Vicor DCDC modules in parallel.....way of getting tighter current sharing?

[ocset]

Hi,
Sorry I wasn’t going to post, but then  I thought…hang on..Vicor is good…people are going to want to know about Vicor…and also, I offer a solution here too….(admittedly I am not quite sure that the module will allow this, and its datasheet doesn’t  say).

So……..We wish to put two Vicor  DCM3623 modules in parallel and assure sharing of current between them. (40vin and vout =  48v at 10A total;   5A from each vicor DCM3623 module)
The DCM3623  module assures sharing by having a current limited output….if any DCM36723 starts to ship too much current, then its current gets limited, and its Vout drops, and the other vicor module then starts sharing the load.

The problem is that DCM3623 current limit can be anywhere from 100% to 143% of nominal Iout (6.67A) …that’s not good enough sharing for us…..therefore we have put a current monitor on each Vicor DCM3623  output and an error amplifier so that we can limit the iout of each module to 5A……………….So the attached schem should give 48Vout at 10Aout from 40Vin….5A coming from each vicor DCM36723 module…..hopefully nice sharing.

We simply make the  external  error amplifier  pull down the TRIM pin of the DCM3623.
The DCM datasheet doesn’t state enough detail on the TRIM pin for us to know if this is workable?

Do you have an opinion of whether or not it is workable?
The datasheet (pg8) does say the bandwidth of the TRIM pin  is 30Hz, but that doesn’t necessarily tell us if the attached schem would  work.

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

[David Hess]

For modules which include remote sense pins, I have seen current sharing enforced by driving remote sense.  The same should work by driving trim.  The same technique works with integrated regulators by driving either remote sense or adjust.

Note that the current could be sensed at the input or output.

[digsys]

There are countless ways of current sharing by driving Vset / VTrim Inputs, even LM317s can be made to share perfectly. IMO, you don't need ultra-fast closed loop sharing timing. I often slow down channel response rates to a 100mS or longer. We're only talking a few % > 5% imbalance occasionally, heck, I sometimes even just use Polyfuses on each output, heat-shrinked together back to the VTrim / Vset, so 30Hz response is HEAPS :-)
There are many application circuits showing all sorts of sharing techniques.

[ocset]

Thanks,
Page 14 of the below document shows basically what we want to do…

http://www.vicorpower.com/documents/application_notes/an_Achieving_High_Accuracy_Voltage_Regulation_DCM.pdf

.....I notice that the feedback compensation capacitor C1 is 2.2uF. If this capacitor was much lower than 2.2uF then the whole thing would go unstable, would you agree?
(this being due to the 30Hz control Bandwidth of the TRIM pin)

[digsys]

To be honest - I don't often follow "recommended values". I often work out my own, do my own limits testing based on how / where I'm going to use the "device".
When component manufacturers write up application notes, they make sure, that for pretty much ANY use the device goes in to, it will stay stable, no matter what. PLUS, they leave plenty of tolerance / headroom. They can't afford to get things "just right". All suggested circuits are just that ... suggested.
What it means is, IF you use what they suggest, or the "average" that most engineers use, you should never have a problem. Naturally, if you have a cr@ppy layout, or bad impedance paths etc etc , you'll come unstuck
My answer would therefore be - Go with recommended (until you've designed a few 100 and feel more confident) .. especially if you're asking the question 
One PS: When going that route - they do "occasionally" screw-up, and that's where forums etc are very handy.