# Large numbers of paralleled DCDC modules?...more than datasheet recomends

1. ## Vicor Bus Converter module

I hope its OK to set this thread up as a question point for Vicor power modules?
(saves doing lots of different posts)

Hi,

We wish to use two BCM4414xH0E5035yzz SMPS modules in parallel for a 3.3kW load.
Vin = 800VDC, Vout = 48VDC, 70A.

BCM4414xH0E5035yzz datasheet

http://www.vicorpower.com/documents/...5yzz-VICOR.pdf

We must be careful that input and output trace resistances (“RINPUTx” and “ROUTPUTx”) are matched, otherwise the modules won’t share current.

However, as seen in the BCM4414xH0E5035yzz datasheet (Page 6), the Module’s internal output impedance, “ROUT”, can be anywhere from 7 milliOhms to 20 milliOhms !!!!!!

From page 2 of Vicor App Note on current sharing……..

Vicor AN:016 Current sharing
http://www.vicorpower.com/documents/..._appnote16.pdf

From AN:016, page 2, Kirchoff’s Law tells us……

I1.ROUT1 + I1.K^2.RINPUT1 + I1.ROUTPUT1 =
I1.ROUT2 + I1.K^2.RINPUT2 + I1.ROUTPUT2

So….

I1(ROUT1 + k^2.RINPUT1 +ROUTPUT1) = I2(ROUT2 + k^2.RINPUT2 +ROUTPUT2) ……Equation 1

Now……
ROUT1 = 0.003
ROUT2 = 0.02
RINPUT1 = 0.1
RINPUT2 = 0.1
ROUTPUT1 = 0.005
ROUTPUT2 = 0.005
K = 1/16

Now find how different are I1 and I2………………i1/12 = 3.02
This is horrendous!!!!!!!
I1 is 3x i2

….And that was assuming input and output trace resistances being exactly matched…which they are not likely to be…if I make the ROUTPUTs 0.003 and 0.02 respectively, then I1/I2 = 6.32 !!!!!!

Can you please confirm that Vicor module sharing_in_parallel is basically a non-runner?

•

2. ## Re: Vicor Bus Converter module

No, that's likely thermal variation just as a mosfet has widely varying Rdson yet is widely considered appropriate for paralleling.

You're citing a Vicor document which details how to parallel Vicor modules and includes a case study showing 14% worst case light load sharing variation (dropping to 1% at full load) of 7 paralleled modules....I think its safe to say you can parallel 2 of them...just follow the app note.

3. ## Re: Vicor Bus Converter module

No, that's likely thermal variation just as a mosfet has widely varying Rdson yet is widely considered appropriate for paralleling.
Thanks, but it does say that those "min typ max"'s for "ROUT" are at one given temperature.

The example given in their AN:016 page 2 looks very unusual...they show ROUTPUT1 and ROUTPUT2 as being equal, and they make ROUT1 and ROUT2 equal to each other...which they may not be.

I do agree however that seven or two exactly matched modules can be paralleled with perfect current sharing if input and output trace resistances are equal.

But There is no way of buying "matched" modules.

4. ## Re: Vicor Bus Converter module

Ok fair but consider the same thing is true for any mosfet yet we parallel them all the time with the expectation of better than 2:1 sharing. Steep Rdson thermal gradients are one reason this works.

When a manufacturer says you can do something you usually can. I'm building a board right now that's expecting to parallel Vicor modules....

You do need to be careful about using identical part numbers. They uprev their parts sometimes with small variations.

5. ## Re: Vicor Bus Converter module

Thanks, i am being pessimistic in considering the situation of opposite ends of the tolerance spectrum occurring together...(ie one module with 20mR ROUT, and the other module with 3mR ROUT.)...but the datasheet effectively says this can happen......and these modules are \$830 dollars each...and the mis-sharing that would happen in such a case is a killer for our application.

- - - Updated - - -

Also, the attached pair of open loop sync Buck’s in parallel has better sharing properties than the Vicor modules. (pdf schem and LTspice sim attached)
As you may see, one sync Buck has ten times more trace resistance at its output than the other…however, the current sharing ratio is 0.775/1.14. …ie 0.68 to 1
…For a ten times difference in output resistance, that is a lot better than the vicor modules.

Another point about vicor modules in parallel that don’t share, is that one will likely trip on its output overcurrent …this is because for a 35A vicor module (ie BCM4414xH0E5035yzz), the overcurrent trip is at 37A.
This is why Vicor modules must share very equally…but with the datasheet stated tolerance on ROUT…the necessary degree of sharing may well not happen.

BCM4414xH0E5035yzz datasheet
http://www.vicorpower.com/documents/...5yzz-VICOR.pdf

6. ## Re: Vicor Bus Converter module

if you can test Rout, then you can match them within a reasonable range.
say, instead of 3 to 20, bin them in small steps
3 to 6, 7 to 10, 11 to 14, 15 to 18, 19 to 20
units within a range are more closely matched than a random selection

7. ## Re: Vicor Bus Converter module

Thanks, I think testing ROUT would be the way to go.
The only way to do it would be to puit it on load and measure its vout, then put it on no_load then measure its vout….then the voltage difference in the Vout's equals the Load current (from when loaded) x ROUT.
Would you agree this is the only way to measure ROUT in a Vicor BCM4414xH0E5035yzz module?

BCM4414xH0E5035yzz datasheet
http://www.vicorpower.com/documents/...5yzz-VICOR.pdf

Also, would you agree that Vicor BCM4414xH0E5035yzz modules in parallel must be extremely well matched, otherwise the one with lowest ROUT would trip out on its overcurrent trip?
This is because the nominal maximum current for the module is 35A…but the overcurrent trip level is 37A

Sorry to ask so many Qu’s, I wasn’t going to ask, then I realised these may be useful to other Vicor users.

•

8. ## Re: Vicor Bus Converter module

Vicor seems to suggest that the modules can be well parallel operated relying on typical figures. They just suggest to balance external wire resistances. At the same time, by giving a wider range of Rout and Itrip parameters, they don't guarantee anything.

It's your decision if you want to build a professional product with these devices.

If you go for it, you should clarify your doubts with Vicor FAE and sales. Do the modules require Rout binning and routine test of load sharing?

Instead of determining a theoretical Rout value, I would check current sharing at e.g. half and full load and set a limit value for current unbalance.

1 members found this post helpful.

9. ## Re: Vicor Bus Converter module

Thanks, We’re still evaluating different motor voltages, and how many motors to use, so it may be 48V…or may be 96V.
Is it acceptable to stack the outputs of two Vicor BCM4414xH0E5035yzz modules on top of each other? (to give 96Vout) I see no reason why not, since they are isolated outputs…..just that we will need isolated control bias supplies, but that’s not a problem for us.
We are thinking of running each Vicor module at half nominal power, so as to avert current sharing problems as discussed.
So for example , we may choose to drive four 96V motors at 3.2Kw a piece….thats 33Amps per motor………..so that would come from a parallel combo of two stacked Vicor BCM4414xH0E5035yzz modules stacked on top of each other…ie 4 x BCM4414xH0E5035yzz modules altogether.

BCM4414xH0E5035yzz datasheet
http://www.vicorpower.com/documents/...5yzz-VICOR.pdf

10. ## Re: Vicor Bus Converter module

as FvM said, talk to the Vicor FAE
the other thing is to build it and test it.

try two matched units to see is screening and binning works
try two severely mismatched units for worst case

as FvM suggests: check current sharing at e.g. half and full load and set a limit value for current unbalance.
you might also check current sharing at no load, 10%, 20% etc once - never hurts to know

1 members found this post helpful.

•

11. ## Re: Vicor Bus Converter module

Thanks, but i'm surprised not to hear more support for actully measuring ROUT?....( method as in #7)

12. ## Re: Vicor Bus Converter module

What do you expect? Confirmation of ohms law?

Besides Rout variation, there can be also small deviation from nominal 16:1 voltage ratio, e.g. due to variation of transformer main and leakage inductance. Thus measuring the actual current sharing under load seems more reliable. But as said, you are buying a product with assured paralleling capability. Thus it's the vendors job to tell you how.

13. ## Re: Vicor Bus Converter module

Call Vicor and let us know. I've talked to them about paralleling although I hadn't latched onto this particular spec to ask about specifically (my mistake).

As mentioned I recall finding language saying if you do parallel it needs to be an exact part number match (can't find it now) - they periodically make small PN changes. My suspicion is that in practice the same part numbers are well matched but the wide R range gives them flexibility to change the product in the future while adhering to the original spec. And/or they may just not be interested in accurately characterizing this particular spec.

14. ## Re: Vicor Bus Converter module

Thanks,

A contractor has now given us a Block diagram for a power supply for a single_motor_drone power supply using vicor modules……

Its 50V at 240Amps to the motor/Drive.

This ultimately comes from an 820VDC source. ( no isolation needed from 820VDC source)
However, there is a 320 metre cable between the 820VDC source and the device…..go_and_return resistance is 9.5328 Ohms.

So (at the other end of the cable) the block diagram has eleven Vicor BCM modules in parallel, with a Vout of 36.6V at 370 Amps………this is then fed to thirty-nine(!!!) Vicor DCM modules in parallel to give the 50V at 240Amps for the motor/drive.

From reading all the Vicor literature, I would say this is no_can_do, would you agree?

AN:016 from vicor states that more than ten BCM modules in parallel becomes “difficult” due to the implemented input and output trace impedance mismatching and also non-uniform cooling of the modules….interestingly it avoids saying that intrinsic ROUT is a problem for parallel operation

The DCM3623T50M53C2T00 datasheet states that no more than eight can be put in parallel….so 39 looks out of the question. However, a read of Vicor AN:030 makes it look like an unlimited number of DCM modules can be paralleled….however, page 6 of AN:030 warns that if current sharing at high temperature is uneven enough, then all the paralleled DCM modules could fail…..however, we can mitigate that scenario by using big enough heatsniks with blown air…..so we cannot find out why only eight DCM modules can be paralleled?

The Vicor PRM modules appear to offer the ability for unlimited numbers of modules to current share in parallel….just that one has to buffer the “share” signal from the single PRM “master” so that it can be transported to all the PRM’s used as “slaves”.

DCM3623T50M53C2T00 datasheet
http://www.vicorpower.com/documents/...53C2yzz_ds.pdf

BCM4414xH0E5035yzz datasheet
http://www.vicorpower.com/documents/...5yzz-VICOR.pdf

PRM48AF480T400A00 datasheet
http://www.vicorpower.com/documents/...T400A00_ds.pdf

15. ## Re: Vicor Bus Converter module

820VDC down to 50V at 240A looks to be a good fit for a standard H bridge converter with 1200V SiC mosfets

this is easily done with planars and current doubler output

The thing is to have soft control on the output to allow for the long cable 320m you need a soft start and the right reaction to load steps ( i.e current jumps ) on the 50V so that you can handle the 15A in the cable - i.e soft turn down if the load is suddenly removed - and with somewhere for the cable energy to go.

We have done 16kW systems with 1km of cable

Vicors need a lot of caps - which can cause you grief in a long cable sytem - do you partic need low height?

kind regards,

1 members found this post helpful.

16. ## Re: Vicor Bus Converter module

One major question: Why parallel all of them when you probably have 4 or more motors? Give each motor its own bus rather than sharing a single gigantic one.

1 members found this post helpful.

17. ## Re: Vicor Bus Converter module

do you partic need low height?
The cable length is 320metres
One major question: Why parallel all of them when you probably have 4 or more motors? Give each motor its own bus rather than sharing a single gigantic one.
Thanks, i'll recheck this, it was told that it would be one motor, but i cant see how that works....it would then need a tail motor like a helicopter?

820VDC down to 50V at 240A looks to be a good fit for a standard H bridge converter with 1200V SiC mosfets
Would you mean paralleling the FETs, ie all of them, to spread dissipation, since its 12kW output.
Or do you mean parallel several full bridges?

•

18. ## Re: Vicor Bus Converter module

we can happily parallel 3 or 4 SiC devices for each switch in an H bridge ... are you looking at Vicor just to keep the height down?

19. ## Re: Vicor Bus Converter module

Height isnt a problem no.
I dont know why Vicor, but partly because their guys like vicor and they used it on a much lower power drone.

Also, since there is no requirement for isolation, what about eight x dual cascaded bucks in parallel ?(with sync rects)
One error amplifier signal for all of them so none hog current

20. ## Re: Vicor Bus Converter module

how does a drone become relevant here? and if your buck converter has one of the primary switches fail ... ?

--[[ ]]--