Announcement

Collapse
No announcement yet.

KDMPower MIPC MI-X8775CD [PCB WF-C rev:E]

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

    KDMPower MIPC MI-X8775CD [PCB WF-C rev:E]

    Remember that cheap $15 PSU that dmill89 posted a while back here?
    https://www.badcaps.net/forum/showthread.php?t=72229
    .
    .
    So I was randomly browsing on eBay (as always) and saw a listing for supposedly the same “775 Watt” PSU with starting bid at $1. Description stated “for part or repair”, with the owner/seller mentioning the PSU had blown/shorted. Surely I could spend a few $$ more and get the same PSU as dmill89 right from Amazon brand new and (probably) working… but where is the fun in that?!

    So I got this one instead for $13 shipped to my door ($0.99 for the PSU and the rest was shipping. ) If dmill’s PSU was able to handle 350 Watts without croaking, it really made me wonder what was wrong with this one. I got the unit nicely packed, and it honestly looked pretty much unused with no dust anywhere:


    On the outside, it appears nearly the same as dmill89’s PSU: standard black finish, non-modular unit with a 120 mm fan and voltage selector switch (i.e. no AFPC.) Only difference was really that mine appeared to have a see-through fan with LEDs.

    And here is a bottom-shot:

    I think one way to tell if you’re dealing with a cheap PSU is when you see those stamped PCB stand-offs on the case. On the positive side, at least there are vents on the back of the PSU to (hopefully) avoid hot spots.

    Here are also all of the connectors this PSU has:

    Aside from the standard motherboard stuff (20+4 pin ATX, 4+4 pin 12V CPU power), this unit also has 2x 6+2 pin PCI-E 12V power, 6x SATA, 3x Molex, and 1x floppy/Berg. The output wires are a mix of 18 and 20 AWG, with the ATX connector, 4+4 pin 12V CPU, and first 6+2 pin PCI-E connector being 18 AWG. The 2nd 6+2 pin PCI-E is simply piggy-backed from the 1st one, but with 20 AWG wire. The Molex and SATA connectors are also all 20 AWG. I do like how they arranged the drive connectors though, with 2x SATA followed by a single Molex on each string. And all wires are UL listed, at least.

    OK, let’s have a look at the label finally.

    It appears like this is not the same PSU model as dmill’s. Whereas his was an MI-08775V2 and rated for 875 Watts, this one is a MI-X8775CD and *only* rated for 775 Watts… like we are going to believe that one, too. What’s funny is that label keeps peeling off by itself. While I was taking the case shots above, I had to re-apply pressure on it at least several times to stop it from rolling onto itself and falling off. It’s as if it knew it was telling me lies and so was shamefully trying to cover itself, fall off, and hide somewhere.

    On a different note, I’m surprised the shipping box didn’t get blown away by the wind when the unit was delivered to my house. There’s not much weight to this PSU. Probably a good half of it is from the output cables. On the other hand, all those output cables scare me - imagine someone trying to plug in a beefy video card with two 8-pin PCI-E power connectors! Technically speaking, since there are supposedly two 12V rails rated at 24 Amps each and no mention of the total max combined current, one could assume this unit is capable of 48 Amps (or 576 Watts) on the 12V rail. But maybe that is possible on this unit, after all. What do I know if I haven’t opened the unit yet? I suppose that part comes next.

    But first, would it B OK if I posted a picture of the fan? (Oh look at me trying to be funny now. )

    This one is a rated for 0.12 Amps - just a tad lower than the one in dmill’s PSU. When I hooked it on 12V directly, it does move quite a bit of air, though. And the blue LEDs are very bright and pretty. (Perhaps that also explains the lower rated power of the PSU - those LEDs need a lot of juice, so they had to lower the rating on the unit. )

    Anyways, here is the unit inside.


    Right off the bat we can see… well, there isn’t a whole lot to see, really. Initially, I thought this might be a 300-350 Watt capable unit. Sure the input filtering is missing, but I expect that to a degree in any cheap PSU. What really got me, though… before I had the unit opened and was looking through the vents, I was honestly very excited, because I could see two TO-3P transistors on the primary heatsink and a PDIP-16 chip on the secondary. This got me thinking the PSU must use the same double-forward design as the 875W unit, but perhaps also with added OCP (due to that PDIP-16 IC.) Then after opening the unit, I looked at the parts on the primary heatsink and noted the “13009” text on what I thought was one of the MOSFETs. Surely this quickly extinguished my excitement.
    - That ain’t no MOSFET! That’s a BJT.
    And only then I saw the obvious - that there are 3 transformers in the middle. The realization hit me like a rock on the head… this is yet just another half-bridge PSU. Looking up info on the SDC2921 PWM controller confirms this 100%.

    Ah well, at least the thing has a bridge rectifier and a properly-sized (tall) 35 mm main transformer. The heatsinks are also not bad. Certainly we can’t expect 775 Watts from that. Heck, does a half-bridge design over 500 Watts even exist as far as ATX PSUs go? If yes, I don’t want to know how much one of those weights or what kind of BJTs it has on the primary. It certainly won’t be anything like you see in the pictures above. But anyways, let’s continue forward.

    I think I see what blew up in there.

    Interestingly enough, the fuse (a free e-cookie for anyone that can find it ) was intact. I suppose this part (and perhaps others) blew up to save the fuse, as usual. 5VSB should still be good, though. It uses a 3313D IC (whatever that is, I can’t seem to located a datasheet) and has the typical 50V, 47 uF “startup” cap.

    Time for an autopsy on the primary, so out goes the heatsink.

    We have not one but two charred small “components”. Initially, I thought these were diodes, but the PCB silkscreen suggests they are resistors. I get quite different readings for the two, despite both being connected the same way. Next to these, you may also see two small 1-Ohm resistors (one in front of Q15, and the other to the right of Q16 Base) - these are blown open. One even appears to have a hole / burn mark on it. The reverse-polarity diodes might be OK, though. Will check if/when I decide to dig deeper into it.

    Next, here’s our beloved MOSFETs…. Oops sorry, I meant BJTs.

    They look OK, but they are NOT OK. The left one is completely blown open on all pins. The right one has a short between Base and Collector. The Emitter has decided it won’t talk to these two in any way - completely open in any direction. Given the no-name manufacturer (or at least I can’t seem to find it), no telling if these are knock-offs or real 13009. I suspect it may be the former. In any case, the MIGHTY fuse defeated them.

    Speaking of which, I think this picture just barely shows the fuse… among other parts.

    For the primary caps, we get a pair of CapXon LP, 200V, 820 uF… in 22x42 mm size. IMO, that is suspiciously small for that capacity. I shall find out later. Also interesting to note there is a 2nd spot for a thermistor (which is bridge by a jumper, since there is one already) as well as an empty spot for a connector for PPFC coil. As for the blue cap on the left of the CapXon’s: that should be a Y2-class cap between ground (earth) and negative (-) primary-side bus… but surely enough, it ain’t Y2-rated and rather just a standard 2.2 nF, 2 kV ceramic cap. In terms of EMI/RFI, this PSU certainly won’t pass… and hence why we don’t see any FCC or UL marks on the label.

    On the other hand, the PCB doesn't seem so bad, at least on the solder side:

    The soldering is very good, overall. Spacing between primary and secondary might be a bit questionable in a few places... but overall, nothing too serious, it seems.

    continued in part 2 below due to 10k char limit...
    Attached Files
    Last edited by momaka; 11-17-2020, 12:23 AM.

    #2
    KDMPower MIPC MI-X8775CD [PCB WF-C rev:E] - part 2

    Of course, who cares about safety, so long as this PSU can do 775 Watts, right?
    Well, can it? Let’s look at the secondary side.

    We get one STPS2045CT rectifier for 3.3V and same for 5V rail. Meanwhile, the 12V rail gets 2x MBR20100CT in parallel. So the label is a complete LIE in terms of the current capabilities on each rail (but I think we expected that already. ) On the other hand, this is probably one of the few cheapo PSUs that I’ve seen with much better rectifier configuration on the 12V rail than the 5V rail. But that also means the 5V and 12V rectifiers are not tied together in any way, so I suspect this PSU probably does not have good cross-regulation.

    The windings for the 5V rail on the main output toroid inductor sign a similar tune. In the past, most half-bridge PSUs would typically get 2x thick wires for the 5V rail and a single wire for the 12V rail. In this unit, both the 5V and 12V rail appear to have the same wire thickness on the output toroid… and it’s not very thick, mind you. :\ IMO, it won’t be able carry even 24-25 Amps… let alone the current of “dual” 24-Amp 12V rails. The output toroid core is also the lower-tier Micrometals-52 and hence is going to be a little more loss-y. Thus, I expect this PSU probably had mediocre efficiency. And with all of that said and shown above, I don’t think this PSU is realistically capable of more than 250 Watts continuous DC due to those output toroids. Maybe it can peak at 300 Watts, but maybe not. I wouldn’t suggest anyone who has one of these to push it that far. And even if the output toroid was bigger, we also have to account for the output ripple, which is where the output electrolytic caps come into play.

    On that front, I think it’s clear this unit won’t do too well there either. Each rail has only 2x 1000 uF caps (though at least the cap spots are for 10 mm diameter.) There are also PI coils for each rail - yes those tiny ones. The -12V rail gets none, though. It only has a single 470 uF cap. Meanwhile, the 5VSB has a 16V, 1000 uF cap and a 10V, 470 uF cap. All output caps are by HANGCON… or is it KANGCON? You tell me.

    OK, maybe that picture didn’t come out too well. But either way, those output caps probably won’t allow for more than 150 Watts total before the ripple goes out of spec. As a side note, we can see on the above picture the caps are dated early 2019 (3rd week, I presume?) So this PSU is, at most, maybe almost 2 years old. Judging by the lack of dust, though, I don’t think it has seen much (if any?) use. In fact, it still has that “new China smell” to it. I suspect this may even be on of those cases where the user just built a new system with this PSU, powered it On, and the thing blew up almost instantly or in very short order.

    Another interesting thing I find is that the SDC2921 PWM IC datasheet says the IC can be configured to provide OPP. Tracing my unit, I spotted an arrow mark on the PCB silkscreen pointing to three resistors labeled “OPP” (and the resistors are installed.) Now, whether or not it was done properly, that’s another question. Perhaps it was, and maybe the primary blew because those 13009 BJTs were dodgy. Or maybe the Base drive circuit was not built with good enough parts. Or maybe the OPP is incorrectly set to sky-high levels.

    Whichever the case, now I have me another PSU to repair… or should I?
    Part of me does NOT want to, because I have one nice Enermax PSU that’s awaiting an ATX wire harness. On the other hand, these cheapo H-bridge PSUs are usually super easy to repair. In fact, if there is one good thing about this PSU, it’s how easy it is to find components and work on. A lot of functions are even labeled on the PCB with text. Probably a great PSU for anyone that wants to learn how to repair/troubleshoot electronics.

    I guess I’ll sleep on it for a while and then decide what I shall do with it. Knowing me, I most likely will end up fixing it with parts from “lesser” PSUs (I’m looking at you Raidmax RX-380K!) But we will see. Don’t stay tuned for this one - it will likely go in the back of the queue on my list of projects.
    Attached Files
    Last edited by momaka; 11-17-2020, 12:26 AM.

    Comment


      #3
      Re: KDMPower MIPC MI-X8775CD [PCB WF-C rev:E]

      Originally posted by momaka View Post
      Interestingly enough, the fuse (a free e-cookie for anyone that can find it ) was intact. I suppose this part (and perhaps others) blew up to save the fuse, as usual.
      I saw it in this photo, do I get the cookie?
      Judging by this better view of the fuse I'm not sure if I would rate it as a fuse or more a wire link?
      "The one who says it cannot be done should never interrupt the one who is doing it."

      Comment


        #4
        Re: KDMPower MIPC MI-X8775CD [PCB WF-C rev:E]

        Originally posted by Per Hansson View Post
        I saw it in this photo, do I get the cookie?
        And we have a winner!
        (Then again, it appears you were also the only participant in that game. )

        Originally posted by Per Hansson View Post
        Judging by this better view of the fuse I'm not sure if I would rate it as a fuse or more a wire link?
        Why not both?

        Well, it is a legit fuse, after all. But it's one of those small, short ones. Yes, it's rated for 250V, but it's a glass fuse and not heatshrinked. So I imagine there is a chance that it could arc-over in the case of a hard short-circuit. Then again, if the BJTs blew open before it did... lol!

        Also, another thing I noticed today (though I forgot to take a picture): on the secondary side next to the 12V wires, there is silkscreen "12V1" and "12V2" that points with arrows how the supposedly "dual" 12V rails should be connected. But looking underneath on the PCB, there are no separate traces for two 12V rails. It's all coming from one copper trace. So this PSU was never even capable of having dual rails to begin with by design.

        On that note, if I do end up fixing this, I will probably remove one of the 20 Amp rectifiers from the 12V rail to use in another PSU. A single 20 Amp rectifier on the 12V rail should still be more than enough. But anyways, those are all just "rough ideas" for this PSU. First I'm going to try finish fixing up my better PSUs. Finished recapping two HiPro's this week and have an old Delta, FSP, and Bestec lining up in the queue. Maybe even a cool Topower/Zumax I posted a while back.
        Last edited by momaka; 11-20-2020, 11:28 PM.

        Comment


          #5
          Re: KDMPower MIPC MI-X8775CD [PCB WF-C rev:E]

          Ahh yes of course I forgot, the fuse is for a 875w PSU with half-bridge design of around 70% efficiency so of course that explains the thick wire inside it

          But can you explain one thing for me, you say this is a half-bridge design but that it uses 13009 BJT transistors for the half-bridge.
          The design examples I found when I Google show the half-bridge topology using two MOSFET switches.
          Is it just a misnomer for ATX power supplies or what is up with that?
          "The one who says it cannot be done should never interrupt the one who is doing it."

          Comment


            #6
            Re: KDMPower MIPC MI-X8775CD [PCB WF-C rev:E]

            Originally posted by momaka View Post

            What's funny is that label keeps peeling off by itself. While I was taking the case shots above, I had to re-apply pressure on it at least several times to stop it from rolling onto itself and falling off. It's as if it knew it was telling me lies and so was shamefully trying to cover itself, fall off, and hide somewhere.
            [/SIZE]
            They didn't even put good quality glue on the back of the label!



            Now, seriously, I am impressed. This psu is so much better than those Hantol/Powertech/t&p that I posted some months ago.

            It even has proper 35 transformer, primaries that I hope they would at least be 560uF real C and secondary side with proper output filtering on the main voltages. I mean it even has coils!

            These are the psus I really like to play with. I would add input filter, check primary caps, try to fix the psu problem and replace all secondary capacitors with 2200uF caps. And 5vsb always gets 2x1000uF japanese low esr caps to be safe.

            And then play with the fan controller.


            Originally posted by momaka View Post
            On that note, if I do end up fixing this, I will probably remove one of the 20 Amp rectifiers from the 12V rail to use in another PSU. A single 20 Amp rectifier on the 12V rail should still be more than enough. But anyways, those are all just "rough ideas" for this PSU.
            Nooo!

            Having enough headroom on the secondary side actually helps a lot with the efficiency and the total power the psu can handle. It also helps tuning the fan to be slower and quieter.

            Comment


              #7
              Re: KDMPower MIPC MI-X8775CD [PCB WF-C rev:E]

              Originally posted by Per Hansson View Post
              But can you explain one thing for me, you say this is a half-bridge design but that it uses 13009 BJT transistors for the half-bridge.
              The design examples I found when I Google show the half-bridge topology using two MOSFET switches.
              Is it just a misnomer for ATX power supplies or what is up with that?
              Well, you CAN build a half-bridge PSU with MOSFETs too... or probably just about any switching device, if designed accordingly. That goes for other topologies too.

              Here is, for example, one H-bridge PSU posted recently that appears to use MOSFETs:
              https://www.badcaps.net/forum/showth...91#post1003591

              As to why cheap PSUs continue to use BJTs for half-bridge designs... that I'm not 100% sure about. I suspect it has to do with the fact that the design is well-understood and easy to copy (after all, half-bridge schematics have been widely-available on the internet for a long time now.) I also suspect it has to do with the fact that H-bridge PSUs can be built with readily-available dirt-cheap parts. BJTs from the 1300x family are one of those parts just like LM358, LM317, 555 timer, and etc. - i.e. parts that aren't really the best in what they do, but very simple, cheap, and universal, so too good to go away / get replaced by something new.

              As a result of using these cheaper BJTs, the switching frequency is also typically lower... which works out fine with the rest of the PSU too, because then you don't need very low ESR caps or output toroid rated for higher frequency. On that note...

              Originally posted by momaka
              The output toroid core is also the lower-tier Micrometals-52 and hence is going to be a little more loss-y.
              That should be Micrometals-26 in post #2.
              Micrometals-52 is the better core (slightly higher frequency and permeability specs) - typically colored green with blue on one side.

              Originally posted by goodpsusearch View Post
              They didn't even put good quality glue on the back of the label!
              Actually, it might be the paint on the case that's the problem. It has a very fine eggshell finish, so probably most labels won't stick to it properly.

              On that note, I didn't see a warranty sticker anywhere. Not that I was expecting this PSU to have one anyways.

              Originally posted by goodpsusearch View Post
              Now, seriously, I am impressed. This psu is so much better than those Hantol/Powertech/t&p that I posted some months ago.
              That is true.

              When you put it that way, it does seem like a PSU worth fixing up. And indeed I think I will.

              Looks like it really is time to break apart that Raidmax RX-380K crapjob I liked above and use its parts in better PSUs. The only reason I've been saving it is because it uses the most simplistic H-bridge design with a KA7500 controller and discrete transistors for the protections. A long while back, The_Unique posted a tutorial/writeup how to roughly convert such PSUs to adjustable ones and it's been on my mind to try that for a long time now (but just among many other projects, so that's why I never got to it.)

              Originally posted by goodpsusearch View Post
              These are the psus I really like to play with. I would add input filter, check primary caps, try to fix the psu problem and replace all secondary capacitors with 2200uF caps. And 5vsb always gets 2x1000uF japanese low esr caps to be safe.
              Well, as I've come to find out, getting an input filter for these PSUs is the hardest part - at least the EMI/RFI chokes, that is. Sure I have some scrap TV boards that do have proper EMI/RFI filters... but a lot of times, those don't fit (usually too big.) And getting new filter parts (at least reputable ones) is not exactly cheap (some places ask as much for a filter as this PSU costs. )

              I could get a bunch of these, though, and wind my own chokes since I already have the wire:
              https://www.ebay.com/itm/5X-22mm-x-1.../152506497515?

              The other thing is... when it comes to recapping, I'm always low on cap stock by the time I get to my "lesser" PSUs. Main problem is (well, IDK if it's a problem, really ) I've acquired too many good Delta and HiPro PSUs, and those take precedent when it comes to getting recapped. So if by any chance I get any leftover caps, only then the lower-tier PSUs may get recapped.

              Originally posted by goodpsusearch View Post
              And then play with the fan controller.
              I guess I forgot to mention, but this PSU already has one. Not sure how well it works, though. Probably worth looking into, as you noted... once the PSU is fixed.

              Originally posted by goodpsusearch View Post
              Nooo!

              Having enough headroom on the secondary side actually helps a lot with the efficiency and the total power the psu can handle. It also helps tuning the fan to be slower and quieter.
              I know.
              But it won't help too much in this case. The single not-so-thick wire on the main PS output toroid for the 12V rail will very likely start pushing the limits around the 15 Amp mark. At 20 Amps (if the core can even handle that), I suspect it will get very hot. So there's no point in keeping too much headroom on any of the rectifiers if the output toroid wires (nor the core itself) can handle that kind of load.

              The other thing that throws me off is how the primary blew up even with (supposedly) 13009 BJTs. If those are truly 13009 and not cheapo knock-offs, then I expect at least a few components on the secondary (like the toroid inductor) would have seen some heating from the large load the PSU was tasked with. But that doesn't seem to have happened, so I think there is also a possibility there may be something not quite right with the design. The OPP should have also acted before the primary BJTs blew (and it didn't). So for these reasons, I don't see why the secondary should be so over-built, when it really is the primary that's going to be my limit again. I actually rather the secondary side rectifiers go first, as that usually triggers the short-circuit protection and the PSU usually just shuts down. Usually.

              Comment


                #8
                KDMPower MIPC MI-X8775CD [PCB WF-C rev:E] - the autopsy results

                OK, since I do think I am going to end up fixing this PSU... first thing that needs taking care of is the blown components on the primary.

                Before removing anything, though, I decided to draw a rough schematic of the BJT drive circuit. Well, I did that... and perhaps I got a little carried away.


                I just wish I didn't use a scrap piece of paper, as I though this would only be a rough schematic that I could later throw away. Now I don't feel like throwing away this garbage paper (at least until I re-draw / copy it.)

                Basically, the above partial schematic of the PSU contains the BJT drive circuit on the primary, the BJT driver circuit on the secondary, and even the OPP circuit traced out (since it was connected to the BJT drive circuit on the secondary - might as well do that too, right. ) With that said, please don't assume it to be 100% accurate. After scanning it and just as I was ready to upload it here, I already found a mistake (which I corrected in Photoshop.) If something doesn't add up, please feel free to point it out.

                Anyways...
                So what does the autopsy reveal? Well, I wrote that on the schematic paper above, but here it is in text format too:
                - R50 and R52, which were the 1-Ohm resistors, are completely OPEN.
                - R51 measures 1.4x KOhms... but from what I can see, the first band on one side of its color code is Red. So no idea what this is, but the resistance likely doesn't match. Thus, this resistor is bad for sure.
                - R53 measured 0.875 KOhms. However, upon cleaning some of the soot on it with my finger, it started reading 0.93-0.95 KOhms. I can't read its color bands, but it's positioned exactly the same way as R51, just for Q15 instead of Q16, so it likely should be the same value as R51. So R53 is bad too.
                - Either D14 or D12 is bad (I can't tell which one had the bad part, as I removed all parts first, then measured them aftewards and lost track which one is which.) I suspect it was D14, since R53 seems to have more damage.
                - And finally, one of the C945 driver transistors on the secondary, Q12, has a low resistance reading (about 100 Ohms) on BE junction... so that one is bad too.

                And that's all I've found to test bad so far - 6 parts total (or 8, if you count the two 13009 BJTs.) Hopefully the SDC2921 didn't take a hit, because the base of Q12 (and Q11) are directly connected to it. With Q12 blown, though, that is a possibility. Of course, it could be that Q12 blew first, causing the primary BJTs to run only "one-sided" - i.e. either only Q15 or Q16 switching, causing the PWM to up the pulse until Q15/Q16 blew. But more likely, since both 13009 BJTs were blown... I think they blew first, which then sent a huge pulse of voltage on the secondary of the driver transformer, taking out Q12 and the rest of the parts on BJT drive circuit on the primary. Whatever, I guess we will see when I replace all the bad parts.

                Originally posted by momaka View Post
                For the primary caps, we get a pair of CapXon LP, 200V, 820 uF… in 22x42 mm size. IMO, that is suspiciously small for that capacity. I shall find out later.
                And their real capacity is....
                .
                .
                .
                .
                *drumroll*
                .
                .
                .
                .
                .
                625 uF only! (well, at least one of them is - I didn't remove both.)
                I checked several times too, just to make sure it wasn't a mistake on my meter. It wasn't. The meter measured that one "820 uF" CapXon LP every time at 625-628 uF. That's not even 680 uF worth of capacity! More like halfway between 560 uF and 680 uF.

                ESR-wise, that cap showed up as 80 mOhms, which I suppose is OK. Vloss was 0.5% or less.

                That said, even if these caps are 625-ish uF, that's still plenty of capacity... for a 300-350 Watt PSU. But I guess the primary couldn't even do that much.

                Well, that's all I have for now.
                Attached Files
                Last edited by momaka; 11-24-2020, 12:21 AM.

                Comment


                  #9
                  Re: KDMPower MIPC MI-X8775CD [PCB WF-C rev:E]

                  BJTs from the 1300x family are one of those parts just like LM358, LM317, 555 timer, and etc. - i.e. parts that aren't really the best in what they do, but very simple, cheap, and universal, so too good to go away / get replaced by something new.
                  When Motorola introduced their Switchmode series BJTs - 2N654x in TO-3 (actually, TO-204) and MJE1300x in TO-220 - they were the first BJTs optimized for use in switchmode power supplies. So they were some of the best at what they did ...... in the late 1970s. Similarly, LM358, et al, were top-notch parts, when National Semi first introduced them, and now they're still too good at what they do to obsolete them. I don't know whether On Semi still makes the MJE1300x series anymore.

                  I suspect you are correct about their persistence in very low-end PSUs. They are part of a suite of inexpensive, well known parts - TL494-clone PWMs, 13007/13009 BJTs, very mature ferrites (inexpensive and available from lots of sources), Micrometals -26 powdered iron cores, last-for-a-few-years junky sort-of-low-impedance output capacitors.

                  And that's all I've found to test bad so far - 6 parts total (or 8, if you count the two 13009 BJTs.) Hopefully the SDC2921 didn't take a hit, because the base of Q12 (and Q11) are directly connected to it. With Q12 blown, though, that is a possibility. Of course, it could be that Q12 blew first, causing the primary BJTs to run only "one-sided" - i.e. either only Q15 or Q16 switching, causing the PWM to up the pulse until Q15/Q16 blew. But more likely, since both 13009 BJTs were blown... I think they blew first, which then sent a huge pulse of voltage on the secondary of the driver transformer, taking out Q12 and the rest of the parts on BJT drive circuit on the primary. Whatever, I guess we will see when I replace all the bad parts.
                  My guess would be that one one of the 13009s shorted, took out the other, and then the resulting big current pulse blew back through the proportional drive transformer, T2, into the Q11/Q12 circuit.

                  BTW, your SDC2921 is a PWM-supervisor combo IC. The PWM circuit looks similar to the old SG3524 type voltage-mode PWM. The supervisor provides PG and the output voltages window detector can shut down the PWM in case of a UV or OV fault. https://pdf1.alldatasheet.com/datash...C/SDC2921.html
                  Last edited by PeteS in CA; 11-24-2020, 01:23 PM.
                  PeteS in CA

                  Power Supplies should be boring: No loud noises, no bright flashes, and no bad smells.
                  ****************************
                  To kill personal responsibility, initiative or success, punish it by taxing it. To encourage irresponsibility, improvidence, dependence and failure, reward it by subsidizing it.
                  ****************************

                  Comment


                    #10
                    Re: KDMPower MIPC MI-X8775CD [PCB WF-C rev:E]

                    Originally posted by PeteS in CA View Post
                    When Motorola introduced their Switchmode series BJTs - 2N654x in TO-3 (actually, TO-204) and MJE1300x in TO-220 - they were the first BJTs optimized for use in switchmode power supplies. So they were some of the best at what they did ...... in the late 1970s. Similarly, LM358, et al, were top-notch parts, when National Semi first introduced them, and now they're still too good at what they do to obsolete them.
                    Yeah, you're right. I should have said it differently. Those parts were indeed the best back when they came out... but today they aren't. Of course, they are still more than good enough, so that's why we probably won't see them get phased out anytime soon. Not to mention their availability and price.

                    Originally posted by PeteS in CA View Post
                    BTW, your SDC2921 is a PWM-supervisor combo IC.
                    Yes.
                    A lot of H-bridge PSUs now use similar chips that are essentially "decked out" PWM controllers with built in UV/OV/SC protections. SG6105 is another popular one, as are AT2005b, ATX2005, and WT7514L. Deer/Allied's "chip of the year" (i.e. 2003, 2005, and 2012) PWM controller is another one I think - good luck finding a datasheet for that one, though.

                    Anyways, that aside, I haven't made any other progress on this PSU yet. Also working on several other old computer stuff right now, so I do whichever one I feel in the mood for. For this one, I just have to pull the parts from the Raidmax PSU.
                    Last edited by momaka; 11-29-2020, 09:22 PM.

                    Comment


                      #11
                      KDMPower MIPC MI-X8775CD [PCB WF-C rev:E] - repaired! (part 1)

                      I was about to throw the towel on this PSU yesterday. Funny how I said H-bridge PSUs are easy to repair… well, they still are. But sometimes little things can throw you off. With that said, the moral of the story below is that you should NEVER trust in-circuit measurements to be accurate. I already know that, but this one tricked me pretty well.

                      So what happened? (Oh gosh! Here goes a book again!)

                      After identifying the blown parts in post #8 above, I dug out the remains of that Raidmax RX-380K PSU from my closet (it's just the PSU PCB with wire harness - case was re-used to provide a home to a nice 250W HiPro PSU that's been installed in another PC for a good few years now.)

                      Before pulling any parts from it, I decided to draw a partial schematic, similar to the one above for the KDM PSU, so I can compare any design differences and try to figure out what R51 and R53 should be. And here is that:


                      Going by that, we see the Base pull-down resistors for the primary-side BJTs are 2.7 KOhms each. But I noted there were indeed some differences in the BJT drive design: instead of two 1N4148 series diodes (D11 & D12 or D13 & D14 in the KDM PSU), the Raidmax PSU had only one 1N4148 diode with a 22-Ohm resistor in series. Moreover, the Base biasing electrolytic cap is rated for 1 uF (instead of 10 uF in the KDM PSU.) The secondary side of the driver circuit also had a few differences, but those don't matter as much.

                      Thus, I was in a dilemma whether to copy the design of the Raidmax PSU by taking parts directly from it and placing them in the KDM PSU, or try to rebuild the KDM PSU how it was originally. Of course, that still doesn't answer the question of what R51 and R53 should be in the KDM PSU (and moreover, the Raidmax PSU is a crap Sun Pro build, so I wasn't too keen on copying it.)

                      As such, I dug out more junk H-bridge PSUs from storage… namely, these two CWT ISO PSUs:
                      https://www.badcaps.net/forum/showthread.php?t=39370
                      https://www.badcaps.net/forum/showthread.php?t=49205

                      As low-end as they may be, at least they were designed by CWT (and the fully-built versions of these ISO PSUs are actually quite decent.) I looked inside them, and they both had 2.7-KOhm resistors used for the BJT Base pull-down. So I guess 2.7 KOhms it is! Time to start gathering and replacing parts. Heck, I decided to look through my junk pulled parts boxes to see what I have so I wouldn't have to pull parts from the Raidmax PSU. There, I found two 1N4148 diodes (to replace both D12 and D14, just in case there is a mismatch between my pulled 4148's and the cheapo ones in the KDM PSU), two 2.7 KOhm ¼ Watt resistors, and two C945 NPN BJTs to replace both Q11 and Q12 as a pair, just in case. I suspect most of these parts came from a gutless junk JNC PSU I pulled to pieces over a decade ago. Examples of that PSU can be found in these posts, under the names of Meico and Frontier:
                      https://www.badcaps.net/forum/showpo...80&postcount=2
                      https://www.badcaps.net/forum/showpo...2&postcount=33

                      Surprisingly, the only thing I didn't have in my parts bin was the 1-Ohm resistors needed for R50 & R52. But then I found a bunch of MOX ¼ Watt ones on a scrap Panasonic CRT TV board I picked from a TV on the side of the road also around a decade ago (which, BTW, I used parts from to build my first headphone amplifier, which I still use today.) Talk about recycling junk, huh!

                      Anyways, all BJT drive parts installed and ready to go.

                      Funny fact: note that the pulled 1N4148 diodes from my crappy JNC actually have a bigger case than the ones originally installed in the KDM PSU. Makes me wonder how low-budget the KDM 4148 diodes must be. Probably some no-name Chinese crap that barely meet the spec sheet of a 4148 diode… if we are even that lucky!
                      Also note my beautiful handiwork of the resistors with soldered-on extensions leads. Seemed like the easiest way to make those 1/4 Watt resistors fit into the tiny 1/8 Watt spots on the PCB. Whatever, it's a cheapo PSU - no one should care.

                      Now the last part: the primary BJTs. Rather than busting out my vacuum desolder pump iron again and pulling the entire primary heatsink from the Raidmax PSU, I decided I'll just remove the primary caps, unscrew the BJTs, and remove them one by one. Besides, I wanted to measure the primary caps of that PSU once more (I did that many years ago in-circuit with a MicroESR, IIRC… so I wasn't sure if the measurements were 100% right.)

                      On that note, here you can see those “330 uF” Metacon caps measured again:
                      https://www.badcaps.net/forum/attach...1&d=1608943994
                      Yeah, they really are 200 uF. What a gyp!

                      I also wanted to remove them to show you this:
                      https://www.badcaps.net/forum/attach...1&d=1608943994
                      Ah, don't you just admire that beautiful Raidmax / Sun Pro craftsmanship! Seriously, I have no clue how they got the leads on those transistors to get so crooked. I feel 0% bad about taking parts from this turd-of-a-PSU. Actually, on that note, after taking the primary 13007 BJTs out, I also decided to take the C945 driver transistors. That's when I noticed that one of the driver pins on the KA7500 measured 4 Ohms to ground. This PSU suffered from bad caps on the 5VSB at one point, making the secondary-side auxiliary rail grossly overshoot and burn resistors going to the driver circuit and KA7500 IC… so there's a good chance that IC is toast too. Luckily, the C945 transistors were OK. Ah well, it was a from the factory. Let's not shed any tears.

                      Back on track… mounted the 13007 BJTs from the Raidmax onto the heatsink of the KDM PSU and test-fitted them, but no bueno! -Their leads were just ever so slightly shorter to go through the PCB on the KDM PSU. Thus, more hacking was needed! I pulled a bunch of dead TO-220 devices from my ultra-uber-junk-bins and clipped their leads - hey even dead parts can be useful for something!!! Soldered those on and this was the result:

                      It sure ain't pretty, but whatever. I just need the crap to work!

                      So after everything was installed and ready for a test, I plugged the PSU through a series resistive device (60 Watt incandescent light bulb), in case anything decided to blow. As expected, though, nothing did. 5VSB came up normal, as did the secondary-side auxiliary rail, which I measured around 13V with no load on the 5VSB and about 14V with a 0.5 Amp load on the 5VSB. Conditions could hardly be more ideal. Thus, I shorted PS_ON to GND and…
                      .
                      .
                      .
                      Nothing?

                      I tried again a few more times, with and without various small loads on the 5V and 12V rails, just to make sure that wasn't the issue. I also switched my series current-limiting device to a 450-Watt heater from a dishwasher, in case the high-impedance was an issue (can be with some PSUs, as I've found over the years.) But nope! PSU was still dead as a door nail, 5VSB aside. At most, the 5V and 12V would climb to ~10 mV, but that's it.

                      The first thing that crossed my mind: that SDC2921 IC must have died. After all, one of the driver C945 BJTs attached to it was shorted (likely from the voltage spike that propagated through the driver transformer when the primary BJTs blew.) Things weren't looking too good anymore.

                      Part 2 coming in a few minutes... I really need to learn how to stay within the 10k char limit, lol.
                      Attached Files

                      Comment


                        #12
                        KDMPower MIPC MI-X8775CD [PCB WF-C rev:E] - repaired! (part 2)

                        But then I had an idea. What if I use another working H-bridge PSU and take the driver pulses from it to try and drive the C945's in the KDM PSU? Seemed like a legit idea, so I tried it. The guinea pig PSU was this gutless Cyberlink PSU - another ultra-low-end CWT unit that I didn't mind loosing to an experiment, if something went south. That didn't happen, though. The way I wired it: the c945 Base-drive signals were taken from the Cyberlink PSU and dumped via jumper wires onto a breadboard, where I had a pair of C945 transistors with their Emitters and Collectors wired to the KDM PSU.
                        Why do it like that?
                        - Because there's no way to tell if the SDC2921 drive pins (which should be open-collector outputs) might try to “fight” or short the Base drives signals from the KDM PSU to ground. So this way, the Base of the C945 BJTs on the breadboard becomes only driven by the Cyberlink PSU and not connected to the KDM PSU board.

                        Once everything was wired, I connected PS_ON to ground only on the Cyberlink PSU. This should have made voltages also appear on the output of the KDM PSU, even if there was a fault on its output somewhere. But it didn't happen! The KDM PSU just made a slight “tick” noise, and that was it, with the output rails again going up no more than a few mV.

                        This didn't make sense! The KDM PSU should have done something more. But it didn't! After thinking for a while, at least this re-affirmed me that perhaps the SDC2921 may not have been the cause (which is good, because I found only one listing on eBay for this IC, and it was for a lot of 5 for $6 total - not expensive, but I dislike buying ICs that I would never end up using.)

                        Back to the drawing board and schematics… I suspected the driver transformer next. Thus, I looked into my transformer/inductor parts boxes and found the driver traffo from that gutless JNC/Meico PSU I mentioned above. What a trooper - it keeps on providing parts I need! This is why I DON'T throw away even the most gutless of PSUs.

                        I extracted the driver transformer from the KDM PSU and connected the one from the JNC in there temporarily with some wires on the back side (and taking good care to make sure the windings were all hooked in the correct order - which I did by looking at the original PCB of the JNC PSU… which now houses a fuse and some high-power diodes I use for other experiments nothing can R.I.P. here, lol! ).

                        While at it, I also put the original driver transformer of the KDM PSU next to the JNC one, just to show the differences. Note how much smaller the driver transformer of the KDM PSU is with its 16 mm core (vs. 19 mm on the JNC one.) And I thought that JNC PSU was “gutless”! What's more insulting is that the JNC PSU was only rated for 235 Watts, yet still has a bigger driver transformer, BJT resistors, and diodes, than the KDM PSU. This re-affirms my notion that the KDM PSU is really just another sub-300 Watt H-bridge unit. As such, no point in keeping the 2nd parallel MBR20100CT rectifier on the 12V rail, because this PSU just can't provide that kind of power anyways. So I removed it to upgrade other PSUs. In all honesty, even the Turbolink PSU shown above is starting to look good compared to this PSU.

                        Anyways, even with this new driver transformer setup, the KDM PSU was still dead, regardless if I tried to use the onboard SDC2921 or Cyberlink PSU to drive the C945 BJTs. Really, the only new difference with the JNC driver transformer was that the PSU made a slightly louder “tick” noise and the output voltages could jump up to the ~100mV range. But for all practical purposes, the PSU was still dead.

                        Like I said, I was about to throw the towel after this. I kept checking all components after each test above, to make sure nothing had gone bad, and nothing did. Clearly, however, if the PSU isn't working with external or internal base drive pulses and a replacement BJT driver transformer, something else was hiding from me. I know the JNC PSU driver transformer is good, because that PSU suffered from a single-transistor (with no feedback) 5VSB circuit going bad and killing the DBL494 power circuits.

                        Eventually, I decided to check the protection diodes across the C945 driver transistors (Q11 & Q12) one more time. The readings (in-circuit) didn't seem too unreasonable, though. The forward voltage drop was slightly low at ~0.29 to 0.3 mV (given they are 1N4148's), but I considered it “passable” due to other circuit components. Meanwhile in reverse direction, the voltage drop was a little over 1.1V across each diode, confusing me that perhaps there is indeed a circuit path somehow through the two 2.7-KOhm resistors connected between the secondary-side Aux. rail and the Bases of Q11 & Q12. Furthermore, when testing these diodes in reverse-bias in resistance mode, I was getting about 1.3 KOhms - which is very close to approximately ½ of the 2.7 KOhm resistors. But once Q11 and Q12 were removed and I re-checked the measurements of these diodes again, it became clear that a circuit path CANNOT exist and they should read open, regardless of those 2.7 KOhm resistors. So I desoldered those diodes and sure enough, one was showing a resistance out of circuit of about 1.x KOhms both ways! Yup, that booger got me!

                        So let that serve you (and me) as a lesson that component measurements in circuit cannot be trusted, even if things do look pretty normal.

                        Once I replaced both of these protection diodes (because, why replace one and risk more headache when I can replace both) and soldered back Q11 and Q12… VOILA! PSU was back to life, even with that jurry-rigged JNC driver transformer. Of course, I replaced it with the original KDM driver transformer after this. Sure enough, the PSU continued to work. Finally!

                        With that said, this KDM PSU is still far from being finished and ready to use in a computer. Obviously, it still needs EMI/RFI filters on the input. But I won't waste any time doing that before testing the PSU with a decent load first (likely by using a Ni-Chrome heater wire.) I only tried the PSU with a light load so far: ~1 Amp load on the 5V rail and 0.18 Amp load on the 12V rail. I have to say, though, I wasn't impressed at all with the voltage regulation - the 5V rail was fine ~5.1V, but the 12V rail was over spec @ 12.74V. Switching the load around (for a ~1.6 Amp load on the 12V rail and ~0.12 Amp load on the 5V rail), the 12V rail was in spec this time @ ~12.3V as was the 5V rail (barely) at 5.21V. At least the 3.3V was OK both times. Still, as I suspected, the lack of coupling between the rectified 5V and 12V rail transformer windings (a better group-regulated design) made the 5V and 12V rails cross-load pretty easily. I'm curious to see how this PSU will do with a real load. I still suspect it won't do too well… but maybe with a bigger load, it will. Who knows!

                        Anyways, that's all I have for now on this PSU. At least it's back to working condition.

                        Comment


                          #13
                          Re: KDMPower MIPC MI-X8775CD [PCB WF-C rev:E]

                          Thanks for sharing this repair
                          9 PC LCD Monitor
                          6 LCD Flat Screen TV
                          30 Desk Top Switching Power Supply
                          10 Battery Charger Switching Power Supply for Power Tool
                          6 18v Lithium Battery Power Boards for Tool Battery Packs
                          1 XBox 360 Switching Power Supply and M Board
                          25 Servo Drives 220/460 3 Phase
                          6 De-soldering Station Switching Power Supply 1 Power Supply
                          1 Dell Mother Board
                          15 Computer Power Supply
                          1 HP Printer Supply & Control Board * lighting finished it *


                          These two repairs where found with a ESR meter...> Temp at 50*F then at 90*F the ESR reading more than 10%

                          1 Over Head Crane Current Sensing Board ( VFD Failure Five Years Later )
                          2 Hem Saw Computer Stack Board

                          All of these had CAPs POOF
                          All of the mosfet that are taken out by bad caps

                          Comment


                            #14
                            Re: KDMPower MIPC MI-X8775CD [PCB WF-C rev:E]

                            Thank you

                            Comment


                              #15
                              Re: KDMPower MIPC MI-X8775CD [PCB WF-C rev:E]

                              You're welcome guys.
                              I still haven't made much progress past fixing it. Still cooking up a plan for a PSU load tester (or rather, I just need to get down to building it - I have most of the parts already... it's just too cold in the garage this time of the year now, so that's what's kind of keeping me back a little.) Once that is complete, though, I think this might be the first PSU to go through it... well, we will see.

                              Also, the 20 Amp rectifier I stole from its 12V rail - that one has worked really nicely in an older Delta 300 Watt PSU, which originally only had a 16 Amp fast recovery rectifier. Now the 12V rail is about 0.1 to 0.2V higher under load on the Delta PSU. Before, it used to drop down to 11.8x Volts under moderate load. Now it's 11.95V or more. So a nice little improvement there. I don't think this KDMPower PSU will miss it, TBH... because again, I doubt I will be able to pull so much current from the 12V rail before the primary blows up.

                              And on that note, I might also steal the PCI-E power cables from it for a better Enermax PSU I am currently fixing up.

                              Comment


                                #16
                                Re: KDMPower MIPC MI-X8775CD [PCB WF-C rev:E]

                                Originally posted by momaka View Post
                                Still cooking up a plan for a PSU load tester (or rather, I just need to get down to building it - I have most of the parts already...
                                Please post your progress on this project I would be very interested to see want you come up with

                                Thanks
                                9 PC LCD Monitor
                                6 LCD Flat Screen TV
                                30 Desk Top Switching Power Supply
                                10 Battery Charger Switching Power Supply for Power Tool
                                6 18v Lithium Battery Power Boards for Tool Battery Packs
                                1 XBox 360 Switching Power Supply and M Board
                                25 Servo Drives 220/460 3 Phase
                                6 De-soldering Station Switching Power Supply 1 Power Supply
                                1 Dell Mother Board
                                15 Computer Power Supply
                                1 HP Printer Supply & Control Board * lighting finished it *


                                These two repairs where found with a ESR meter...> Temp at 50*F then at 90*F the ESR reading more than 10%

                                1 Over Head Crane Current Sensing Board ( VFD Failure Five Years Later )
                                2 Hem Saw Computer Stack Board

                                All of these had CAPs POOF
                                All of the mosfet that are taken out by bad caps

                                Comment


                                  #17
                                  KDMPower MIPC MI-X8775CD: assembled with input filter (v1)

                                  OK, so I was cleaning up some of my other open/unfinished projects the other day and decided to put this KDMPower PSU back together too. Surely I could solder the input wires and slap it shut again. But it bothered me to leave it with no input EMI/RFI filter. Plus, I bought a few cheap common-mode EMI/RFI chokes a while back from eBay and wanted to test them. These are really tiny (14 mm OD, 8 mm ID, 5 mm thickness) with rather thin wires (20 AWG?) I already checked that they can handle up to 3 Amps of AC current without melting (they did get warm, though not as bad as when I tested them at 5 Amps of AC current. ) So I figured if there is a crappy PSU worthy of these cheap chokes, it would be this KDM PSU. And to go with that, I also dug out a 0.22 uF X2-class cap that I pulled from an ancient CRT TV board over a decade ago.

                                  As soon as I got to installing these parts, I ran into a problem/dilemma. Remember that tiny fuse hidden under the NTC thermistor?
                                  https://www.badcaps.net/forum/attach...8&d=1605593160
                                  Well, the manufacturer of this PSU was so cheap that instead of installing jumpers to bypass the missing EMI/RFI choke, this is where they placed the fuse - saving time and components that cost nearly nothing.

                                  This meant that if I wanted to install a choke, I'd have to move the fuse.
                                  - But where to?! Sure there is a spot on the PCB for it, but it's for a vertical fuse… and I didn't want to solder ugly extension leads to the existing fuse to make it fit there. Not only that, but I would also have had to heat-shrink the fuse to make sure it couldn't arc-over if it blew. And overall, it seemed too close to the NTC… which is where things get even more interesting, as the NTC is not connected on the same trace/line as the fuse, and rather on the opposite one. So it would be preferable if they are spaced a little further apart.

                                  Anyways, after a lot of pondering and thinking about something so trivial on such a POS PSU, this is the crude and basic primary EMI/RFI filter I whipped together:

                                  In the end, I decided not to mess with reusing the original “mighty” fuse … which by the way, is only rated for 3.15 Amps (though it is a slow-blow type), and yet it still managed to defeat the original no-name cheapo “13009” BJTs, which blew open first. Sure makes me wonder how crappy the rest of the parts in this PSU must be then. Anyways, that 3.15 Amp fuse got replaced with an even more “super-powered” 4 Amp, 250V round fuse. I went with that choice because the round fuse lead pitch fit perfectly the holes on the PCB, and because I have a bag of these (salvaged from a trash can at my last job - why anyone would throw away good fuses is a mystery to me. )

                                  Yeah, replacing a smaller fuse with a bigger one is never really a great idea. Then again, if I was to actually follow the PCB marks on this PSU, I'd have to have installed either an F8A or F6.3A 250V fuse. Ha! Screw that! But you know what is more worrisome than this now? Look at the Live and Neutral labels. See anything out of the ordinary? If not, here's a visual hint:
                                  https://www.badcaps.net/forum/attach...1&d=1614405269

                                  How about now?
                                  .
                                  .
                                  .
                                  .
                                  If you thought “swapped Live and Neutral”, then you got it right.
                                  On the above picture, you can see the hole that is labeled for the Live wire takes on the top trace, goes through the NTC thermistor, and then to the rest of the PSU (bridge rectifier.) Meanwhile, the lower hole is labeled for the Neutral wire, which connects to one of the holes for the fuse. So they designed the PCB for the fuse to go on the Neutral! I mean, this is not a super-huge deal (especially in Europe, where with the directional ambiguity of the Schuko plug, Live and Neutral can be swapped around.) But in North America it sort of is, especially since the power switch is a single-pole type and on the Live only.

                                  Of course, I noticed this PCB design “mishap” only after installing the fuse and finishing the input filter. Again, it's not a terribly huge deal, but it sure made me mad, wanting to smash the PSU. For a moment, I thought I might have to do some trace cutting and other rewiring to get everything done properly. But then I figured an easier solution:

                                  Probably not the best picture to show it, but you can see my easy solution: take a permanent parker and just change the Live and Neutral labels, then connect wires accordingly.
                                  There! Now the fuse will be on the Live wire.
                                  After dong this, though, I came to another realization: the voltage selector switch would now be wired to the Live as well. Again, this is not a big deal and the PSU would work fine like that too (all it does is it bypasses two diodes in the bridge rectifier and connects the Neutral directly to the “mid point” of the two primary electrolytic caps.) However, I've noticed in other PSUs it's just standard design that the voltage selector switch is always connected to the Neutral (probably due to the switch's proximity to the case?) So to fix this new “issue” too now, I disconnected the voltage selector switch wire from point “SWB” on the PCB and instead drilled a new hole in front of the bridge rectifier on the opposite (now Neutral) trace.

                                  While still on the discussion of PCB design/layout, let's look back to this again:
                                  https://www.badcaps.net/forum/attach...1&d=1614405269
                                  So you see how the distance of the optocoupler and transformer pins between the primary and secondary side is at least 5-6 mm, as are the primary and secondary traces where the input wires are soldered to? This is to prevent any arc-over from primary to secondary. However, look at the trace after the fuse and the secondary-side trace next to it - they are much closer together. This is yet another design flaw of the PCB! There's a good chance this unit may not pass a high-pot test… and probably why we don't see any safety agency marks on the label.

                                  There's more PCB goofiness, though…
                                  Remember how I mentioned there's a spot on the PCB for a connector for a PPFC coil?
                                  https://www.badcaps.net/forum/attach...8&d=1605593160
                                  Well, look at the bottom (solder) side of the PSU again in post #1.

                                  This connector is placed on the positive rectified line after the bridge rectifier. That means if a PPFC coil is installed, it would be fed rectified DC current instead of AC… which doesn't make much sense, IMO. Moreover, in North America or any place with 110-120 V AC line, this location for the PPFC coil makes it quite *useless* - at least for ½ of the AC wave cycle. Why? Because with the voltage doubler circuit active (i.e. 110-120 V AC operation), only when the “upper” input cap is charging is when the PPFC coil will work. For the “lower” input cap, current will not go through the PPFC coil as it charges. Therefore, installing a PPFC coil in a PSU like this and using it in a country with 110-120 V AC mains will make the PPFC coil very ineffective. So in essence, the manufacturer of this PSU clearly designed the PPFC coil only with 220-240V countries in mind… and if I had to guess, that means mostly for Europe, since that's one of the few places where PFC is required for electronics above a certain power level. Meanwhile, I also checked a HiPro and a Bestec PSU with PPFC coils, and those had them connected to the AC line right before the bridge rectifier, as I expected.

                                  Anyways, I know that is a lot of writing and discussion for such a cheap POS PSU. But the more I dig in there, the more flaws and shortcomings I find. Sure the PSU can still work at the end of the day, despite these. But it goes to show that such cheap PSUs are not just about using less and inferior/smaller components. Rather, it seems that they also lack a good deal in terms of design.

                                  Up next, I will be doing some “unconventional” load tests. So feel free to stay tuned for those. Oh, and if some of you may be wondering why I didn't put any Y2-caps in the input filter - don't worry! I will go back in there and revise/improve it, if this PSU manages to survive the tests. As I stated in the beginning of this post, I just wanted to put the PSU back together to clear my projects area a little. On that note, as I was putting it together, I finally had enough of that silly label peeling itself off. So I tore it off and did this:

                                  Yes, label now resides inside the PSU. I was going to throw it away, but figured why not slap it on the inside as an extra insulation between the primary components and the case. The way I have it in there will also not allow it to peel itself off anymore, as it's held by the fan on the top and the insulation sheet in the middle and bottom. Plus, it can stay to remind me of how over-rated this PSU really was. As such, I will probably need to whip out a new label… but that will come at the end, after testing and all. Heck, I still haven't changed the original Hangcon / Kangcon caps, and those will certainly have to go if I want to put this PSU in a PC.

                                  So there's more to come.
                                  Attached Files
                                  Last edited by momaka; 02-26-2021, 11:58 PM.

                                  Comment


                                    #18
                                    KDMPower MIPC MI-X8775CD: load-tested and still running… for now (part 1)

                                    As promised, I did some load testing on this PSU. The results were better than I expected in some ways, and just as bad as I thought they would be in others. But before I go off digging in numbers, let’s talk about the test “setup” first, accompanied by a visual (picture.)


                                    Oh, look at that perrrty blue fan! The MI-X8775CD does look good, I give it that. But that’s not the focus of the picture. Rather, it’s that red multimeter showing the AC line input voltage going to the PSU and the stainless steel pot behind. No, I won’t be dropping this PSU in water while running, if that’s what you were thinking. The whole plan of the “setup” is to have a heating element in series with the PSU input line – same thing, actually, as having an incandescent light bulb to limit the current. However, the reason for going with the heating element is to allow more power through… and the water to keep it cool. This way, I can load-test the PSU (to an extent) and not worry if something goes wrong with it, as the series device (heating element) would limit current (and power.) And since the series device causes the AC line going to the PSU to drop slightly (or more than slightly, depending on load), I can see at what input AC voltage (if any) the PSU would shut down. After all, ATX PSUs are rated for a certain input voltage range. Going below this range (low AC voltage, simulating a long brown-out condition) should NOT damage the PSU if it’s designed properly, and the PSU should shut down / refuse to start. But did it? We shall find out later.

                                    In the case above, I used a 500 Watt heating element inline with the AC going to the PSU. It might seem like a very high limit, but remember that this is only if the PSU is completely short-circuited on the AC side, allowing full voltage to be present across the heating element (and hence, full power.) Otherwise, if the PSU is to operate, then the highest power it can draw with that series element is when the AC line is equally split in half between it and the heating element, yielding about 125W from the series device and 125 Watts from the PSU. But AC line split in half equally means the PSU would have to be operating at 60V input. Clearly, that’s a bit too low, as most SMPSes are typically specified to work at no less than 95-100 V AC input, regardless of design. So if the PSU is to see at least 90-100V on its input, then it shouldn’t be loaded too much.

                                    Again, going back to the picture/case above, I had a 12V, 20 Watt halogen bulb on the 12V rail of the PSU and an incandescent bulb drawing 1.5 Watts from the 5V rail. With this DC load (approximately 21-22 Watts total), you can see how the series heating element was dropping voltage on the AC line, leaving the PSU to work with only 111V AC. The AC at the wall was in fact 120V, though:
                                    https://www.badcaps.net/forum/attach...1&d=1614667911

                                    So where does the stainless steel pot come into play? Well, it was full of water… but again, not for dropping the PSU in it while it’s powered on. Rather, I had the heating element placed in it (it is a submergible type, BTW), so that it could stay at a steady temperature and yield a more steady resistance… which I measured at around 29.5 to 30 Ohms. Knowing the resistance of the heating element along with the AC line voltages with and without load allows me to estimate the current (and power) the PSU is drawing on the primary (this is something a Kill-a-Watt could do too, BTW.) To get that, we take the no load voltage (120V) and subtract the load voltage from it (111V), giving us 9V across the 500 Watt heating element. This means the current going through it must be 9V / 30 Ohms = 0.3 Amps… which is also the AC current the PSU is drawing on the primary, since it’s in series. At 111V AC input, the PSU is drawing approximately 111V x 0.3 Amps = 33.3 Watts with the 21-22 Watt DC load on the output. Efficiency-wise, that comes out to 22 / 33.3 = 66%... which is downright terrible. Of course, keep in mind many PSUs have poor efficiency at such low loads. The regulation, on the other hand, should have been passable at this low load… and it wasn’t, to say the least:


                                    Red multimeter shows the 12V rail and the gray RS one shows the 5V rail. While the 12V rail was still in spec (albeit a little high), the 5V rail was low and outside of ATX spec at 4.5V. Looks like the PSU needs a bigger load on the 12V rail.

                                    Of course, the above load test was just more for me to see if the PSU works without blowing up (as I haven’t really properly tested it ever since it was fixed) and to see if the numbers added up… which they did, as I confirmed with my Kill-a-Watt meter also hooked to the same setup. So up next I added a bigger load on the 12V rail and repeated the same test.


                                    Red and gray multimeters show the 12V and 5V rails are now in spec (though the 5V is just barely so, but that’s another item for discussion.) The black one shows the current draw on the 12V rail: about 5.66 Amps into a resistive load bank. Actually, that resistive load bank measures about 1.8 Ohms, so the current should have been a little over 6 Amps. But due to all of the resistance from the jumper leads I was using, there was a lot of voltage drop on the jumper leads’ wires and thus the current was lower. Nonetheless, this is still more than 3x bigger load than the first test at around 71 Watts total… which is getting somewhat close to the 125 Watt limit at half AC voltage. Speaking of which, let’s see what the AC we got now.


                                    Ooof! 89.9V That’s a very low AC input voltage. When I saw this, I was surprised the PSU even ran with normal voltage regulation on the output (well, as normal as it gets for this PSU. ) I even shut it off after a few seconds to check if anything was getting hot inside, but it wasn’t. The PSU was actually running OK at this low voltage… which is not really a good thing. At this low of a line voltage, it should be shutting down. However, I think a lot of half-bridge designs (especially cheap ones) have this problem, and will continue to try running even at abnormally low AC input until something cooks.

                                    In any case, I proceeded with running the same calculations as I did above. Now we can see the AC current was about 1.02 Amps, making the AC-side power consumption about 91.73 Watts. So with approximately 71 Watts of load on the output, the total efficiency was now 77.5%... which is not bad at all for a cheapo half-bridge PSU like this, and especially given the low AC input line at not even 90V. As you can see, increasing the load to more normal levels does change efficiency quite a bit.
                                    Attached Files
                                    Last edited by momaka; 03-02-2021, 12:57 AM.

                                    Comment


                                      #19
                                      KDMPower MIPC MI-X8775CD: load-tested and still running… for now (part 2)

                                      After the above tests, it was clearly time to step it up. Unfortunately, my current load test setup with all the jumper leads was not longer up to the task. Clearly, I needed to make something that would drop less voltage. A few days later, I cobbled something together with a 20-pin ATX connector from an old motherboard, a 4-pin 12V CPU connector, and a 6-pin 12V PCI-E power connector. Soldering good thick wires to these (16 AWG) and adding proper nut connectors at the end allowed me to draw the full 6 Amps of current from each heating element in my load bank.

                                      As such, I tried the following load setup next: 2x 6 Amps (12 Amps total) from the 12V rail, 2.5 Amps from the 5V rail, 1.74 Amps from the 3.3V rail, and of course a 1.3 Watt load on the 5VSB + a 1.2 Watt load from a 12V mm fan running @ 5V to cool the load bank. I didn't take pictures for this one, but the total load (given the nearly same output voltages from the PSU) yielded about 172.6 Watts of draw from the DC side. Of course, this wasn't without one very important change: I swapped the 500-Watt heating element for a 1400-Watt heating element (a mini toaster oven with heating elements configurable for either 700W or 1400W, depending if broil or bake setting was chosen.) This change allowed me to draw the 172.6 Watts from the output with the AC input line staying at a steady 94.5V. Doing all of the calculations again, I came up with 219 Watts for the primary side and efficiency moving up a notch at 78.8%. Again, this is not bad at all. With the efficiency still increasing, this means the PSU is far from its limit.

                                      Before moving to the next test, however, I ran the above config for about 5 minutes straight and then unplugged and opened the PSU to check temperature inside it. To my surprise, the primary heatsink wasn't hot at all like I expected it to be with the low AC line voltage, and neither was the output toroid. Both were just slightly warm. Granted I only ran the test for about 5 minutes and my room temperature around this time of the year is a steady 17.5-18°C (63-64°F.) So looks like the PSU still had headroom on those. The secondary heatsink, on the other hand, was quite warm… though still nowhere near burning hot. (I guess goodpsusearch was right in suggesting to leave the extra rectifier on the 12V rail… though I still think the PSU would do better with one bigger rectifier on the 12V rail vs. two smaller ones.) Of course if the PSU is ran all day with that load and at higher ambient temperatures, I wouldn't be surprised to see the secondary heatsink reach 65-80°C. The hottest component, by far, had to be… guess what? - The NTC inrush thermistor. I couldn't even keep my finger on it. I can see now why they add relays in parallel on high-power PSUs – over 2-3 Amps, and the power dissipated by the NTC becomes a lot greater than what the secondary coil of a relay would use (not to mention high power NTCs could get expensive.) Apart from the NTC, the small CM choke I added was also very hot. Not sure if this is due to the AC line passing through it or because of its proximity to the NTC thermistor. My guess for now, is the latter.

                                      In terms of voltage regulation, the PSU did fairly OK now. 12V rail was still around 12.2V. Meanwhile, the 5V rail actually improved a little by dropping down to 5.17V. And the 3.3V rail was a bit high at 3.41V, but still in spec and IMO a good result, considering it was the least loaded of the main rails at 1.74 Amps. With that said, I think the designer(s) of this PSU probably “cheated” a little and intentionally wound the 12V rail to output slightly higher voltage all the time so that the PSU would do OK in any 12V-heavy PC. I didn't do a 5V-heavy cross-load to see how that will go, but I probably eventually will. My suspicion, of course, is that this PSU won't do well at all with that...

                                      Last test consisted of me trying to pull 3x 6 Amps (18A total) on the 12V rail with still the same 2.5 Amps on the 5V rail, 1.3 Watts on the 5VSB, and 1.2 Watt fan (3.3V rail left unloaded this time.) All in all, that would have been about 230 Watts from the DC side. But considering the AC input was already at ~95V with the 172-Watt DC load above, I expected the AC line should drop below what the PSU can function at normally. And sure enough, that's exactly what happened… except the MI-X8775CD did not shut down and tried to continue to run… which is NOT a good thing at all! Because the input AC line was clearly too low (varying between 50 and 80V AC ), the PSU appeared to be pulsing on and off at very high frequency… to the point where the output on the 12V rail was reading around 8-10V and the 5V rail was reading 2.5 to 3V. Ugh! Imagine what this could do to your PC hardware! There was also a nasty loud 120 Hz hum noise emitted by the PSU. It looks like the primary was self-limiting after all, but the PSU just wouldn't latch and stay off. As long as I had the PS-ON signal jumped to ground, it would try to keep running.

                                      Suffice to say I did not run the above test for more than a few seconds. After all, I didn't want to burn out the primary BJTs again, as I don't have any more replacements currently on hand. I'm pretty sure that if I didn't have the series heating element and if the PSU was left to run at such low AC line, it would likely have blown up pretty quickly. I opened it up briefly after this test, and just for the few seconds of abuse, the primary heatsink was starting to feel a tad bit warm, while everything else was still cool.

                                      I'm pretty sure I could pull 18 Amps from the 12V rail if I connected the PSU directly to the wall. But given the abuse I just have to the PSU, I didn't want to risk it. Instead, I ran some numbers and calculated that I could pull additional 3 Amps on the 12V rail (from the test where I had it at 12A) if I could somehow boost the AC line after the series heating element. This is where a UPS transformer came to the rescue. Wired with its primary and secondary windings in series and primary powered from the AC line, this brought a nice 12V AC boost on the PSU input … which at no load was close to 134V AC (meaning each 200V cap inside the PSU was sitting at close to 190V )

                                      So I configured the PSU for about 15.5 Amps on the 12V rail, with the other rails same as before (2.5A on 5V, 1.3 Watts on 5VSB, and 1.2 Watt fan on 5V)… thus, about 203 Watts DC load. Then I jumped PS-ON to ground. The PSU buzzed again with that 120 Hz hum, but only for a short moment, after which the output voltages came up normally and PSU started running OK. At this DC load, the AC input going to the PSU read 84.3V with the boosted AC voltage. Due to the UPS transformer boosting the voltage (and not knowing exactly how much, though it really should have been 12V AC), the calculations for the power draw of the PSU on the primary and its efficiency may not be entirely accurate. But what I came up with is about 60% efficiency at this load… which may be right, given the very low AC input at 84.3V that barely allowed the PSU to run this test. But hey! It survived that one as well. 5V rail was still the same at ~5.15V, while the 12V dipped only a slight bit to 12.08V.

                                      Overall, I'm glad to see the PSU manage up to 200 Watts of DC load, with most of that on the 12V rail too… not to mention the silly-low AC input line. I'm just not so excited about the fact that it doesn't shut down with a low AC input. Ideally, it should refuse to power on at anything under 95-100V AC. Kind of makes me wonder if such circuit could be implemented and whether I should attempt that with this PSU. Also makes me wonder how other half-bridge PSU will fare with such a test. Another point to mention is the 5V rail regulation was not that great either. Looks like the manufacturer clearly built this PSU with a 12V-based PC in mind. So it's quite possible this PSU will do terrible with an old Athlon or Pentium 3 PC. And lastly, I would also like to determine, if possible, where the OPP protection is set on this unit and also if other protections (such as short-circuit and over-voltage) work too.

                                      I knew this PSU would be another “fun” repair project, but I didn't know it's going to kick off so many ideas for so many other projects / tests.

                                      Comment


                                        #20
                                        Re: KDMPower MIPC MI-X8775CD [PCB WF-C rev:E]

                                        changing the subject slightly,
                                        have you ever cut and stripped one of the centech probe leads??
                                        i doubt it - or you wouldnt have put 111v through one!

                                        i cut one to put clips on the end - after i saw the "conductor" i threw the cables in the bin!!

                                        Comment

                                        Working...
                                        X