LC resonance program

Have you tried DOSbox? I prefer this build:

Yours works, but I like this one:

Just do it in Javascript. Never going away, always portable!

Tim

Seven Transistor Labs, LLC Electrical Engineering Consultation and Contract Design Website:

Have you tried DOSbox? I prefer this build:

Yours works, but I like this one:

And dosemu works in 64-bit now too.

Cheers

Phil "Freelance 4.0 forever" Hobbs

You have some nice calcs there.

You link to

which is interesting. We're considering designing a PCB that will have a lot of surface power dissipation, a several-kilowatt polyphase switcher, and thermals dominate the design. But I think we'll build a physical mockup, so we can play with fans and things.

I'm thinking a chunk of dremel'd copperclad FR4 with lots of heat sources. Maybe surface mount resistors, maybe just run a bunch of current through the copper.

Instead of using a calculator program, or - horrors - doing the math myself, I often just Spice things and fiddle, or use a feedback loop to force the solution. One nice thing about that is that I can comment the model it and save it in a project folder for future reference.

Convection is usually your biggest problem. If you use the calculator to find the properties of a rather wide, high current trace, you'll find the dissipation goes something like 8mW/K/mm^2, which is more-or-less the good old fashioned rule of thumb (howsabout that?).

Using *lots* of vias, and a thermal pad to an aluminum mounting plate (with subsequent heatsink if need be), can do wonders. There are 300W

FR-4 isn't very conductive (we've had this thread before), so pour all layers (improves in-plane heat spreading by... uh, >4x I think?), and use vias (improves thru-plane conductivity by up to 50x I think?).

Speaking of filled vias (as came up in that thread), I wonder if lead-free solder is more conductive than leaded. Hmmm... looks like 63 vs. 53, not a big enough difference that you'd want to count on it. And about 1/8th that of copper, but again, you don't get terribly much copper in the barrel.

Anyway, as for ~1kW levels, I'd strongly recommend going with THT parts bolted to a heatsink, but especially if you can get efficiency up there, it'll still be okay to construct with this method.

Use lots of screws... don't want that big board bowing out against the thermal pad strips beneath it...

Most of those calculators, I tossed together because I was doing something at the bench. SPICE is great when the datasheets are correct, but when they omit important information (e.g., RLC parameters, or s-parameters), or outright lie, I have to dick around with coils and caps forever. Being able to calculate forward and back helps save iterations and get a better idea what's going on.

The worst data I saw, if you were wondering: an E180F/6688 (for a little tube radio project) claims 11pF in the datasheet (including dynamic effects!), but I measured 23. That really screwed up the RF amp design, to the point where I doubt it's worth even having it anymore...

Tuning filters is awful, so I've been using a lot of toy SPICE models on the project -- same idea. Constructing filters is just as bad, but I've made out pretty well between making things modestly adjustable and using some surprisingly accurate and useful old-timey data. (Good luck finding data on coil coupling constants, for whatever geometries -- unless you want to go and build a freaking FEMM model or what have you, you're kind of stuck. But they knew that, back in the day, too, and painstakingly compiled tables, curves and nomographs to help the every day engineer. Good stuff.

(On topic relevance, category: hare-brained ideas. Using COTS spring coils for coupling RF, isolation, etc. You're making a double-tuned filter, inevitably. The bandwidth of which (and therefore, data rate) depends on coupling and impedance. Coils of known geometry can be positioned and made resonant, give or take some possible adjustment for positional and electrical tolerance. Which might be done automatically, hmm. Doing it in a couple-transistor circuit, of course, is the harder part.)

Tim

Bolting boards to cold plates is a nuisance. I'm thinking about mounting some number of fans on spacers above the parts, blasting straight down onto the top of the board. So I wouldn't care much about the FR4 vertical thermal conductivity, or vias. I have no clue how to analyze that, so we'll just make some mockups, push fans around, terrmal image everything.

A lot of surface-mount heat sinks might help some.

I spent way too much time tweaking a lowpass filter, for an ARB, to get decent response using standard value parts. So we bought the NuHertz software, and it did it in seconds.

Coupled surface-mount inductors? Interesting. I wanted to use coupled inductors, on opposite sides of a pcb, to get fast 6 KV isolated gate drives. We didn't get the job, so I didn't try it.

Coilcraft has some nice parts, including some high-current super-wideband inductors, good for bias tees and stuff.

The 'old rule of thumb' for convective surface rise here is;

delta degC = mw / cm^2

Can there even be a rule of thumb for dissipation, without some kind of arbitrary rise limit assumed? Anyways, it looks like there's some kind of numerator/denominator issue with your 'something like'.

The convective estimation assumes matchbox-breadbox sized regular shape in free air - no fan or passers-by.

Yeah, you missed the K^-1, hahah.

Yup, nothing too small or too large, unobstructed, unforced. From what I've seen of PCBs, chassis and such, it works out surprisingly well.

With even a small fan, the power can go up by 4x easily. That's, for example, 5A in a 20 mil trace.

Tim

That's good. I'd like to have all such scripts in one place. I miss a good R divider calculator (given the desired division ratio and current, please tell me what would be the best configuration in, say, E24 when, say, 4 resistors are allowed) and a comparator-based Schmitt trigger (give Vcc, V_lo, V_hi and the E series, returns the resistor values).

This guy has very useful SMPS inductor calculator:

Best regards, Piotr

Yup, got that one on my bookmarks! If you go up a level to

Regarding inductors, Stroobandt's is the best I know of for solenoids, outside of (again) cranking your favorite FEM for a while. (And even then, I would expect it's still more accurate until you use a fairly fine mesh, which probably takes quite a bit of time to compute!)

The trace width (Chemandy's has grounded CPW -- though I haven't found a diff grounded CPW yet) and filter (passive or active) calculators are probably of broad interest, too.

Tim

I wrote a program to design voltage dividers, using parts that we have in stock.

This has saved me an enormous amount of time, and kept us from buying more parts to stock.

Yes, Tim, many of us have seen and admired your website. Are you going to show us the source code, to understand the benefits and issues from going with Javascript?

I think he just did! The Javascript functions such as "res5_Update" in that RLC.html file appear to be what you're asking about.

In other words, the web page is the source code.

:) for example, control-u in firefox.

Could you, then, re-state your 'something like' so that it makes sense?

(deltaT ~ K, deltaT ~ 1/area is a simple check)

RL

Yea--

if the constant is k == mW / (K*mm^2), then: P = k * A * dT

So power goes up with area (sounds right), and up with temp (yup).

Or conversely, dT goes up with power, or inverse with area. Just rearrangements.

Tim