-
Hi,
We need to put 1200A into a 12V load from a 30V input.
Our client just wants to do a single 1200A buck converter switching at 10kHz.
He believes this may be the cheapest way.
This will mean a 15uH Buck inductor capable of doing 1230Apk..the ripple in that
inductor is only then 52Apkpk...which means a low gradient current ramp..not great for current mode control.
But yes we can add ramp to solve that. However, this inductor cannot be wound on any OffTheShelf core in the world.
The core material would have to be designed, and then we would have to find somebody capable of manufacturing it in small quantities.
As such, this is a terrible idea.
Better, we suggest , is to do ten 120A Bucks in parallel at 10kHz. The inductor would be a 60uH inductor wound with 10 turns
on a High-Flux torroid (two torroids stacked).
The torroid would be HF601147-2
The 10 turns would be 280 strands of 0.56mm enammelled copper wire, which gives a DC resistance of 0.9mR.
This gives conduction loss of 13W (core loss is only 2W).
Spec'ing out inductors, which are so big that no offtheshelf cores exist , is a terribly bad idea, we believe.
So we are stuck with picking the biggest core thats available offtheshelf, (HF601147-2) and just using that, and
adding as many converters in parallel as it takes to meet the load spec.
In the world today, the High Flux HF601147-2 is about the biggest offtheshelf core in the world.
So we must use that.
Do you agree this is the only realistic way forward? I mean getting your own personalised "torroid-the-size -of-a-toilet-seat" made specially is super expensive and not worth it.
The HF601147-2 datasheet is on page 79 of the "Micrometals Alloy Powder Core Catalog 2021"
https://s3.amazonaws.com/micrometals-production/filer_public/2f/ed/2fedd6eb-44d0-4834-a442-8222486f0b77/micrometals_alloy-en-2021.pdf
-
Is that the peak current or continuous?
Well there is no real limit. just use a larger core.
You need to use a multi-phase topology for such a huge current, but does it have to be a buck converter? -
just stack more cores?
-
Thanks, its the continuous current
-
1200A@12V, so ~15kW into 0.01 ohms total load impedance? Ouch! That is a monster conductor at the output, and not that much better at the input! Output caps will be comedy, especially if you get on your usual no electrolytic trip. 1 milliohm is 10% loss (1.5kW) in this thing!
I would be having serious conversations with your customer about why the idiot is trying to run that much power at such a stupid voltage (I might be more polite depending on how much money I thought I could extract!).
30V up the snout, so about a 2:1 ratio allowing for losses.
Multi phase for me for sure, apart from anything else that custom transformer is available but it is going to COST you a fortune unless you are buying thousands of the things, much better to do a multiphase thing, even with a LOT of phases that can reasonably be done with litz or tape winding on a conventional core, maybe a big ETD core with a gap, or such. Cheap, off the shelf and the sort of thing you could get someone like Pace to build you a dozen of.
12 phases at 100A, 1200W is a LOT more then 12 times easier then one at 12000A in reality. 1mR of losses per phase here is 10W, 120W total, not 1.5kW, the savings in cooling are significant.
Personally I would be looking for a topology that made better use of the magnetics then a buck running at that low ripple will, zero crossing PSFB maybe? Hell maybe just do a full bridge jobbie with a fixed 2:1 transformer ratio? Simplest is bestest. -
1.2 mHy 5kV CONTINOUS
#4 RECT , ~ 1M l, 1M dia NO core
j
-
I would be having serious conversations with your customer about why the idiot is trying to run that much power at such a stupid voltage (I might be more polite depending on how much money I thought I could extract!).
It's not a totally crazy amount of current. I've worked with magnet power supplies that could supply over 300 amps at a few volts. We just used big honking cables. Not 1200 amps, but lower voltage and a similar load resistance. And the VRM on a threadripper motherboard can deliver over 300 amps at 1V, so a fraction of the load resistance implied by 1200 amps / 12 V.
Of course 15 kW is much more power than either of those, and in any case they use lots of phases to avoid needing silly filter capacitors or inductors.
-
Before you try finding the largest inductor you can find I would investigate if 1 buck regulator is indeed the lowest cost option because that might not be the case at all. Stored energy in stray inductance in various places will be very difficult to snub with "normal" components.
100 nH @ 1200 A is 72 mJ which has to go somewhere. Unlike the inductor which can be build with off the shelf parts such snubber capacitors probably don't exists.
-
Quote
Stored energy in stray inductance in various places will be very difficult to snub with "normal" components.
Thankyou, this is indeed one of the mains reasons that single stage high power , hard switched SMPS's are very very rare.
When the hi side fet turns off, that stray inductance energy needs somewhere to go. As you allude, big beefy turn off snubber.
The other reasons high power , high current , hard switched smps is so rare is that its hard to get the switching node and high di/dt current paths low in dimensions....meaning
one has a big noise issue on one's hands. ..which can destroy the operation of the power supply's controller.
This is why doing even hundreds of paralleled converters is generally the better way.
-
A single phase 1200A buck converter, or 10 120A in parallel is not practical.
I repeat, you need a multi-phase buck regulator, not just many converters in parallel. Going from what I've read, 1200A would need 30 phases, with each inductor and switch carrying 40A. That's not a massive current and is much easier to deal with. It will also men a higher frequency than 10kHz can be used, thus a much smaller inductor, assuming there isn't some other reason why you have to use that frequency.
https://www.allaboutcircuits.com/technical-articles/pros-and-cons-of-multiphase-dc-dc-converters/
https://www.ti.com/lit/an/slyt449/slyt449.pdf
https://www.ti.com/lit/an/slva882b/slva882b.pdf
https://www.infineon.com/dgdl/Infineon-Digital_multiphase_controller_dual_phase_operation_in_multiphase_buck_applications-ApplicationNotes-v01_00-EN.pdf?fileId=5546d46272e49d2a01737bf04c75581c
If you opt for a transformer and a push-pull topology, then I still think you'll need polyphase inverter, but not so many phases, as the current in the primary will be lower and the duty cycle higher, although it's not something I'm familiar with. -
Quote
I repeat, you need a multi-phase buck regulator, not just many converters in parallel. Going from what I've read, 1200A would need 30 phases, with each inductor and switch carrying 40A.
Thanks, couldnt have out it better myself...though i once offered a few hundred 20A converters in parallel for a 5000A spec.....and got trashed off the job...they said they just wanted one stage!
They also wanted the current to be tightly controlled , it was quite varying...and all the variation during a 100ms pulse!....and the switching frequency of the single stage that they said they wanted was just 2khz!!!!......there were only 200 switching cycles over the 100ms pulse length!!
Havign said that, ive worked on electroplating power supplies which give out 250A from a single stage half bridge......10V, 250A.
But the primary side voltage is 600V......so not so high current there.
The secondary side current didnt go through any transistors...just isotop diodes in parallel which were soft switched on turn on and off.
It was a half bridge with no output inductor, and just the leakage on the transformer was thus used. Output current was discontinuous!!!! (gave the zero voltage/current turn on and off of the diodes)
For whatever reason, they had used primary side series bus caps which were massively underrated for the ripple, and they were failing LR & C, but custs never minded and just happily bought new units.......the company even had a youtube video showing the custs how to replace the failed units!
...i think if you have transistors hard switching the current, then 40A is about the most you can do. -
As I alluded too, some very basic back calculations reveal that a single stage is impossible to implement with the inductor not being the most difficult part of the exercise. It's rather strange that you've started from the inductor.
-
Did you ask them why it must be one stage?QuoteI repeat, you need a multi-phase buck regulator, not just many converters in parallel. Going from what I've read, 1200A would need 30 phases, with each inductor and switch carrying 40A.
Thanks, couldnt have out it better myself...though i once offered a few hundred 20A converters in parallel for a 5000A spec.....and got trashed off the job...they said they just wanted one stage!
They also wanted the current to be tightly controlled , it was quite varying...and all the variation during a 100ms pulse!....and the switching frequency of the single stage that they said they wanted was just 2khz!!!!......there were only 200 switching cycles over the 100ms pulse length!!
Havign said that, ive worked on electroplating power supplies which give out 250A from a single stage half bridge......10V, 250A.
But the primary side voltage is 600V......so not so high current there.
The secondary side current didnt go through any transistors...just isotop diodes in parallel which were soft switched on turn on and off.
It was a half bridge with no output inductor, and just the leakage on the transformer was thus used. Output current was discontinuous!!!! (gave the zero voltage/current turn on and off of the diodes)
For whatever reason, they had used primary side series bus caps which were massively underrated for the ripple, and they were failing LR & C, but custs never minded and just happily bought new units.......the company even had a youtube video showing the custs how to replace the failed units!
...i think if you have transistors hard switching the current, then 40A is about the most you can do.
Tell them it's only way to do it.
Failing that. They sound like pain in the bum. You're better off turning the job down. It would be foolish to take on a job with such impossible requirements. -
Quote
As I alluded too, some very basic back calculations reveal that a single stage is impossible to implement with the inductor not being the most difficult part of the exercise. It's rather strange that you've started from the inductor.
Thanks good point,,,starting from the stray reactance energys is a great start. Turn off snubbers become crucial and interesting. -
184" UCLBL 1940s cyclotron
Manhattan Proj Calutron racetrack 1942 for U235 separation for A bomb
j
-
Does the output need to be DC?
For a one off, simply getting a custom 50/60hz transformer would might be the cheapest, lowest risk option if they can accept AC?
If you have a 3ph AC mains supply then these sorts of output voltages / currents aren't that hard, assuming the solution doesn't have to be very portable (because it'll weigh 300kg... lol)
-
A place I used to work made some really big inductors. Took hours to run them down, or they would overheat. They were made from niobium-tin alloy but the cores were quite tricky as they had to keep the wire under liquid helium or they'd stop superconducting.
Happy times.
-
Damn... I was hoping this was a real question. Bummer.
-
This is what I was referring to above, when I mentioned three phase transformers. M1 to M6 are the inverter and M7 to M12 the active rectifier. I would use four of these circuits in parallel, each offset by a phase angle of 30°. Note this circuit is bidirectional and can feed power from the secondary, back to the primary.
-
-
The largest "off the shelf" high frequency cores that you can get are amorphous iron core halves, which you can join multiple of for even higher power.
-
Cutler Maine ? Still in operation, possibly one of the most impressive VLF operations I have ever heard of.