Old American saying: If it ain't big enough then get a bigger one :-)
I try to avoid snubbers where possible. They cost efficiency, most of the time. At least when compared to regenerative methods.
-- Regards, Joerg http://www.analogconsultants.com/
I don't have experience with IGBT but in general the commutation frequencies go down the larger the plant. Inductances become larger, capacitances becme larger, ringouts and spike become slower.
-- Regards, Joerg http://www.analogconsultants.com/
The energy is a lot higher. That's why I was hoping Tim would respond.
He raised the issue of good layout and low inductance causing a problem with switching transients. How do you kill a good layout and make the problem go away? Adding stray inductance in the layout seems like it would only make things worse. Certainly the ringing would go to lower frequencies. Does that help by slowing the risetime into the diodes?
IRF makes some really big switching semiconductors. 80 Amp at 1,200 Volt is enormous. Put that into a 250uH inductor and you have some significant energy to handle when you try to switch.
How do they handle it? Snubbers? I'm not sure that would be the way to go. It would waste a huge amount of energy and things would get hot.
It is a basic problem whether you are handling milliwats or megawatts. What do you do with the energy when you need to switch the current into an inductor?
Mike
Yeah, maybe he can show us an example.
If the diode is a big old fat one where you need a 1/2" socket wrench to tighten the contact lugs you may not have much choice.
It is also something that regulators may not let us get away with anymore, at least not if the snubbers burn off a lot while there is no load. Snubbers remind me of big "clean" Diesel motors. They had to raise the exhaust temps so sometimes they deliberately made the engine less efficient.
I try to dump it into a rail from where it can be re-used.
-- Regards, Joerg http://www.analogconsultants.com/
Hey, Mike -
Where did you get the 250uH? Did I miss something? If so, disregard.
John S
You can only do that to one rail. VCC.
In fact, if it's a standard diode to VCC, you have no other choice. That's where the energy is going to go whether you like it or not.
The energy that goes through the other diode to ground is simply lost. Is this lost energy accounted for in the efficiency calculations?
I noticed playing with Larkin's amazing capacitor bypass patent that a little miller on the mosfet dramatically changed the conditions at turnoff. It had a huge effect on all the waveforms. I told Larkin he would not like it.
But whenever something surprising like this happens, I wonder if it could be turned to out advantage. Maybe it is pointing the way to another solution to the problem.
Mike
Thanks John. Good question. I'm sorry for being so sloppy and not giving the reference each time I need to use it. The file is
I wonder how big a 250uH inductor would be that can handle 80 Amps? And where do you find a power supply that can deliver 1,200V and 80 Amps?
These guys are not playing games.
Mike
If you're talking about sloshing magnetizing current into catch diodes to Vcc and to ground, like my circuit, it's symmetric. And no energy is lost by running current through ideal diodes and inductors.
I don't like Miller capacitance? I never knew that.
When I use 270 ohm gate resistors in by little converter, I get really nice smooth trapezoidal edges. What's not to like?
Keep in mind that you are simulating, and I am breadboarding.
There's probably an combination of gate resistance and the pseudo-resonant-snubbing capacitance that optimizes efficiency and noise. But it's pretty good now, with just the gate resistors. I may tweak it when the real PCB is built, just for fun.
I do have other problems to solve on this board. Like getting the 250 MHz ADC data into a cheap Cyclone FPGA, and making the CAT6A equalization work, and getting rid of the heat, and stuff like that.
John
[...]
No, that was Gus Portokalos. He didn't like the Millers, at least not until after Toula and Ian were married.
It makes heat which you probably can't use on that board. And it isn't green to do that :-)
[...]-- Regards, Joerg http://www.analogconsultants.com/
Not that bad. Even a 5mH/60A is only sround 30lbs, and 8" by 3.5" (round):
And
For example, in this market:
Although it is customary to first buck down to something lower, like
600V, and go from there.-- Regards, Joerg http://www.analogconsultants.com/
Great movie.
He didn't like the Millers, at least not
Miller Lite is ok if you're hot and thirsty and just need something wet to drink.
I'm running about 75% efficiency, 24 VDC in, 9 volts out at 4 watts. The fets are cold. I'm losing about a volt in the bridge rectifiers, over 10% right there. There's loss in the transformer, too. Not bad, overall.
John
Yes, I could probably watch it five more times and not get bored.
Yuck ... Gordon Biersch isn't too far from you guys and there you can get some real stuff.
I believe you are running a DRQ127 330uH at around 100kHz. That's a bit tight, they are only around 100V-usec. So some saturation and core losses there.
-- Regards, Joerg http://www.analogconsultants.com/
I don't like the GB stuff. Too hoppy for my taste. Lately we've been drinking Widmer and Stella and Purple Haze, abd Harp, translucent stuff like that. I don't trust liquids that I can't see through.
My wife does drink Guinness. Can't imagine why.
Some of the brit "pub" beers are OK, like Boddington and Speckled Hen. Both have the nitrogen thingie inside.
150 KHz, and the drive into the transformer is 24 volts p-p. Lots of margin.John
So do I, love it. On my way back on Friday I stopped at Trader Joe's and stocked up on Trappiste and Porter. The Porter has the translucency of tar.
[...]
So where's the remaining 10% of losses? All in the wire?
-- Regards, Joerg http://www.analogconsultants.com/
years ago I did a hack. Needed an isolated low current reference source, so there was this spare TTL FF in the circuit not being used.
I did something like this with a couple of minor things added.
+-----------+. ,- PULL ^ .--o--. | I(SENSE) )|( ___ -|J S Q+--+ ___ )|( -|___|+--+----+|> | -|___|-+--+. ,- | -|K R Q+--+ | )|( | '--o--' | | )|( | +--------|-+-' '- | | --- | PULSE CAP --- | + | +--+--------------------+ (created by AACircuit v1.28.6 beta 04/19/05I placed all the components I needed on a daughter board and stuck it in there vertically attaching it to the FF that was in the circuit already.
Of course I have a full wave rectifier on the other side with cap.
Jamie
Yeah, but those'll cost as much as the transistors themselves, and they look like excellent resistors with a nasty tempco. With the 140A peaks I've estimated on the previous boards, if they'll even handle peaks that high for any length of time, the voltage drop will literally be as high as a junction diode's. They would be excellent on the new boards, but since I've reduced dI/dt, silicon junction diodes are still fine.
FWIW, we first tried 8A, then 16A rectifiers. Both died, even though the average current is very small. 30A rectifiers survived. I'm guessing electromigration or hot spots caused them to burn up -- worked fine for a few minutes, then....tick!*
*Incidentially, the frequency of Psst..BANG! has dropped precipitously since I joined. Careful design and debugging, combined with the handy use of semiconductor fuses, keeps the mortality down to 1-2 transistors per failure, and limits the energy so well the cases are rarely cracked, let alone charred and fragmented. Have yet to lose an AC rectifier. Yeah, the fuses cost about as much as the transistors, but you only lose two or three at a time (i.e., the fuse and a transistor or two), not the whole power system! :)Not completely true. Big slow snubbers that make trapezoidal waveforms usually do, but just a tad, enough to roughly double the edge time, slows down the waveform enough to let the transistor turn off better, while storing a minimum of energy in the snubber. Saw an article somewhere that shows a loss minima between rising snubber loss and falling switching loss. It doesn't amount to much, maybe 20% of the total switching losses, but that's more than 10%.
In my case, at 400kHz and 700V swing, there probably isn't much time for a "lossless" (resonant) snubber. Might be something to try later on. But the sims say only 100W in the resistor, out of 5kW per half bridge, which isn't bad. The transistors burn more, so I'm not concerned.
Tim
-- Deep Friar: a very philosophical monk. Website: http://webpages.charter.net/dawill/tmoranwms
Ohh, that's nothing. Our sister company is setting up purchasing on some
1200V 600A bricks, and those are the small ones. The melting guys are playing with 2400V 600A bricks right now, and they go up from there. :) We don't use anything that big, but the 1200V 300 and 600A modules are handy for
Searching linears site on the title, it looks like it is AN122, August
2009.-- John Devereux
f .
The EDN article itself is dated Jan. 8, 2009, visible in a different view of it:
Nice article. The above view also describes an avalanche pulse generator (tr < 400pS), mercury-wetted relays (50V into 50 ohms, tr=3D250pS), and, back on John topic, using the LT1533 ultra-quiet edge- controlled switcher.
Making each figure of the article an individual .PDF needing downloading maximizes the uselessness of getting it from EDN.
-- Cheers, James Arthur
I have never understood why they do that. Someone higher up on the editorial board must have dozed off a long time ago :-)
-- Regards, Joerg http://www.analogconsultants.com/
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.