Help: Looking for guru input on upgrading an S12II-based PSU

[em132]

The supply in question is an unused/"NOS" Antec EA-500 "80-Plus" (not the "Bronze") with Seasonic's early "500W" S12II board. While I don't yet have this supply in hand, I fully expect it to be filled with OST RLS and RLX. I plan to replace these with FR/FS Panos (mostly).

Of course, I also plan on replacing the big kahuna on the primary side. To that end, I'm currently holding a pageant to determine who'll wear the tiara. Here are the four finalists:


The ravishing "Lady in Red" is the current crowd favorite, largely because of her slight "Hours" advantage and the fact that her hi-freq Ripple rating is 100kHz (!). The underdog CC is closest to the original's capacitance and still has a respectable hi-freq Ripple rating of 50kHz. The two Rubys have more junk in da trunk, which might help with hold-up times, but I'm wary of the effect this could have on inrush Amps, and whether the main PFC coil could handle it. Your thoughts will help crown the winner. (And yes, momaka, adding a film 450VDC 0.22µF "lil' helper" is also in the cards.)


As for the all-important secondary swaps, I want to put in some Pano goodness wherever I can. However, I did read up a bit on PSU recaps and have taken heed that too low an ESR can negatively affect SMPS regulator circuits, and that some PSUs are more prone to this than others. Is this the case with S12II boards? For instance, the +12V "pre-coil" (C502) is a 3,300µF D12 RLX with an estimated ESR of 21mOhms (please correct if that's wrong). I have a few D12 Panos that would fit nicely in there: 3,300µF, 3,900µF & 5,100µF, all with yummy 10kHr. & high ripple ratings. Thing is, their ESRs are all in the 12 to 15mOhm range. With regards to the 2 "post-coil" "dual rail" caps (C503 & 504), the originals are 1,000µF D8 RLS with an estimated ESR of 41mOhms (again, please correct me if that's wrong). D10s supposedly can squeeze in these spots, and I had planned on putting 2,200µF FS's in there, but those have an ESR of 18mOhms. Would there be a problem? Same ESR question/concern would apply to the other rails:




I also plan on addressing the issue of "The-Lil'-Q901-That-Could" and its proclivity for caramelizing its immediate surroundings. My 1st go-to hack would be to slip one of these on the toasty little sucker. However, I'd be open to just swapping it with an electrically identical part in a more thermally-friendly TO-126 or TO-220 casing. Suggestions?


Also, this part right here:


I'm guessing that's a MOV? If so, is it a worthy candidate for a more robust upgrade? And (again) if so, what specs would you recommend?


Finally, your thoughts on using "SMPS rated" ceramics instead of 'lytics in the high(-ish)-voltage/low-capacitance spots (e.g. 25V-50V, 1µF-22µF)?

TIA!

[em132]

Alright, so you can forget about that last point (ceramics).

And I now know that's an MOV in the 3rd pic., and that I should just choose one with a 275V (or 300V???) "max. AC Volts" rating.

I'm about 95% set on using the WE as the primary cap.

That still leaves me with my ESR questions/concerns and what larger equivalent NPN device might fit at Q901.

[dmill89]

While there certainly isn't any harm in replacing the primary cap (as long as what you replace it with has an equal or higher capacitance and voltage rating, within reason of course), all the Seasonic built Antec Earthwatts PSUs I've come across have had a Japanese cap on the primary so it is probably fine as long as it isn't an insanely high-hour unit. When it comes to the RLS caps you're probably close enough ESR wise to be ok, for the RLX though that is a pretty significant difference, I'd definitely want to put it on an oscilloscope with a load before using it in a system with that big of a change. I've recapped a 430W EarthWatts (Seasonic built), but I used Chemi-Con KY and KZEs that were pretty close in ESR to the originals(and what you'll generally find in "real" S12IIs), so I can't comment too much on how drastically lower ESR caps affect these.

[momaka]

The supply in question is an unused/"NOS" Antec EA-500 "80-Plus" (not the "Bronze") with Seasonic's early "500W" S12II board. While I don't yet have this supply in hand, I fully expect it to be filled with OST RLS and RLX. I plan to replace these with FR/FS Panos (mostly).

Of course, I also plan on replacing the big kahuna on the primary side. To that end, I'm currently holding a pageant to determine who'll wear the tiara.

I'm about 95% set on using the WE as the primary cap.

If you feel confident in it, then go for it.
I personally haven't seen WE electrolytic caps widely used in PSUs yet. So with not much of a track record to refer to, I prefer to stick to the proven Japanese brands, whenever possible / available.

But again, since this is a NOS PSU, the original primary cap should have pretty much zero "wear" on it. So perhaps just adding a PP or ceramic "helper" cap in parallel with it should do the trick. Speaking of which, this helper cap doesn't have to be exactly 450V or 2.2uF. This is what I went with, since this is what I had in stock. I think anything in the range of 0.047 uF to 4.7 uF (or even higher capacitance? more is better?? ) should do the trick. 450V rating might be a bit low, actually, depending on the specific PP cap specs. 500V or 630V should do nicely, though.

By the way, I'm curious to know what is the original primary cap in your unit.

In my EA-500, it was a 400V 330 uF Hitachi HP. I Replaced it with a 450V, 500 uF Nichicon 85C cap and it's still OK (though admittedly, I haven't used the PC much in which this PSU resides since ~2019.)
I *think* there shouldn't be a problem even with the highest capacity caps from your list... even though that large Rubycon @ 680 uF is ~2x more than the original.

To be on the safe side, perhaps stick with the 390 uF UCC (or 470 uF WE if already determined on such.) I'm not sure how much the hold-up time can be improved with added capacity, since it seems that CM6800 -based APFC designs are programmed to turn Off operation of the APFC rather quickly after the AC is removed, which then disables the main PWM switching. At least from my experiments, I have observed that the hold-up time is still very short, even with a very light load on the PSU output. On PSUs with no or passive PFC using similar forward-converter topology, the hold up time is usually much longer and proportional to the load.

As for the all-important secondary swaps, I want to put in some Pano goodness wherever I can. However, I did read up a bit on PSU recaps and have taken heed that too low an ESR can negatively affect SMPS regulator circuits, and that some PSUs are more prone to this than others. Is this the case with S12II boards?

Given that the PSU uses OST RLX caps (supposedly motherboard grade... though as you can tell, they are rather lame in terms of specs) and that the output inductors use Micrometals type -52 core material, I suspect Panny FR and FS caps should be fine. In fact, I think I used a Nichicon HN somewhere (5V or 3.3V rail) in my EA-500 as well as Panny FM 6.3V 2200 uF caps in both my EA-430 and EA-500 units. For reference Panny FM @ 6.3V 2200 uF has the same ESR and RC ratings as the 6.3V 2700 uF FR, 10V 2200 uF FR, and 10V 2700 uF FS.

What's interesting is that the OST RLX series datasheet does not have an entry for a 10V 2200 uF cap. But going by case size (10x20 mm) and the 10V 1800 uF cap entry on the RLX datasheet, this suggests the ESR of the 10V 2200 uF RLX should be around 18 mOhms ESR. As such, the above mentioned caps should be fine.

For instance, the +12V "pre-coil" (C502) is a 3,300µF D12 RLX with an estimated ESR of 21mOhms (please correct if that's wrong). I have a few D12 Panos that would fit nicely in there: 3,300µF, 3,900µF & 5,100µF, all with yummy 10kHr. & high ripple ratings. Thing is, their ESRs are all in the 12 to 15mOhm range.

Well, again, note that a 3300 uF cap is not listed in the table of 16V caps. So going by the case size of the 16V 2200 uF RLX and guesstimating a little, the ESR should be somewhere around 16-19 mOhms... which isn't too far from 12-15 mOhms.

With regards to the 2 "post-coil" "dual rail" caps (C503 & 504), the originals are 1,000µF D8 RLS with an estimated ESR of 41mOhms (again, please correct me if that's wrong). D10s supposedly can squeeze in these spots, and I had planned on putting 2,200µF FS's in there, but those have an ESR of 18mOhms. Would there be a problem?

I think the ESR of the D8 16V 1000 uF RLS is might be slightly lower than 41 mOhms. Note again that there is no entry for a 16V 1000 uF part in the RLS datasheet. The same case size only lists a 680 uF part. Thus, for 1000 uF, the ESR should be a little lower, but probably not more than the 38 mOhm of the 10x16 mm part... In any case, let's not fret over 3 mOhms of difference.

Now, I can't say if going with the 2200 uF FS's and lowering the ESR that much would cause an issue or not. My guess would be, no. But since I didn't replace the post-coil RLS caps in both of my units, I can't say with 100% certainty.
To be on the safe side, why not go with the 16V 1000 uF FR or 16V 1200 uF FS? (Both have 8x20 mm case size option.) Their ESR is "close enough" to that of the RLS. Same goes for the 10V versions of FR and FS.

And I now know that's an MOV in the 3rd pic., and that I should just choose one with a 275V (or 300V???) "max. AC Volts" rating.

If PSU was NOS, just leave the "old" one. It hasn't seen any voltage spikes, so it will be fine.
Might want to put heat-shrink tubing around it, though (it helps contain exploded pieces, should the MOV catch a high-power spike and short-out.)

That still leaves me with my ESR questions/concerns and what larger equivalent NPN device might fit at Q901.

For Q901 (H8050 in my unit... or just *8050 NPN BJT), just about any NPN BJT that meets or exceeds the V_ceo, V_cbo, and I_c (continuous Collector current) ratings should work here, since the transistor is used as a linear regulator. Thus, something like 2SD882 (or equivalently, any *d882) should be fine, provided you intend to put a heatsink on it (since it still has to dissipate the same amount of power as the original TO-92 BJT and likely would still get very hot.) You could also attach it to the secondary heatsink (but just beware that you will need a silicone insulation pad, since most d882 transistors have an exposed metal back that is tied to the Collector pin.) Alternatively, you could also go for a TO-220FP (fully insulated) transistor and directly attach it to the secondary heatsink without the need for a silicone pad. I don't have any suggestions for a generic TO-220FP substitute off top of my head. But looking through my parts bin (or rather my parts index), I have a 2SD1266 NPN BJT, which is probably overkill but should still do the trick.

[em132]

Thank you both for the replies.

It looks like I won't be getting this supply until mid/end-April. My cousin, who was supposed to bring it to me last week, was forced to reschedule his trip. Oh well, at least it gives me more time to plan this project.

If you feel confident in it, then go for it.

I decided not to use it after all (I eliminated all four contestants, actually). I looked up the specs of that HP3 (safe bet that's what's going to be inside mine) and it lists an already good ripple rating of 2.19A @ 120Hz. It would be dumb to "upgrade" to a cap with a ~12% lower ripple. Thankfully, I also have a 470µF VXH on hand that's rated at an only slightly lower 2.14A--much more of a legit "upgrade." And I agree about going overboard on the capacitance front--470 should be more than enough at this power level. Besides, that 680 VXT beast would be of better use elsewhere at some point.

Speaking of which, this helper cap...

I have on hand a 0.33µF/450VDC/110°C, and a 0.1µF/630VDC/105°C; both are Panasonic ECW-F(D). Which would you use?

But going by case size...

That's how I arrived at my guesstimates too. Within a specific series, imp./ESR (and even RC) do seem to be directly related to case size.


To make this easier (and since there's no rush), let's try this approach:

RED = Original
GREEN = Candidates (9k/10k = hours @ 105°C, ripple given in A @ 120Hz/100kHz)


+5V & +3.3V
C402/5: RLX, 2,200µF, 10V, D10, 20mΩ

FS, 2,700µF, 10V, D10, 18mΩ, 10k, 1.98/2.47

C403/6: RLS, 1,000µF, 10V, D8, 41mΩ

ZLJ, 1,000µF, 16V, D8, 41mΩ, 10k, 1.18/1.96
FS, 1,800µF, 10V, D8, 30mΩ, 9k, 1.25/1.56

+12V
C502: RLX, 3,300µF, 16V, D12.5, 21mΩ

ZLJ, 3,300µF, 16V, D12.5, 18mΩ, 10k, 2.56/3.66
FS, 3,300µF, 16V, D12.5, 15mΩ, 10k, 2.55/3.19
KY, 3,900µF, 16V, D12.5, 17mΩ, 10k, 2.51/3.35
FS, 3,900µF, 16V, D12.5, 13mΩ, 10k, 2.9/3.63

C503/4; RLS, 1,000µF, 16V, D8, 41mΩ

ZLJ, 1,000µF, 16V, D8, 41mΩ, 10k, 1.18/1.96
FS , 1,200µF, 16V, D8, 30mΩ, 9k, 1.25/1.56

-12V
C506: RLS, 1,000µF, 16V, D8, 41mΩ

ZLJ, 1,000µF, 16V, D8, 41mΩ, 10k, 1.18/1.96
FS , 1,200µF, 16V, D8, 30mΩ, 9k, 1.25/1.56

C507: RLS, 220µF, 16V, D6.3, 130mΩ

ULD, 270µF, 16V, D8, ??mΩ, 10k, 0.23/0.33

+5Vsb
C122: RLX, 3,300µF, 10V, D10, 22mΩ

FS, 2,700µF, 10V, D10, 18mΩ, 10k, 1.98/2.47
FR, 3,300µF, 10V, D12.5, 18mΩ, 10k, 2.08/2.6

C123: RLS, 220µF, 16V, D6.3, 130mΩ

ULD, 270µF, 16V, D8, ??mΩ, 10k, 0.23/0.33
ULD, 220µF, 10V, D6.3, ??mΩ, 10k, 0.15/0.21

----------
There... Have at it!

If PSU was NOS, just leave the "old" one

I was thinking of using something with a higher joules rating. I found a Panasonic that might fit the bill:

ERZ-E14A471S1

It's 6.6mm thick, so it should fit.


And finally, re. Q901...

I still have a 0.25A ADDA I pulled from a SS-350ES a while back (same fan used in the EA-500). I found that this fan pulls about 0.08A at 4.1V (the base voltage from the S12II fan controller). The fan I plan to use (Sanyo 109R0812E402) only pulls 0.05A at that voltage. This means the poor lil' Q902 should have about ~36% fewer watts to contend with at low loads (0.21W vs 0.33W). I think using the slip-on heatsink should suffice as a remedy in this case. Yes, it'll still run hot, but I doubt it'll be as prone to brown its surroundings.

[em132]

Almost set on how I'm going to proceed.

Started testing and cherry-picking the caps I'm going to use. The big 470µF VXH tested at a low-ish 400µF--within the 20% but still, why do the big Rubys test so low so often?


Gonna use the Panasonic FS 2,700µF at C402 & C405...

the Ruby ZLJ 1,000µF at C403, C406, C503, C504 & C506...

the Nichicon ULD 270µF at C507 & C123...

and the Panasonic FR 3,300µF at C122.


The only question now is what to use at C502, the ZLJ or the KY. I'd really like to use the Chemi-Con due to its higher capacitance, but it measures a bit high at 4,360µF--still within spec, but I recall reading in other threads how lytics that measure high is not always a good sign (though it IS rather tall (39mm), which might explain the healthy capacitance). The Ruby's measured capacitance is pretty much dead-on with its rating.

[em132]

Cousin came in early, last Thursday!

The PSU was still shrink-wrapped. Unboxed it and plugged it in using one of those LCD testers and a spared HDD... Seems to work OK. Cracked it open and sure enough there's that Hitachi on the primary. The rest of the 'lytics is an even mix of OST and Pce-Tur. C502 (RLX) is bulging slightly despite having never been used . Regardless, it's getting a FULL recap (minus the small ceramics). Also, that "yellow" gunk is starting to look mighty brownish and brittle. so I guess I'm going to have to play dental hygienist again.

Still on what to use at C502. I REALLY want to use that big Chemi-Con, but am leery about its high measured capacitance.

[momaka]

Almost set on how I'm going to proceed.

Looks good to me. 👍

The big 470µF VXH tested at a low-ish 400µF--within the 20% but still, why do the big Rubys test so low so often?

I find this quite common with large HV caps. Not sure if it's something to do with how (my cheap) meters test these or if the capacitance really is that low.

The only question now is what to use at C502, the ZLJ or the KY. I'd really like to use the Chemi-Con due to its higher capacitance, but it measures a bit high at 4,360µF--still within spec, but I recall reading in other threads how lytics that measure high is not always a good sign (though it IS rather tall (39mm), which might explain the healthy capacitance). The Ruby's measured capacitance is pretty much dead-on with its rating.

The KY is reading only about 12% higher in capacitance, which is indeed quite well within spec.
If the cap is new, this, this might be normal. However if the cap was used (and possibly abused?) before, perhaps check the leakage current. High leakage current is one thing that can make a capacitor show higher-than-normal capacitance on a meter.

To check the leakage current in a simple DIY manner, connect a 1k resistor and a multimeter (in current metering mode) in series with the cap positive terminal, and connect all of this to a DC power supply - preferably as close to 16V as possible. Let the cap charge for a minute or three and then read the current going through the cap. This value is the leakage current. According to KY datasheet, the leakage current should be less than 3 uA or 0.01CV, whichever is greater. In your case, 0.01CV will likely be greater, since C=3300 uF and if the test DC voltage is 15V, then the leakage current for the cap should be less than 0.01*3300*15 = 495 uA.

I also suspect your KY cap is fine... but if you want to take the extra step above, it can't hurt, right?

The PSU was still shrink-wrapped.
...
C502 (RLX) is bulging slightly despite having never been used .

For many older aqueous-based low-ESR caps, especially non-Japanese brands, it's actually worse when they sit on the shelf unused for a long time rather than being used in a circuit (and taking a charge.) The low-voltage offerings (4V and 6.3V) seem to suffer from this even more.

So yes, this is not uncommon sighting.

I bet you it will read really (abnormally) high capacitance too. That's a sign of internal high leakage current, which happens when the electrolyte starts to react with the foil.

Also, that "yellow" gunk is starting to look mighty brownish and brittle. so I guess I'm going to have to play dental hygienist again.

Oh, fun!
It really only matters that you remove it from the primary side, though (as that's where the high-voltage can cause more of an issue if it leaks somewhere it shouldn't.) The APFC toroid is definitely one place. IIRC, I partially unwinded and rewinded the APFC toroid on my EA-430 just to get rid of the glue, as it looked like it could go through the varnish eventually.

Still on what to use at C502. I REALLY want to use that big Chemi-Con, but am leery about its high measured capacitance.

Well, it can't be worse than an already bulging RLX, can it?
But hey, if you want peace of mind, then just go with the ZLJ. Going with the original-spec capacitance of the circuit is a pretty safe bet already.

I still have a 0.25A ADDA I pulled from a SS-350ES a while back (same fan used in the EA-500). I found that this fan pulls about 0.08A at 4.1V (the base voltage from the S12II fan controller). The fan I plan to use (Sanyo 109R0812E402) only pulls 0.05A at that voltage. This means the poor lil' Q902 should have about ~36% fewer watts to contend with at low loads (0.21W vs 0.33W). I think using the slip-on heatsink should suffice as a remedy in this case. Yes, it'll still run hot, but I doubt it'll be as prone to brown its surroundings.

Using a lower-current fan is a good way to curb this. However, also compare the CFM ratings of the two fans, if possible. Loosing CFM due to a slower-turning fan will make the PSU heat up more, which will in turn bump up the voltage going the fan, which will in turn bump up the current it uses. So Q902 might not necessarily run cooler, except at the starting / cold PSU temperatures.

If you have any small Schottky diode, you could do what I did and help Q902 by injecting 5V (minus a Schottky diode voltage drop) from the 5V rail to the fan (+) terminal. I actually didn't even use a Schottky diode, IIRC, since I didn't have a small SMD one to fit where I put it. But even with that, the fan was getting about 0.1 to 0.2V higher voltage, IIRC, and helping Q902 slightly, at least at the lowest starting speeds.

[em132]

More info/pics to come...

(More testing to be done... Going very well so far.)

[dmill89]

More info/pics to come...

(More testing to be done... Going very well so far.)

That appears to be the Delta built version of the Antec Earthwatts rather than the Seasonic built one (not that it matters much as both were decent units), the U-shaped slide off outer casing (that contains the top, back, and bottom with separate sides) is the giveaway (as is the green paint, they switched to Delta before going to the green paint, so some of the Delta built units are grey as well, but I've never seen a green painted Seasonic bult Earthwatts.


Here's one of the Seasonic built ones (an EA-380 but they're all the same basic design, just beefier components on the higher wattage ones):

[em132]

Oh, it's an S12 board alright. Look closely at the heatsink arrangement... It doesn't match the Delta boards.
I really went on this thing.

[dmill89]

Oh, it's an S12 board alright. Look closely at the heatsink arrangement... It doesn't match the Delta boards.
I really went on this thing.

Interesting, so I'm assuming some kind of "Frankenstein" with a Seasonic board in a Delta case and different heatsinks? Or is it some very uncommon version of that PSU (it doesn't match internal pics of any of the "normal"/common variations when I looked them up)?


"Standard" EA-500 (Seasonic):



EA-500D (Delta):



EA-500D "Green" (Delta):

[em132]

Seasonic EA-500 board (S12II) + casing from a dead--yet still new-looking--EA-380D from the late 2000s (had to "ovalize" the board's mounting holes a tad to make them align with the casing's tabs). I think the whole thing looks majorly badass with that black finger guard.

Seasonic's trademark "fingered" heatsinks are still there; they've just been "augmented" with cheap pieces of heatsink stock (from AE), cut to size and bolted and glued with thermal glue. Trust me, they are REALLY stuck on there.

I also wanted to modify the whole thing to function solely as an SKU equivalent supply for Lenovo 12VO boards (with the 14-pin headers)... And since the S12 is group reg'd and that there are no 3V5V lines going to the motherboard, I had to internally preload the +5V at around 3.2W (using a 35W 7.5Ω TO-220 resistor) for the regulation on the 12V rails to remain within ATX specs (assuming a ~280W max total load on the 12V rails), in case I were to ever leave the SATA power cables unused. If you look closely you can see the resistor mounted on the middle heatsink, immediately behind the fan. One nice benefit from this Lenovo-specific mod is that it let me cull a bunch of cabling from the secondary side, allowing that whole area to "breathe" easier.

I also swapped the AC filter board for an enclosed 250V/15A Shurter module, put in a lovely Arctic P8 Max fan (with a custom "Antec" hub sticker designed by yours truly), a removable 10A BS1362 fuse (mounted via 1/4" fuse clips), a 350 Joule MOV, a 10ga grounding cable (got the idea from those old Tagan PSUs), and some sleeving on the MB, CPU and grounding cables.

But I'm getting way ahead of myself here. I still want to test its stability and how it handles crossloads before I come up with a more elaborate write-up.

[dmill89]

Seasonic EA-500 board (S12II) + casing from a dead--yet still new-looking--EA-380D from the late 2000s (had to "ovalize" the board's mounting holes a tad to make them align with the casing's tabs). I think the whole thing looks majorly badass with that black finger guard.

Seasonic's trademark "fingered" heatsinks are still there; they've just been "augmented" with cheap pieces of heatsink stock (from AE), cut to size and bolted and glued with thermal glue. Trust me, they are REALLY stuck on there.

I also wanted to modify the whole thing to function solely as an SKU equivalent supply for Lenovo 12VO boards (with the 14-pin headers)... And since the S12 is group reg'd and that there are no 3V5V lines going to the motherboard, I had to internally preload the +5V at around 3.2W (using a 35W 7.5Ω TO-220 resistor) for the regulation on the 12V rails to remain within ATX specs (assuming a ~280W max total load on the 12V rails), in case I were to ever leave the SATA power cables unused. If you look closely you can see the resistor mounted on the middle heatsink, immediately behind the fan. One nice benefit from this Lenovo-specific mod is that it let me cull a bunch of cabling from the secondary side, allowing that whole area to "breathe" easier.

I also swapped the AC filter board for an enclosed 250V/15A Shurter module, put in a lovely Arctic P8 Max fan (with a custom "Antec" hub sticker designed by yours truly), a removable 10A BS1362 fuse (mounted via 1/4" fuse clips), a 350 Joule MOV, a 10ga grounding cable (got the idea from those old Tagan PSUs), and some sleeving on the MB, CPU and grounding cables.

But I'm getting way ahead of myself here. I still want to test its stability and how it handles crossloads before I come up with a more elaborate write-up.

Nice, with those modified heatsinks it kind of looks like one of the old pre-OCZ PC Power & Cooling Silencers (also SeaSonic S12II based but heavily modified).

This is a 750W PC Power & Cooling Silencer, of course this is built on the larger version of the platform (and some components are relocated to make room of the massive heatsinks), but there are definitely some resemblances, I haven't seen any good pics of the smaller Silencers (all I could find were pics of a stripped board missing the primary caps and a heatsink for a 500W unit) but based on the layout and heatsink style would have likely looked pretty similar to what you put together.

[em132]

Here, for your pic collection (do those innards look familiar at all?):



PIC CREDIT

EDIT: Might as well post a few more pics of my newborn while I'm at it:

[momaka]

Nice work! That is one helluva custom unit!!
The white silicone on the new parts (caps) looks as if this was completely done at the factory... and why those shots in post #9 above also fooled me to think your unit wasn't a Seasonic S12II build, just like dmill89 thought too.

[em132]

So, it's been a short while since my last post. (Hey, life happens; contrary to popular belief, one can't survive on caps alone.)

Anyway, I finally got around to firing this thing up and... I'm glad to report that my house did not explode (whew!). Seriously, you never know after a custom job like this. I left it running for about an hour... No sparks, no magic smoke, no funky smell, no space-time portals unexpectedly opening up... All good!

I then (loosely) tested how jumpy the rails are when connecting/disconnecting four HDDs on the SATA plugs, using my trusty UT61E's Pmin/Pmax function. First I made sure to load the 12V with a 7Ω resistor (on the motherboard plug). Worst results I got was ~0.4V drops on the 5V when attaching each new drive. The 12V rail hardly budged. No doubt the preload resistor is helping quite a bit here.

Next step was putting it in its new home: a modified Lenovo M58 case. PC booted up nicely (again, whew!). I ran a couple stress apps (Prime95 & StressMyPC) for an hour straight: No BSODs (once more, WHEW!), CPU & GPU temps never broke 70°C and the air coming out the PSU barely got warm. Ambient temp was ~25°C/78°F. I'm happy with these results.

Mission accomplished. This PSU should (SHOULD!) be stable and should (SHOULD!) last for years (then again, it might wind up nuking the entire US East Coast ).


Summary of mods/upgrades:
- Delta-made case from an Antec EA-380D "Green"
- Arctic P8 Max 80mm fan (presumably with two NMB ball bearings)
- Schurter 15A filter module
- 10A BS1362 fuse mounted in 1/4" clips
- Panasonic 350J MOV
- Kemet X2 cap
- Rubycon VXH 470µF/450V/105°C (5k hrs.)
- Panasonic ECW-F(D) 0.33µF/450VDC/110°C "helper" cap
- Swapped ALL other 'lytics with a mix of CC, NC, Panny and RC (all 10k hrs. )
- TDK/Epcos 1µF/630V PFC/primary film cap
- Replaced the 3 other small fims caps with Vishay units
- Added a Stackpole Elec. 7.5Ω/35W power resistor to preload the 5V rail
- Slip-on heatsink on Q902
- Increased thermal mass on all 3 main heatsinks
- Conversion of secondary output to Lenovo's 12VO/14-pin board specs
- Added 10ga ground cable
- Cable sleeving
- And last but not least, a fancy new label


The rest of my pics:




Thanks again to dmill89 and momaka for the advice; I'm glad I kept the ESRs as close to the originals as possible (one less thing to worry about).

Oh, and in case anyone's wondering, it now weighs in at a fairly hefty 3.8 lbs.

[sam_sam_sam]

To check the leakage current in a simple DIY manner, connect a 1k resistor and a multimeter (in current metering mode) in series with the cap positive terminal, and connect all of this to a DC power supply - preferably as close to 16V as possible. Let the cap charge for a minute or three and then read the current going through the cap. This value is the leakage current. According to KY datasheet, the leakage current should be less than 3 uA or 0.01CV, whichever is greater. In your case, 0.01CV will likely be greater, since C=3300 uF and if the test DC voltage is 15V, then the leakage current for the cap should be less than 0.01*3300*15 = 495 uA.

Thank you for explaining how you do capacitor current leak testing and what the parameters are and what to look for