Fake 400W PSU?

Archived from groups: alt.comp.hardware.homebuilt (More info?)

On Wed, 16 Mar 2005 16:55:50 +1100, Franc Zabkar
<fzabkar@optussnet.com.au> wrote:

>Earlier I posted some test results for my generic "400W" ATX PSU. Its
>+12v and +5V regulation was so poor that I was compelled to look
>inside.
>
>My first impression was that the PCB looked no more substantial than
>that of a 200W PSU. The caps were very small, and the inductors didn't
>seem very big either. The caps in the pi filters were absent, and the
>chokes were replaced by links.


Sure, I've observed this over the years... don't recall the
generic labels but I had old 200W PSU that were virtually
identical inside to newer PSU that only had a label swap.
They didn't upgrade the PSU, they upgraded the label!

>
>There was only the one fan, with a sleeve bearing.
>
>The PCB had only the one IC which initially raised some concerns.
>However, I have since discovered that this IC integrates several
>functions - the PWM controller, a voltage supervisor with a Power Good
>output, a PS-ON input, and what may possibly be a 3.3V regulator. The
>part number of this chip is obscure - there is a "2003" in large type
>and "IFGN3M0317E" in a much smaller font.
>
>The caps were all 105 degC low ESR types
>(http://elec.luxon.com.tw/Product%20Word/LZ.zip), except for the two
>big 330uF 200V filter caps on the primary side which were 85 degC.

I can tell you that the average 300W PSU has 470uF and
higher primary side caps. These would be typical came-with
case budget PSU, a typical 400W would be nearer 680-1000uF.


>
>The +5VSB rail was tightly regulated by a TL431 precision reference.
>
>The +3.3V and +5V rails were each rectified by a S10C40C dual diode,
>and the +12V rail was rectified by two FR302 fast recovery diodes.
>
>According to the datasheets, the FR302 is rated at 3A, and the S10C40C
>is rated for 10A.
>
> http://www.mospec.com.tw/pdf/fast/FR301-FR307.pdf
> http://www.mospec.com.tw/pdf/schot [...] 0C60CE.pdf


You weren't supposed to notice that! Yeah, it's an
impossibility that some generics could approach their
labeled capacity for more than long enough to drain the
output filter caps, which isn't too long considering those
caps are often a fraction of the mfd values found on better
units. Luxon caps though are among those seen to fail most
often on boards, I'd easily prefer just about any other
name-brand cap. There is a miracle invovled though, that
they seem to put these generics together SO cheaply.

>
>This would mean that the +12V rail would have a maximum rating of only
>6A, and the +5V and +3.3V rails would be limited to 10A each. The
>combined wattage for these three voltages would therefore be 155W.
>Allowing 2A for the +5VSB rail, and maybe 10W for the -5V and -12V
>rails, this gives a total wattage for the PSU of only 175W. Having
>said that, I believe these ratings may be relaxed if the voltages at
>the transformer windings are much higher than the output voltages of
>the respective rails.

Often one telltale sign of an overrated PSU is the size of
the transformer. I suppose they could just switch 'em at
higher frequency but if that was such a simple solution then
I'd expect others to do same, so again transformer size
would matter relative to better PSU.


>
>The PSU was installed in the case with the PCB at the top, ie
>component side down. I would think this would impair the cooling.

Most are though, it's quite rare for it to be upright even
when there's only one fan, unless the case itself was odd
and reversed the screw mounts such that this caused that
orientation.


>
>The PCB has an LC EMI filter on the mains input,

LOL, you seem to have a "high-end" generic, many of 'em just
put in jumpers there even though they still use a PCB with
the filter option.


>but strangely there
>is an additional, almost identical filter on a daughter PCB attached
>to the input IEC socket. This would make sense if the output IEC
>socket were wired to the output of this filter, but instead it is
>wired as follows:
>
> EMI EMI
> AC I/P o->-|- filter ----o--- filter --- rectifier
> | #1 #2
> AC O/P o-<-|
>
> |<- daughter ->|<----- PSU PCB ------>>
> PCB
>
>
>IMO, a more sensible configuration would be as follows:
>
> EMI EMI
> AC O/P o-<--- filter --|-o--- filter --- rectifier
> #1 | #2
> AC I/P o->-------------|
>
> |<- daughter ->|<----- PSU PCB ------>>
> PCB


Usually I see smaller inductors and film caps on the
daughter boards, so they'd cover different frequencies.

>
>Printed on the PCB is the following legend:
>
> F5A/250V (for 180W -- 235W)
> F7A/250V (for 300W -- 350W)
> F6.3A/250V (for 250W)
>
>The installed fuse is rated at 5A. :-(

Might be just as well, considering 5 x 110V, 550W * .7
efficiency would be 385W, higher than the expected output
potential.

>
>The PCB is labelled "Y-B200 ATX Ver 2.8" which also suggests a 200W
>rating.
>
>This appears to be the circuit diagram:
>
>"LC-B250ATX ch. Y-B200-ATX ver. 2.9 JNC Computer Co."
>http://electro-tech.narod.ru/schematics/power/JNC_Y-B200-ATX.zip


You're doing pretty well to find that. Even so, it doesn't
list the component values on many critical parts, a generic
based on that *could* be sufficient for a cheap system if
only made a little beefier.

>
>The label on the cover reads as follows:
>
> L&C Technology Inc
> Model: LC-B400ATX
>
> +3.3V +5V +12V -12V -5V +5VSB
> 28A 40A 17A 0.8A 0.3A 2A
>
> +5V & +3.3V combined load 220W
> +5V & +3.3V & +12V combined load 380W
> Total output is 400W max.

Don't forget the "with MTBF of > 50K hrs @ 80% load and 25C"
part.


>
>On the Net I found the following references:
>
> http://www.anscase.com/powersupply.htm
> http://www.anscase.com/psu-b400.htm
>
>http://www.telecommander.com/pics/links/powersupplies/l&c400wp4amdpsrt/l&c400wp4amdpsrt.htm
>
>http://www.amd.com/us-en/Processors/DevelopWithAMD/0,,30_2252_869_1039^4038,00.html
>
>These suggest that there is a genuine L&C LC-B400ATX PSU which is
>certified by AMD and which probably lives up to its specifications. At
>the very least it has two ball bearing fans.

Err, the telecommander link reads "dual ball bearing cooling
fans" which looks (from the picture) to mean one fan, dual
bearings in it. Frankly I wouldn't even trust that, but
it's possible, looks to be a "Globe" brand fan in the
picture they provide which is often seen in the cheaper
PSU... though still better than a cheap sleeve fan which
seems subject to seize up far too early in it's life.

I don't think we can assume AMD certifies this L&C model let
alone what you may have there, if it's the same. The AMD
list of of manufacturers who have submitted *something* in
the past, not necessary that all their products would be
considered acceptible by AMD. AMD used to have a more
comprehensive list including models and combined 3V + 5V
ratings, but they took that page down maybe 3 years ago and
it would be less applicable today regardless, since their
newer lines of CPU are more likely powered by a board using
12V for CPU power... making even the old more detailed list
obsolete.

As for the "AMD Recommended" line on the (telecommander
link), it seems more like stretching the truth (to put it
kindly) than any assurance of anything, let alone that AMD
has ever had one of these to test.

Archived from groups: alt.comp.hardware.homebuilt (More info?)

On Fri, 18 Mar 2005 08:21:58 +1100, Franc Zabkar
<fzabkar@optussnet.com.au> wrote:

<snip>

>>You weren't supposed to notice that! Yeah, it's an
>>impossibility that some generics could approach their
>>labeled capacity for more than long enough to drain the
>>output filter caps, which isn't too long considering those
>>caps are often a fraction of the mfd values found on better
>>units.
>
>I had considered this when I saw just how small these caps were. If we
>assume a conservative switching frequency of 40kHz, and a load current
>of 10A, then, in one half cycle of the oscillator, a 1000uF cap would
>discharge by ...
>
> dV = I x (0.5 x dt) / C
> = 10 x 0.5 x 25 x 10^-6 / (1000 x 10^-6)
> = 0.125V


This supports my accusations that often the generic power
supply will continue working but the motherboard caps may be
overly stressed. Often I hear of boards that don't
"normally" fail due to caps, having such failures.


>
>As for the ratings of the diodes, I had previously seen many AT PSUs
>where the diode ratings were a little less than the claimed output of
>the respective rail. For example, looking inside two generic 200W
>PSUs, I see a Mospec S15D40L dual diode rated for 15A, whereas the
>claimed output of the +5V rail is 20A.
>
>A third generic 230W unit (Power Win, model PW230,
>http://www.power-win.com/) appears to be more honestly rated. It has a
>10A dual diode on a +12V rail rated at 9A, and a 30A diode pair on a
>+5V rail rated at 23A.
>
>BTW, one of the generics has a single heatsink with the AC-side
>chopper transistors at one end, and the DC-side diodes at the other
>end. Scary!

I don't recall any that bad but sometimes after a second or
two I'll just throw 'em away... not needing to know "how
bad" but rather that they fell below the minimal level I'd
accept for use.

>
>> Luxon caps though are among those seen to fail most
>>often on boards, I'd easily prefer just about any other
>>name-brand cap. There is a miracle invovled though, that
>>they seem to put these generics together SO cheaply.
>
>My old socket 7 motherboard uses this brand in its Vcore regulator.
>The capacitors still look healthy after many years. IME, bad caps find
>their way into all sorts of equipment, both high end and low end. I
>suspect caps are one product where a particular manufacturer may get
>it right one day and wrong the next.
>
>Having said the above, this post appears to confirm your findings:
>
>"Blowing up 100's of Power Supplies":
>http://groups-beta.google.com/group/sci.electronics.repair/msg/8e892d0c5d0bf0c4?dmode=source
>

I don't recall their spec sheets but in some cases it seems
the fault is in the parts selection process, that the Luxons
used aren't suitable for the circuit. Just seems odd that
if/when such a poor decision is made, so often the cap
they'd end up using was a Luxon... or CapXon. I mean, if
one wanted to substitute an inferior cap for the job they
could pick anything at all. More of the ones I've seen fail
on boards are Turquoise-green/silver-striped jacket though,
maybe "LZ" series. Formerly I think they were black with
gold jacket, or perhaps the preceeding family they replaced
were. Many of the Luxons I've seen in generic PSU are black
jacketed. Don't recall model numbers but I've also seen the
black-jacketed Luxons on boards where they wouldn't be
subject to so much ripple and they didn't fail.


>>>The PSU was installed in the case with the PCB at the top, ie
>>>component side down. I would think this would impair the cooling.
>>
>>Most are though, it's quite rare for it to be upright even
>>when there's only one fan, unless the case itself was odd
>>and reversed the screw mounts such that this caused that
>>orientation.
>
>All my AT PSUs had the PCB at the bottom. However, it occurs to me
>that mounting the PCB upside down may reduce the tendency for dust to
>settle, so maybe it's a good idea???

Maybe, hard to speculate whether it makes a difference
compared to the different heat densities when upside down.

I suspect it's more of a modular design decision though,
that since some do have a fan on the bottom they just make
them all in same fashion whether the particular unit uses a
bottom fan or not.


>
>>>The PCB has an LC EMI filter on the mains input,
>>
>>LOL, you seem to have a "high-end" generic, many of 'em just
>>put in jumpers there even though they still use a PCB with
>>the filter option.
>
>Maybe the PSU is a reasonably designed 180W unit. After all, a
>Volkswagen does not become a bad car simply because someone puts a
>Porsche badge on it.

Maybe, though to a certain extent I've not seen the decent
lower wattage units scaled back as much due to wattage as
the higher wattage units merely being limited by board
real-estate. Have a Mitac 200W from a Compaq system that
has 470uF primary caps, pairs of 2200uF caps for 5V _and_ 3V
across inductors, and a pair of ~ 1500uF (guessing, they're
covered in that infernal rubber cement) on 12V even though
12V is only rated 6A.


>>>The installed fuse is rated at 5A. :-(
>>
>>Might be just as well, considering 5 x 110V, 550W * .7
>>efficiency would be 385W, higher than the expected output
>>potential.
>
>The *suitability* of the fuse was of no concern to me. More important
>was the fact that by using a 5A fuse the manufacturer has confirmed
>that the intended application for the "multipurpose" PCB is as a "180W
>-- 235W" PSU, not 400W. Perhaps a genuine 400W configuration of this
>same PCB would have bigger caps, larger diodes, taller transformers
>... and a bigger fuse.

Yes I think it would... if a genuine 400W version even
exists.



>>Don't forget the "with MTBF of > 50K hrs @ 80% load and 25C"
>>part.
>
>You have one, too? :-)

I saw that on one of your links, though I might have one of
those PSU sitting in a pile somewhere, except the most input
*filtering* I usally see on 'em is a thermistor for inrush
limiting, sometimes.


>>Frankly I wouldn't even trust that, but
>>it's possible, looks to be a "Globe" brand fan in the
>>picture they provide which is often seen in the cheaper
>>PSU... though still better than a cheap sleeve fan which
>>seems subject to seize up far too early in it's life.
>
>Mine has a "Ruilian Science & Technology" fan.


That sounds familiar, I think Channel Well puts those in
their low-end units, or maybe it ws "Zirulian" or something
similar with a Z at the front.

The sad part is that for all those that die, seems most
often it's either the output caps or the fan that fails. If
they'd sent an extra $1 on parts they might be significantly
more viable for cheap systems... not that I'd want one in
any system I relied upon but it's just a shame that $1 could
make the difference between very early failure and a unit
that could power a budget, relatively-energy-conservative
system for it's entire viable life.

Archived from groups: alt.comp.hardware.homebuilt (More info?)

On Wed, 23 Mar 2005 07:22:36 +1100, Franc Zabkar
<fzabkar@optussnet.com.au> put finger to keyboard and composed:

>Have you ever seen a 30A or 40A dual schottky
>barrier diode in a TO-220 package?

I've answered my own question. Mospec make 30A TO-220 schottky parts,
but not 40A.

See http://www.mospec.com.tw/diode/schottky.htm

I'm left to wonder what rectifies the +5V 40A rail.


- Franc Zabkar
--
Please remove one 's' from my address when replying by email.

Archived from groups: alt.comp.hardware.homebuilt (More info?)

On Wed, 23 Mar 2005 02:31:22 GMT, kony <spam@spam.com> put finger to
keyboard and composed:

>On Wed, 23 Mar 2005 07:22:36 +1100, Franc Zabkar
><fzabkar@optussnet.com.au> wrote:

>>Another clue as to the PSU's origin was a small code on the label.
>>Mine has a "LC34XA" in the bottom RH corner, whereas the TT PSU had a
>>"TTnnXA" (I can't recall the nn digits). I'm guessing this could be
>>some kind of OEM ID.
>
>Well I may have something similar they put into a mATX
>casing, it's code is AL20XB, is an Allied/Foxconn 200W. It
>has what looks to be same size tranformer as in the
>following picture, but labeled "EI-33MBS". The controller
>chip has "2003" on it instead of "2001" though, so this may
>be the next generation design. I"ve even less crazy about
>the Foxconn units because their metal casing has metal
>plated with something that roughly resembles extremely thin
>galvanization but quickly oxidizes to a white rustlike
>coating.

Mine is the same.

>It's rated as
>
>3.3V - 17A
>5V - 12A (70W combined 3V+5V)
>12V - 13A (ATX 2.03 unit, has P4 4-pin connector)
>
>170W combined 3+5+12V rating
>A pair of 330 uF primary caps, manually switched input
>voltage.


>I'm having a hard time making out the diodes, not a lot of
>space inside. Looks like MOSPEC F16C20C, 40N03P (both
>TO-220) and a TO-247 part I can't make out. I'm somewhat
>impresed that that used a TO-220 7905 in it though, with a
>tiny heatsink no less!

The F16C20C is a 16A 200V fast recovery twin diode, while the 40N03P
is probably a 40A 30V part. Based on these data I'd say your PSU spec
looks genuine. One caveat is that the "F" parts have a higher forward
drop than Schottky diodes (about 2x) which means they would dissipate
a lot more heat.

>It does have a fairly good AC filter circuit considering
>it's mATX, a couple of chokes, film caps, 4 mov before
>rectified.

Are you confusing MOVs with noise suppression caps?

>>Here is a photo of my fake PSU:
>>http://members.optusnet.com.au/~fzabkar/LC-B400ATX.jpg (361KB)

>I don't recall ANY
>moderately good name-brand 300W PSU with such a small
>transformer.

I have a 300W AT PSU (Skynet Electronics, model SNP-930T) in my
Aridyne server case. Everything looks so much bigger in it. The +5V
rail even has a trimpot for adjustment. The two mains caps are 1000uF,
and both negative rails have heatsinked 79xx regulators. The unit is
rated at 5V@30A, 12V@12A. The diodes are 30A and 25A schottky parts,
and the fan is a Matsushita Panaflo. My only concern is that sometime
in the future I may get a visit from a Terminator. :-)


- Franc Zabkar
--
Please remove one 's' from my address when replying by email.