High Power SMPS's done with cheapness and ease and speed.

[ocset]

Hi,
I wasnt going to post.....then i thought...hang on, this may be  of use to others...

Is this the “cheap, quick and easy way” to do high power SMPS’s?

The attached shows two  identical Two Transistor forward converters in parallel…..each is 1.2Kw (50vout and 24iout)
If any one trys to hog all the current…its current limiter kicks in and stops it from doing so.

Many more can be similarly paralleled to build up a high power SMPS

No feedback loop fighting.
Nice thermal spreading.
No interconnecting electronics needed to ensure sharing.

What’s not to like?
(attached is the pdf schem and LTspice sim)

[fcb]

It simulates OK. Well done.

Would V2 and V7 sources be the same in the application? Add some capacitance to the front end and offset the phase (make a slight tweak to the timing network values).

Is this 800V source ESR correct? Isn't this coming off a really long wire?

[ocset]

Thanks yes i forgot the 9.5 ohms source impedance.
Yes the sources would be the same one.

I suspect that when on say 70% load……with say 10 SMPS’s in parallel like in the way of the top post…..then maybe one unit might be totally off…and if a sudden 100% load transient occurs…..that “off” unit (with its error amplifier railed low) may take some time to get “compus mentus”, and so sudden transients may not be as well served.

[filssavi]

It will never work well even with 2 converters in real life, what  will happen is that the mismatches in parameters will make sure that you have a bunch of circulating current killing your efficiency
You basically have 2 control loops trying to keep a different constant voltage to the same node.

Also the non synchronous switching will probably cause problems (although the transoformers inductances will save you  from blowing stuff up probably)

There is a reason if for high power LLC resonant is the default

[ocset]

You basically have 2 control loops trying to keep a different constant voltage to the same node.

Thanks, yes…but I don’t see why  thats a problem…..one will do the regulation, and the other  one will see its error amplifier rail low….but due to the current limiter in there……in fact, one  could be current limiting, and the  other one doing voltage loop regulation.

The described method is the modus operandi of many off the shelf SMPS’s.

Also, the Vicor   DCM3623x50M53C2yzz  DCDC module works in exactly the way that I have described…and its datasheet says its good for paralleling up to 8 in parallel, just as I have done in the top post.

Vicor   DCM3623x50M53C2yzz  DCDC module datasheet:
http://www.vicorpower.com/documents/datasheets/DCM3623x50M53C2yzz_ds.pdf

what  will happen is that the mismatches in parameters will make sure that you have a bunch of circulating current killing your efficiency.

..Thanks, as you know the LLC (which was mentioned elsewhere in your post)  is known for its oft-times high primary circulating current…depending on the "m"  value selected.

[fcb]

If you've only got a limited number of channels why not seperate the switching from the regulation - essentially drive all the half-bridges from the same PWM signal (via transformers or opto's or POF fibres).

So basically one master channel doing all the voltage regulation and current sense, and then a bunch of identical half-bridge stages slaved to that master.  One loop to rule them all.

[ocset]

Thanks, I admit that one controller is better…..(eg one error amplifier for multiple units etc etc)……though in some ways I like the idea of a totally independent unit, that doesn’t need its control linking up to other units…where the local ground may not be the same due to current flows, etc……and indeed, different units are likely to be on different PCBs and a single error amp signal would need wiring between units.....for two units maybe a single  feedback signal is best, but for a great many units,  we like the idea of multiple independent units, each self contained....ie as in the top post.

[SilverSolder]

Do most of the downsides go away if all the parallel SMPSs output through a diode into the shared load?

[ocset]

thanks but  one smps can still hog even if they all have diodes at their output.
Also, as you know, paralleling is for high current, so an in line diode isnt really wanted.

[fcb]

Your design doesn't require isolation right?

Why not just build a big a$$ synchronus buck converter.

[ocset]

Thanks, yes, thats a good consideration....we also thought of paralleled Sync Bucks.
...and also open loop sync bucks with fixed D switching to knock the 820 volts down to about 400v or so...then  add further regulating  converters from there to give the 48V-50V at 240A

But The time frame means Vicor ready made modules are the top solution of the co.

[fcb]

Thanks, yes, thats a good consideration....we also thought of paralleled Sync Bucks.
...and also open loop sync bucks with fixed D switching to knock the 820 volts down to about 400v or so...then  add further regulating  converters from there to give the 48V-50V at 240A

But The time frame means Vicor ready made modules are the top solution of the co.

I'm confused.  You mention:

Vicor DCM3623x50M53C2yzz  DCDC module datasheet: http://www.vicorpower.com/documents/datasheets/DCM3623x50M53C2yzz_ds.pdf

In a previous post on this topic, these are 50VDC input max.  Do you mean some other Vicor converter?

[ocset]

Thanks yes,  we will have a 820VDC input, which we will drop down with a vicor BCM module (divide by 16) ....to 51v...then the DCM3623 modules  will  run off that.
As they draw current, there vin will decrease further due to the cable  resistance.

The DCM3623 will still run up to 55v.