the gutless, bloated, and fried power supply hall of shame

[c_hegge]

Yes. AFAIK, Logisus only use Keerda as their OEM.

[Pentium4]

Is there any specific reason why they screwed the ground wire to the PCB and not the case?

[momaka]

^ Yes.
They save one screw. One screw saves them probably $0.001 + 20 seconds of labor. 1000 PSUs saves them $1 + 20 x 1000 = 20 000 seconds of labor. 20 000 seconds of labor is about 5.5 hours. If they have to pay someone $2/hour to add that screw, that's ~$11. Add $1 from the screws saved and they have saved ~$12 total. IT TOTALLY MAKES SENSE!

[mariushm]

The metal screw goes through touching both the ground earth ring and the metal case on the other side.

This also puts pressure on the pcb and on the other side of the hole, there should be a bit of solder around the ground ring on the pcb which makes sure there's direct contact between the case and the ground trace.

Like this board has.. see the corner holes and the solder:

02.jpg

It's really not worse than the other method of screwing it by itself.. the only flaw is there is now washer to prevent the screw from loosening itself from vibrations:



But most cheap psus don't have these anyway, or thread locker liquid.

[tom66]

You may laugh about it but doing it that way saves maybe 15 seconds per worker, which is a couple of cents -- it all adds up -- but it also increases throughput (reducing costs further.) And think about this: by making the earth strap part of the PCB installation process you make it harder to screw up. Worker can "forget" to install earth strap but he'll have lots of washers left. I don't care about cost saving like this (it's kind of an art.) It's when it gets to cheap or insufficient components, like poor quality caps, that it is a negative thing.

[c_hegge]

^
I agree. Not using an extra screw hole for the earth isn't really that much of a problem compared to cheaping out on the components or leaving them out.

[ben7]

Originally Posted by c_hegge

^
I agree. Not using an extra screw hole for the earth isn't really that much of a problem compared to cheaping out on the components or leaving them out.

Yep, lol I agree too.

It doesn't change the way the PSU functions, and it won't fail in a few months like under-spec'd parts.

It doesn't look as good, but hey, the average PC user dosen't even know what the inside of a power supply looks like xD

[momaka]

^ Yeah, I must say I agree with that as well.

Worst PSUs IMO are the cheap ones that have a fancy painted case, sleeved cables, and a shiny box because then they skimp badly on the components inside to bring the price down. I rather get a plain gray box with semi-decent parts in it than the other way around. And I could care even less for the box. JG and other review places seem to complain sometimes when the box is "not good enough". I say who cares, it's a box. Not like I'll be storing my lunch in it anyways.

[Behemot]

Originally Posted by momaka

Not like I'll be storing my lunch in it anyways.

But dinner maybe?

[momaka]

^ Only if it's fast food.

[Behemot]

Well, may be, no? Heard that crap is pretty popular in US…it's spreading like cancer here, too…

[tleu8472]

FSP Group FSP300-60PNA(PF)

- Yihcon brand caps on the secondary - the 3300uF ones are blown with leaked and rusted electrolyte
- LXK (?) brand 680uF caps on the primary - still look OK with no visible bulge or leakage
- Some rust on top of the PFC coil

Is it worthwhile fixing this unit by replacing all of the secondary caps and giving it a good clean? Does the rust on top of the PFC coil matter?

If possible, I would like to use it to power an old Athlon 64 (S754) system.

Any thoughts would be appreciated.

[mariushm]

Yes, replace secondary capacitors (don't use very low esr capacitors, panasonic fc, nichicon he should be good) , clean the dust .. rust shouldn't be an issue.

It's a good 300w power supply, not lying about what it can do and will be ok for Athlon 64.

[Behemot]

This one will be OK for almsot everything

Yihcon, what the heck is FSP using lately?!

[Evil Lurker]

Here is a real winner, a Tiger-USA branded PSU, model ATX-450W P4.

From the label, its rated @

+5v 35.0A
+12v 16.0A
-12V 0.8A
+3.3V 24.0A
+5VSB 3.0A

Adding up the numbers and taking into consideration the model number on the label would lead one to believe that this is a 450w rated unit. However, after opening up the unit and taking a look inside it is obvious that the label is a flat out lie as to the ratings.




The first glaring thing missing is the input filtering section. The only things this unit has are 2 Y-caps... no common mode chokes, no X-caps, no MOVs or surge protection, just a fuse and a wimpy NTC thermistor. As from my understanding, this does not meet the EU filtering requirements so the CE markings are blatantly forged.

A closer look at the PCB reveals even more horrors.



Aside from the no-name input filtering caps we see that the unit has undergone the 4 diode bridge rectifier substitution using 2A05 2 amp rated diodes. So if the unit is running off of a 125v supply, then the unit might pull 450 watts for a brief period of time or so before the diodes go into meltdown. Because of this factor alone, there is absolutely no way that this unit can do its rated power output.

But wait there is more...



Looking at the rear of the PCB we can see that the primary switching transformer uses an EI-33 core. Not only that, the core seems to be not fully wound... most SMPS transformers that I have seen the core will barely slip over the windings. I'm not totally 100% for sure if this core could do the rated power output, but if I were a betting man I'm pretty sure it there is no way in hell it could. Furthermore, the heatsinks are some very thin wimpy design. I will however award bonus points for the switchers, a pair of generic 13007 fast transistors simply because they have a metal tab and can dissipate more heat than the typical isolated solid plastic cased transistors I have encountered in other units.

In the output section, we can see that the unit has a bunch of small no-name caps and relatively small filters with some very suspicious looking brown stuff on several of the caps. Additionally we see that there are three rectifier diodes, two MOSPEC s16c40c for the 3.3v and 5v rails, and a F12c20c for the 12v rail. Lacking is the typical LM7912 negative voltage regulator seen on most other units, so I'm guessing that if the unit does have a negative 12 volt rail, its output is unregulated. Looking at the datasheets for the rectifiers to my horror the F12c20c is a 12 amp rated part with a 150ns reverse recovery time and a Vf of 1.3V... again the rated power output on the label is complete bullshit. I'm thinking that provided the PSU works at all the efficiency of this unit has to be absolutely terrible. The S16c40c's aren't that much better... at least they are Schottky diodes. These have a Vf of around .~70 under full load, reverse leakage doesn't seem too bad, and there is no reverse recovery time listed. Again, these are 16 amp rated parts so the label is a lie. on a personal note, just like the PSU manufacturer, I don't trust MOSPEC as far as I can throw them. Last but not least are the output wires... they are labeled as being 20 gauge, but they seem very thin compared to other wires I have seen. It may just be the insulation rating, but honestly they look more like 22 gauge wires to me.

The PSU is controlled by what appears to be a generic TL494 chip and a LM339 quad comparator. If it wasn't for the fact that the ICs have 494 and 339 on them along with a bunch of other random numbers and letters I don't thik I could identify them they are so generic. Also I didn't pay too much attention, but as far as I can remember there aren't any zeners on the secondary side, so I'm guessing the unit has no undervoltage/overcurrent/short circuit/whatever protection.



Checking out the 5VSB circuit I can see that it uses an FR302 diode. These are simply fast recovery diodes with a reverse recovery time of 150ns and a forward voltage drop of 1.2v. As best as I can remember, a standard 1N series general purpose rectifer has about the same efficiency and specs. I don't have a scope, but I'm pretty sure the ripple has to be way out of spec on this unit.



Now then, normally I would simply chunk a hunk of crap PSU like this, however due to the lack of protection circuits and its simple controller and no visible signs of any components letting out the magic smoke, I decided it might have some potential as a bench PSU for other things so I decided to conduct an autopsy. The first thing I checked were the no-name ultra generic output filtering caps with the brown junk on top. To my shock and amazement, all the electrolytics on the board passed ESR testing and capacitance values with flying colors with the exception of two... the primary filtering caps.



When I yanked the primaries off, I noticed that they didn't weigh much compared to other primary caps of the same value I have yanked off other boards. The PSU was supposedly manufactured in 2009 according to the markings found in various locations, so they should be OK right? Wrong. Testing these caps revealed why the PSU was disposed of...



Each of the caps had a capacitance value of less than half the rated specification. This coupled with the fact that I could find no defective silicon leads me to believe that the PSU if mechanically fine, it simply could not maintain sufficient output voltage to keep up with load demands.

So since I basically had a nutless wonder that should work I decided to do a bit of experimentation with component swapping since the PCB has additional holes for other component configurations. I have yet to get it all put back together since I mangled a couple traces getting everything off, but obviously the first thing to go were the primary caps and diodes... replaced them with a couple of used Rubycon CE series caps of identical value 0.8 ESR and nearly dead on spec capacitance wise along with a GBU606 bridge rectum fryer. I considered swapping the main switching transistors with a couple 13009's but I decided to keep them since I figured they would need a bit of adjustment in component values to drive them properly. The crap caps since they all tested good were put back on since I don't have any good caps in the same specs. Also the secondary rectifiers had to go... I replaced the 12v rectifier with an STPR1020CT which has a lower Vf, reverse leakage current, and about half the reverse recovery time as the MOSPEC diode. I considered using a STPR1620CT, but thought that would be a waste of silicon. For the 3.3v and 5v rails I slapped on a couple STPS2045CT's... these have a higher current rating. Also considered were STPS3045 in TO-247 package but I decided again, a waste of silicon considering the size of the transformer. The only other thing I have swapped out is the 5VSB rectifier diode with a 3 amp 100v ultrafast part. Not for sure if it would work, but I might change it out to a SB560 diode.

If the unit comes to life once again I plan on swapping out the controller and comparator with some name brand IC's and see about modding the feedback loop to change the output voltage and swap out the output wires for something else... I wouldn't dare put this PSU into a PC. Also, no input filtering for now since I don't have a scope and can't measure what frequency the noise that needs to be filtered so I don't know what component values to use.


Well that is about it for now for this PSU.

[Wester547]

Originally Posted by Evil Lurker

I don't trust MOSPEC as far as I can throw them.

In all seriousness, I see MOSPEC rectifiers used in many more commonly known PSUs, such as Hipros and Huntkeys (Huntkeys are not fantastic but they are still far better than the gutless of gutless wonders), albeit of a much higher rating and size and they do fine (or as well as more commonly used parts from say, General Semiconductor, STMicroelectronics, Lite-on Semiconductor, etc...). I think MOSPEC schottky rectifiers are good. You can't expect 16A parts to do double their rating in a crappy PSU even with bridge topology. :P And I don't think it would hurt to put a TO-247 part in that PSU since they dissipate heat better than TO-220 parts but of course they are more expensive as you made of note (it would be cheaper to use a pair of TO-220s connected in parallel). There's no +5&+3.3V combined rating on the PSU label that I can see either so I too would be suspect of the labeled ratings. Those heatsinks are also pathetic, to be expected of gutless PSUs. I see a polypropylene capacitor in that PSU also, which I bet isn't safety-approved. The PSU doesn't have much filtering output and input wise either, in any regard, which cannot help its ripple suppression and voltage regulation (assuming there is any), and nevermind noise.

It might be a good PSU to completely overhaul and repair, though it may not be worth the time or effort depending on what it's needed for either.

[Behemot]

I think this is POS, good only for . If you ask me, this thing can't do more than 300 W. Looking at the transformer, maybe not even that. It's just waste of components for resurecting this one.

[c_hegge]

I give it 250W without blowing up (maybe 300W for a few seconds at a time), but maybe 150-200W with the ripple in spec.

[Pentium4]

Kinda looks like a Sun Pro. I don't think that thing could do a watt more than 200 and I think that's generous. Not with those input caps. By the way, what are the brands of the input and secondary caps?

[c_hegge]

Originally Posted by Pentium4

Kinda looks like a Sun Pro.

It's a Kaisen (http://www.gzkaisen.com/ks_en/)