Eaton 5115 rackmount UPS repair help?

[paulbt]

Hello to everyone!
It is my very first topic opened here in the repair section.
At my workplace the UPS from our rack stopped working a few days ago. If the power button is pressed or hold, no relay clicks inside and there is no output. It is not my job to repair it, but it is part of my hobby, so I took it home and tried to troubleshoot it. So I will describe what I did and what I've found so far without having a schematic or even a block diagram at hand. I have no experience with any UPS, but I tried to use my common sense and my humble power supply experience I have so far to understand the modules and how they connect between each other.
- first I visually checked all electrolytics, they all seem fine, I did not tried to unsolder the biggest ones to check them with a LCR meter, but maybe I will do it in the next days;
- the batteries seem to be ok at around 38V from 6x6V lead-acid;
- then I plugged it in the wall socket and check the 230V which goes to some relay contacts which either connects it directly to the output or to the HV oscillator powered by the batteries
- checked with an external 12V supply all the relays - they all click and their contact state (short or open) seems to change as it should;
- checked all the THT diodes, resistors, MOSFETs, coils, common mode filters, MOVs, X and Y caps - nothing suspicious 
- checked all the SMD bipolar transistors and diodes - nothing as well 
- checked if the power button works and it does very well. I followed back the two traces that it shorts together - they lead to the other side of the board to some diodes and smd bipolars
- visually checked for broken traces and then used the continuity tester on a few of them that seemed burned - nothing again   
- then with an external 38V bench PSU connected instead of the batteries I decided to check the supply voltage of the THT ICs (opamps, smps controllers) on the board. No voltage on any of them. No current consumption from my PSU. I followed back the VDD trace and they all lead to either the input or the output of a TO220 type LM7805. Indeed it's input and output read 0V both.
- then I did reverse engineering to the circuit nearby the 5V regulator (please see drawn picture below) and I've found out that there is no voltage at the input of 7805 because the switching MOSFET is not working because the IC that controls it has no supply because it's supply is connected to the input of the 5V regulator... 

So my questions are:
- why the switching IC is supplied with voltage from the same circuit? It is like a chicken & egg loop, it needs some external initial impulse in order to start working.
- how can I find the rootcause of these missing voltages? the MOSFET, the HF transformer and all the other discrete components from the drawing are all fine. I'm not sure about the switching IC (UC3843AN).
- should I look somewhere else on the board?

PS: Related to my first question, I tried to hook another external 15V supply to the input of the 7805, 38V supply connected as batteries, transformer connected, no 230V input connected and the UPS somehow was alive, but no 230V output, the fan was working, the buzzer was beeping some error once every few seconds, some relays were clicking repeatedly and the current consumption from the 38V supply was around 100-200mA (not CC limited).

 

[Runco990]

This may be not at all a problem on your UPS, but I'll tell you MY Eaton story.

Thing was dead and pretty much like yours.  No stand by 5V, batteries were good.  I broke my head on this thing for weeks.  Then I put it in a corner, as there are no schematics.

Eventually I picked it up again and tried to figure out where the stand by was being created.

Long story short.... the fault was a broken HAIR FINE trace on the board in the feedback loop for the 5 volt supply.  This break was caused by copper corrosion on what later turned out to be TWO traces that were incredibly thin and right in the path of the intake fan on my unit.

Over time, they invisibly corroded and opened.  I found them only by ohming out traces.  The eventual fix was to run wire wrap jumpers.

The UPS has been running fine since.

I'd suggest to at least consider that possibility as I have seen this a few times. 

[fzabkar]

https://www.onsemi.com/pub/Collateral/UC3842A-D.PDF

Typically with ICs such as the UC3842AN, there is a high value resistor between Vcc and the rectified mains. This provides the start-up power for the IC. The IC then begins oscillating and sends pulses to the MOSFET. When the supply starts up and begins to regulate, it provides a regenerated Vcc from a secondary winding of the transformer. This is required because the startup resistor cannot provide enough current to keep the Vcc capacitor charged when the IC is in run mode. If this secondary supply doesn't come up, such as is the case when the output is overloaded, then the Vcc supply decays and the IC switches off. The Vcc capacitor then recharges via the resistor and the cycle repeats. Some intelligent ICs might only "hiccup" a few times before giving up and shutting down until the next power cycle.

Be aware that these startup resistors (approx. 100K) often go open for no apparent reason.

[paulbt]

https://www.onsemi.com/pub/Collateral/UC3842A-D.PDF

Typically with ICs such as the UC3842AN, there is a high value resistor between Vcc and the rectified mains. This provides the start-up power for the IC. The IC then begins oscillating and sends pulses to the MOSFET. When the supply starts up and begins to regulate, it provides a regenerated Vcc from a secondary winding of the transformer. This is required because the startup resistor cannot provide enough current to keep the Vcc capacitor charged when the IC is in run mode. If this secondary supply doesn't come up, such as is the case when the output is overloaded, then the Vcc supply decays and the IC switches off. The Vcc capacitor then recharges via the resistor and the cycle repeats. Some intelligent ICs might only "hiccup" a few times before giving up and shutting down until the next power cycle.

Be aware that these startup resistors (approx. 100K) often go open for no apparent reason.

Very nice explanation! I downloaded the IC's datasheet only to read what it mainly does and to see the pinout. I didn't read more about it and I didn't looked for a typical application circuit. I may have figured it out before posting here. When I will get home I will do some more reverse engineering and look for that pullup/startup resistor. I will also check for more thin traces that may be broken. 

Thank you both for your suggestions ! 

L.E.: I was thinking to send an email to Eaton customer support to ask for the schematic, what are the odds that they will actually provide me the document?
 

[SoundTech-LG]

The Eaton Legal Department is terrified. Not going to allow release of documentation.

[fzabkar]

Warning:

If the PSU doesn't start, there may be rectified mains voltage (340VDC or 170VDC) sitting on the bulk capacitor. This will give you a nasty zap if you don't discharge it. I use a 100W incandescent bulb for this purpose.

[paulbt]

Warning:

If the PSU doesn't start, there may be rectified mains voltage (340VDC or 170VDC) sitting on the bulk capacitor. This will give you a nasty zap if you don't discharge it. I use a 100W incandescent bulb for this purpose.

Good morning!

Yes, I have some "experience" from the past with HV electrolytics, big X2 caps, heatsinks from cheap computer PSU or even touching mains 230V with the back of my hand while probing around).

Yesterday I did some more reverse engineering and I found that the VCC pin of UC3843 is not connected directly to mains via a resistor but connected to some power smd resistors - to collector of a pnp low power bjt - the emitter is connected directly to +BATT - the base is driven to pulldown resistor via the phototransistor from an optocoupler - the diode of the optocoupler is connected in series with a zener to a bridge rectifier - the AC inputs of the bridge are connected to the EMI/RFI filtered 230V input via some limiting power smd resistors on both sides. I have checked each component in this path and they were all ok. Then I have connected the AC 230V to the board and I saw about 10V at VCC pin, but VREF pin was dead (0V). I took the IC out and supplied it again with an external 10V from my PSU and it was the same problem and 3mA consumption. I tried with the other identical IC from the board and the consumption was about 10mA and there was a steady 5V at VREF pin.

Today I will order replacement parts and I hope that the UPS will work after the repair.

I will keep you updated, thank you all for the discussions!

[donwulff]

Resurrecting this thread because the final post says "I will keep you updated" 

I find myself now in similar situation, with a EATON PowerWare PW5115 1500i. Though tower model, not rackmount, that finally gave up the ghost last week and was decomissioned. Both the charger & inverter modes were still working, just in the Automatic Voltage Regulation (ie. line-powered) mode it would start continuous beep with all lights on and the lowest battery light blinking. Batteries were fine though. Nothing on the serial port. I figured I'd try to fix it as soldering practice. Verdict: My soldering skills suck. (Learned new thing about my hobbyist JCD Soldering Station: Screw the plastic part on FIRST, before attaching the soldering tip & connecting the metal nut. The plastic part came loose and the threads kept slipping off...)

First note that, the user manual (or in fact any user manual I could find) doesn't cover that specific fault indicator. Continuous beep & all lights on is internal fault, but nothing about lowest light blinking? Maybe it's an error in the user manuals.

Second, this is the top search match for PowerWare PW5115 schematics, but they're nowhere to be found on Google. However, for anybody who came here looking for those, AliSaler has complete set of schematics for PW5125. I find these are close enough to the PW5115 to be incredibly useful, though of course the component codes are entirely different, so you'll have to match to the circuit. https://www.alisaler.com/ups-powerware-pw5125-1500va-schematic-diagram/ for mine, but look at the "related articles" for rackmount RM's of different sizes.

Initial assessment of the failed UPS was that almost all of the electrolytics were reading in 2-30k ESR range, which is way out of spec. This isn't too surprising, as with three batteries in series the charging voltage is over 40 volts, and all the electrolytics are Jamicon/Teapo 1000h rated for 50V. A common rule of thumb seems to be to size them at twice the voltage for prolonged use. It's possible to get 63V condensators with the same footprint, so that's what I did for the 1uF, 10uF, 22uF, 47uF. There's 470uF at 35V which I'm upgrading to 50V. There's a 50V 100uF, 470uF, 1000uF and 3300uF (which you should discharge before touching) which aren't so ridiculously out of spec so letting them be for now. After removing the smaller caps from the board, many of them are reading 80 ohms or even open circuit, so I guess it's really been hanging on by a thread and at this point you might choose to just replace all electrolytics.

Electrolyte on my board labeled as C823 (47uF 50V) right next to the 5v regulator 7805 (probably heats a whole lot) was completely open circuit, though, which I presume was the cause of the problem. I had really hard time tracing the circuit on the board. In the above linked schematic the same component is C27 (Negative side connected to 7805 Pin 1, 2K resistor across to two rectifier diodes from transformer pins 1 and 3). It filters the 24v power-line which seems to only power the Automatic Voltage Regulation relays, hence why it would fail when trying to switch on AVR power.

Replacement BOM for this specific model:
2x 50YXS470MEFC10X20 CAP ALUM 470uF 20% 50V Radial Rubycon (Third one for 470uF 50V on board?)
10x 63ZLJ47M6.3X11 CAP ALUM 47uF 20% 63V Radial 6.3x11 Rubycon (Incl. one on control board)
3X 63YXJ22M5X11 CAP ALUMN 22uF 20% 63V Radial 5X11 Rubycon (Incl. one on control board)
8x 63YXJ10M5X11 CAP ALUM 10uF 20% 63V Radial 5x11 Rubycon (Incl. five on control board)
2X 860020772005 CAP ALUM 1uF 20% 63V Radial Wurth Electronics (Incl. two on control board)

Tried to pick highest use-time caps available for the footprint, didn't match frequency or other parameters beyond capacitance and one voltage class higher.

50V 100uF on control board, 470uF, 1000uF and 3330uF not replaced (yet). Working fine and as others have noted less audibly noisy now. Likely many of the relays were constantly switching on and off with the unfiltered DC, shortening their lifetime as well. Possible way to determine when an UPS is about to fail.

Additional thought: Has anybody reverse-engineered the firmware of these UPS:es? I doubt there's much one might want to (safely) tweak with these, but even possible official firmware updates have been taken off the net. At least having a nice serial protocol (newer versions come with UPS) would be bonus, otherwise it does what an UPS should. Right now I can't even tell if it's supposed to give anything on serial port or not...