class-D amp so far

Yikes!

Here's what I have so far. The servo seems to work enough to protect the big caps, but this simulates at 11 hours per second, so it's a tad hard to tune. The integrator could be slower.

It will also consume my hard drive with the .raw file if I let it run very long.

I can kick this off, go to Loews and Bevmo, take a nap, and then see how it's doing.

I bet it could be done pretty well--if the nonlinearity is symmetrical, you're in the middle of the B-H curve. If the core is starting to saturate on one half-cycle, you'll have some curvature, and if you're really getting into trouble, there'll be a big spike. A few-turn sense winding would be the bee's knees, but you already have the transformers.

Of course, since you have rich customers, you don't have to sweat the BOM cost that much--why not wire the caps up as a nonpolar? (It would only take four times as many.) ;)

Cheers

Phil Hobbs

There are many ways to measure the magnetization level of a transformer cor e.

Flux Balancing Control of Ungapped Nanocrystalline Core-Based Transformer i n Dual Active Bridge Converters

US patents

"Flux-Gate: The flux-gate principle is a well-known concept for current measurement [14]. This measurement principle can be ada pted in order to build a flux-density transducer, as presented in [15] and shown in Fig. 4(d). In this concept, an E-type core is utilized whereby the primary and secondary windings are placed around the core?s middle leg while two addition al windings are placed in one of the ?I? parts of the core, as depicted in Fig. 4(d). With this arrangement, a dc bias in the flux-density would result in a change in the inductance in the terminals of the auxiliary winding, thus enabling monitoring of the flux density in the core."

With DC current bias this probe coil on E-core:

use 2 toroidal cores:

Some nice links !

I really like the Google OCR old books on archive.org as well.

Great histories !

The DC blocking capacitor only creates the illusion of correct operation of the transformer. ?apacitor causes transient problems. A circuit with a capacitor in the primary coil will fail if the secondary i s loaded an incandescent lamp with a diode.

of the transformer.

is loaded an incandescent lamp with a diode. Also, the cap and transformer form a series resonant circuit that may give unexpected issues.

I can sell that.

How? It simulates nicely, and the rev A board works... just not enough brute power. We will need to parallel enough caps to manage the ripple current.

I don't plan to do that; the load is symmetric three-phase. But if an asymmetric load did saturate the transformer, my brute-force dumb amp will just power through.

I'm simulating now, with a .step on the sine wave frequency. Time for a nap or a hike or something while that runs.

Mike quit Analog and has hinted at a next-gen simulator after LT Spice. I think it ought to use an Nvidia card to get us, say, 1000x the speed of running on a PC.

I want part value sliders and I want to see waveforms change instantly.

They are stock Coilcraft parts now; anything else would probably have to be custom. One of my guys is doing a test to see how hard we can overload them for 15 seconds, and how air flow affects theta. Putting more wire on one core might be thermally disadvantageous.

The caps are small and cheap. I should make sure they won't fry from ESR heating, but they only see the switcher ripple current, which isn't huge. I could use four.

Given that I want my amp to output maximum possible power for 10 seconds or so, I probably want fets with low Rds-on, like 10 milliohms maybe, to reduce I2R power at the expense of capacitive switching losses, which won't increase with power.

More interestingly, I want my heat sinks to have maximum mass, without much regard for the classic degsC/watt spec. Air flow won't matter much in 10 seconds. So a heat sink should be a solid block of aluminum. Fins just waste the opportunity for mass.

Annoyingly, the people who sell heat sinks don't know much about this stuff, so seldom list the weight of their parts. And they are slow to answer questions, as in days so far. I'll have to get some and weigh them.

Conveniently, 1 watt dumped into 1 gram of aluminum rises about 1 degc per second. The electrical analogy is 1 farad, so simple circuit theory or Spice can do my thermal transient modeling.

Similarly, I care about the thermal mass of caps and inductors, not their usual ratings.

Copper has similar specific heat capacity as aluminum, but due to the higher density, the copper slab dimensions can be much smaller. In addition, the heat conductivity is much better, so the heat spreads quickly through the slab.

Attach a normal (aluminum) heatsink to the copper slab to remove the _average_ heat power to the environment.

tirsdag den 10. november 2020 kl. 14.53.13 UTC+1 skrev snipped-for-privacy@highlandsniptechnology.com:

just order aluminium bar or plate from a metal supply?

This would be easy, for two TO247 fets:

It's an Aavid heat sink and two aluminum nitride insulators, parts that we have in stock. I need to do the measurements and math on all that. I don't think that four of those heat sinks would fit on my board, but maybe.

If it's borderline, I could lay out the board so that we could substitute a big machined aluminum block if we had to.

tirsdag den 10. november 2020 kl. 17.11.31 UTC+1 skrev snipped-for-privacy@highlandsniptechnology.com:

an aluminium bar in between the fets can screw a heat sink on top if needed

sorta like this:

We could tap the top of my heat sink and bolt a block there. I have the height.

Maybe I have enough mass to get over my surge. Gotta weigh a heat sink.

The Aavid sales rep contacted me. Useless. He doesn't even know what he sells. Their part numbers are 96 digits long and the data sheets are horrible.

I can use four of the little ones, same net thermal mass.

and they are not made by Aavid!

It's looking like this:

The big fets (on 4 heat sinks, 16 grams each) and two big toroidal inductors are on the bottom. There's a fan on the bottom too, blowing right onto those parts.

All my available layout people are busy, so I'm doing this myself. There is a huge advantage to having the circuit in your head when you do a layout.

Funny how, if you push stuff around aimlessly for a while, the pattern sort of snaps itself into place. Of course, I have to option to change the circuit, and they don't. Two of the big inductors don't want to fit, so they are gone. I'll just blow a lot of air on the survivors.