Paul and Sabrina's Cheap 3 Phase Inverter (AC Controller) with Field Oriented Control

[cts_casemod]

Originally Posted by Astro

Most of this inductor discussion is going way over my head but i am curious. Isn't an inductor at it's simplest just a coil of wire? Couldn't we make our own? I know there is probably a lot of sophisticated and precision manufacturing involved to get every last percent of performance from an inductor but couldn't we devise a way to DIY something that gets us close?

You most certainly can. Im building my own test inductors for my charger, but they dont come cheap and there is a lot of work involved in researching suitable cores, frequency ranges, inductances and most importantly finding the right supplier. The cores look the same but they have very different characteristics, to which I would rather be backed up by graphs as the one posted by Paul, that clearly dictate the inductor specs at any given current range.

At high frequencies one needs to accound hysterisys, skin losses and the inductance variation with current, so the original simulations may not prove exactly the same results as expected on a DIY setup. More, it is impossible to guarantee that a particular build is exactly the same, the windings are all re-arranged equally, air gap is correct, etc, etc.

[P-hack]

I've been playing with the spreadsheet too and think I see, possibly, it working @300a with one $40 (in bulk) 165mm 26u sendust toroid @50khz and 3a/mm^2: CS1625026E - Sendust - Powder Cores | MH&W Direct with 20 turns @17uh effective, 72% duty cycle. CSC . Will take about 7lbs of 26 gauge magnet wire (litz would be nice), some sort of 3d printed wagon wheel insert/seperator would be helpful for winding and reducing capacitance. The strandiness will need to be leveraged because the window fill factor is considerable.

I'm not sure about welding cable, there are some studies on unvarnished stranded vs litz though (more headhurt). I would lean towards stranded magnet wire at a minimum, even if the strands cant share current perfectly, they will share heat load fairly well.

[MPaulHolmes]

That CH-200 is a gapped core. He just got back to me about it.

[P-hack]

here is a simulation @ 50khz & 17uh effective, 144v in, 510v/70a out. Average switch current is 180A (240 rms), peak at 310A. Havent looked into switches much yet., any thoughts? This approach would be like rewinding a small motor though. About $100 worth of 26 gauge magnet wire, though I'm using a somewhat conservative 3a/mm^2

Edit, I've got a 6MBI100L-060 that can do %72DC @20k, so I should look at 3x interleaved I recon and see what happens to price/wire at 1/3 the current.

edit2: ouch on highflux http://direct.mhw-intl.com/ch1016060e.html

[P-hack]

For some reason I'm able to get 3 of these @ 25khz to work with no red numbers 3x interleaved, 20 turns apiece, DC = %72/3. CS740060E - Sendust - Powder Cores | MH&W Direct 3 cores and 5lbs of wire. peak switch current is like 350A though average is 65A, still way too peaky for my 6mbi100 and a little fast. Though I've been watching valery wrestle with inductors for a while, so I am probably not thinking of something.

We could use an experiment though, simulating that ch200 on the edge at 300A 120uh:
@ 10khz duty cycle @%72, peak inductor current is 290A, min is 210 (~28% ripple), avg is ~250a, switch current avg=180A, peak @ 293A. Diode peak @ 295A, avg 70A. I arbitrarily put 1000uf in front and behind.

Paul, if you feel lucky with that ch-200, and have a suitable switch and diode (or halfbridge) and 300+amp 144v source, the test load is 7.2 ohms (510v/70A). Maybe something like $5 water heater elements (8 ohms total eventually, start higher) in a bucket of water for a dummy load, dunno what you have at your disposal, but there isn't a lot of room for error on the load on the low side of 7.2 ohms and if the blows open, you will be at thousands of volts in no time with a constant duty cycle applied (use an over voltage shutdown at a minimum, say 540v, overcurrent wouldn't hurt either, and it does have a bit of overshoot (740v?) if you just slam it on at %72 DC, so roll it on)... Big Dummy Load « Pro Sound Training

[MPaulHolmes]

I've got some 600v 600amp IGBTs laying around. I have 144v source, but I don't know about at 300amp. I could do 48v at 300amp at the moment. I don't have a load for it though. I used to have some 12 gauge NiChrome wire. But I lost it when we moved.

[cts_casemod]

For me such values of inductance are way to low as it would be a pain to get rid of the EMI on the AC line.

Obviously different scenarios. But the lesson learn is that it could perhaps be wise to increase inductance a bit, perhaps to 100 or 200uH as this would make things easier on the components, specially at anything below 100KHz.

Looking at those values of inductance I would say it could actually be a good idea to go discontinuous, avoiding the diode reverse recovery with consequent mosfet stresses, or the need for a SiC diode altogether.

Vallery is using some new choques on his 25KW chargers, which handle 25KW continuously, below 20KHz and if memory serves me right, are designed to be water cooled.

I remember he mentioned, at one point, that the now uses a type which is capable of operating at high temperatures, but I don't recall nothing about inductance values. Perhaps he could be a good choice of knowledge regarding this matter.

[P-hack]

fyi, @paul, if you change vIn to 48v and the load to 2.3 ohm, you should get about 161v out @ 70A (still 72% DC), which should look pretty much the same to the inductor (and is a saner first test).

@cts, the ch is rated @ 120uh, if it can handle the overcurrent.

[MPaulHolmes]

I was looking at resonant soft switching. Lee Hart was helping me out with understanding it on the EVTech list. A standard "fast switching" igbt module that's rated for 1200v 300amp is $172 new
Powerex Product: CM300DY-24NFH

and can do 70KHz if you add the Zero voltage switching. It doesn't look that complicated. You just add a capacitor and inductor, and some sort of circuit for inferring when the zero voltage crossing is happening. It can actually make the IGBT switching losses go to ZERO (epsilon haha).

Going to the 300amp SiC module would mean being able to probably switch at hundreds of KHz. If only it were cheaper.

[P-hack]

re Zero-current turn-on and zero-voltage turn-off, that sounds like a fidgety bunch of headhurt maybe sortable for a fixed load. Think I'll wait for SIC to become more affordable on that one.