Announcement

Collapse
No announcement yet.

ATX power supply keeps blowing PFC mosfets — Cooler Master RS-550-ACAA-E3

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

    ATX power supply keeps blowing PFC mosfets — Cooler Master RS-550-ACAA-E3

    This is an ATX PSU rated for 550W powering 24/7 a computer load of about 200W, fed by a line-interactive 1500VA UPS. Mains is 230V.
    1. First issue: the PSU blew a mosfet on what I think is the active PFC boost converter, shorting its G D S pins together causing a short across the + and - lines produced by the rectifier on the 230V mains. The surge protector of the house tripped. When this happened I found that the UPS was unexpectedly disconnected from the wires of its battery. The UPS wasn't damaged, but I suspect this spontaneous detachment of a battery wire caused a condition that blew the power supply, because the UPS doesn't provide output voltage without its battery connected. The blown mosfet is a 16N50C3 in TO-220, replaced with an IRFP460 in a bulkier TO-247 which I had ready. The characteristics are mostly similar except particularly for the slower slew rate and higher gate capacitance of the IRFP460. The PFC PWM seems to be generated by a CM6800 which according to datasheet has a max frequency of 250ĸHz. The IRFP460 can switch up to about 3MHz.

      After this replacement the short between + and - was gone.
      Pictures shows two IRFP460 screwed to the input-side heat sink that replace two blown 16N50C3.

    2. Second issue: after I replaced the blown mosfet, I plugged the PSU to mains and tested the standby voltages on the bench. I had 5V VSB and the 5V power-on voltage from the green wire. All stable so far. I shorted green to black to turn on the PSU with a load of 3 x 12V fans (almost 1A total) plus the PSU's own fan. No load was on the 5V and 3.3V lines. For about a second it seemed to work fine and the fans powered on, then the other 16N50C3 on the same PFC circuit blew up, shorting + with - as before. The replacement IRFP460 physically next to it was undamaged.

    3. Current situation: I cleaned the PCB as I suspected that metal dust / fragments from cold solder joints may have caused a short in a control pin around the CM6800, then replaced the second 16N50C3 with another IRFP460 and tested again. I have a wall switch to quickly toggle mains power to the PSU on/off, to quickly cut power after repeated short tests of about 1 second each, so to prevent another blow up. Green remains shorted to black with a jumper wire. This time the dummy load is an old ~20 watt hard disk (1.2A on 12V + 0.75A on 5V). With the wall switch on, I measure ~12V between yellow and black and ~5V between red and black. The measures are approximate as the multimeter doesn't have time to stabilize before I turn the PSU off. The hard drive doesn't begin to spin up even after a second. The heat sink with the two IRFP460 becomes warm to the touch just after a second, and the 400V bulk capacitor of the boost converter remains discharged to 0V. It has a 1MΩ resistor in parallel so it would need several minutes to discharge.

      With only a multimeter I tested all discrete components on the PCB for continuity. There are no shorts across components and all diodes' forward voltages are within expectation. BJTs tested fine for continuity, resistors are all within specs, the 3 optocouplers at the bottom center of annotated_2.jpg tested ~1.2V for the diode side, the transistor side isn't shorted. The integrated rectifier (visible in annotated.jpg, left side of heatsink) is fine. No capacitor is shorted, and I can observe the electrolyte ones charging up slowly under the multimeter's own voltage. The transformers don't short primary against secondary. The white glue has not become conductive due to deterioration.

      I suspect what is broken is the PWM driver IC (CM6800) mounted on the vertical board close to the bulk capacitor. I tested its pins for shorts and there's none. One of its outputs connects to the base of both BJTs in front of the burnt/replaced mosfets, with one of these emitters connected to both mosfet gates thru a zener (unknown voltage) in parallel with a 22Ω resistor. The pair zener+resistor is replicated twice, once for each mosfet gate. (see annotated.jpg). It seems both mosfet gates are held open by the IC, conducting + straight to - via the boost converter coil (see annotated_2.jpg), preventing the capacitor from charging.

      There's a second PWM IC, TNY278PN, which I'm unsure of its relation to the faulty circuit.


    I found both replacement ICs (CM6800 and TNY278PN) on ebay for ~5€ each, I may order them based on your recommendations.

    ----

    Images description:

    20230124_185848.jpg and 20230124_185924.jpg show front / rear of the disassembled heat sink on the output side of the PCB.
    Attached Files
    Last edited by davidebaldini; 01-25-2023, 03:38 PM.

    #2
    Re: ATX power supply keeps blowing PFC mosfets — Cooler Master RS-550-ACAA-E3

    Mistake one: if Mosfets are paired, replace both at the same time with the same part numbers. So don’t mix match them. They blow, because one was switching a little faster than the other. Same thing happens with diodes in parallel.

    The TNY27x is your standby power. So this would be always on, whereas the PFC only runs when the PSU is turned on.

    Comment


      #3
      Re: ATX power supply keeps blowing PFC mosfets — Cooler Master RS-550-ACAA-E3

      Attached annotated PCB layout.

      The IRFP460 G D S pins are marked. Green, yellow, purple and black circles indicate jumpers. The two PNP transistors in front of the replaced IRFP460 are marked, with two pairs of zener+resistor connecting both gates, one pair per gate. Both BJT bases are connected to the CM6800 board via the thru-hole resistor (on the PCB components layer, not visible here) that crosses over the SMD R9 on the tracks layer.
      Attached Files
      Last edited by davidebaldini; 01-26-2023, 09:06 AM.

      Comment


        #4
        Re: ATX power supply keeps blowing PFC mosfets — Cooler Master RS-550-ACAA-E3

        Originally posted by CapLeaker View Post
        Mistake one: if Mosfets are paired, replace both at the same time with the same part numbers. So don't mix match them. They blow, because one was switching a little faster than the other. Same thing happens with diodes in parallel.

        The TNY27x is your standby power. So this would be always on, whereas the PFC only runs when the PSU is turned on.
        Noted. The overheating I got after replacing only one 16N50C3 seemed exaggerate to be explainable only by a mismatch of response time between the two mosfets, as the remaining 16N50C3 exploded after less than 1 second of uptime. After I replaced this one too with another IRFP460, I'm still getting a warm heat sink after just one second.

        Comment


          #5
          Re: ATX power supply keeps blowing PFC mosfets — Cooler Master RS-550-ACAA-E3

          Use a 100W incandescent light bulb in series with the unit for testing:
          If it is on full bright the PSU is shorted but it wont blow spectacularly since it is only passing 100w to the bulb...

          Second: are you sure the main bulk filtering capacitor is ok?
          If it is very low on capacitance the PFC stage will explode...
          "The one who says it cannot be done should never interrupt the one who is doing it."

          Comment


            #6
            Re: ATX power supply keeps blowing PFC mosfets — Cooler Master RS-550-ACAA-E3

            Originally posted by Per Hansson View Post
            Use a 100W incandescent light bulb in series with the unit for testing:
            If it is on full bright the PSU is shorted but it wont blow spectacularly since it is only passing 100w to the bulb...

            Second: are you sure the main bulk filtering capacitor is ok?
            If it is very low on capacitance the PFC stage will explode...
            I wanted to use an incandescent light bulb in series with the PSU, but this is Italy, the incandescent ones have been illegal for 15 years, importing is troublesome and takes months for customs clearance. If I build a proper test bench I'll look into automotive halogen bulbs, which are still sold. But so far, during the short 1 second tests, the surge protection breaker in the house hasn't tripped with the new pair of IRFP460, which are larger and more robust than the original mosfets, so currently the PSU is only causing an overload but not a complete short between mains. I think this overload is what's overheating the mosfets, and when the old ones blew, probably because of overheating, they shorted across the mains, tripping the house surge protection.

            I didn't notice anything wrong with the bulk capacitor for what a continuity test can show on a multimeter, except that it remained at 0V after every test, whereas it should have charged up.

            Comment


              #7
              Re: ATX power supply keeps blowing PFC mosfets — Cooler Master RS-550-ACAA-E3

              you can get incan lamps usually with decorative filaments that look ancient.
              not 100w though, usually 40 or 60 at most.
              so 2 in parallel would work

              Comment


                #8
                Re: ATX power supply keeps blowing PFC mosfets — Cooler Master RS-550-ACAA-E3

                Originally posted by davidebaldini View Post
                I didn't notice anything wrong with the bulk capacitor for what a continuity test can show on a multimeter, except that it remained at 0V after every test, whereas it should have charged up.
                If you leave the PSU in standby the cap should charge to 325VDC (PFC is off in standby) how long do you still have 5VSB present on the ATX connector if you power the PSU off?
                "The one who says it cannot be done should never interrupt the one who is doing it."

                Comment


                  #9
                  Re: ATX power supply keeps blowing PFC mosfets — Cooler Master RS-550-ACAA-E3

                  Originally posted by Per Hansson View Post
                  If you leave the PSU in standby the cap should charge to 325VDC (PFC is off in standby) how long do you still have 5VSB present on the ATX connector if you power the PSU off?
                  I ran the test you suggested. With the PSU on bench w/o any load attached, with power cord attached to mains, in standby, the 5VSB measures 5.10V and the bulk capacitor reaches 310V, equivalent to 220V rms, which is expected for the Italian grid. The mosfets don't overheat while in standby.

                  As soon as I switch the mains off, the 5VSB begins to drop. It reaches 3V one second after the mains is off, and continues to drop. The bulk capacitor is at 0V 10 seconds after the mains is off. I couldn't manage to measure it sooner. I suppose this exceedingly fast discharge is not the correct behavior, because the 1MΩ resistor in parallel to the capacitor seems to be specifically assigned to the task of discharging it once the PSU is powered off, and it would take minutes to discharge it.

                  Comment


                    #10
                    Re: ATX power supply keeps blowing PFC mosfets — Cooler Master RS-550-ACAA-E3

                    Originally posted by stj View Post
                    you can get incan lamps usually with decorative filaments that look ancient.
                    not 100w though, usually 40 or 60 at most.
                    so 2 in parallel would work
                    +1
                    Actually, go to an appliance parts store (if you have one such in Italy... I know they do exist other parts of South-Eastern Europe) and ask for appliance light bulbs. Typically, those are 40W incandescent. Most often found in ovens (since LED and CFL stand no chance to work at such high temperatures.)

                    Originally posted by davidebaldini View Post
                    I wanted to use an incandescent light bulb in series with the PSU, but this is Italy, the incandescent ones have been illegal for 15 years, importing is troublesome and takes months for customs clearance.
                    Any appliance with a heating element in the range of 100 to 300 Watts will do the trick too... or two 500-1000 Watt in series (thus reducing the voltage at short-circuit across each to 1/2 and as such, reducing the power rating to 1/4... so 2x 1000W elements in series will act the same as one 250W element.)

                    Originally posted by davidebaldini View Post
                    I didn't notice anything wrong with the bulk capacitor for what a continuity test can show on a multimeter, except that it remained at 0V after every test, whereas it should have charged up.
                    Unfortunately, for those big caps, only way to test them is with a capacitance + ESR meter.

                    Originally posted by Per Hansson View Post
                    how long do you still have 5VSB present on the ATX connector if you power the PSU off?
                    You know, I find this test doesn't always work too well on new APFC PSUs, especially with CM6800/CM6802. It seems that many of these are designed to cut off all power just 100 ms (or whatever the hold-up time is set to with the circuit design), including the 5VSB. Both the ThermalTake Smart M Series M850W and Corsair CX750M (both CWT PUQ platforms) shut off the 5VSB and main power almost immediately after being disconnected from the AC, even after installing a much bigger cap in the TT M850W PSU. It appears these PSUs have an AC detection circuit. Once AC is cut off, 5VSB is disabled within less than half a second, turning off the APFC and main PWM supply too. Main capacitor remains charged quite high for a very long time, though. Got a nice 370V DC shock across my chest the first time when I didn't expect it (was holding the primary heatsink with one side and accidentally touch another part that was on the +B). Gave me a good jolt... and a bit painful too. LOL.

                    Comment


                      #11
                      Re: ATX power supply keeps blowing PFC mosfets — Cooler Master RS-550-ACAA-E3

                      I found the bulk capacitor is defective. I desoldered it from the PCB to test it independently from the circuit. I charged it to 10V with a dedicated bench PSU thru a 210Ω resistor; for as long as the bench PSU leads were clamped to the cap, the multimeter read 10V across the cap. As soon as I unplugged one clamp of the bench PSU from the cap, without first switching the bench PSU off, the voltage dropped to 0V.

                      I did the same test with two other used capacitors I had as spares from an old recap job, (see picture) and they held the charge for minutes after the clamps were removed, as it should be. I tested this broken cap for DC resistivity with the multimeter, and even when discharged to 0V it's over 2MΩ, which is the measurable upper limit for this multimeter.

                      I suppose what might have bursted the cap was an excessive switching frequency on the PFC, or perhaps the UPS rammed it when its battery spontaneously unplugged; no idea. This PSU was working fine until a few days ago, it even had excellent hold-up time against the UPS, which needs 3ms to activate the inverter when power goes out.

                      I may cram these two spare caps into the PSU chassis and wire them in series, to replace the original cap; they are 200V 105°C 680uF Teapo, so I'd have an equivalent 340uF cap. But perhaps there might still be something else broken on the PCB?
                      Attached Files
                      Last edited by davidebaldini; 01-26-2023, 10:59 PM.

                      Comment


                        #12
                        Re: ATX power supply keeps blowing PFC mosfets — Cooler Master RS-550-ACAA-E3

                        If you wire the Teapo caps in series, make sure to add their respective divider/balancer/bleeder resistors as well. Typically these are 270K to 500K. Lower resistance = better balance of voltage between the two caps, but also slightly more heat dissipation (so make sure the resistors have appropriately high power ratings.)

                        There shouldn't be anything else wrong with the board if you found the PSU appeared to work for the short 1 test intervals... but again, wire up a series current/power limiting device. If incandescent bulbs are not available, use halogen ones. The "long-tube" high-power ones (200-500 Watts) I don't think are outlawed, because they don't have anything equivalent in LED versions. So go to your hardware store and check. If not, then use an appliance with a 300-500 Watt heating element (and no "smart" electronics - i.e. the heating element can be controlled by a physical switch and not a relay or MCU.) "Mini" toaster ovens, hair dryers (on low setting), and sandwich makers are good for such use.

                        Finally, consider adding a 0.1 to 1 uF PolyPropylene (metal film) capacitors in parallel with the "main" cap in the PSU. The PP cap can take on some of the high-frequency ripple current from the APFC and possibly give the electrolytic caps an easier time. It should also provide as a "backup" path for the APFC inductor to dump its energy, in case one of the electrolytic caps opens. A 400-450V PP cap might work, but I suggest using one rated for 630V, just in case.
                        Last edited by momaka; 01-27-2023, 12:01 AM.

                        Comment


                          #13
                          Re: ATX power supply keeps blowing PFC mosfets — Cooler Master RS-550-ACAA-E3

                          That's a good test davidebaldini!
                          5VSB should not be dropping linearly as mains is disconnected, so that confirmed that cap is bad.
                          Moreso that it read 0v after just 10 seconds, well, it could have been caused by the 5VSB using too much current to operate at idle but you ruled that out by removing the cap and testing it out of circuit, good work!

                          Momaka: I had no idea such circuits existed that cut off 5VSB after 100ms of blackout, I will try to make a mental note of that!

                          P.S davidebaldini: If you use a incandescent light bulb for testing take care that to run APFC it will need to have a pretty high rating, see the linked post below for an example.
                          (I had to use a toaster as an example to run that unit).
                          Btw here is a link to a Swedish store selling that incandescent light bulb if you need it...

                          https://www.badcaps.net/forum/showpo...1&postcount=77
                          "The one who says it cannot be done should never interrupt the one who is doing it."

                          Comment


                            #14
                            Re: ATX power supply keeps blowing PFC mosfets — Cooler Master RS-550-ACAA-E3

                            Thanks for the suggestions. Today I prepared a "replacement kit" inspired by momaka's suggestion using components I had at hand, with a voltage divider between the two electrolytic caps and a film capacitor in parallel (see schematic) rated 100nF 630V. The 4 resistors are 220ĸΩ 1W which will replace the original 1MΩ resistor on the PCB. The corrugated cardboard box in the picture contains the two caps, which will be squeezed between the two heat sinks leaving about 4mm of clearance, so hopefully the cardboard will insulate sufficiently from the heat.

                            Tomorrow is test day.
                            Attached Files

                            Comment


                              #15
                              Re: ATX power supply keeps blowing PFC mosfets — Cooler Master RS-550-ACAA-E3

                              Originally posted by Per Hansson View Post
                              Moreso that it read 0v after just 10 seconds, well, it could have been caused by the 5VSB using too much current to operate at idle but you ruled that out by removing the cap and testing it out of circuit, good work!
                              Agreed, that was a good test and indeed the cap should not have dropped to 0V, even if the 5VSB does continue to run. I don't think I've ever seen those 5VSB offline ICs be capable of using all of the energy in a big cap. Usually at least 10-20V will remain for a few seconds after they shut off. The bleeder resistor will discharge that to lower levels, but it will take a few more moments. In any case, I have not found these caps to be at exactly 0V very quickly after turning off. So that was a good tell-tale the cap was no healthy.

                              Originally posted by Per Hansson View Post
                              Momaka: I had no idea such circuits existed that cut off 5VSB after 100ms of blackout, I will try to make a mental note of that!
                              It's strange. I don't necessarily think all modern PSUs have them. But it appears that anything with CM6800 or CM6802 might. The other one that I recall doing this was a Rosewill Valens 600, though I don't remember if that one used a CM6800/6802. Not sure if one or both of my Enermax PSUs did it too. But I remember at least one more unit than the two CWT PUQ platforms did it. Very annoying, IMO.
                              In contrast, the hold-up time on some of my OEM Delta and HiPro PSUs (without APFC) can be almost a good second for the main PS running with a load. And most of my HiPro PSUs will run the 5VSB without load for at least 15 seconds (some even almost 30.)

                              Originally posted by Per Hansson View Post
                              P.S davidebaldini: If you use a incandescent light bulb for testing take care that to run APFC it will need to have a pretty high rating, see the linked post below for an example.
                              (I had to use a toaster as an example to run that unit).
                              +1
                              Toaster would do well too.
                              Might have to force it / clamp it down to stay On, though.

                              Originally posted by davidebaldini View Post
                              Thanks for the suggestions. Today I prepared a "replacement kit" inspired by momaka's suggestion using components I had at hand, with a voltage divider between the two electrolytic caps and a film capacitor in parallel (see schematic) rated 100nF 630V. The 4 resistors are 220ĸΩ 1W which will replace the original 1MΩ resistor on the PCB. The corrugated cardboard box in the picture contains the two caps, which will be squeezed between the two heat sinks leaving about 4mm of clearance, so hopefully the cardboard will insulate sufficiently from the heat.
                              Schematic looks good.
                              However, I would advise against running such long wires between the caps, the resistor divider board, and the PSU board itself. Such long wires would have terribly-high ESR and ESL at high frequencies. The PSU still might work for a test... but be careful. With such long wires, there will be lots of RF and EMI that other neighboring components could pick up and possibly malfunction.
                              This is how I did it for an Enermax PRO 82+ 525W:
                              https://www.badcaps.net/forum/attach...4&d=1541443598
                              https://www.badcaps.net/forum/attach...3&d=1541443598
                              ... though probably worth noting that I removed the APFC coil and had much more space afterwards. But in my case, it was a conversion from APFC to no APFC... and it worked, but not so great - PSU can't do full rated load or even half-rated load. Shuts down a little after 200W by my guesstimates. So that's one possible adverse affect to removing the APFC circuit.

                              The thread can be found here:
                              https://www.badcaps.net/forum/showthread.php?t=72147

                              In any case, if you can, consider shortening the wires from the caps to the PSU's PCB.
                              Last edited by momaka; 01-27-2023, 09:32 PM.

                              Comment


                                #16
                                Re: ATX power supply keeps blowing PFC mosfets — Cooler Master RS-550-ACAA-E3

                                It works!

                                This is how it looks now just before I closed the chassis and called it done


                                I did the standby test which yielded 5.1V SB (purple wire) and 4.8V PS_ON (green wire). This time VSB remained stable and precisely regulated at 5.1V for ~30 seconds after I toggled off the wall switch, while in the previous test with the broken cap it began to drop immediately. The bulk capacitors now retained some charge for ~40 seconds past power off.

                                I shorted green to black with the same 20W hard disk as a dummy load. It spun! No overheating from the two IRFP460 and both heat sinks remained cold. Red wire had 5.05V and yellow 12.1V, both within spec. Too bad I couldn't test for ripple with a scope. These voltages remained stable for over a second after toggling off the wall switch, with cooling fan energized and HDD spinning at full speed (estimated on noise) which is expectable as the bulk capacitors hold around 50 joules at 380V. I left it running a few minutes and all remained stable and cold. I put it into a storage box, vacuumed the mess in the workshop, and I'm done with it!

                                Confessions: I was too lazy to fix or trim the long wires... as I already glued them to both boards (the PCB and the "kit" board) and taped them for extra strength, but at least I routed them above the heat sink, somewhat away from EMI sensitive stuff; there's about 3cm from the wires to the CM6800 itself, hopefully it's enough. The resistance of these wires should be negligible, around 10mΩ [1], only the center wire of the voltage divider is thinner to help with identification, as it's not supposed to bear high currents and I didn't have a choice of wire colors, only of gauges. But I'm not sure about the ESL. One day in the following weeks I'll take it out of storage and mount it into my workstation to properly test it under full load tripping the UPS.

                                Thanks for all the valuable help!

                                @momaka: I'm enjoying so much your story of the crazy Enermax mod

                                ----

                                [1] estimated for an equivalent 30cm copper wire with a 0.5mm² section.
                                Attached Files
                                Last edited by davidebaldini; 01-28-2023, 03:03 PM.

                                Comment


                                  #17
                                  Re: ATX power supply keeps blowing PFC mosfets — Cooler Master RS-550-ACAA-E3

                                  Originally posted by davidebaldini View Post
                                  The bulk capacitor is at 0V 10 seconds after the mains is off. I couldn't manage to measure it sooner. I suppose this exceedingly fast discharge is not the correct behavior, because the 1MΩ resistor in parallel to the capacitor seems to be specifically assigned to the task of discharging it once the PSU is powered off, and it would take minutes to discharge it.
                                  Originally posted by davidebaldini View Post
                                  I found the bulk capacitor is defective. I desoldered it from the PCB to test it independently from the circuit. I charged it to 10V with a dedicated bench PSU thru a 210Ω resistor; for as long as the bench PSU leads were clamped to the cap, the multimeter read 10V across the cap. As soon as I unplugged one clamp of the bench PSU from the cap, without first switching the bench PSU off, the voltage dropped to 0V.
                                  As soon as I read about the capacitor not holding charge for even 10 seconds, yup, it's dead Jim. That's the reason why your PFC stage blew in the first place. Nice that you have located the fault without blowing more parts.
                                  Originally posted by PeteS in CA
                                  Remember that by the time consequences of a short-sighted decision are experienced, the idiot who made the bad decision may have already been promoted or moved on to a better job at another company.
                                  A working TV? How boring!

                                  Comment


                                    #18
                                    Re: ATX power supply keeps blowing PFC mosfets — Cooler Master RS-550-ACAA-E3

                                    I took this power supply out of storage yesterday to mount it back on the workstation were it blew up. Apparently it's working fine, it's been running overnight at full CPU load, no apparent problems so far, and it holds the UPS switching time. I didn't do any further bench tests the way I'd like, it would have been interesting to measure the temperature of the replaced mosfets since they have higher gate capacitance and lower slew rate, but those heat sinks are large anyway, and after a day nothing smelled hot from the rear airflow. Spartan, but I didn't have the time.

                                    I would have replaced the shorted mosfets with the same part number (16N50C3) but the price was absurd, 8€ for two pieces on Ebay, and anything cheaper that I found on Ebay couldn't bear the necessary current. Screw that.

                                    Comment


                                      #19
                                      Greetings:
                                      Please could you tell me what is the value of the ZD1 component on the Cooler Master RS-550-ACAA-E3 PCB?

                                      Comment

                                      Working...
                                      X