swim. They like it that way and the upperclassmen/grad student guys they're farming to assist with their personal research projects like it that way, too. ok.

Feynman has been gone many years now, he wasn't the end-all-be-all of physics and IMO there are better options in pushing the year of our Lord

One of the better YouTube series on circuit design at the transistor level is this series from the department of engineering at IIT Madars by one K. Radakrishna Rao, which ironically enough uses some of Jim Thompson's op amp designs as analysis exercises.

Rather it's this series on analog IC design

Interesting guy.

He did a lot in 103 years.

I have his "Fields and Waves in Communications Electronics", also with Whinnery and Van Duzer, 1965.

Same book, different edition, I think.

They have a lot of interesting stuff that you don't get in the more physicsy books such as Jackson, stuff like variational methods and conservation of complex power.

Harrington is like that too. Collin talks about stuff like using the axial component of the field (E or H) in a waveguide as a potential from which the other components can be derived.

Cheers

Phil Hobbs