load-sharing DC power supplies (Paralleled DC outputs)

I am writing this because of some battles I'm having with the switcher PSU's for a Neve Capricorn digital recording desk.  The "DSP rack" has two 5 Volt/200 Amp switching power modules with DC outputs which are hard-wired/bolted to each other via large/thick copper bus bars.  The service manual info is kinda vague on how to set the voltage trims (like "each has to be within 0.1 VDC of the other"), so I've been in contact with AMS-Neve tech support for assistance.  (IE, under what conditions are the trims supposed to be set...idling?  Under load?  If under load, how can I "iso" one from the other?  etc  etc)

In the past, I've encountered dual PSUs in things like large PA desks, but each PSU was isolated from the other via high current diodes.

So, I now realize I need some knowledge as to how two (or more?) PSUs can have their DC outputs literally paralleled and yet not "fight each other".  I guess large computer mainframes/servers use this same sort of load sharing, but I've never looked deeply into how it is supposed to be done correctly.

In a way, it all reminds me of the issues when using jumper cables to parallel the 12v batteries between two automobiles....

Best,

Bri

Brian Roth said:

So, I now realize I need some knowledge as to how two (or more?) PSUs can have their DC outputs literally paralleled and yet not "fight each other".

Click to expand...

Hi,
I believe the rectifier diodes on the secondary act as iso-diodes.
Regards,
Milan

Logic suggests they would easily release smoke if out of trim, but hypothetically if the PS is high current and relies upon the load for the pull down, it only needs to regulate how much current it supplies to regulate the output voltage. In that scenario, trimming the two supplies output voltage will mainly adjust how equally they share the common load between them.

Depending on the nature of the failure mode, one could pretty seamlessly pick up the load if the other dropped out ... or not.

This all sounds a little scary to me, but a 5V supply would waste around 10% in diode sharing.

JR

Usually the supplies should be tested under load, but.....  I would await for tech support's answer since it is an application for their system...  The only other thought would be to see if the power supplies are COTS and designed to be paralleled per their datasheet (if COTS) and read any thoughts the manufacturer may have had...

If it is one big diode load share, then other protection mechanisms (hopefully designed into the supplies) would kick in; for example, if one 200A supply is drawing too much because it is feeding the load *and* the other supply because it is much worse than -0.1V from the other, then hopefully its current limit would engage, resulting in not-desirable behavior in the desk (like a brown out et cetera)... 

One would hope....

Good point. If they both are set to current limit at slightly more than 1/2 the nominal output draw, the hotter voltage output will current limit and dip down to meet the lower voltage output until no longer in current limit and both are roughly supplying 1/2 the current. ... still a little scary, but if they are designed to be operated that way it should be OK.

JR

0.1V at 200A is only 0.5 milliohm, so the bus bars do the load sharing.
Power supplies are not treating sink and source the same. The ballasts can source 200A, but may not be capable of sinking much current. If you feed a 5V PSU with 5.1V PSU, the former's regulation will turn off the ballasts, until the latter's voltage drops.

I am not quite familiar with the term "ballasts" or perhaps I know of it by another name?

If I think it is what I think it is (dangerous for my feeble mind), then yes, we are perhaps getting into buck mode in one direction (mains HV to DC output LV), with proficiency in only that direction, and thusly ballast could just mean the output capacitance for the filter caps (or pi-esque networks for filtering et cetera)?

If so, then the "stronger" supply would in effect just prop up the output filtering capacitance voltages in the "weaker" supply while the switching portion of the "weaker" supply upstream of said filter caps output would just coast (perhaps based on local feedback sensing before the caps?)...

Interesting thought indeed...

I would definitely wait for tech support and in the meantime if curious, look for a data sheet if these supplies are COTS (Commercially available Off The Shelf)....

Just guessing here, but a load drawing hundreds of amps may not need an active pull down,  so the PS just modulates the current out to hit target output voltage. Without current limit the higher voltage PS gets all the load, with similar current limiting around 1/2 draw, they could share. ... or not.

JR

The ballast is the series element in a linear regulator.
In SMPS, it's somewhat different, but basically its a piece of circuitry that is capable of sourcing current but not sinking.

disconnect the buss bar, calibrate each supply with a 1 amp load and i bet they will work fine when bootstrapped.

CJ said:

disconnect the buss bar, calibrate each supply with a 1 amp load and i bet they will work fine when bootstrapped.

Click to expand...

Damn!  CJ, that is a great idea, and if my Dale 8 Ohm/250 Watt non inductive load resistors were here and NOT 250 miles away from this current project site...  lol.

This 5 VDC at 400 Amps stuff kinda scares me....I could arc weld with that sort power....

Best,

Bri

All the SSL parallel configs use diodes to block each supply from the other, and you calibrate them as CJ suggests, one at a time exactly the same with the same load.
Don't the manuals have a cal procedure?  Neve were always more oblique than SSL starting with Vseries.
Mike

sodderboy said:

All the SSL parallel configs use diodes to block each supply from the other, and you calibrate them as CJ suggests, one at a time exactly the same with the same load.
Don't the manuals have a cal procedure?  Neve were always more oblique than SSL starting with Vseries.
Mike

Click to expand...

No diodes at all.  The "+" and "-" bolts from each of the two supplies are attached to a large/thick copper bar.  And, I could not find any cal procedure in the manuals...hence my email to AMS-Neve.

Best,

Bri

They were never the same company after RN bankrupted it the first time.  Nothing compares to Neve manuals from the 70's.  Or for that matter dbx and Urei manuals.
Mike

Brian, I would temporarily connect them to a lesser load (by removing a number of modules or by connecting a dummy load) and use an amp clamp to measure and match the two feeds to the sum point. If the buss bars are too big for the amp clamp you could temporarily use a couple lengths of heavy gauge cable.

David Kulka said:

Brian, I would temporarily connect them to a lesser load (by removing a number of modules or by connecting a dummy load) and use an amp clamp to measure and match the two feeds to the sum point. If the buss bars are too big for the amp clamp you could temporarily use a couple lengths of heavy gauge cable.

Click to expand...

Dear David,

Unfortunately, the (piss-poor) mechanical design will not allow that.  I wish I had taken some pictures because it's tough to describe.  The two PSU modules slide into "cubby holes"  Their two protruding bolts DCV output bolts on each module are attached to the two closely-spaced, large copper bars with nuts, and THOSE bars then bolt onto a pair of copper "stubs" that protrude from the bottom edge of the card cage above.

Gaining access to the PSU modules is a huge PITA, involving removal of the disk drive bay, and then the panel with the 240V power inlet and circuit breakers (with multiple wires attached).  It's one of the worst designs I have ever seen!

So, the only choice appears to be to extract the modules, connect a temporary mains cable, and set the output voltages with the modules loaded into some sort of dummy load.

Best,

Bri

Maybe an option would be to measure and compare current in the PSU mains cables? The efficiency of the supplies should be very similar, so it would least get you in the ballpark.

Or might there be some way to "sample" the DC current by connecting wires to two points on each buss bar and measuring the voltage drop?

David Kulka said:

Maybe an option would be to measure and compare current in the PSU mains cables? The efficiency of the supplies should be very similar, so it would least get you in the ballpark.

Or might there be some way to "sample" the DC current by connecting wires to two points on each buss bar and measuring the voltage drop?

Click to expand...

Dear David,

I eventually found a problem which was causing one of the supplies to shut down.  Each of the 5V/200A PSU modules has a fan mounted on the back side of the module (opposite end from where the +/- 5VDC bolts protrude and are connected onto the copper bus bars).  It was impossible to see whether the fans were functioning without removing the module (or a bunch of sheet metal covers in the rack).  With each module removed and powered, the bad fan was found and replaced.

Sidebar:  "Going by ear" to "detect" the fans with the modules installed into the rack is impossible, due to the large cooling blower for the entire rack!  That blower looks like a somewhat smaller version of the blower in my house's furnace!!

I have not yet received an answer from Neve re. their recommended procedure to trim the modules to have "less than 0,1 VDC difference" between the modules.  I ended up doing some "test and try" with the modules unloaded, since the "dummy load" suggestions earlier in this thread hadn't been posted while I was at the site, 250 miles from my home-base.

I ended up trimming each PSU to 5.40 VDC no load, which resulted in 4.95 VDC across the bus bars with both supplies installed and running the entire rack.  I wanted to hit a bit over 5VDC, but it takes two guys around 20+ minutes for each dismantle/tweak/reassemble iteration.

Sigh....

As I mentioned in a previous posting, I didn't think to take any snapshots of the "innards" with the disk drive bay removed, and the power inlet/breaker/power monitor box removed (and dangling by a bunch of wires).  All I have is a shot with everything assembled....attached.

Best,

Bri

PS, here's a snapshot of the entire DSP rack...

Bri