Announcement

**needtoseepics** · 10-15-2013, 08:07 AM

Re: Dressing up (a "450W" ATX supply) for Halloween -- YueLin YLP-013

Here are the before pics, where I'd just taken the PCB out of the housing.
3x from the top, and 3x from the side. I can't take anymore, 'cause it doesn't
look like this anymore (more pics to follow):

Attached Files

**needtoseepics** · 10-15-2013, 08:22 AM

Re: Dressing up (a "450W" ATX supply) for Halloween -- YueLin YLP-013

I got the primary input filter from this Wan Nien Enhance Electronics supply,
and this is where I also got that 6Amp bridge rectifier that I was considering
installing in the YueLin. See what I mean about the pcb near the rectifier itself
and the input coils all seemed fine, but obviously they've had quite the "burn in"
load test (pun intended).

The capacitors on this board are Jamicon and Teapo. The Jamicon seem fine,
but the Teapo's were bulgy. Is Jamicon better or worse than Canicon? No idea.
I removed all the Micon caps from the YueLin, but there are still some Canicons
on it... would these Jamicons from this blown Wan Nien be better than Canicon?

Attached Files

**needtoseepics** · 10-15-2013, 08:31 AM

Re: Dressing up (a "450W" ATX supply) for Halloween -- YueLin YLP-013

Here is the input filter stage installed on the YueLin, with also two "Y rated caps"
in the X cap locations (is this a bad idea.. let me know if you think so). These Y caps
came from the SGI's Lucent supply. Also you can see in this image the thicker
input wires I've installed to the PCB.

Attached Files

**needtoseepics** · 10-15-2013, 08:49 AM

Re: Dressing up (a "450W" ATX supply) for Halloween -- YueLin YLP-013

Here you may see the recap itself, with one additional point.. when I took that
decapped output stage pic, I had only removed the micons, and not yet the
canicons from the output stage. I ended up putting those two "Low ESR"
Canicon 470uF 16V to replace two Micon 470uF 10V on the 5VSB circuit.
I don't know if that's a good idea or not, but I figured I'd remove Micon entirely,
and so this leaves me with only Canicon (presumably not-the-best) and
Nichicon (from a Lucent supply SGI used in 1997ish, so not the newsest,
but also, not fake like apparently we have to worry about when buying "new"),
with only Canicon on the main PCB, and all the Nichicon on the perfboard.
(perhaps I'll replace the canicons with out-of-spec other parts, or eventually
buy new (hopefully not fake) exact-spec replacements, but I'm not done yet,
and just doing output stage here).

So this is where I made it to, and it does power on, and that's how it looked
when I measured the voltages above (in initial thread post). I think the next
thing to do would be replacing the silicon parts (bridge rectifier 4A -> 6A first),
but I was thinking I should probably hit it with a scope first to get an idea
of what's going on. So might be awhile before I return to this thread.

(hint: this is where the comments should happen)

Attached Files

**needtoseepics** · 10-16-2013, 08:26 AM

Re: Dressing up (a "450W" ATX supply) for Halloween -- YueLin YLP-013

Alright, this morning I did a scope test using an HP 16500B. For a light load,
I attached one hard disk, and two additional fans to the YueLin YLP-013.

I took two shots each with the scope set at 1ms, and 10ms. The A1 channel
is connected to the 5V output line, and the A2 channel the 12V output line.
Each channel is offset/centered by/on it's respective "correct" voltage,
5V for A1, and 12V for A2, and the scale is set to 250mV, so each is slightly
under "1 tick" away from center.

I notice nobody's commented yet... perhaps this will spawn some replies.
If there is another test you'd recommend I do to have more information, let
me know, and if I can try it, I will. I think the results here look good, so I'll
likely hook this up to a pc later today and see how it goes, then perhaps try
my bridge rectifier upgrade to the 6Amp heatsinkable model. We'll see.

Attached Files

**ben7** · 10-16-2013, 09:26 AM

Re: Dressing up (a "450W" ATX supply) for Halloween -- YueLin YLP-013

You really should put in better secondary rectifier diodes, better primary side switching transistors, and possibly a better transformer.

I personally don't like the idea of mounting the capacitors like that, but, whatever works for you ... xD

**needtoseepics** · 10-16-2013, 09:22 PM

Re: Dressing up (a "450W" ATX supply) for Halloween -- YueLin YLP-013

Alright, I've picked up a few more "bad" supplies, and I suspect I got
lucky with regards to a compatible transformer, and NPN power
transistors. Now, the YueLin's original primary switching parts are
labeled:
0107
P13009-0

And although I don't know who makes them, I notice that there is an
FJP13009 "High Voltage Fast-Switching NPN Power Transistor" from
Fairchild Semi. which these are likely similar to.

The PCB has wider drilled spacing than what is used, so it will fit the
TO3P package, but the stock part is the TO-220 package, so I ought to
be able to use any NPN power transistor in either of those packages,
and, presumably, if for no other reason than heat-dissipation due to
increased external surface area, I should consider parts in the TO3P
package to be "better". Now, one of the AT supplies I've got has two
parts labeled SKC4138 in the TO3P package, which seems to be SanKen
brand, and according to:
https://cdn.badcaps-static.com/pdfs/...6fe02e2b70.pdf
on page 5, it says C4138 is good for 500Vcbo, 400Vceo, and 10Amps.
Of course, since I don't know the specs of the original part, I don't know
if that's better or worse, but it is in the larger package size, so,..upgrade?

Also, the SGI/Lucent supply contains 2x International Rectifier 541L2
parts. I'm not actually sure if these are even compatible, or even if
they are NPN transistors or not, but they are in a TO3P package, and
it's I suppose possible they are compatible. I'm of course more
likely to fit the SK parts since I at least have some information
about them, and coming from an AT supply it's easy to tell they were
being used for the same thing, whereas with the 3x pcb's inside this
Lucent supply, it's hard to tell what's going on where, but it's possible
that the IR parts are better than the SK and could be used. Any idea?

With regard to the output stage, I think these International Rectifier
30CPQ150 parts I harvested out of the Lucent supply (yeah, they used
IR all over the place in it) will work as "Schottky Rectifier", I looked up the
original parts, the MOSPEC F12C20, and ?PANJIT? SB3040ST, and
it seemed like a match, with the IR part having an improve rating.

Also... like I said, that Lucent supply has several PCB's and uses ribbon
cable and thicker "power cable" connecting them, so I guess that's why
I started thinking "maybe it's okay to mount capacitors on a different PCB
from the control circuitry", since Lucent is reputable (though this particular
supply is dead, hmm... red flag?), and this supply they did that way is
rated for 623W, and the physical dimensions of it are massive compared
to those of a standard "box" sized AT/ATX supply.

@ben7 --
Thanks for the response by the way, but now I feel the need to ask:
Is your reason to dislike this mounting method technical in nature
(because of increased impedence, or extra heat now that the transformer,
heatsinks, and, additionally these caps, are all being in the same place),
purely cosmetic, fire hazard or what?

(now I'll have a lot of desoldering to do before I can continue, but it seems
like I ought to have enough usable parts in this ever-increasing scrap heap
of power supplies to go through a few more stages of this dress-up)

**kc8adu** · 10-16-2013, 10:29 PM

Re: Dressing up (a "450W" ATX supply) for Halloween -- YueLin YLP-013

those long thin leads totally defeat the purpose of low esr caps.
just get proper caps and install on the board.panasonic fm are a good choice.

**needtoseepics** · 10-17-2013, 08:56 PM

Re: Dressing up (a "450W" ATX supply) for Halloween -- YueLin YLP-013

So here are the parts I have harvested so far, and might be good candidates for
upgrade use. I don't really know how to tell if one transformer is better than another
one, but the size of the one I've extracted, labeled DVE E401201-014, is slightly
larger than the one that's installed by default in the YueLin. The pinout is at least
two pins on the input side, and six pins on the output side, with the "out-the-top"
center tap to ground. The spacing on the two input pins is a bit closer together
I thnk than the one that is installed by default, which itself is labeled El-33 4.5mH EL.

Supposing I can "make it fit", I'm guessing it's worth a try and would be considered
an upgrade?

The schottky rectifiers I collected are the (above mentioned) International Rectifier
30CPQ150 units. I also collected these International Rectifier 541L2 parts which
I couldn't find any information on. I'm hoping they are switching transistors, does
anybody know for certain? If not, I do have those SK C4138 which I still haven't
desoldered yet, so, I'll collect them if nobody knows what these 541L2 parts are.

And, as I mentioned before, I've got this KBU 605 6Amp bridge rectifier, which I'll
be testing first because I don't know if it's good or not (you can see where the
char marks are on its legs perhaps in the photo) since the supply I pulled it from
had a burned up input stage. I did pull the stock 4Amp bridge rectifier, A KBL406,
off of the YueLin by now, but I haven't put this one on (need to clean the pins off
a bit more I think and bend them about before I can get it to fit in)

So, that's where I'm at so far.

@kc8adu, how low is the resistance of low-esr caps, because if I calculated it right
(I used the "Wire Resistance Calculator" from cirrus, found here:
http://www.cirris.com/testing/resistance/wire.html
) 22AWG wire for 1.2ft (7in there, 7in back) ought to be adding 0.019 Ohms of
resistance, in series obviously, to the capacitor. So, is this a considerable amount
for these components? Everything I notice seems to say "Low ESR", but.. low,
compared to what? I haven't seen any hard numbers like, for instance, is 1 Ohm
considered "High ESR", is 0.1 Ohm High ESR? If 0.019 Ohms is enough to push
a Low-ESR component out of that status, then I suppose that must be the case.
At what specific numerical value does a component become "Low ESR", and, if
I know that, then I could add 0.019 to the ESR of such a typical component,
and see if it is still low or not.

Not to say I wouldn't buy "the right" parts, but even if I do, with shipping and all
it kinda doesn't let me do anything for a while, and at least at the moment I can
test the other upgrades, and the supply seems usable, and within ATX spec.
From what I was reading, ATX spec says 50mV ripple is allowed/permitted, and
the voltages must be within 5% of the "correct value", so, looking at my scope
results, I'm very much "within spec", at least for this simple test (which I realize
is not representative of an actual computer system with activity), so I don't really
see a problem, yet. If you can propose a test that will demonstrate that my supply
has a technical issue caused by this layout, I'll do my best to carry it out, and
perhaps adapt things to work around the issue.

If not, I'm gonna assume you all just have an issue with the cosmetics of an
additional PCB

Attached Files

**kaboom** · 10-17-2013, 10:59 PM

Re: Dressing up (a "450W" ATX supply) for Halloween -- YueLin YLP-013

Originally posted by needtoseepics View Post

Here is the input filter stage installed on the YueLin, with also two "Y rated caps"
in the X cap locations (is this a bad idea.. let me know if you think so). These Y caps
came from the SGI's Lucent supply. Also you can see in this image the thicker
input wires I've installed to the PCB.

It's not going to filter nearly as well.

The common mode chokes need a low common mode impedance where those X caps normally go. The Y caps are only a few thousand picofarads at most = higher Xc = higher common mode impedance. The chokes then won't "dominate" with their Xl, resulting in more RFI leakage from the supply.

Use caps that are at least .1u, preferably .22u.

Originally posted by kc8adu View Post

those long thin leads totally defeat the purpose of low esr caps.
just get proper caps and install on the board.panasonic fm are a good choice.

^This.

You've got minimal load on there in your test. The output inductor (after secondary rectifiers, but before filter caps) and filter caps have minimal ripple- duty cycle is way down, ripple is (normally) relatively low.

Some large resistors, for 5 and 3, and a few bulbs for 12, will really test it. Don't push it too far, with those wimpy heatsinks. Too thin, and too little surface area, to adequately cool those devices.

**LLLlllou** · 10-24-2013, 06:09 PM

Re: Dressing up (a "450W" ATX supply) for Halloween -- YueLin YLP-013

Originally posted by needtoseepics View Post

Is Jamicon better or worse than Canicon?

That's analogous to asking "what tastes better, cat turds or dog turds?"

**cheapie** · 10-24-2013, 07:08 PM

Re: Dressing up (a "450W" ATX supply) for Halloween -- YueLin YLP-013

Originally posted by needtoseepics View Post

From what I was reading, ATX spec says 50mV ripple is allowed/permitted, and
the voltages must be within 5% of the "correct value", so, looking at my scope
results, I'm very much "within spec", at least for this simple test (which I realize
is not representative of an actual computer system with activity), so I don't really
see a problem, yet.

You're actually being harder on it than the spec says. (Yes, I sat down one day and read the ATX12V (version 2.2) power supply specs. I was bored, okay?) The spec calls for some amount of capacitance at the load (10uF lytic and 0.1uF ceramic caps), and allows 120mVp-p ripple on +/-12V. See section 3.2.6.

Attached Files

ATX12V_PSDG_2_2_public_br2.pdf (310.2 KB, 32 views)

**Pentium4** · 10-24-2013, 07:08 PM

Re: Dressing up (a "450W" ATX supply) for Halloween -- YueLin YLP-013

Originally posted by LLLlllou View Post

That's analogous to asking "what tastes better, cat turds or dog turds?"

To be fair, I'd rate Jamicon quite a bit higher than Canicon. But I still wouldn't use them in a power supply.

I'd install the filtering caps traditionally, but it looks like you have a nice project going here. Good job man!

**Th3_uN1Qu3** · 10-25-2013, 03:49 AM

Re: Dressing up (a "450W" ATX supply) for Halloween -- YueLin YLP-013

Jamicon capacitors are decent for general purpose applications. They are completely unsuitable for use in a SMPS.

**kc8adu** · 10-25-2013, 06:05 PM

Re: Dressing up (a "450W" ATX supply) for Halloween -- YueLin YLP-013

looks like several psu makers missed that memo!
im looking at you bestec!

Originally posted by Th3_uN1Qu3 View Post

Jamicon capacitors are decent for general purpose applications. They are completely unsuitable for use in a SMPS.

**Pentium4** · 10-25-2013, 07:43 PM

Re: Dressing up (a "450W" ATX supply) for Halloween -- YueLin YLP-013

Yeah, Bestec loved using GP Jamicon caps. How did they manage to keep the ripple low then with non low ESR caps? Maybe the PI coils did most of the work?

**Wester547** · 10-25-2013, 07:59 PM

Re: Dressing up (a "450W" ATX supply) for Halloween -- YueLin YLP-013

^ Well, for the +3.3V rail, Bestec seems to use a unique design that has one ferrite coil after the first (input) capacitor and another after the output capacitor, which seems to reduce the ripple voltage even more. That being said, the toroid for the +3.3V has a bit of a different method than the group regulated +5V/+12V toroid you see in most PSUs. It uses a saturated coil which means less squandered power and better crossloading. As for why Bestec get away with GP capacitors from Jamicon and Teapo and still manage to keep the ripple voltage low? Well, the toroidal coils and ferrite coils do a huge part of the filtering (unlike in flyback topology which lacks the toroids) so there isn't actually a tremendous difference between low ESR and GP capacitors in the output of a SMPS (not that I would recommend using GP capacitors at all, but extremely low ESR capacitors aren't necessarily a benefit either). The use of -52 material instead of -26 material seen in older PSUs also suggests a higher switching frequency which also means you can get away with smaller components and inductors. It also depends heavily on how well the feedback loop is designed (some designs are so bad that they have the feedback after the output filter...) and how well the PSU is cooled and ventilated, of course.

**Pentium4** · 10-28-2013, 04:05 PM

Re: Dressing up (a "450W" ATX supply) for Halloween -- YueLin YLP-013

Originally posted by Wester547 View Post

^ Well, for the +3.3V rail, Bestec seems to use a unique design that has one ferrite coil after the first (input) capacitor and another after the output capacitor, which seems to reduce the ripple voltage even more. That being said, the toroid for the +3.3V has a bit of a different method than the group regulated +5V/+12V toroid you see in most PSUs. It uses a saturated coil which means less squandered power and better crossloading. As for why Bestec get away with GP capacitors from Jamicon and Teapo and still manage to keep the ripple voltage low? Well, the toroidal coils and ferrite coils do a huge part of the filtering (unlike in flyback topology which lacks the toroids) so there isn't actually a tremendous difference between low ESR and GP capacitors in the output of a SMPS (not that I would recommend using GP capacitors at all, but extremely low ESR capacitors aren't necessarily a benefit either). The use of -52 material instead of -26 material seen in older PSUs also suggests a higher switching frequency which also means you can get away with smaller components and inductors. It also depends heavily on how well the feedback loop is designed (some designs are so bad that they have the feedback after the output filter...) and how well the PSU is cooled and ventilated, of course.

Interesting. Too bad Bestec got such a bad reputation from the 250-12E because they make some pretty good power supplies. I may try an experiment recapping a Bestec with Rubycon PX to see if it can keep ripple maintained with the higher quality GP caps

**needtoseepics** · 11-03-2013, 10:48 PM

Re: Dressing up (a "450W" ATX supply) for Halloween -- YueLin YLP-013

I found two (different) X2 capacitors, one labed IEC384-14II
55/100/56X2 250V~, then on the backside (frontside?) it's labeled
OKAYA RE224 0.22uF K 275V~ iP, and black in colour. The other one
is labeled MEX/TSC MKT GPF .1 M 250V~X2 40/85/21 211, then in the
bottom left is 565-1, blank on back side, and orange in colour.
Due to spacing, I've placed the shorter (orange) one by the bridge
rectifier, and the slightly taller one (black) one closer to the
input fuse. I assume these are a better match than the Y caps I had
before.

I've installed the 6Amp bridge rectifier, and replaced the Canicon
470uF 200V 85'C caps with identically equivalent Rubycon capacitors,
labeled: 200v470uF(M) NEG. +85'C USP. These are of course used, and
harvested out of (yet another) junk supply. But presumably used
Rubycon is better than used Canicon?

Also I swapped out the primary-stage switching transistors with the
SK C4138 parts, which are in the larger package size, and the
secondary-stage schottky barriers with the international rectifier
30CPQ150 parts, with two of the IR parts on the right, and the one
original part that was the larger package size went to the far left,
as was done in that other thread.

I took pictures of this process, which I'll attach here, but then
after taking the pictures, I did some more stuff that I don't yet
have pictures of. Specifically, I also swapped the Y caps for some
larger blue ones I pulled from another supply, and recased into a
Sparkle Power box which has a much larger vent area, and used the
fan from the Sparkle since it is in much better condition (The
sparkle had a big burn mark on the pcb and apparently it got hot
enough to desolder one of the large resistors which the burn
is centered around, 'cause it just fell out and was bouncing around
in the box, hmm.. no wonder it didn't work). I went ahead and soldered
an A/C power input connector to the YueLin board that I harvested from
(yet another) supply, and put the appropriate plug on the switch in
the back, along with a Teapo X2 cap direct on the inlet, like I've
seen posts on here with people doing that -- is this a problem since I
now have 3x X2 caps? two on the board, and one on the inlet, or is
this fine to do?

I ran the reboxed YueLin overnight with a single harddisk load, and
with no issue in the morning, I put it in a machine (Athlon X2 4200,
4GB, Quadro FX 3500, Broadcom GigE), and ran it for a day in Linux
with glxgears running. Didn't seem to have any trouble, so I suppose
at this point I can think again about the transformer, or possibly
ordering some correct output filter caps. That transformer I have
harvested didn't have its input pins in the same spots as the holes
are drilled on the YueLin pcb. I assume the two input pins must be
"the same function" as the two holes on the pcb provide, so, if I
mount it at an angle, with the output pins in the holes, and run
some wiring to the transformer's input pins from the pcb holes,
should I expect this would still be an improvement? At least if
I do it now, I know it worked before, whereas if I had done it at
the same time as all the other stuff and it hadn't have worked, I
wouldn't have known how to assign blame.

I might be ordering some "proper" output capacitors, but I guess
I need to do more reading/pricing before I make a decision on what
exactly to get. Whenever I open it again, I'll take pics of how it's
looking now with the recase, etc.

I was gonna try that whole "lamp" suggestion, but I'm just wondering
are we talking 60Watt light bulbs on the 12Volts line, or what does
"lamp" mean? Obviously I didn't try it yet, and just stuck it in a
computer system for a load testing (and didn't scope it since the
silicon upgrade yet).

Attached Files