Announcement

Collapse
No announcement yet.

Corsair CX750M [Model 75-002019] - burned output toroid inductor

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

    Corsair CX750M [Model 75-002019] - burned output toroid inductor

    Looks like I may need a little help from the PSU experts (or anyone really! ) I have a Corsair CX750M (Model 75-002019) that I picked up for free about 4 years ago. This is the PSU:



    So here’s what’s strange about this one: it appears to work normally (normal output voltages) and any PC is stable with it. However, after a while (typically 10-20 minutes, but time can vary depending on the load), there is a sweetish smell of burned magnet wire insulation emanating from the PSU. Upon opening the PSU after this happened (many times), I was able to confirm the smell is coming from the output toroid inductor, which is insanely hot at that point. Attached at the end of the post are pictures of the whole PSU, just for reference. However, this picture probably shows it all:

    Note the insulation on the top of the toroid is very easy to chip. A few temperature tests showed me it easily went past 100°C, if not even 150°C.

    Funny thing is, I used the PSU for a good while as a “test” PSU for non-critical hardware (typically only for very brief tests, so that the output inductor couldn’t overheat too much.) I just couldn’t understand how that burned output inductor was still letting the PSU work… but somehow it did.

    Now here is the big question: did the output toroid inductor overheat because it was defective? Or was there another component that’s causing it to overheat? Or perhaps the toroidal core was improperly selected for the task?

    While I was searching for pictures on the internet to see if I can find an image that shows what the core of a working CX750 looks like, here is something even more interesting I found:
    https://linustechtips.com/topic/8908...ng-and-repair/
    ^ Apparently, it’s a thread on LinusTechTips forums that looks to have been started in 2018 with some people chiming in as late as 2019 and 2020, all with the same type of failure: burned output toroidal inductor on the CX750M (but also a few other PSUs that use this same CWT platform.).

    According to jonnyGURU there, the issue was due to a manufacturing defect with the output inductor on early units and was supposedly fixed on latter units. I don’t know how true that is. However, going by the info in that thread, the digits before the zeroes in the S/N of my CX750 should indicate the manufacturing date code. In my case, that’s 1537, which would be 37th week of 2015 – just one week later than the O/Ps. I got this Corsair PSU in the Spring of 2017, so that’s likely when the failure occurred. Person I got it from stated his PC would work most of the time, but then start cutting out after very long sessions. And he said the smell was really worrying him.

    Now, I haven’t really done an in-depth test of the components in there. But I did a quick check on the “Sync Card” – the little daughterboard that sticks out right behind the inductor. This board holds the MOSFETs that are used for the 12V rail synchronous rectification. Picture in link form (hi-res):
    https://www.badcaps.net/forum/attach...1&d=1632104825
    So far, everything checks out OK… or at least none of the MOSFETs are shorted and all of their Gate driver resistors are OK, along with a few other parts I checked while in there.

    Here is also a high-res picture of the secondary side PCB of the PSU (just in case anyone needs it or wants to trace connections.)
    https://www.badcaps.net/forum/attach...1&d=1632104825

    Since I can’t seem to find anything wrong so far, I think a good starting point would be to replace/rewind that output inductor, and then go from there to see how the PSU behaves. I *think* the core of this inductor should be black in color on all sides, as it appears to be very similar to this ThermalTake M850W PSU. Basically, both of these are Channel-Well Technology / CWT units (PU-Q platform, if I’m not mistaken.) From what I understand, when these ferrite cores go above a certain temperature, they may loose their specs. So, rewinding the old core will likely result in the windings to burn again. Therefore, I need to find a replacement core.

    The big question is: where and what specs do I look at to find a proper replacement toroid core? Obviously the windings are custom to the PSU. But that doesn’t worry me as much, as I will just copy the design of the old core. What I have learned from previous posts on BCN (big thank you to PeteS on that matter, specifically) is that many of the cheap gutless PSUs (but also a few good ones) use Micrometals “26” -type core material for their output inductors – those yellow cores with white paint on one side. Then there’s Micrometals 52 -type core: light green with blue paint on one side. This one is typically seen on many OEM PSUs and often used in place of -26 cores on the better-built cheapo PSUs brands, due to having very similar permeability specs, but lower losses. So where does that put these toroids with black cores? From my search online, I found the following PDF uploaded on one website:
    https://www.badcaps.net/forum/attach...1&d=1632105150
    According to that, the core should be a “-45” type.
    This seems to agree with the diagram on the product finder on this Micrometals page:
    https://micrometals.com/products/product-finder/
    However, could anyone confirm if my findings are correct?

    Besides core type, we also need to know the size. In the case of my CX750M, the output toroid measures about 35 mm for the outer diameter (OD) without the windings. The hole diameter (or inner diameter, ID) of the toroid is about 20 mm. And the thickness of the toroid is about 11-12 mm. That 35 mm OD converts to about 1.377 inches… so perhaps we can say it’s a 1.3” OD inductor? I’m almost certain, for example, that the output inductor for this BFG Tech BFGR550WGSPSU ATX power supply goes by the part number of T130-26 (meaning it has 1.30 inches OD and uses -26 core material.) Putting the two side by side for comparison (CX750 core is one on the right):

    … I think the CX750M toroid is also a T130 part. Does that mean I should look for a T130-45 part? Yes/No? This is the T130-45 datasheet, if that's of any help:
    https://www.badcaps.net/forum/attach...1&d=1632105150

    Let’s assume the above is correct for a moment. Now the question becomes, where can I buy a replacement toroid core like this? I already found a replacement for the T130-26 toroidal core for the BFG PSU here. However, I don’t see this seller (or any other place online I’ve looked at so far) to have T130-45 cores in stock. Moreover, since that seller seems to be selling T130-26 cores in bulk, I do wonder if I could get away with using that in the Corsair CX750M PSU… or two core in series, if one isn’t enough to get a high enough inductance? If not, I also wonder if a T131-26 toroid core (that the seller also has in stock) would work. T131-26 appears to be the same thing as the T130-26, but with smaller ID / thicker core basically, so I imagine it produce higher inductance too (provided I can even fit the windings on it.) It's datasheet is attached here:
    https://www.badcaps.net/forum/attach...1&d=1632105150
    The only reason I mentioned the T131-26 is because going back to that PDF above, the T131-26 appears to have a very similar Al (permeability) value to the T130-45 core (116 vs. 105). The initial permeability (u_i) seems to be a bit lower (75 vs. 100), but perhaps not dissimilar enough to cause problems?

    Alternatively, I also found this toroid core on Digikey:
    https://www.digikey.com/en/products/...002701/8599642
    It’s not a Micrometals part, but rather made by Fair-Rite Products Corp. The part number is 5961002701 and Digikey P/N is 1934-1595-ND (should the link above ever stop working for some reason.) Initial permeability on this one is 125 and Al is listed as 140. But when it comes to inductor specs, I don’t have any idea which parameters are important here and which are not… or how much I can deviate with these.

    So if I can somehow find a good replacement core for the output toroid inductor, I think I should be able to wind my own inductor. Of course, the big question is, once more, if I have properly identified the old core… and if so, would any of the inductor cores I pointed to above actually work? Any advice would be greatly appreciated here!

    And yeah, I know… it’s an older PSU that probably isn’t worth fixing, as it isn’t anything special. However, I would like to take it as a fun challenge.
    Attached Files
    Last edited by momaka; 09-19-2021, 08:48 PM.

    #2
    Re: Corsair CX750M [Model 75-002019] - burned output toroid inductor

    Micrometals -26 material - powdered iron, 75 permeability, THE standard material long used in switching power supplies with switch frequencies under 50KHz (e.g. those using BJTs). Yellow body with a white side.

    Micrometals -52 material - powdered iron, 75 permeability, very commonly used in switching power supplies with switch frequencies over 50KHz, due to lower core losses compared to -26 material. Green body with a blue side.

    Micrometals -45 material - powdered iron, 100 permeability, core losses similar to -52 material. All black.

    This webpage gives some simple comparisons for Micrometals power conversion core materials, https://www.micrometals.com/products/materials/pc/ .

    The -45 material is higher cost than the -26 and -52 material. But the advantage that may have been the reason it was used is that its A(L) is higher than the other two materials. In practical terms, that means getting the same inductance with fewer turns of wire (lower copper losses and possibly allowing use of a larger gauge of wire, lower copper losses due to skin effect) or a higher inductance for the same number of turns of wire.

    One of the caveats with powdered iron cores is that somewhere between 100C-120C the core's magnetic characteristics start to deteriorate, leading to thermal runaway and effective total loss of magnetic characteristics. My WAG is that the cooling of the Corsair P/S was inadequate (or fan failure?) or compromised by the computer being in a confined space (or simple dirt), leading to the core getting too hot and thermal runaway.

    So, using -26 material would result in higher core losses, not a good thing. Using -26 or -52 material would require recalculating the number of turns, and if it was a coupled inductor recalculating the ratios of windings' turns.
    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


      #3
      Re: Corsair CX750M [Model 75-002019] - burned output toroid inductor

      I would look for CCA wire being used and airflow could be restricted with the plastic guards.
      From dealing with chinese magnetics manufacturers, they have terrible quality control on the material. The annealing, the composition, the alloys - are all over the place. If a batch is bad, they don't really care.

      Mag Inc has cores as well. Maybe try a sample request.

      Comment


        #4
        Re: Corsair CX750M [Model 75-002019] - burned output toroid inductor

        Fair-Rite and Mag Inc. produce ferrite materials, and Mag Inc. produces molypermalloy powder and other specialty powdered-materials (as well as tape wound cores, a whole different aminal).

        Micrometals' closest competitor for powdered iron cores I'm aware of is TSC Pyroferric, http://www.tscinternational.com/mainpyro.html . Their -75 material crosses to Micrometals' -26, and their -LL material to -52. TSC Pyroferric doesn't seem to have a cross to -45 material.
        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


          #5
          Re: Corsair CX750M [Model 75-002019] - burned output toroid inductor

          For lower core losses, you may want to go with Micrometals T130-52 (green/blue) with Al=75 (more turns for any given inductance than -45, thus increasing resistive (Cu) losses). You may be able to use thicker wire or bifilar winding to compensate for the increased resistive losses. Bifilar or Trifilar is better due to skin-effect in the conductor.

          Comment


            #6
            Re: Corsair CX750M [Model 75-002019] - burned output toroid inductor

            Thanks for the help so far, everyone!

            So I guess we can safely say the original core is indeed a T130-45 toroid.
            That moves us to step 2: finding a replacement, and preferably something with a more sane price, in addition to actually being readily available... which the T130-45 toroid is not any of these. (At least I can't seem to find one for sale anywhere from the "traditional" parts stores.)

            Originally posted by redwire View Post
            From dealing with chinese magnetics manufacturers, they have terrible quality control on the material. The annealing, the composition, the alloys - are all over the place. If a batch is bad, they don't really care.
            That's pretty much what jonnyGURU was suggesting over at the LTT forums link above - he thinks the inductors were wound too tightly and then developed a short-circuit in the windings.

            I'm personally not 100% convinced that's the truth, though. Indeed it might be like you suggested instead - that the output toroid just wasn't cooled adequately or that the toroid was of poor quality.

            Originally posted by redwire View Post
            Mag Inc has cores as well. Maybe try a sample request.
            Thanks for the link!

            Actually, I ran into mag-inc while looking at that first Ebay seller's store that I linked to above, but forgot to post about it in my post above. Apparently, he has this too:
            https://www.ebay.com/itm/261329798516?

            ^ it's a Mag-Inc / Magnetics Kool-Mu series toroid core, P/N 77548-A7. Datasheet attached here:
            https://www.badcaps.net/forum/attach...1&d=1632371706
            This toroid has the same 1.3" OD I need, while the initial permeability is rated for 125 (slightly higher than that of the -45 core of 100), and Al value is 127 +/-8% (again, about 25% higher than that of of -45 core.) To me at least, that seems close enough to the specs of the T130-45 core. The only thing I'm not sure about and don't really know how to find out from the datasheets, is the core loss. The -45 datasheet lists the maximum core loss as 71 mW/cm^3 (@ 100 KHz and 130 Gauss), whereas the Kool-Mu 77548-A7 datasheet lists the core loss as 750 mW/cm^3 (@ 100 KHz and 100 mTesla). Seems that difference of almost 10x is quite high... but I don't know how the different peak fields (130 Gauss = 0.13 T, and 100 mT = 0.1 T = 100 Gauss) affect these measurements.

            So what do you all think about that Kool-Mu core? Possible replacement contender?

            Originally posted by PeteS in CA View Post
            This webpage gives some simple comparisons for Micrometals power conversion core materials, https://www.micrometals.com/products/materials/pc/ .

            The -45 material is higher cost than the -26 and -52 material. But the advantage that may have been the reason it was used is that its A(L) is higher than the other two materials. In practical terms, that means getting the same inductance with fewer turns of wire (lower copper losses and possibly allowing use of a larger gauge of wire, lower copper losses due to skin effect) or a higher inductance for the same number of turns of wire.
            Awesome, thanks for that link!

            It appears that, aside from the higher u_i for -45 core, it isn't that much better in any of the other aspects. In fact, the info above even shows that -45 core has overall lower frequency limit than -26 core... but still lower loss, despite that. Thus, a -52 core may be a good possible alternative here, provided I add a few more turns to get it to the right inductance.

            But again, we are back at theory vs. practice here. The problem is, I can't find a T130-52 toroid anywhere...
            Actually, hold that thought...
            .
            .
            I can:
            https://www.ebay.com/itm/183925967793?
            The only complaint I have with that one is the price.
            Otherwise, that may seem like the safer router... or not?

            Originally posted by PeteS in CA View Post
            My WAG is that the cooling of the Corsair P/S was inadequate (or fan failure?) or compromised by the computer being in a confined space (or simple dirt), leading to the core getting too hot and thermal runaway.
            I can't say if the PSU's cooling (at least from the factory) was inadequate here or not. However, the fan certainly wasn't failed in this unit, and neither was there much dust accumulation. The PC case the PSU came from also had good ventilation. Owner was the original buyer and only had the PSU in that system (I asked.) Moreover, the PSU wasn't used anywhere near its full rated load. (IIRC, system specs for that PC were something along the lines of an average i5 CPU, a GTX770 video card, and 2 HDDs... so nothing that could push the PSU too hard - maybe around half to 3/5 of its rated specs, max.)

            So I think that means either the output toroid just didn't get proper cooling due to poor PSU design *OR* the output inductor had a manufacturing defect (that for whatever reason took a long while to show up? )

            Originally posted by PeteS in CA View Post
            So, using -26 material would result in higher core losses, not a good thing.
            Probably...

            While, I don't know much when it comes to inductors... and any leftover knowledge that I have from my college days is certainly more rusty than a 100 year-old sea mine... but I've been reading on the topic a little bit (and probably not really gotten anywhere with it, other than enough to be cocky here ), so here's what's on my mind:
            I understand it correctly, a physically larger inductor usually has a larger effective magnetic path length, Le. This directly results in lower magnetizing field, H (if all other constants are the same, of course.) And because u = B / H... or rather B = H * u ... that means both the lower H and lower initial permeability, u, will result in lower operating B... which again, if I understand correctly, is what affects (lowers) core loss.

            Thus, I've been wondering about using the T131-26 core - it's slightly bigger (by having a smaller ID, making it thicker), which should result in it having an overall lower magnetizing current H... and thus lower B_peak. Additionally, due to being bigger, its Al value is actually very similar to that of the T130-45 core (116 vs 105 nH/N^2).






            Originally posted by PeteS in CA View Post
            ... and if it was a coupled inductor recalculating the ratios of windings' turns.
            It is a coupled inductor... but it only contains two discrete windings: one for the 12V rail and one for the -12V rail. The -12V rail is generated by the induced field from the 12V rail. What the O/P found on the LTT forum thread above was that both windings on the output inductor had 12 turns.

            By playing around with this (unofficial) calculator and entering the parameters from the T130-45 datasheet (and using a few fudge factors, like choosing 10 AWG wire for the windings), it suggested that the inductance of the original T130-45 toroid with 12 turns, should have an inductance of about 19.6 uH and Al = 136. Seems about right, I think?

            Originally posted by PeteS in CA View Post
            Using -26 or -52 material would require recalculating the number of turns
            Right...

            So I wanted to see what would happen if I kept all of the parameters the same with the above online calculator, but only changed the permeability to 75 (for either a -52 or -26 type core). The results suggested that I would need about 14 turns to keep the same inductance, regardless if it's a T130-26 or T130-52 core.

            More interestingly, I tried the parameters for a T131-26 core and also a much larger T175-26 core (both of which are readily available to buy, hence the reason for this.) With these, I don't even need to tweak the number of turns to keep the same inductance. And for both, the Al value came out nearly identical (135 vs. 136 vs. 136 for the T130-45.) So I think those may also be possible replacements, despite being cheaper -26 cores. Moreover, both will still fit inside the PSU - yes, even the T175 would.

            The only question that still remains is what kinds of core loss I'll be looking at if I tried these alternative cores. Again, the choices are:

            Magnetics Kool-Mu 77548-A7:
            https://www.ebay.com/itm/261329798516?

            Fair-Rite Products Corp. P/N: 5961002701 (Digikey P/N: 1934-1595-ND):
            https://www.digikey.com/en/products/...002701/8599642

            T130-52:
            https://www.ebay.com/itm/183925967793?

            T130-26:
            https://www.ebay.com/itm/254557664194?

            T131-26:
            https://www.ebay.com/itm/251887121050?

            T157-26:
            https://www.ebay.com/itm/251850983867?
            .
            .
            .
            At least those are the choices I have that are readily available right now. Just wondering which one of these I should try experimenting with first, though. I'm leaning towards the -26 cores, as they seemed to be "tried and true"... and I can always try to re-use them in some other PSU, if bad comes to worse. The T130-52 would be nice if it was cheaper... but I'm also considering it nonetheless. I feel like the Fair-Rite and Kool-Mu would be the more "risky" alternatives.
            Attached Files

            Comment

            Working...
            X