Repairing YM-3591AAR power supply (HP P/N 81-00000051)

[Zdomici]

We have had a HP P2000 G3 drive array system in my workplace since 2010. In 2014, one of its power supplies failed and we bought a new one for about 250 USD (the system wasn't covered by warranty anymore). But this replacement power supply failed in 2015, a little over year after its purchase. In the meantime, the P2000 systems became obsolete and the price of new PSU has jumped to hefty 500 USD. We have found cheaper ones on Ebay, but they were either used or the sellers didn't ship to my country. So my colleagues asked me to look and try to repair them. I was successfull and I figured I should share my findings online.

WARNING! The repair should be done only by skilled technician with appropriate education in electrical engineering! Improperly performed repair may result in electric shock (and even death) or destruction of your equipment! Also, disassembling the power supply automatically voids its warranty, possibly even warranty of your entire P2000 system.

All 3 power supplies (1 good and 2 bad) we have are model YM-3591A, option AR, revision A. Interestingly, all have SA060Axxxxxxxxxx serial numbers, despite they were bought at completely different times. Both faulty units reported these errors in the System Events interface:

Disk enclosure event: fault. (enclosure: 1, WWN: 500xxxxxxxxxxxxxx) power supply 1, power supply status: Over voltage, DC failure

The orange/amber LEDs on the units also lit up when the errors appeared. Initially, the errors randomly disappeared and reappeared, but after a weeks, they became permanent. I measured the output voltages with multimeter and oscilloscope, but all voltages and ripple were well within tolerances, even under heavy load. Thus I concluded that PSUs themselves work fine and only their over-voltage detectors are faulty. However, the detectors are very complex (several ICs surrounded by about 100 passive components), so I tried to find their schematic online. I didn't find one, but in desperate attempts to google it I came across a public FTP server of the PSU's manufacturer, 3Y Power Technology in China. Among other stuff, it contained an Excel table with fault analysis of several dozen PSUs returned via RMA. From it, I learned that the leading cause of failures in SA060Axxxxxxxxxx production run PSUs were only 3 components: resistors R441 and R453 and fuse F401. Also, their failures were almost always indicated by orange/amber LED, just like our 2 bad PSUs. These components were hidden under thick white silicone (?) putty, which I had to remove. They were rather hard to find, so I'm attaching photo with their location.

-Resistor R441 has marking 83B, under EIA-96 code it means it has 7.15 kohm.
-Resistor R453 has marking 79B, so it has 6.49 kohm.
-Fuse F401 has marking 0 and I think the manufacturer simply replaced it with 0-ohm resistor to save money.

In our case, both PSUs had faulty R441. When I measured them on boards with ohmmeter, they had roughly 10 kohm, which was immediately suspect - a resistor connected in parallel with some other components should always read same or lower resistance than its nominal value. But to be sure, I also measured all 3 components in the good PSU:

-Resistor R441 measured 4.2 kohm
-Resistor R453 measured 3.9 kohm
-Fuse F401 had under 1 ohm.

That confirmed our R441s were really bad. I didn't have 7.15k resistors in 0603 package on hand, so I replaced them with 8.2k and 56k in parallel. After that, the PSUs worked normally again, no errors, no orange/amber LED.  Personally, I think the faults were caused by that nasty silicone putty. The bad resistors I desoldered measured open, but they didn't look burned; most likely they cracked internally. The putty probably exerted some added mechanical force on them and they couldn't handle it during thermal cycling. That would explain why the PSU errors sometimes disappeared. Unfortunately, the putty covers many more SMD components, which means that such cracks could develop in parts other than listed in the Excel table.

And here are some other things I learned about the PSUs:

-The PSUs contain Atmel ATMega8L microcontroller which is used to communicate with the P2000 chassis. In one PSU that oscillated between good/faulty states particularly often, this firmware became corrupted. The P2000 chassis then listed the PSU as "unknown device". Fortunately, the firmware wasn't locked inside MCU's Flash memory, so I could "transplant" both Flash and EEPROM contents from the good PSU to the bad one. The PSU has 6-pin Atmel ISP header to connect the programmer, I'm attaching a photo with pin functions. But keep in mind that all our PSUs were from the same revision and even same production run, I have no idea what may happen if you "transplant" the firmware between different ones.

-You can find standalone 24C64 serial EEPROM connected to the microcontroller. Presumably, it is used to store PSU's serial number (so it won't get deleted in case of MCU firmware update).

-PSU output voltages are a bit unusual, there are +12V, +5.1V and +3.36V standby outputs. See the picture of the main connector.

-As became the norm lately, the +5.1V voltage is produced from +12V by DC/DC converter.

-When you connect the PSU to the mains voltage, only the 3.36V standby voltage appears at its output. The ATMega8L MCU is also powered by this voltage. To start the main voltages, you need to short the starting pin to ground (similar to ordinary PC PSUs). The starting pin is a bit shorter than others, so it is easy to find.

-The PSUs can be run with no load without any apparent problems; I run them idle for at least 5 minutes and nothing bad happened.

If you encounter other faults in these (or other HP) PSUs and manage to repair them, please let us know in this forum. I spent quite some time figuring this all out, so don't be lazy and share your findings, too!

[Serdg]

Came to repair the power supply YM-3591AAR.

Serial number SA060Axxxxxxxxxx. Replaced the faulty Resistor R411. For parallel resistors, the 8.2 kω and 56 kω. Now when the switch unit is in the power mode (standby) is lit green and yellow led. With the closure pin PSU start to the housing, the power supply starts, the yellow led goes out. All voltages are normal. Bus +12 volts loaded to a resistor of 15.7 Ohms 30 Wat. When open (pin PSU start) the power supply stalls. In what direction to look? May be required to close the control compensation of the supply of tires to the relevant motherboard pins? Where are they? Is there a diagram of the response of the connector?

[Zdomici]

@Serdg: Unfortunately, I can't tell you what is wrong with your PSU without looking at it. But the Excel table I've attached to my first post lists many other problems and their possible sources (faulty components). Try to check them. Also, the resistors in "my" PSUs cracked because of white putty that was covering them. You may try to remove it from the entire board and check all components underneath it.

As far as I can tell, there is no publicly available documentation of what the other pins on the main connector do.

[Serdg]

Thank You Zdomici !

One cooler was not working. Replacement with serviceable solved the problem.

YM-3591AAR

[Alexij]

Hello, Zdomici!

your investigation was very useful . In my case R441 also came to "opened cirquit" state and I had to replace it.

But my case is a bit more complicated. I had two power supplies: one with "amber LED" sympthom (#1), and the other one 

with completely destroyed PFC section, broken fuse e t.c (#2).  After replacing the power FETs of PFC I've got 390VDC

at capacitors but PS still failed to run. I supposed the problem was in mesonine sub-board with PFC and PWM

chips and replaced it with good one from the "amber LEDed donor" (#1). The problem remained. Later I've found that 

one of the sync rectifier FETs was dead, replaced it, and PS became alive. So I made a conclusion that the mesonine board from #2

was OK and soldered it to #1.

     When I saw your post I thought that now I'll repair #2, replaced R441 but all remained the same :(

Probably mesonine was not OK.  

    I'm writing this long story for 3 reasons

1. you asked to report other cases with this PS. Now you know about rectifier FETs :)

2. do you know what is the "NDF" reason in the chinese chart that you kindly provided?

3. do'nt you think it is rather strange that there are som many problems with simple resistors in low current circuits?

Maybe you have some ideas concerning it?  I have only one: both problematic resistors are located close to the header that

connects t-sensor to the board. It is connected with "thick" wire. The intensive airflow may cause some vibration of that wire

which is passed to the header -> board and destroys the resistors.    

[Zdomici]

Hi Alexij, I'm really glad my post has helped other people. I'm rather curious how long will your PFC and rectifier repairs last. Over the years, I've repaired several different SMPS for consumer electronics, but usually they didn't survive for long. For example, I successfully replaced semiconductors for the main voltages, but a few months later, something else failed (standby voltage supply or filtering capacitors, for example). Consumer stuff is simply designed for low cost, not for longevity. On the other hand, YM-3591AAR are server PSUs, so maybe they won't suffer from this.

Anyway, about your questions:

1. Thanks!

2. I believe NDF stands for No Defect Found. This may cover situations when PSUs showed faults in actual disk arrays (due to heat, vibrations, etc.) but the faults didn't appear when they tested the returned PSUs in 3Y factory.

3. Like I wrote in the original post, I think the resistor faults were caused by white putty on some parts of the PSU. Or maybe the resistors were from a bad production batch. That would explain why 2 resistors caused so large percentage of faults. But yes, your vibrating cable hypothesis may be true, too.

[Djovani1]

Hello!
A faulty resistor P278 is detected (thanks to the table).
I suspect that this is not the only problem, because The resistors themselves do not break.

[Djovani1]

Has been restored after replacing one resistor!

[Serdg]

Hi! I have a question. Your material repair unique! I want to raise the voltage of 12 volts . Make it to 13.8 volts. The fans are spinning at maximum rpm. I want to lower the revs. Many server power supplies allow you to do this. It is necessary to manipulate with the external pins (pins) But are the other manufacturers. On the forums write a lot. I use the power supply for their needs. You have the appointment of the external pins? All the best!

[Zdomici]

Like I wrote earlier, I was unable to find schematic or pinout anywhere. If you find it, please post it here.

If you want to increase the 12V output voltage, you can also do it the hard way: You can analyze the 12V converter and identifiy its main switching control IC. Then you can find its datasheet and see how its feedback loop is connected. It should be possible to increase the voltage by changing components (usually resistors) in the feedback loop. However, I doubt it will be that easy - the PSU has extensive over-voltage protection circuits which will intervene and stop the PSU if you raise the output voltage too much. Also, there is a small 12V->5V converter and it may not like the increased voltage, either.