The Eurocases PSU is full of bad caps - Hec ones. Mine has even tougher job - it has to run entierly fanless. Yet powering my server is not so hard task, the mobo ECS P6STMT and 14W 1200MHz VIA C3 Nehemiah CPU aren't so demanding... Yet there is 1G of ram - SDRAM. (more images of my fanless PSU mod can be found there: http://ax2.old-cans.com/s.php?p=silentpc&id=27&c=8&d=1&v=v2 ) (and this fanless PSU entered uninterrupted service 8. 17. 2006 and it is still running flawlessly!) Now let's list all the caps used for anyone who want recap that thing: Eurocase PSU ------------------- 2x 680uF 200V d22x46 (can be up to d25x75) 1x 1000uF 16V d10 4x 1000uF 10V d10 (can be 6.3V) 3x 470uF 16V 3x 47uF 50V (max 23V on it) 2x 22uF 50V (max 11V on it) 1x 4.7uF 50V (max 5V on it) 4x 1uF 50V (max 0.16V on it) I have absolutely no idea at first, where to get these big caps: But later that become clear Now the other ones. There is many and only 1000uF ones for filtering the output. And unnecessary 10V ones - for the 5 and 3.3V rails. At least always two of these 1000uF are in parallel to filter better, but... is not 2000uF too small still? The blue ones are Hec too, just 16V ones (one 1000uF as well, as two other 470uF ones - third 470uF 16V one is black...?!). I'm the only one who get the feeling that 1000uF cap aren't nowhere near enought for filtering the 12V rail??? Now let's take a look how the output, that killed the G-Luxon caps and my mosfet on my board, looks, shall we? Not very good, I would say. Note that the spikes sometimes exceed 0.2V ... How can one rate such output in terms of quality on the scale 1 - 10...? So, now about recapping. I sort of giving up ATM for the big ones, but from the small ones, these 1uF and 4.7uF cap's voltages suggested that I could happily use a 16V ceramic SMD 1210 cap there. 10uF one. And I did: After replacing these caps I run the PSU on dummy load for about half a hour and none of these ceramics are damaged, so I think that it is safe to replace them with any 16V caps - ceramics preferably - it last, hehe Now the rest - as Rainbow suggested: I think it will be safe to replace the 22uF 50V caps with a little bit higher capacitances, so the end list should look like that: Eurocase PSU ------------- 2x 820uF 200V d22 - Panasonic TS-ED or Samxon EP 1x 2200uF 16V d10 Samxon GD 4x 3300uF 6.3V d10 Samxon GD 3x 470uF 16V Samxon GC 5x 47uF 50V Samxon GF 5x 10uF 16V ceramic SMD ones I took the time to properly measure with oscilloscope the MAXIMAL spikes I find on the caps. It is for the 47uF ones as follows - 27V, 26V and 15V. Both these 22uF ones has 15V maximal spikes. (in fact, is was 14.7V and such, but I rounded it up... ) Conclusion - a 35V cap is enough there. I won't not recommend to stretch 25V caps there, for safety, but they proably should hold it - maybe Now the filtering. After my bad experience with the PSU I wanted to make sure that nothing bad will happen again. So, what should I do? Block (bridge) each of the big output caps with 10uF 16V ceramics? Will that help enought? Or should I also add in parallel 100nF 16V ceramics to this? What should I do to make the absolutely best results :?: I think I know, but since I got THERE these PSU in my machines, I better ask what will be simply the best solution As for the imput ones, I better take a look and ask. As long as I see there a Hec cap, I'm not happy. What about at http://digi-key.com 493-2519-ND - the Nichion LS cap LLS2D681MELZ ...? It is d22x40 suxxka, 3000h rated at 85°C and can deliver 2480mA (680uF / 200V as need) There are also a Nichicon GU type sold, but this one deliver only 1780mA, while rest of specs are almost same (105°C, tough) Nichicon LS 820uF 200V d25x35 Load life is 3000h at 85°C and nice ripple - 2790mA But d25 and I need d22... Panasonic TS-ED 680uF 200V d22x45 Load life is 3000h but 105°C and ripple - 2420mA Samxon EP 820uF 200V d22x50 Load life is 5000h at 105°C and ripple - 2180mA And a intresting remark. I knew beforehand that my server got BAD CAPS inside. Yet I waited till problems arise with the recap. Now problems arised and my mobo is dead. So, my upgrade was: 3.3V - 330% (2x 1000 to 2x 3300) 5V - 330% (2x 1000 to 2x 3300) 12V - 220% (1000 to 2200) Futhermore I blocked each of these bigger caps with 10uF 16V Murata 1210 Ceramic cap, witch bring the total ceramics used for the PSU recap to nice number 10 now (remember, I already replaced five 1uF and 4.7uF electrolyte caps by these SMD ceramics soldered from back of the PCB, of course) That should make my PSU run great, fanless for five, better six years - I hope. We could even took that as the fanless test of Samxons GD, hehe Futhermore I think that the PSUs aren't that critical on the low ESR caps, so I could get away well using these GD ones - for sure they are at least 10x better that the Hec ones, so... Now it is time to show some pics, tough my work has yet to be completed. So, now my PSU output filtering caps are Samxons there: 4x Samxons GD 3300uF 6.3V. More that enough I think, but it should not stress the regulation more, but IMHO less. Give the mosfets more room to breath with the increased capacity This is the D10 16V 2200uF Samxon GD cap - nice suxxka, is not it? (BTW, the gold color is exactly the same, as on Pannyes FM... coincidence? ) 470uF 16V Samxons GC. Supporting 10uF 16V ceramics caps That should do the trick And yep - after they killed my mobo, I did not trust the Hec caps at all. So - off they go from my PSU! Okay, singe I got my caps, I made the finish of recap of my Eurocase PSU happen. So, in the end, I used all Samxon caps for the end filtering, replaced all the small caps with 10uF SMD ceramics and used five 47uF 50V Panny FC caps as well, as two Panny TS-ED 820uF 200V imput filtering caps: Pretty, aren't they? However the results I got aren't pretty, except 5V rail, tough. On very mild load, 1A from 12, 5 and 3.3V rails it show this: 3.3V rail 5V rail 12V rail And on the top of that, when I added to the mild load two harddrives on the PSU (200G WD and 80G Seagate), the ripple increased notably, hiting 70mV in the worst case on 12V rail. Also the great 5V rail get bad to 26mV, same goes for 3.3V rail, witch doubled it's ripple to 50mV. That is weird. I blocked every the Samxon caps and still getting such crap? It won't help that the new Chieftec PSU got 250mV ripple - it is full of Teapos, so, no wonder... (and Chieftec indeed suxx) Better PSUs could hold these ripple in reasonable limits, for sure anything over 50mV aren't good, so... I did a little experiement - I removed these 10uF SMD ceramic blocking caps on the end caps. And after removing of the ceramics, the ripple got notably better! Eurocase 350W ripple with ceramics: 12V - 70mV 5V - 26mV 3.3V - 50mV Eurocase 350W ripple w/o ceramics: 12V - 54.4mV (11.99V) 5V - 18.5mV (5.11V) 3.3V - 53mV (3.338V) Conclusion. Adding ceramics into PSUs is counter-productive. Indeed they could filter the voltage a bit better, however they does trick the feedback into knowing it delivering still good voltage (the ceramics could hold and even deliver the charge for it, but not for the whole powered PC) and therefore dramaticaly increase the number of switching need as well, as increasing then ripple on load dramaticaly. Don't do it. It is probably still okay to add these ceramics somewhere around the PSU connector on the PC mainboard, as from such distance they could not do harm to the voltage regulation, yet I'm not now entierly sure about that anymore. Anyone care to explain me what I find out? PS. sadly I did not measured the ripple when there are the original Hec caps well - but I fix this on the two other mine Eurocase 350W PSU's waiting for recap Todays I recaped my second of my there Eurocase 350W PSUs. This one is not modified in any way (in fact it was for the first time I ever opened it) and I did what I did to the fanless one - mixed Panny / Samxons recap. End caps Samxons, rest Panny and these 4x 1uF and 1x 4.7uF caps replaced with 10uF SMD ceramics 16V ones. All as with the previous modded fanless PSU - oh, okay, I run out of the Samxon GC 470uF caps, so I used for the there the Panny FM ones - no big deal, I think. These Pannyes are just taller that Samxons, that it is. The better specs of GC Samxons over FM Pannyes aren't important, I believe. Anyway, as I promised, I took the measurments on my dummy load (1A from each 3.3, 5 and 12V rails) + two HDDs first, and then after recapping. PSU was entierly same as the first one, no bulging of the Hec caps visible. There was a lot of dust and stuff, tough. First time opened from 2003 - go figure Voltages and ripple measurments Before recapping on dummy load 12.09V - 16.3mV 5.14V - 16.8mV 3.34V - 6.9mV Now this if ficking amazing! :?: :?: This is great output voltage! I won't expecting that from anything less say Enermax 620W Liberty! How this is possible from such crappy PSU at all?! After recapping on dummy load 12.08 - 47.3mV 5.14 - 24.1mV 3.339 - 24.1mV WTF?!?! I worked so hard and all I get if WORSER results?! Now that make me angry! I don't want my work turned out to worser results that was before! WTF?! Who can explain me this, I don't know what to think... After recapping on Abit ST6-R with Prime stress test running 12.08 - 49.6mV - 4.0kHz 5.14 - 13.6mV - 8.9kHz 3.339 - 8.7mV - 1.8kHz Whoa! That is MUCH better! The 5V line is great, amazing! Now why?! The measurments was taken from the HDD connector (12 a 5V rails) so any caps on board can't play much role - not to mention that this old board did not have 12V support line and take all from 5V line, yet still don't have any input filtering caps (!), so... why any how? I did not understand what is going on, unless... ...unless the fact, that we overshoot these filtering caps at the end (4x 1000uF to 3300uF and 1x 1000uF to 2200uF) means that simply the PSU need some serious load to produce clean woltage now? Anyone who know better or can tip a reason why ripple after recapping on my dummy load does suxx and when runing real computer with this config: ABit ST6R (raid) 1680Mhz Tualatin (120x14) 1.625Vcore 512MB SDRAM 2-2-2-5 120Mhz 160G Maxtor 8MB cache (6Y160P0) PCI GF4 MX440 64MB 80/166 DVDRW NEC DV-2510A (8x DVDRW) IDE 100MB zip Edimax 9130TXA 100MB NIC ...it show very good voltage. 13mV ripple on used for Vcore 5V rail is great and the 8.7mV on mainbard is simply outstanding Almost 50mV ripple on 12V line aren't very good, tough it is just 12V line and used mainly for the other devices that mainboard, so, acceptable. ATX standard says that 3.3 and 5V rails ripple has to be bellow 50mV and 100mV is allowed for 12V rail. And my last third Eurocase 350W PSU was recapped just recently. Used all Samxon caps - Samxon EP 820uF 200V d22 for the input voltage filtering: And the rest was done by Samxons too. Samxons GF 47uF 50V was used to replace the Panny FC ones And the PSU is working wery well now. I did not bother to measure the ripple, as the different input caps could hardly has impact on the output riple and so the 47uF ones changed from Panny FC to Samxon GF The only odd thing is - and that was the case with all previous PSUs as well - noise The PSUs after recap become noisy for like a week or sometimes longer before the coils somewht "adjusted" their whinning to new PSU parameters, lol This does never happen in Enermax PSUs or Antec ones. It is simply a faulty design... But I now know that there are WAY worser PSUs. One word - Mercury.