tripled coupled inductors

You're missing the underlying symmetry of the problem, though.

Check the current waveform.

Once you have a nice linear ramp, whose amplitude is always absolutely defined by the history of PWM applied in the circuit: you only need control PWM to regulate that current.

Then, and only then, can you think about voltage regulation!

ANYTHING ELSE leads to the same fundamental problems you've seen before -- things exploding, poor regulation (of current or voltage), no limiting, whatever. You can put all kinds of pretty or dirty hacks on top of it to try and control those conditions, but the best methods are simply equivalent. You might as well go right for the most elegant solution in the first place! :-)

Tim

What I have observed with the 4 uH inductor in series with the transformer primary limits the current to 48V/1.26 ohms = 38 amps at 50 kHz. This provides current limiting during charging of a large value output capacitor, as well as into a near short of 3 ohms. Here are some waveforms:

Into 5k:

Into 3 ohms:

Into 100 ohms:

And with 25% duty cycle PWM:

So, it seems to me that it should be OK to use the full 50% duty cycle (maybe 49% with dead time) without any sort of fancy current or voltage feedback and PWM regulation. The output voltage with a 5k minimal load is about 370 VDC, and into 100 ohms it drops to about 235 volts. That is well within the range for a 240 VAC VFD. The thing I don't like about the 25% PWM is the ringing during the off times.

At the 25% duty cycle, it takes 227 watts for 195 watts out (142V), for

I was not seriously considering the saturable core reactor for regulation, but more as a concept.

Paul

The problem with your simulation is that the symmetry is perfect, and the real world is not

So you won't see flux walking in the simulation, but you will see it in the real world with destruction to follow

To avoid flux walking you need to run the forward converter in voltage mode and add a series capacitor to the primary winding.

You need to look careful into the maximum rate of change of the duty cycle to avoid being forced to only use half of the available flux swing

Cheers

Klaus

I tell them that sometimes. But now speaking of skin effect, why can't they make a connector that connects to the center of a solid wire ?

I mean, think of the cause of skin effect. It must be due to the fact that you are almost certainly connecting to the skin of the wire. Well what if there was some kind of clamp that would make it connect to the center of a solid wire ?

If that is plausible I bet it can be sold to somebody, but we would need demonstrable results.

Den mandag den 29. august 2016 kl. 00.03.35 UTC+2 skrev snipped-for-privacy@gmail.com:

it has nothing to do with where you connector to the wire

It's an interesting question: if you launch a current into the centre of a wire, how far does it go 'til the skin effect asserts itself?

Dunno, myself, with the data given. Cheers

Phil Hobbs

Romex is great, but the wire spacing will add a little inductance on long runs. Might cost a dB at 20 KHz. Using all three wires would help a little. I actually have some Romex measurements somewhere.

We had a real problem with long runs from NMR gradient drivers to the coils. The customer used a pair of RG8 coaxes, one for the current and one for return; lots of loop area and inductance. We made our own low-L coax from mil-type teflon insulated stranded wire - very thin insulation - and an outer braid, then shrink tubing.

There are some very cheap bifalar transformers around. It doesn't seem to increase cost much. Bifalar winding greatly reduces leakage inductance (and increases capacitance, and reduces breakdown voltage.)

At high frequencies, solid wire can have more skin losses and more eddy-current losses than stranded. Sometimes that matters.

...

Hi Phil

That does exist. A redesign of the avionics power supply of the Gripen (Swedish fighter plane) based on similar principles was at least considered. It looks that they went as far as flight testing...

Apparently one can achieve higher reliability with saturable magnetics (as opposed to conventional PWM-based switching supplies) as well as higher efficiency in certain cases.

Dimitrij

Thanks, Dmitrij, but that's the adventurous PE Schoen, not I. I stick with LM2594s and such, myself.

Cheers

Phil "Still searching for an honest SMPS" Hobbs

I'm considering using a mess of TPS54302's in the power supply of the current gadget. It's synchronous and runs at 400 KHz spread-spectrum so the EMI situation should be OK. It has internal soft-start. Input voltage extends to 28, unusual these days.

I could run each switcher output to the ENABLE of the next one to sequence the startups, so not all try to start up at once.

I'm skeptical about getting 3 amps out of the SOT23/6 package, so I have some coming in to test. They are so cheap I could use lots of them.

OK, so I had already introduced some imbalance by adding a 10 milliohm resistor to the source of one of the bottom legs of the FWB. Now I changed the PWM so that one cross-pair has an ON time of 9.25 uSec while the other remained at 9.5 uSec, with a 50 kHz waveform with 20 uSec period. That is about 2.5% imbalance, which is pretty extreme. Here are the results:

There is definitely some significant imbalance, with about 25 mVDC on the transformer primary and 23.5 amps DC. The transformer primary has 1 milliohm resistance and the series inductor L4 has 10 uOhm. Yet the input power is just 596 watts for 549 watts output for 92% efficiency. Granted, the transformer's inductance and that of L4 may suffer from so much DC current, but if the series inductor does not saturate (which it won't if it's air core), the circuit will still limit currents to safe values. Also, I could read the DC level of the common low-side resistor (10 mOhm) and use that to shut down the PWM if it is excessive.

This has become OT for this discussion, so I should probably add to my other thread on the DC-DC converter. Even better, perhaps I should just build this thing and test the hell out of it. ;)

Paul

I would expect it to take effect within a length of a couple of diameters. It's matter of the eddy currents and their magnetic fields interacting.

Perhaps instead of Litz wire, a cable made of multiple layers of alternating thin conductors and insulation would work as well, although the signal may couple to the outer layers through capacitance eventually. The effects are minimal at audio frequencies, but above 10kHz they are definitely measureable and perhaps detectable by an audiophool's sensitive ear.

Paul

Maybe. Or it could be much farther. I don't recall having seen any calculations either way.

The point of Litz is that every strand has the same percentage of its length at the surface as every other, so (ideally) there's no difference in the series impedance as a function of position in the cross-section.

It has less capacitance than tape windings, which are otherwise similar.

Cheers

Phil Hobbs

The skin of a solid wire IS a connector to the center of a solid wire.

Keep in mind, use of commodity materials is critical to low cost magnetic design.

A transformer using core and bobbin shapes found in flourescent lamps was once quoted to me at Cdn$0.02 in 10K quantities, even though it had a few inches of 3xFEP solid wire in the build. I had to ask for confirmation on that.......

RL

Coilcraft LPHR8045-223ME appears to be 1:1:1.03, if I'm interpreting correctly...

Unfortunately they don't list a price, just a web-form for getting a quote. Ick.

Cheers, James Arthur

Well what if there was some kind of clamp that would make it connect to the center of a solid wire ?

Now I think you're pulling my leg.

I suspect you could sell it to audiophools. No demonstrable results required, just hype.

The point of saying "for audio, I don't think it matters" is at audio frequencies the skin is thick. As the frequency goes up more of the current is pushed from the center of the wire to the outer diameter.(the skin)

For a 12 gauge wire this page shows a (very) worst case of a 1 to 1.1 ratio of Rdc to Rac at 10kHz and 1 to 1.3 ratio of Rdc to Rac at 20kHz. However then concludes with "Actual measured increase in AC Resistance due to Skin Effect at 20 kHz is less than 3%."

Here's a blurb from Belden.

Nice find! I like the 1.5kV rating, appropriate for projects where I need a flying 24V supply. The high leakage inductance and high insulation rating argue against trifilar winding, making me hopeful for a lowish input-output capacitance. Otherwise I'm going to be winding my own.