Melb Cup PSU fun

[steenss]

Hi everyone. Nice site. By way of intro I thought I'd post a mod on a PSU I got from a colleague who was having issues with his PC. I gave him a CM RS-430-PCAR (Hipro 400W) in exchange for this Hytec HYT-550A:

hk400-1.jpg (180.68 KiB) Viewed 29307 times

Secondary side caps had me initially thinking they were all Teapo SC given the 2x 3300uf 10V on the RHS (3.3V):

hk400-2.jpg (229.98 KiB) Viewed 29307 times

On inspection, we see these lovelies:

hk400-3.jpg (128.27 KiB) Viewed 29307 times

I replaced the following:
3.3V 2x Teapo 3300uf 10V -> UCC 3300uf 6.3V KZG
5V 2x KSC 2200uf 10V -> UCC 2200uf 6.3V KZG & 3300uf 6.3V KZG
12V 3x KSC 1000uf 16V -> UCC 1500uf 16V KZG
5vsb KSC 2200uf 10V -> UCC 2200uf 10V KMG
1x KSC 1000uf 10V -> UCC 1000uf 16V KZG

I used a 2200uf on the inside of the 5V Pi filter because the shorter can size let it fit better next to the output coil. 3.3V output is via 2x12A shottky diodes, 5V is a single 30A, & 12v are 2x16A. Not too far off its label...

[steenss]

Here's the completed job:

hk400-4.jpg (243.22 KiB) Viewed 29306 times

As you can see the OEM is Huntkey & is built on the ever suspect HK400-12AP. The ERL35 transformer & 680uf primary caps are a little tight for my liking, but at least it does employ a 6A bridge & full complement of input filtering. Coil whine has virtually gone & output voltages seem pretty good under load.

[c_hegge]

It looks great. Nice job recapping it. However, Chemi-con has known manufacturing defects with all KZG and KZJ capacitors, so they may not hold up any longer than the original caps. KZE and KY would probably be better replacements and they aren't affected.

[steenss]

Heh, I thought someone would comment on the KZGs. I've got a few & didn't want to use better caps on this older topology PSU. This will at best be a test bench PSU. I got around to a number of re-cap projects today - all precipitated by my dieing home server motherboard. Damn you nVidia...

One thing I didn't initially notice was the T8A fuse rating on the label. That's serious juice for this PSU. The secondary 12V rails are also genuinely split. Now all I need is a load tester & a CRO.

[c_hegge]

Fair enough then. The generous fuse rating is used to allow for the inrush current when the two big capacitors charge. That often draws a lot of current, but only for a second. Usually there is an NTC Thermistor in the input filtering, which limits the initial inrush current. Some manufacturers cheap out by using a smaller NTC Thermistor (or in some cases, leave it out altogether) and using a higher rated fuse that will be OK with the inrush.

[shovenose]

Nice job on the recap
Sometimes (always?) I put in the caps I have; I hardly ever buy capacitors new.

[steenss]

Thanks fellas. It does employ NTCs, the fuse is also overspecced due to the half-bridge primary for 110V mains. BTW, the PSU is starting to chirp a bit. So much for the KZGs . Feedback loop?

[c_hegge]

A chirp or whine usually means that the feedback loop voltage is oscillating. That can be caused by failing caps (or inappropriately spec'd caps).

[steenss]

Thanks fellas. It does employ NTCs, the fuse is also overspecced due to the half-bridge primary for 110V mains. BTW, the PSU is starting to chirp a bit. So much for the KZGs . Feedback loop?

Correction. It's a single forward primary with provision for passive PFC. Must've had half-bridge fever...

I was rebuilding my home servers & using this PSU for some board testing. The oscillating stages (coil whine & chirping) was getting worse. The system (C2Q) wasn't prime95 stable with mild OC, all memory banks full, unless mem was run 1:1 ratio. OC seemed OK with 2 modules. Had the iron out for another project so tore down the PSU & replaced all KZG with Rubycon MBZ from a circa 2005 motherboard. I used symmetrical 2200uf caps on the 5V this time. I also replaced the smaller value PCB electrolytics. These were BH branded SH series 105C, 2x220uf 16v, 1x47uf 50v, 1x22uf 50v, 2x10uf 50v & 2x4.7uf 50V with Teapo SEK 105C. Didn't bother with the daughter card. Those fast recoveries on the 12v secondary also looked more like 18As, but hard to make out without removal. Result is a silent PSU. Load voltages 12.01v, 5.03v, 3.32v, & sb 5.05v. OC with all mem channels (DS RAM) is prime stable. 1600 FSB @ stock volts is prime stable. Looks like it's a good OCing C2Q. 24C ambient, idle temps ~27C-29C, prime95 ~44C-47C. Air cooled with Tuniq Tower @ 1400RPM. Mostly pointless as I'll probably use Antec Neo Eco 80+ & underclock/undervolt. Just shows how bad/mediocre caps ruin the show... (Edit:Corrected temps)

[c_hegge]

Yeah. It's a real shame that Chemi-con didn't get it right with the KZG and KZJ. A lot of manufacturers were caught out with them, thinking they were doing the right thing and using good caps.