3 kW output power with wide voltage range and autoranging output power

[dutch66]

Hello everybody,
I try to find the best switching topology for a slightly unusual power supply. The power supply will be used in scientific experiments where large currents get shut off regularly. Ideally the power supply would have the following characteristics:

Input from single phase 120 V or 240 V outlet. In reality the unit will be driven by a generator.
Variable output voltage up to 1000 V.
3 kW output power independent from output voltage. It should deliver 3 A at 1000 V to 45 A at 66V and below.
Input PFC.
Quick changing load. The load can be simulated with a resistor and an H-bridge. When switched on the power supply sees a high resistive load that gets rapidly turned off when the H-bridge switches to it's off stage.

I am in an early research phase for the project. I see a lot of work on power supplies that use a first PFC-boost stage, followed by a phase shifted full bridge stage. But these designs typically have a relatively small range of output voltage. Also most commercially available units seem to have a fixed maximum output current. An exception seem to be units from marway.com that output more current at lower voltage.

I wonder whether a combination of several different boost voltages followed by a phase shifter full bridge will get me to the desired requirements.

If anybody knows some literature example or has some ideas on a usable topology I would be thankful for a hint.

[capt bullshot]

For a wide output voltage range at high power levels, usually a buck (non-isolated) or forward (isolated) topology is preferred according to my knowledge (that doesn't include latest and fanciest technology). That's what I've seen in (not the latest) commercially available wide and auto ranging power supplies.
Modern stuff, I believe, uses different topologies, at least synchronous rectification to create bidirectional operation (so it can act as a load, too).

[Faringdon]

Also....

For this, you can use......

Two parallel Boost  PFC stages…...make them share 10ms at a time, as this is easiest. So one has 10ms half sine, then the other, and so on.
Then put in  a 3kW LLC converter for isolation. Vin = 400VDC, Vout = 400VDC.
Then have 3    Paralelled buckboost converters to go from  the 400Vdc to either 1000v, or the 66V and below.
There are many ways to paralell the 3 buckboosts, eg

1...Parent child (parent sets vout, children copy  parents output current)
2...Single error amp on vout , feeding each power stage’s  pwm comparator (digitise the error voltage if you want, then DAC it back at each pwm comparator)
3….Active vout droop
4….Simply clamp each SMPS at its nominal max current (this is variable in your case, but for any specific case, you can clamp it)
5...Transconductance error amps in each buckboost, then connect their outputs together, so its like a single error amp.
6...UCC39002 based share bus.....or do a DIY version of this...ie tweak each  error amp ref voltage till you get sharing....do this iteratively and continuously.
...make one power stage the parent, and just tweak all the others in accordance with this.


You could run the LLC in open loop, f(sw) at upper resonant frequency. Just have good current limiting. Put it in burst mode when at lower power levels.
But to start off, have the LLC in high frequency switching, as you charge up the output caps...have the load switched  out  for this startup period.
The go to f(sw) = upper res freq.
For open loop LLC, choose res components such that the gradient of Vout vs f(sw) for max load is not too steep.
And of course, ensure that upper res freq is sufficiently far from the peak of the Vout vs f(sw) [at max load] graph.

As you know, at the upper resonant frequency, an LLC converter  just gives vout = vin * turns ratio...kind of like a mains transformer.

[asdf336]

The max vout*iout figure of merit is extremely high.  This points to a ranging design with switching series parallel stages. 

You could use pfc->multiple psfb (I have some experience doing exactly that) that can be switched in series or parallel.  I’d first look at that you can do with 650V fets which are commodity. With the common sync rectifier scheme on the secondary the secondary fets will see double voltage plus ringing (ringing is a problem on psfb).  So that’s 250V or so max so maybe 4 stages?

Note the secondaries could share a single primary bridge, controller and possibly transformer which would have significant cost savings at the loss of some optimizations you could implement with multiple primaries.


The other topology mentioned PFC->LLC->Buck Boost is also possible but this is a super wide range.   You’d need giant high voltage fets, SiC probably, hard switching 1000+ volts…..yuck. 

Maybe not out of the question…I did PFC->Vicor->Buck-boost to make 5-300V in a soon to be released product.  But only 100W.     

You could also have multiple LLC secondaries with multiple buck or buck-boost output stages that switch in series/parallel. 

[Mazo]

The need for constant output power would lead me to a flyback or some flyback derived topology,maybe combined with variable frequency and duty cycle control.As the 3kW output power is really high,multiple flybacks to spread the power?(10 "units"->300W per unit which is okayish),ofcourse some interleaving would help the output and input capacitors' life a lot.
Just my 2cents.

[Faringdon]

The other topology mentioned PFC->LLC->Buck Boost is also possible but this is a super wide range.   You’d need giant high voltage fets, SiC probably, hard switching 1000+ volts…..yuck.

Yes i see your point...switching 1000V doesnt sound too great.
I like your idea, asdf336,  about series stages.....
In fact, i think i would go ....
Dual Boost PFC...->400V......Four Two transistor forwards, with secondaries stacked. So each one would have vout from 16-250V. (45A to 3A)
..Then again, thats still a large duty cycle variation for the 2TF's.

...so maybe i'd have the four 2TF's outputting say 50V to 160V.....and then have a buckboost at each 2TF output...to give the 16V to 250V output requirement of each of the 4 stacked stages.

It sounds like this is a one off lab supply, so it can be "done in a shoebox" type of thing.

...so i'd be stacking the 4 buckboost outputs, to give the 65V (and less) to 1000V output

[ogden]

Variable output voltage up to 1000 V.

How variable? In case of science I doubt you are going into mass-productin so, stacking multiple off-the shelf supplies and switching between them makes sense especially when development of actually working & dependable powerful, wide output voltage range power supply will definitely exceed price & resources of said multiple supplies. Just saying. I would rather let my engineers do actual science engineering, not power supply for science engineering...

[dutch66]

First of all many thanks for the comments. I very much appreciate the help!  Another project got in my way for some time and can thank you and answer now.

This is not a one off lab supply. In fact it is neither one off nor lab. It will actually be generator powered and used outside It must be well build and rugged. That being said it could use commercial modules if they can be properly integrated.

The max vout*iout figure of merit is extremely high.  This points to a ranging design with switching series parallel stages. 

The idea to split this up into several smaller supplies sounds very intriguing to me. Even more so if the units could be bought of the shelf. Does anybody have a suggestion for a 0-250V module that can be used in such supply? As we would need to integrate it, it can't be a lab supply more a module with some kind of external control.