I am building a desktop power supply using free parts from an IBM netfinity server from the late 90's. The power supply is built by Magnetek and has a case # of 3722-40-1 and a IBM part number 01K9880 made week 16 of 1999 and rated for 400 watts by Magnetek.
I did my due diligence in regards to searching for information and documentation without any luck. All I managed to find was refurbished and used units for sale. I am assuming this is because IBM is notorious about not sharing information.
Without any schematic available, I decided to try and reverse engineer the power supply. This is where I will tell you, I am no engineer. Most of my experience is that of an electrician and avid do it yourself type. While tracing parts of the board I decided to apply power to the board and take some measurements with my Rigol 1054Z. In the process of doing this I managed to turn the damn thing on!
I was probing the main processor on the power supply a 40 pin XS912AB (no info available) (bottom line number) 80387800400. The second number listed on the bottom of the chip is the same on multiple smps from the same server, the top number is different on the second unit I have and is XS824AB.

When the supply turned on I was probing a ceramic resonator labeled 8.00T next to the processor. As it turns out I probed the output pin of this 3 pin resonator. I was surprised that it turned on. I measured voltage on all of the rails and had 3.3, 5.47 and 12.44 volts and the 12 volt rail dropped to 12.11 volts under a load. I even attached 3 different motors to the outputs to test under a load. Needless to say, they all worked under a load. I also went on to try and figure out what the oscilloscope did to turn the supply on and cannot understand how or why it did. I did further testing and could not get it to start by using a DMM and check the voltages of the resonator. BTW the voltage was something like 2.6 volts on pins 1 and 3 of the ceramic resonator. With the DMM not turning it on, i thought about paralleling a capacitor to pin 3. this works kind of. I attached the positive pin of a small electrolytic cap rated at 16 volts and attached the negative pin to ground on the output of the supply (which is what I used for all voltage measurements against a ground), and nothing happened. Puzzled I tried to attach just the positive pin of the paralleled cap to the pin 3 and then placed myself in the circuit attached to the ground pin (the ground pin was only touched by myself and not the supply ground). By placing myself in the circuit, it turned the power supply on. Once the supply starts, I remove the capacitor from parallel and it seems to run perfectly fine until I remove the AC power from the rectifying board of the supply.

Curious about the results I took it one more step. I used a piece of 16 awg insulated wire and touched it to the pin 3 and nothing happened. I tried it this way because I dont know if the extra inductance in this ceramic resistor is what is causing this to happen or not. So i then stripped the unattached end of the wire and touched it while it was in circuit with pin 3. It works. Now sometimes it starts immediately or "kind of burps to life" if I remove the wire too soon and then goes back to standby mode until I touch it again. But it always starts without fail. There are no changes or modifications done to the power supply other than they are not in the metal case, but instead on my workbench side by side and even have the fan plugged in and laying on top of the heat sinks. There are no variables involved that I can think of other than the supply is on my workbench.

I mention these came from a server. I have already taken two apart just for something to do because I didnt think I would be able to get them to work. I cannot remember I I have any other complete ones laying around I have not taken apart yet. I will look and if I do, will repeat this on it and see what happens. I will also add the server was removed completely working and intact minus the hard drive.

Can anyone explain to me in a somewhat simple way how or why this is working like this? And what I should be concerned with in regards of putting this back in its case with a wire soldered to pin 3 and ran to the outside of the case for me to activate it with?









Moderators note : removed double images

 

Really, no thoughts????

 

Sorry, but your rambling discourse is way too wordy for me to read and comprehend.
If you could condense your observations and questions to a few simple sentences on separate lines, I'll give it a shot.

 

Could be your body capacitance allows a brief current path to ground to kick-start the supply. Can't recommend relying on that as a 'permanent fix', bearing in mind the LETHAL voltages that prevail in parts of the power supply.

 

I have seen crystals that were slow to start. In fact, I had one on a digital signal synthesis device that required me to touch it to get it started after a long period of inactivity. Not sure that applies to ceramic resonators. The MCU's I use require the oscillator to be running and stable before they start.

As for the other symptoms you describe, is the supply a DC to DC converter or does it run from line voltage? It is possible that a "power good" signal is needed for it to start. In other words, there may be two problems: 1) Slow oscillator start; and 2) A power sequencing issue. PC supplies also require a load before starting.

More information about how you have it hooked up, how many wire go the the supply, and so forth might help.

 

Change the resonator, or change the chip...

 

I have seen crystals that were slow to start. In fact, I had one on a digital signal synthesis device that required me to touch it to get it started after a long period of inactivity. Not sure that applies to ceramic resonators. The MCU's I use require the oscillator to be running and stable before they start.

As for the other symptoms you describe, is the supply a DC to DC converter or does it run from line voltage? It is possible that a "power good" signal is needed for it to start. In other words, there may be two problems: 1) Slow oscillator start; and 2) A power sequencing issue. PC supplies also require a load before starting.

More information about how you have it hooked up, how many wire go the the supply, and so forth might help.

I have it hooked up just as it would be in the metal case it comes in. The only exception in the 120 VAC feeding the mains rectifying board does not have a ground wire attached, just the neutral and line are alligator clipped to the pins. Like i said, it is a IBM server hot swap smps. So there is no information about the supply I can find, maybe my google-fu needs improvement.

 

Change the resonator, or change the chip...

I have another chip that came from an identical sister supply, I didnt think about changing it to check if the phenomenon would still work. The resonator didnt survive the removal process from the sister supply. I really am not wanting to "fix" the supply ( i dont think it is broken), I just wanted to figure out how to use it as a bench power supply.

 

Most Atx psus have a Green start up wire, then you can use the 3,5,12V outputs for a bench psu...if you have the smps chip datasheet you can see how it functions,then you can modify it.

 

Sorry, but your rambling discourse is way too wordy for me to read and comprehend.
If you could condense your observations and questions to a few simple sentences on separate lines, I'll give it a shot.

I will do my best, I merely provided all of the information I could because thats what is usually asked for and sometimes needed.

IBM Netfinity server smps removed from retired working server to be re-purposed as a benchtop power supply.
SMPS built by Magnetek for IBM Magnetek code on supply is 3722-40-1 IBM part number is 01K9880 400 watt smps hot swapable (I think)
Intention is to "hack" the unit for reliable DC power.
The supply is out of the steel case on my bench being and being fed 120v AC to the line and neutral pins. There is not a ground attached to the ground pin.
On the large board with the IC and ceramic resonator in question I am able to get the unit to turn on by putting myself in circuit with the ceramic resonator pin that is connected to pin 7.
"long" story short, it reliably starts the power supply every time. Did I find a way to easily hack a IBM supply? Do you see any problems with reliability? I want to solder a wire to this pin and run it out of the case to turn it on when needed just by touching it. I know this sounds terribly unsafe, but I am confident that the small capacitance and voltages of this resonator is harmless. I do not intend to leave it running when I am not present, I have a small commercially made unit I use for charging batteries.
I hope this is short enough and to the point enough, and have a feeling if this is what I posted initially I would have been chastised for being vague.

 

Most Atx psus have a Green start up wire, then you can use the 3,5,12V outputs for a bench psu...if you have the smps chip datasheet you can see how it functions,then you can modify it.

Right, I have one of those and is exactly like you say to get working. This unit came from a IBM Netfinity Server and is hot sawpable. So there is no external wiring harness, and the only wires inside the case are for attaching the 2 boards and cooling fan together. Its very interesting to try to hack a piece of commercial equipment vs. what I would call hobbyist or I guess mass market. It only has 3 DC rails 3.3, 5, and 12 volt outputs. It doesnt have any of those oddball negative DC outputs like the small PC power supplys do.
Like everything IBM it seems to be proprietary and no information seems to be available. Even looking up the IC I get nowhere. It seems like its a ghost chip but must be just house numbers and leads to nothing I can find.

 

I hope this is short enough and to the point enough, and have a feeling if this is what I posted initially I would have been chastised for being vague.

Better for me.
Sorry if I was a bit abrupt, but when I see several very long paragraphs of discussion I sort of blank out.

So it sounds like a resonator problem.
If connecting a small capacitance from the resonator pin 7 to ground or some other point (such as the V+) doesn't work, then I see no problem with bringing a wire out to touch and get it started as a work-around.
Adding a resistor in series with the wire would add a measure of safety. Use the highest value that still allows it to start.

 

Better for me.
Sorry if I was a bit abrupt, but when I see several very long paragraphs of discussion I sort of blank out.

So it sounds like a resonator problem.
If connecting a small capacitance from the resonator pin 7 to ground or some other point (such as the V+) doesn't work, then I see no problem with bringing a wire out to touch and get it started as a work-around.
Adding a resistor in series with the wire would add a measure of safety. Use the highest value that still allows it to start.

Thank you!