Seasonic B12 BC-550 – barely 2 years old and with BAD CAPS already!

[momaka]

I know I've been a little scarce lately (like the last 2-3 years), but I'm still here and still doing my thing with fixing PSUs.

For today's considerations, I have a Seasonic B12 BC-550 [A551bcafh] 550 Watt ATX power supply for you (click on links for full size images).

https://www.badcaps.net/filedata/fetch?id=3591771


https://www.badcaps.net/filedata/fetch?id=3591772

It's a modern ATX unit with fixed (non-modular) cables and an 80-plus bronze certificate. Here's the label:

https://www.badcaps.net/filedata/fetch?id=3591773

All in all, it seems like a decent Seasonic unit. Of course, if I'm making a post about it, you can bet there has to be something that isn't quite right with it… something that I was able to spot through the vents at the time of sale. (Some of you can probably tell where this is going.) But for $5 after almost no bargaining, I couldn't pass it up. So let's open it and see what was wrong.

https://www.badcaps.net/filedata/fetch?id=3591774


https://www.badcaps.net/filedata/fetch?id=3591775


https://www.badcaps.net/filedata/fetch?id=3591776

Well, there you have it - bad caps in a supposedly good brand PSU… and in 2024 no less! One might think that good manufacturers like Seasonic would have learned their lesson by now. But I guess nothing's given.

The bad caps are all from Asia'X (i.e. Fuhjyjin Electronic Co.). And looking at the middle picture above, these caps have date codes of “2209” and similar. I don't know if the last two numbers represent the production month or production week – one would specify the later half of 2022 and the other somewhere in early 2022. Regardless, the fact that the caps were made in 2022 and have already gone bad is a BIG SHAME! That's barely over the 2 year mark… and I'm not sure if they even made it that far, because I got the PSU a few months (around late October 2024) with the caps bulging as seen above. So they could have gone bad before then.

Here's the rest of the pictures of the PSU:

https://www.badcaps.net/filedata/fetch?id=3591777


https://www.badcaps.net/filedata/fetch?id=3591778


https://www.badcaps.net/filedata/fetch?id=3591779


https://www.badcaps.net/filedata/fetch?id=3591780

Also, if anyone has this unit and damaged SMD parts (either due to physical damage or something gone wrong with the PSU), let me know – I have some high-resolution pictures that appear to capture a good deal of them. I can upload them upon request (saving the forum some server space, if these are not necessary.)

Now back to discussing the PSU… the electrolytic caps aside, this appears to be a pretty good build.
It's a modern half-bridge LLC resonant topology (note the lack of an output toroid) with synchronous rectification on the output (a single 12V rail). Filtering is provided by a bunch of 16V 470 uF polymer capacitors (not sure of the brand or series) followed by a few (bulged ) 16V 2200 uF electrolytic capacitors. The 3.3V and 5V rails use DC-DC buck-converter circuits (the two small daughterboards.) Jumping back to the primary side… there's two MOSFETs for the LLC resonator, along with two MOSFETs and a diode for the APFC. There is also a relay to bypass the NTC thermistor once the unit is ON for further efficiency improvement (and I personally don't like to see the NTC thermistor as the sole link between the output of the APFC and the primary side electrolytic filter cap, since a blown/open thermistor there can cause the APFC MOSFETs to go *POW* .) Next, the EMI/RFI filtering – it's all complete, of course, as would be expected from a competent manufacturer like Seasonic.
For the Google/search engine bots, PCB has the following info written on it:
YL ATX0500GQ(TC) Rev1.6
2021/08/02

Now if you are wondering, why did I open this PSU and not send it for a warranty repair… that's because I got my screw driver out and opened the PSU before I even thought about checking the model number online for anything like that. As it turns out, this PSU has a 5-year warranty on it. Even the serial number shows it. So I guess we can say I screwed myself out of that. No regrets, though. Sending the PSU in for warranty from where I am right now (Eastern Europe) might cost more than what the PSU is worth new. And for what, to send me a new/refurbished PSU likely with the same crappy caps inside? I highly doubt Seasonic has any repair centers where someone will bother to sit down and repair the actual problem. More likely, they just have warranty depots to send new units out. If that's likely to be the case, I would rather fix the problem myself. Actually, I'm glad I did as I have some interesting findings to show. But I'll detail those in the next post below. This one will just end it with summary of the parts inside.

ICs:
CM6502UHHX (APFC), CM6901X (LLC PWM ctrl.), Welltrend WT7527V (supervisor), [STR-]A6069h (5VSB PWM-FET IC), two APW7073 (buck PWM ctrl., 1x for each DC-DC daughterboard)

Wiring:
Input side: 600V, 18 AWG
Output side: 300V, 18 AWG
Output connectors: 20+4pin ATX, 4+4-pin 12V CPU, 2x 6+2-pin PCI-E, 4x SATA, 3x Molex

Transformers:
35 mm wide main; 19 mm wide 5VSB


Primary Side:
* three 0.22 uF X2 caps, three 332M Y2 caps, two 102 Y2 caps, three CM chokes, one 14d511k MOV between L-N
* MF72 10d15 NTC + 10A125V relay
* 0.5 uF 500VDC metal film cap (APFC input), 2x 683J 1000V film caps (LLC resonator)
* APFC section: 2x ST16N50AF MOSFETs + 1x LTTH806SDFW diode, all in TO-220FP cases; T106-45 inductor core (toroid)
* LLC/main PS switching: 2x ST16N50AF MOSFETs
* Ltec HP, 420V, 390 uF, 25x46 mm, 105°C
* C61: Asia'X TMX, 35V, 220 uF, 8x16 mm (primary aux. supply filter)
* C66: Asia'X TNX, 50V, 47 uF, 6.3x11 mm (5VSB startup cap)

Secondary Side:
* 5VSB
*** 1x Asia'X TMX, 16V, 2200 uF, 10x20 mm before PI coil
*** 1x Asia'X TMX, 16V, 1000 uF, 8x18 mm after PI coil
*** PI coil: 1R5 (7-turns, 24-AWG, 2.5-3 mm core)
*** 45R20s SMD rectifying diode

* 12V Rail
*** 4x polymers, 16V, 470 uF, 10x11 mm before PI coil
*** 2x Asia'X TMX, 16V, 2200 uF, 10x20 mm after PI coil
*** 2x free 10 mm dia. Spots
*** 4x total, 2x on each DC-DC daughterboard: polymers, 16V, 330 uF, 5x11 mm
*** PI coil(s): 2x 3-turn, 16 AWG, 5 mm rod core
*** 2x(?) 044R5NT MOSFETs for sync. rectification

* -12V Rail
*** 1x ChengX GR, 25V, 1000 uF, 10x20 mm before 7912 linear regulator
*** 1x Asia'X TMX, 16V, 2200 uF, 10x20 mm after 7912 linear regulator
*** SS26 SMD rectifying diode

* 3.3V Rail
*** 2x polymers, 16V, 330 uF, 5x11 mm (on output of DC-DC daughterboard)
*** 1x polymer, 16V, 330 uF, 5x11 mm, close to wires
*** PI coil provisioned but not installed
*** 3N5R0 + 3NA3R4 MOSFETs for DC-DC buck circuit

* 5V Rail
Same exact configuration as 3.3V rail above

* other electrolytic caps:
*** C72: Asia'X TMX, 35V, 220 uF, 8x16 mm (secondary aux. supply)
*** C60: Asia'X TMX, 16V, 220 uF, 6.3x11 mm (fan filter)

[momaka]

Seasonic B12 BC-550 – recapped

Let's start the recap with hall-of-shame -worthy pictures of the Asia'X capacitors.

Ah, lovely!
Except for the two 50V 220 uF ones, all the other ones were completely out of spec, as seen above. Even the -12V rail cap, which is sitting behind a linear regulator (LM7912) and does not see any stress or heat at all, still managed to fail on its own. I think that alone speaks volumes about the quality (or lack thereof) of Asia'X caps (particularly the TMX series.) Seasonic, if you are reading these, PLEASE drop these capacitors from your production floor ASAP.

Now let's talk about the recap. Spec-wise (on paper), Asia'X TMX series are nothing extraordinary – about on par with UCC KY/KYB, Nichicon HE, and Rubycon YXG/YXS.
The 16V 2200 uF parts are rated for 0.030 Ohms impedance (ESR) and 1850 mA of ripple current. With that said, here is what I used for the recap:

5VSB: 1x UCC KY 10V 2200 uF (10 mm dia.) & 1x Rubycon ZLH 10V 1000 uF (10 mm dia. despite the PCB being outlined for 8 mm spot)

12V: 1x UCC KZH 16V 2200 uF (10 mm dia.), 1x UCC KZE 16V 680 uF (8 mm dia.), and 1x Ltec LZG 16V 1000 uF (12.5 mm dia.)
I originally had a UCC KYB instead of the KZH, but needed the KYB for matching the ESR specs of the caps in another PSU I was recapping. Since this Seasonic PSU uses resonant LLC topology and already has polymers on the output, I figured it wouldn't care about the lower ESR of the KZH. Unfortunately, I didn't have another one to match it… and seeing all the empty cap spots (likely for the more powerful version of this unit), I decided to “experiment” a little and put the the Ltec LZG and UCC KZE caps there. I do admit that using Ltec LZG is a bit risky here (especially since it's a pulled cap from an LCD TV PSU board… though it did measure and reform OK). But TBH, despite the fact I don't trust Ltec caps all that much, even they seem miles better than Asia'X in terms of quality. I'm also not too worried if the Ltec cap fails. The 680 uF UCC KZE should be able to hold fine on its own with the 2200 uF KZH.

-12V: here I replaced only the Asia'X TMX 16V 2200 uF cap… with a Nichicon PM 16V 100 uF Why? Because the -12V rail output is generated by an LM7912 linear regulator. For anyone that's looked in the LM78xx/LM79xx datasheets, such capacity is not really required anywhere. Typically 25-100 uF on both the input and output are fine if using electrolytic caps. Actually, with higher capacities on the output, a back-feeding / protection diode must be installed between the input and output pins of the regulator to protect it. So the reduced capacity won't affect anything here. The only slightly more worrisome part is that I left the 25V 1000 uF ChangX cap on the input (as an experiment), which might come back to bite my butt… but we will see.

Some pictures of the recap…
https://www.badcaps.net/filedata/fetch?id=3591796
https://www.badcaps.net/filedata/fetch?id=3591797
https://www.badcaps.net/filedata/fetch?id=3591798

Last but not least, the caps that filter the two auxiliary rails generated by the 5VSB circuit: one on the primary (filtered by C61) and one on the secondary (filtered by C72). Both of these had Asia'X TMX, 35V, 220 uF, 8x16 mm caps. Initially, I replaced both by Rubycon YXJ parts of the same voltage and capacitance rating, and thought all was dandy. But remember how I said in previous post that I have some “interesting findings”? Here it goes. For whatever reason, I always have a curiosity of how the auxiliary rails vary as load is added to the 5VSB. So as if by luck, I decided to check this PSU as well. What I found is that the voltage on spot C61 (the primary side aux. rail) can increase quite a bit as the 5VSB is loaded. How much, you ask? - Enough to go over the 35V rating of the cap in that spot.
The picture below shows the voltage on C61 when the 5VSB is loaded with 2.21 Amps. As you can see, C61 measured 37.4V




Now if we have to be realistic, such scenarios where the 5VSB rail will be loaded so high all the time are probably going to be an exception rather than the norm. In most cases, the 5VSB will likely see loads of 1 Amp or less from a relatively standard PC setup. Still, I think that's no excuse to scrape by with the bare minimum here. Furthermore, it's a Seasonic PSU and I just expect better than this! Clearly a 50V cap should have been used in that spot.

So I had to re-do my recap above, replacing the 35V Rubycon YXJ on spot C61 with a (used) Nichicon PJ 50V, 220 uF cap. I then measured the voltage across this cap once more with various loads, just to confirm what I saw. With no load, C61 was sitting at about 20V. Bumping the load up to 350 mA quickly brought the voltage to close to 24V. And for 1 Amp of load, I got about 30V. Going with a higher load beyond this didn't increase the voltage in such great amounts anymore… but with anything 2 Amps or more, it goes over 35V! I didn't bother to test at the maximum load of 2.5 Amps, because I didn't feel like dragging out my large Ni-Cr heating element. I imagine the voltage will probably be around 38-39V with that load.

On the other hand, the voltage across C72 was fairly stable and didn't increase anywhere that much, so I left the 35V YXJ cap there alone.

For the 5VSB startup cap (Asia'X TNX 50V 47 uF) at spot C66, I used Rubycon YXJ again (of the same voltage and capacitance).

And that more or less summarizes this PSU's recap. The only Asia'X cap I didn't bother with was the one at the fan output connector (Asia'X TMX 16V 220 uF), as the fan controller appears to use a linear regulator circuit and it probably won't matter if this cap goes bad or not.

So there you have it – a new lease of life on this 2022-built Seasonic PSU… with (mostly) good cap brands now.
I suppose the race is on whether the Ltec LZG, ChengX GR, or the input Ltec HP 420V 390 uF cap fails first, given the PSU does have APFC. I didn't put my usual “trademark” film polypropylene cap across the main input cap (to help with the high frequency ripple current), so we shall see how this one turns out.

Adios!

[Per Hansson]

Nice post Momaka, I enjoyed reading it! You should send Seasonic an e-mail about it so they can enjoy it too!
P.S: I was surprised about the -12v rail caps too: is there a back-feeding / protection diode installed between the input and output pins of the regulator to protect it or not?
If not maybe you will find more of these with a shorted -12v rail in the future

[stj]

the diode only exists in datasheets, i have never seen one in a real circuit!

[PeteS in CA]

Given polymer caps' low impedance and where those 'lytics were in the output filter circuit, the 'lytics really should not have been terribly stressed. Those Asia'X caps were pretty awful, and a decent P/S mfr should have known to use decent 'lytics for output filtering.