Teardown: HP DPS-1200FB 12V 100A PSU

[NiHaoMike]

High wattage 1U PSUs have amazed me ever since I got to work with some in a server room. After I found one on Amazon for only $20, I decided to buy one in order to take apart.

It's the highest quality and most complex PSU I have ever taken apart! There's a UCC3895 full bridge controller, two UCC3818 PFC controllers (two phase PFC), a PIC, an EEPROM, some oddball "DNA 1006" ASIC (cryptographic anti counterfeiting measure?), and a Freescale chip. Output rectification is synchronous as expected for that power level, with 6 MOSFETs. The output transformer secondary seems to be made from stamped copper sheet, in a group of three in parallel. The output capacitors are strangely on a secondary board with some resistors and SOIC MOSFETs on them, presumably to quickly discharge them on power off and maybe as a minimum load to keep it stable.

As for how to use it, a jumper from pin 33 to pin 36 will power it on and allow it to be used as-is as a very high power 12V power supply. That can be tweaked up to about 12.8V with an internal pot, good enough for testing automotive electronics but not enough to charge a battery. The output caps are only 16V so they would be marginal at 14.4V anyways.

For the more advanced hackers, it would be trivial to add a dual diode to the transformer output and get 24V or so. Or by doing a little hacking of the secondary winding by disconnecting two of the secondary windings and rewiring them in series to a separate rectifier, get about 48V at 16A and 12V at 33A, perfect for a high powered audio amplifier and a gaming capable HTPC. (That would actually be a very cool project - a gaming PC built into an AV receiver along with a digital amplifier that would put any common receiver to shame.)

[turdferguson]

Do you know how low the voltage can be adjusted? I'm looking for 9.5V-10V.

Thanks.

[NiHaoMike]

Adjust the pot closest to the UCC chip to get it down to about 11.7V. Then add two or three series diodes to drop the voltage.

Note that if you're trying to emulate a 3S Lipo pack (can't think of much other application for 9.5-10V at high current), those are actually around 12.6V full charge so the PSU used as-is would do just fine.

[melson]

Mike,
I would like to adjust this power supply to 14v. I changed the voltage adjust pot to 20k however the over voltage protection shuts the power supply down around 13.7v. Can you tell me how to adjust the over voltage trip point higher? In addition, I am curious as to the function of the other 2 trimmer pots.

Thanks for any thoughts you may have, Mel

[NiHaoMike]

From what I have read, nobody has tried to hack the OVP. I don't think it would gain very much as the output caps are only rated to 16V, meaning that the stock 12V is right on the "recommended" rule of thumb limit.

[codeboy2k]

That's a great deal for $20.  I bought a DPS-700GB for $14 and I thought that was a good deal back then too.

mine is 700W total, 12.15V at 56A, 3.3V @ 3.5A and 10.5V@ 1.5A.

I don't know what the 10.5 V was used for in the server.

These are great units, so much better built than the standard PC PSU crap

[oskarjose]

I have several of these power Supplies that do not give the 12v, the fan starts but the green LED not, know how to fix this flaw

[Aerospace]

I just got one of these supplies and I could not force it on using the single resistor method.  I had to use two resistors. First resistor is a 100 ohm from pin 33 to 12V ground, then I soldered pins 36 & 37 together and added a 560 ohm resistor from these two joined pads to the 12V ground.  System came right up with a green light.  Now to mod it to 13.55V.... Amateur radio stuff....  Anybody figure out to change the OVP level yet??
  I hope this information is worth the case of beer I consumed trying to figure this one out.   

[RedNeckEngineer]

I also have one of these. I don't think it's possible to do a complete tear down without destroying it. It would be nice to find the schematic.

[obd.tech]

Has anyone found a way to up the Voltage output  to somewhere near 13.8V yet?

[djQUAN]

I have a 12V 244A server powersupply. I tore it apart to check for damage (got it from the scrapper by the pound, definitely less than $20) but did not take pics.

Lots of power devices, big transformers, inductors and caps. Everything is huge.

Would take pics if anyone's interested.

[lukaq]

I have a 12V 244A server powersupply. I tore it apart to check for damage (got it from the scrapper by the pound, definitely less than $20) but did not take pics.

Lots of power devices, big transformers, inductors and caps. Everything is huge.

Would take pics if anyone's interested.

Sure, let's take a look. @244A should be pretty solid one

[DrTune]

Hiya,
I reverse-engineered the CPU in the DPS-1200FB and figured out how to read all the internal status stuff (voltage+power in/out, internal temperatures, fan speed etc) over I2C (e.g. with a Raspberry Pi or whatever you prefer). Have not figured out how to turn the output on/off yet.

Python code here (and disassembly of the CPU firmware in the power supply etc)..
https://github.com/raplin/DPS-1200FB

[NiHaoMike]

I ended up modifying mine into a charger for a very large 4S lithium battery - 11.2-16V output voltage range with current limit. First thing I did was change out the output caps for some rated to 25V. Not sure what exactly does the OVP but I did a lot of hacking before I finally disabled it - I think it was the little resistor on the back side of the control board to the right (as viewed from the top with the output connectors to the right) of another small resistor and a somewhat bigger one below it, right on the other side of the board from the trimmers.

That reverse engineering of the firmware is very helpful. My original plan was to reverse engineer the connections to the PIC and then write new firmware for it, but reusing what's there might be a lot simpler. Probably still have quite a bit of hacking to go in order for it to operate with an adjustable current limit and PFC gating/variable bus voltage to boost efficiency. Not having I2C control over the on/off isn't a big deal since I already plan on at least two hardwired logic signals that both must be in their active states in order for the output to be enabled - one controlled by the BMS and the other by whatever SoC I end up using for the wireless control interface.

[DrTune]

Hm I was thinking the PSU looked like a PITA to take apart, due to the soldered PCB edge connectors all over the place; hard to even access the backside of the control board..  Stuff is really packed in there.  I'd like to retrofit something like an I2C DAC and have programmable output voltage (I didn't really dig into what's on that control board but I saw a TI PWM controller which I'm assuming runs the main DC-DC conversion down to 12v, using an opto-isolated feedback loop.

I'm assuming the PIC is largely there to allows i2c reading of everything (which it does rather nicely), plus fan speed control plus monitoring for various error conditions; input UVP/OVP, (probably) output UVP/OVP, over-temperature (it monitors two temp sensors, it looks like one is the input air temp and one is an internal heatsink).  It runs a timer interrupt pretty fast (I think in the ballpark of 2khz) but I don't think it's actually part of the control loop for the PSU other than handling error conditions (and the fan)

Any tips for taking this thing apart nondestructively? Desoldering that fat row of edge pins on the control board looks tedious.

[sundance]

@DrTune:
Thank you very much for your detailed research and work on that PSUs.
I recently got some of them (HP DPS-800GB) and I guess your findings also apply to this type.
What's quite interesting is that there are some/a lot of articles on the internet (like https://pa0fri.home.xs4all.nl/Diversen/DPS-800GB%20A%20Server%20sypply/DPS-800GB%20A%20PSU%20eng.htm) that claim to adjust the output voltage by attaching a simple pot between Pin 32 and GND.
But since Pin 32 is SDA of its I2C interface I wonder how that could be possible...

-sundance-

[james_s]

Server PSUs are great, I use them to power my RC airplane chargers. Many can even be modified so they can be run in series, for my big charger I have a setup like that delivering 24V at up to 1350W. Prices are ludicrously cheap, often little more than the cost of shipping. My favorite so far are some IBM 675W units that are compact and extremely quiet.

[tsman]

But since Pin 32 is SDA of its I2C interface I wonder how that could be possible...

That PSU doesn't have the same pinout. The PSON pin you need to pull low isn't at the same place on your one.

[sundance]

Any idea about the pin layout of the DPS-800GB?

[sundance]

For a start, it would help a lot just to know the PRESENT signalling pin along with the ENABLE pin to start the PSU.
The wide spread hack (connecting pins 31 and 34) to start the PSU does not work on (my) hp ATSN 7001044, which is a direct replacement for the DPS-800GB (I have 3 of those, and there the hack works...) which might give a hint that this is not the proper way to do it (just like with DPS-1200FB connecting pins 36 and 33 with a resistor, which also doesn't work on all PSUs)

So any input on that matter is highly appreciated...

.sundance.

[sundance]

These are my findings so far:

Code: [Select]

Pin 25: n/c                    Pin 40: -9.5V (when activated)
Pin 26: GND                    Pin 39: GND
Pin 27: n/c                    Pin 38: GND
Pin 28: ?                      Pin 37: +3.3V STB
Pin 29: +5V STB ?              Pin 36: +3.3V STB
Pin 30: ?                      Pin 35: +5V STB
Pin 31: #ENABLE                Pin 34: PRESENT
Pin 32: ?                      Pin 33: ?

Like with the DPS-1200, you have to pull-up the PRESENT pin with some 4k7 resistor to +5V STB.
Then to start the PSU, connect the #ENABLE pin to GND.

[overthrowrobotics]

I just made a rough CAD model of this.
https://grabcad.com/library/hp-dps-1200fb-1-1200w-server-power-supply-1

I'm designing a 2oz PCB with a 90 degree connector and probably just a large single connector for handling 4 gauge wire. I use PCBWAY and ordering extra PCBs doesn't cost much if anyone is interested.

[coromonadalix]

here's a thread :
https://www.rcgroups.com/forums/showthread.php?1005309-A-simple-high-quality-12Volt-100Amp-Power-Supply-Part1/page150

[Trurl]

As for how to use it, a jumper from pin 33 to pin 36 will power it on ...

On a different forum, E90 Post, in the thread "DIY - Server Power Supply for Coding":
https://www.e90post.com/forums/showthread.php?t=1838716

the OP of that thread claims that a 22k resistor between pins 36 & 37, and shorting(switching) between pins 33 and ground is (perhaps) a better switching method.


His reasoning, is that "pins (33 & 36) are wired to weak pull-up or pull-down circuits"... "pin 36 is wired to a pull-down resistor, which means it isn't true ground. The PSU isn't connected to a real server socket, so the board won't fight you in trying to pull up the connection, but if you have the option between true ground ... and a pull-down on 36... Might as well choose true ground."

Your insights regarding the above would be much appreciated.

Also, there's a very cool mod video of a slightly different but similar model DPS-800GB in which the gent lifts the OVP circuit's sense pin, then attaches various resistors (including a 2k variable resistor @ 09:50 into the video) between pin 32 and ground to achieve a similar result of raising the output voltage. I especially like the variable method, which is how I've done my own previous mods on various (ATX) PSU units.
https://youtu.be/9--g63ew9HU

Is this possible because pin 32 at least in this different model is associated with a 'voltage divider' connected to voltage regulation or is this related to an entirely different concept/function? Do you think this would work with the DPS-1200FB?

Cheers~

[Trurl]

A good reference on the startup resistors:
https://colintd.blogspot.com/2016/10/hacking-hp-common-slot-power-supplies.html

I completely overlooked it initially as "geek speak", but the author colintd's clarification of the likely assignments or functions of pins 33, 36, 37 etc. was very enlightening!

He states essentially that pin 36("Present") is likely intended for the PSU to sense being "connected to the server" by a 'voltage pull up' when inserted into a server's slot. This 'pull up' of its voltage can be mimicked by connection with pin 37's "Standby 12V" or pin 33's "Enable" voltage ("pulled up to the PSU internal 3V3 bus via a 10K resistor").

But what is really revealing is that he goes onto suggest that the voltages, resulting from the range of usable resistors, between pin33 and pin36 are "unconventional." Apparently the "splitting of 33 & 36, into a pulled up / clamped "present"(36) signal, and a simple pull to ground "enable#"(33) signal (which can also be driven over I2C) is much more consistent with other PSU" he has worked with. Further, "a fixed 22K resistor from "Present"(36) to 12VSB(37), and a separate switch between Gnd and Enable#(33), ... leaves the possibility of software control over the PMBus(Power Management Bus)." In other words, one can do more with "conventional"(as intended) connections(than the method of connecting pins 33 and 36 which 'does the trick' but limits other control features, that would otherwise be available).

Pin 34 is apparently a "Current Monitor" output (60mV/A) so it's got a current meter built-in and a "6:1 resistive(voltage) divider to ground would give 10mV/A, and you could then connect a cheap 0-1V voltmeter module, ideally with programmable decimal place, to directly display current in A" which means adding a bulky current shunt or even an inline Allegro current sensor IC is not even necessary! Dang!

This should apply to many similar model HP Common Slot Power Supplies so this is very useful info!