HP 0957-2183 Power Supply 48V 1kW, how to turn on? - reverse enginerding

[Yansi]

Hi,

I have just acquired this nice puppy for just a couple bucks. It is a HDD array (server?) power supply, HP part no. 0957-2183. It is a 1kW module, 48V output (plus the 12V 5A standby power out). I could acquire many more of those, and you could too - they are abundant it seems and nobody knows what to do with all of them, as there is no info how to use them outside of the server assembly.

Before I attempt reverse engineering the bastard, I'd like to ask first - does anybody have access to any kind of documentation for it? Or, does anyone have any insider info how these operate? What are the signals and pinouts for all the IO pins at the back of the unit?

I have found this module mentioned here: https://www.rcgroups.com/forums/showthread.php?1005309-A-simple-high-quality-12Volt-100Amp-Power-Supply-Part1/page250

However this bodge does not work for this part no., as the red text warns you at the top of the page.

Here is the photo of the name plate.  On some of the PSUs (found photos on the web), there is sometimes a muRata logo in the right topmost corner - probably the OEM manufacturer of the unit?

Thank you for any info.
Y.

[coromonadalix]

Well they issued a warning indeed.

Time to dismantle it, take not of all the ic's   etc ..etc ...  and try to find the datasheets ....

Or register at rcgroup ask for help or send your unit to a knowledgeable people there and they may be able to do something wth it ??

[Yansi]

That rcgroup forum does not seem to have any credibility for me, hence why I am writing about it here, where we can properly discuss engineering aspects of the thing, not just randomly hacking the thing by making assumptions.

The module is already dismantled.   Photo here:

Sorry for external link, I will upload images here later.

Quick but information dense description of the PSU:

What is inside: It is a dual redundant PSU, two AC inputs, two separate active PFC correction circuit (boost topology), two separate DC links (each two caps is one of them). Not really interested in reverse engineering the PFC, there is nothing interesting in it, apart from it looking like it is an off-the-shelf bought module, as the design style is quite different, then the rest of it.

The main DC/DC: each separate DC bus has its separate DC/DC.  The DC/DC converter is a full bridge, with a center-tapped secondary, choke in the center tap, two STTH2003 rectifiers, one per leg. Each STTH2003 is connected as a parallel pair. Each one has a parallel R-L network connected in series with its anode. The little chokes are heatsinked on a common heatsink with the diodes. What the fack is that RL network for?! R is I think 420ohm, 3W. (

)

The H-bridge uses (WTF!) IRF840 low side mosfets driven by a common low side gate drive transformer. High side uses IRFP450, also driven by a common gate drive transformer. Not sure what type of modulation scheme it uses, as the combination of  weak low side and high power high side fets does not make much sense. Combined together with the fact there is a common gate transformer for low and high sides, it means that you can never have both low and high sides turned on at the same time.  O_o 

What I can say more so far?  Probably that I will throw some schematic on a paper in a minute, and that the decoupling caps for the H bridges are way tooo far from the transistors. (those small brown caps in the middle of the board, mosfets are on the large heatsink under the transformers.)

[coromonadalix]

Your sucess will come at finding the main controllers ic(s) specs, feedback loops  etc...  and knowing how to power them on and off ...  newers psu's have I2c communications and even spi lines too in some models


Not arguing
RcGroups helped me a lot in the past for 1kw and 2 kw psus i've bought for cheap prices, i find them reliable, but sometimes, you dont / wont have 100% succes each time

[Yansi]

Not arguing either, just when investing the time to find the feedback loops and controller ICs, why not investing a bit of time to discover the overall concept and DC/DC topology, go through its good and bad to learn something?

So here's the schematic of the main DC/DC. The other one is identical, working in parallel.

Main DC/DC power schematic:

I would note here a couple of things:
a) Having a common gate transformers for low-, and common one for both high-side transistors is quite weird. (I have no idea how this can be worked with).
b)Someone was smoking some pretty stuff when designing the gate signal shaping circuit for the high side. Never seen this bizzare circuit. The aux N-MOS is MMBF170 or equivalent (SOT23 marked as "6Z").
c) no transformer primary DC isolation capacitor is present
d) low side is IRF840, high side is IRFP450.
e) what the hell on earth are doing those LR couples in series with the diodes? (Seems like the small cores are deliberately driven into saturation).

This is some weird configuration.

EDIT: It could probably be only driven as a deadtime controlled bridge: One diagonal on, all off, other diagonal on, all off.  However that still does not explain the asymmetrical rating of the mosfets - as in this case it does not make sense.

[nctnico]

Looks like a forward converter to me. The inductors in series with the diodes are to soften the reverse recovery pulses from the diodes. The reverse recovery snap-off can cause nasty EMI noise.

[station240]

1. Find the 12V STB supply pins
2. find the 0V pins
3. Find any pins with floating voltages that are clearly almost a full voltage, but not quite. (eg 4.8V, 11.8V)
    These are clearly pulled high by a resistor.
4. See what happens if you pull those floating pins low (to 0V).
5. Make a note of what remaining pins are left.

You still need to trace back where the 5K pot should connect to. I suspect it's a fine control, and has fixed resistor in parallel so it still produces 48V without the pot.

[Yansi]

Looks like a forward converter to me. The inductors in series with the diodes are to soften the reverse recovery pulses from the diodes. The reverse recovery snap-off can cause nasty EMI noise.

Well, it definitely is forward, and not flyback. But which forward? How it is driven and why's there different lo/hi mosfets and why are those gate drive transformers connected so weirdly, sharing both low and both high transistors?

1. Find the 12V STB supply pins
2. find the 0V pins
3. Find any pins with floating voltages that are clearly almost a full voltage, but not quite. (eg 4.8V, 11.8V)
    These are clearly pulled high by a resistor.
4. See what happens if you pull those floating pins low (to 0V).
5. Make a note of what remaining pins are left.

You still need to trace back where the 5K pot should connect to. I suspect it's a fine control, and has fixed resistor in parallel so it still produces 48V without the pot.

The pinout of the IOs is likely the same as the other supply on the RC groups. Haven't checked in depth, but it seems really the same. What differs with this HP part number, that ... a relatively good source of information warned me, that this PSU needs to get written something via the I2C port to get turned on. 

What 5K pot?

[coromonadalix]

maybe the i2c is the psu status to the server

you need to find the controller chips behind ... yeah its seems a forward topology but we need to tackle the brain .. i want to see the brain  loll  more photos please

For bus sniffing you have bus pirate ...  but at the same time you need a server to know the protocol / data used ?