Annoying heavy load-failure in Tecnoware Free Silent 500W

[momaka]

Re: Annoying heavy load-failure in Tecnoware Free Silent 500W

BTW I've noticed that when they do that (one bigger and another smaller cap on +5VSB) the bigger one is often placed the closest to the +5VSB transformer, is there any particular reason for that?

Yes.
Long traces = good antennas. Square waves traveling on the output will produce EMI and possibly some RFI as well. Placing the cap as close as possible to the transformer output makes the PSU produce less EMI/RFI, since the noise and ripple on the output is shunted to ground quicker without going through long traces. This also reduces noise on the output that comes from external EMI and RFI.

[TELVM]

Re: Annoying heavy load-failure in Tecnoware Free Silent 500W

It is known that the cheaper the PSU, the smaller the output caps, and viceversa. So one is tempted to suspect that deployed capacitance may be more related to budgetary than to electrical reasons, and that larger capacitance may in fact be a good thing.

About rules of thumb, Seasonic might have their own private ones. Look at this humongous 4700uF KZE (ESR=15) they put on +5VSB inside the Antec Truepower New 650W:



Together with its 1000uF companion downstream of the PI coil that amounts to 5700uF for +5VSB only.


BTW I've noticed that when they do that (one bigger and another smaller cap on +5VSB) the bigger one is often placed the closest to the +5VSB transformer, is there any particular reason for that?

[momaka]

Re: Annoying heavy load-failure in Tecnoware Free Silent 500W

Yeah and now take unknown unit and try to measure power output through the wattmeter. You'd be glad to get 10% error.

I've already done that with a few random SMPSes of mine, including some ATX PSUs, and got reasonable results as I did above. Anyways, we've already had a thread on this stuff, so I won't get into this discussion here.

Their output caps per rail are 2x1000uF or even 2x470uF on some occasions. Putting there more than 2x3300uF is just

Not really.

In SMPS, high capacitance is *always* more favorable than anything else, especially for older designs (half bridge and STF) where the switching frequency is relatively low. (Think about an audio amplifier with a line transformer - those always have huge caps, and that's because at 50/60 Hz, they really are needed to produce a smooth output voltage. And that's the whole idea behind SMPS: increase the transformer switching frequency so that you have have a smaller transformer and smaller output caps. The further you increase the frequency, the further you can shrink down components.) In general, there isn't really any limit to how high you can go with the capacitance. However, keep in mind that the inrush currents will also be larger with larger output caps, so the rectifiers and coils will be stressed more during power-ON sequence. And if the PSU lacks soft-start, then you may even burn a rectifier out. But other than that, the feedback loop will almost always be much happier with the higher capacitance.

Speaking of the feedback loop, the capacitor ESR comes at a distant second. And by that, I mean you don't want to go too high, but definitely not too low either. Keeping it close to original is the best thing to do. (Too high of an ESR means that noise and ripple won't get filtered too well. Conversely, too low of an ESR means that noise and ripple will be filtered very well, but the low ESR of the caps may cause ringing, which is essentially noise and ripple generated from the output filters themselves.) However, if you increase the capacitance, then a decrease in the ESR may not be as critical.

Also, again, this will depend on the SMPS design. Higher frequency designs may be able to cope better with smaller capacity and lower ESR caps. But for an older design, like the cheap PSUs in this thread, higher capacitance will be a lot more important than keeping the ESR low. So for example, if you had to choose between 2x 2200 uF low ESR caps (such as Nichicon HE or Chemicon KY) and 2x 4700 uF general purpose 105C caps, go with the 2x 4700 uF caps. Most manufacturers simply don't do this because with the above example, the 4700 uF general purpose caps will be much larger than the 2200 uF low ESR ones, which means more space is needed on the output side of the PSU (which may not be an option). And in many cases, the 2x 2200 uF caps may actually turn out to be cheaper, despite being low ESR.

As a general rule of thumb, you should replace caps with same capacitance or 1 to 2 sizes bigger.

Not if it's a cheap low quality PSU - for those, you disregard what's in the PSU and simply go by the rule of thumb I listed in post #16.

For example this psu has issues when it is connected with 12V power hungry PCs:
https://www.badcaps.net/forum/showthread.php?t=23996

The situation gets worse if you try to replace the 12V output caps with ones of bigger C.

That's an older Sirtec design. They often seem to have issues with 12V power-hungry systems, especially if you increase the capacitance on the 12V rail (and then even worse if you use low ESR caps). But those are the only PSU's I've seen that do this. Go figure

[goodpsusearch]

Re: Annoying heavy load-failure in Tecnoware Free Silent 500W

First, the esr drops too much. The same reason you should not recap psus with mobo grade caps.

It also has to do with the feedback circuit.

If the total output capacitance of a rail exceeds what the psu manufacturer had in mind when he designed the feedback circuit, then you get oscillations.

For example this psu has issues when it is connected with 12V power hungry PCs:
https://www.badcaps.net/forum/showthread.php?t=23996

The situation gets worse if you try to replace the 12V output caps with ones of bigger C.

[TELVM]

Re: Annoying heavy load-failure in Tecnoware Free Silent 500W

... Putting there more than 2x3300uF is just ...

Why? What would happen if i put 2x 4700uF?

Just want to know the reasoning behind the rule of thumb.

[goodpsusearch]

Re: Annoying heavy load-failure in Tecnoware Free Silent 500W

Dude you need to focus. The psu in the link has nothing in common with those gutless wonders we are talking about here. In those psus you are lucky if they have coils installed between the caps or even space on PCB for them.

Their output caps per rail are 2x1000uF or even 2x470uF on some occasions. Putting there more than 2x3300uF is just

As a general rule of thumb, you should replace caps with same capacitance or 1 to 2 sizes bigger.

[TELVM]

Re: Annoying heavy load-failure in Tecnoware Free Silent 500W

L8 500 is high end?

[goodpsusearch]

Re: Annoying heavy load-failure in Tecnoware Free Silent 500W

Why is that? Where does that arbitrary limit come from?

Be Quiet L8 500W with at least 13200uF (6x 2200uF) on +12V rail:

=> http://www.techpowerup.com/reviews/b...L8_500W/4.html

We are not talking about high end power supplies here. We are talking about old half bridge or 1 transistor forward cheap ass platforms.

[TELVM]

Re: Annoying heavy load-failure in Tecnoware Free Silent 500W

2x2200uF or maybe 2x3300uF is the maximum you can put.

Why is that? Where does that arbitrary limit come from?

Be Quiet L8 500W with at least 13200uF (6x 2200uF) on +12V rail:

=> http://www.techpowerup.com/reviews/b...L8_500W/4.html

[Behemot]

Re: Annoying heavy load-failure in Tecnoware Free Silent 500W

Yeah and now take unknown unit and try to measure power output through the wattmeter. You'd be glad to get 10% error.

[momaka]

Re: Annoying heavy load-failure in Tecnoware Free Silent 500W

In fact, the scenario is worse than you painted (see pic). Those resistors are 0.5 W, definitely!

*sigh* Built to fail, I suppose
Not today, though - you fixed it

Actually, I wondered whether I have to put some load to the 5 and 3.3 rails.

Most older designs like this almost always perform well with a good load on the 3.3V and 5V rails. Not having any there and solely loading the 12V rail really tells you if the PSU is somewhat capable of handling cross-loading. So it's a good thing you tested without load on 3.3V and 5V rails.

Here are the data:
nominal before recap, 21W after, 21W after, no load
12 V 11.9 V 12.1 V 12.4 V
5 V 5.2 V 5.2 V 5.1 V
5 VSB 5.1 V 5.1 V 5.1 V
3.3 V 3.3 V 3.4 V 3.3V
–12 V –11.4 V –11.5 V –11.5 V
The final verdict is of course yours, however :-)

Nothing out of the ordinary here. But I was wondering about that recapped PSU you put in one of the computers. Would really like to see how these PSUs do "in the field" after they are fixed.

Yes, the #42 post (Technoware FAL500FS12) concerns a previous revision of #14 post (FAL501FS12). The The so many posts before yours have convinced me NOT to recap these three PSUs of mine. I will rather replace with recapped, third series (i.e. #14) Technoware PSUs, definitely better designed.

Ah okay, got it.
Well, you can use the version in post #42, but again, you will need some 3300 uF and 4700 uF caps for the 3.3V/5V and 12V rails, respectively. Note that there is lots of space around the caps in that revision, so it will be much easier to use 12.5 mm diameter caps. And for the 12V rail, maybe even a 16 mm diameter cap. That should make it a lot easier to find capacitors with such large capacity. If you have the choice, always go for the bigger capacity. The ESR is important, but not as much. In general, any 105C cap will do.

On the other hand, it's not easy to find big caps as you suggested me which fits the narrow PCB places. In fact, the 2200 uF, 16V, 10 mm have been proved particularly expensive!

Even Behemot's?

Anyways, if it's easier, see if at least one 12.5 mm cap will fit on the 12V rail, besides the other 10 mm cap (for the revision in post #14, that is).

Actually, I did not believe that it was so complicated to do. No 20 A multimeter, sorry.

Well, yes, it really isn't. But if you are trying to get the total power consumption accurately, you will also need to do the 3.3V and 5V rails. (Though it's worth mentioning that many motherboards rarely pull anything more than 20-30 Watts combined.)

And i did it, waiting for a post of yours.

Okay, I will chime in there a little later tonight.

El cheapos are not really much precise, but still much more than measuring through wattmeter.

I don't know . You keep saying that, but I can measure my PSUs just fine. In fact, I just tested a laptop power adapter. Measured the input power with my P3 Kill-a-watt and the output with a multimeter (had a resistive load connected). Got efficiencies in the range of 84-87% and a Power Factor of 0.47-0.52. Maybe not very accurate, but still a decent ballpark, I think.

[everell]

Re: Annoying heavy load-failure in Tecnoware Free Silent 500W

The 500 watt Technoware in post #42 looked very familiar. I have an identical pc board which was marketed in the USA as a 350 watt IMicro Model HP-350-G12S. Mine had a bulging 5vsb capacitor. I added a small coil and another capacitor to the 5vsb because the pc board was marked for C-L-C but the coil and second capacitor were missing.

[Behemot]

Re: Annoying heavy load-failure in Tecnoware Free Silent 500W

You can use clamp meter for higher currents. El cheapos are not really much precise, but still much more than measuring through wattmeter.

[omega]

Re: Annoying heavy load-failure in Tecnoware Free Silent 500W

Dear Momaka, thanks for your wide and thorough post!

Even if that 470 Ohm resistor was rated for 1 Watt, 0.61 Watts is still enough to make it run very hot. A good rule of thumb for resistors is to dissipate no more than half of the rated power (so for 1 Watt, that would be 0.5 Watts). I guess the PSU manufacturers here weren't too concerned with that, though

In fact, the scenario is worse than you painted (see pic). Those resistors are 0.5 W, definitely!

If the voltage were in spec with that 21 Watt car bulb on the 12V rail, chances are, they will be in spec when the system is loaded as well (provided the CPU isn't too much of a load on the 12V rail). Speaking of which...

Actually, I wondered whether I have to put some load to the 5 and 3.3 rails.

What did the voltages measure in that PC when it was idle and under load? Sometimes this can be a decent indication if the PSU is actually capable of running the PC in the long term. Generally, if many of the voltages are too close to being out of specs, that indicates you've likely reached the PSU's limits or are getting very close.

Here are the data:
nominal before recap, 21W after, 21W after, no load
12 V 11.9 V 12.1 V 12.4 V
5 V 5.2 V 5.2 V 5.1 V
5 VSB 5.1 V 5.1 V 5.1 V
3.3 V 3.3 V 3.4 V 3.3V
–12 V –11.4 V –11.5 V –11.5 V
The final verdict is of course yours, however :-)

If yes, I say this revision (in post #42) is even more terrible. One cap per rail means you will have to use HUGE caps on the output to keep ripple and noise low.

Yes, the #42 post (Technoware FAL500FS12) concerns a previous revision of #14 post (FAL501FS12). The The so many posts before yours have convinced me NOT to recap these three PSUs of mine. I will rather replace with recapped, third series (i.e. #14) Technoware PSUs, definitely better designed. On the other hand, it's not easy to find big caps as you suggested me which fits the narrow PCB places. In fact, the 2200 uF, 16V, 10 mm have been proved particularly expensive!

Easy, but less accurate way:
Use a Watt meter (like P3 Kill-A-Watt or similar) to measure the power drawn from the wall. These cheap PSUs have 65-75% typical efficiency, so the DC power drawn by the computer will be only 65-75% of what you measure with the Watt meter - and most of that from the 12V rail.
Example:
PC draws 170 Watts from the wall at full CPU load. Then that means the PC is drawing between 170 x 0.75 = 127.5 and 110.5 Watts of DC power (of which, 85-90% is likely from the 12V rail).

Harder, but much more accurate way:
You will need a meter that can measure at least 20 Amps of DC current continuously (most multimeters say they can measure 10 or 20 Amps only for a few seconds). Splice it between the motherboard and PSU on the CPU 4/8-pin 12V connector to measure the current. This will present the majority of the load on your PSU if there is no medium/high power dedicated video card. In your case, you said you had a GeForce 6200, so add 10-20 Watts for that.

Actually, I did not believe that it was so complicated to do. No 20 A multimeter, sorry.

Probably better to start a new thread about that one so we can keep better track of the discussion. Feel free to link the thread here, though (once it's created.)

And i did it, waiting for a post of yours.

[momaka]

Re: Annoying heavy load-failure in Tecnoware Free Silent 500W

What if I put bigger capacitors in the one-cap-per-rail PSU? Let's say I would double capacitance. If I remember, two same caps in parallel is equivalent to a double capacitance cap.

Correct.
Yes, double the capacitance would be better. As I suggested, 3300 uF should be the minimum for any of the rails. 4700-6800 uF would be a lot better, though.

You can try improving these power supplies but no one here can guarantee to you that they are going to be safe to use on a power hungry and expensive system, such as an Intel quad core computer. I wouldn't.

I know and I agree with what you said earlier - I wouldn't bother to fix such a cheap PSU for such a good PC either. But sometimes the choice is not ours to make.

Actually, I've had a similar scenario happen before when I was a tech. Had a friend call and said some computer at his work was down. When I came, his boss made it clear: either repair the faulty PC for this low amount of money or they'd rather throw it away and get a new one, because this one was cheap and the boss was convinced it wasn't worth spending too much money on repair. Now, the PC in question was a decent C2D or C2Q machine if I remember correctly, so I thought it was definitely worth trying to save it (I'm pretty sure they wouldn't have given it to me, even if I said it couldn't be repaired for this cheap - knowing that place, they'd probably take that computer to the parking lot and take turns to run it over with their work trucks rather than give it to someone). Upon opening the case, I was greeted with a cheap Linkworld PSU that turned out to be the problem. I suggested a good quality PSU, but owner was dead-set on spending a low amount of money on repair. So instead I went with a used/refurb 250 Watt HiPro PSU from a local electronics store. I opened the HiPro to clean it a bit, then checked for bad caps and also the fan. After that, swapped it in and called it a night.

Was it a perfect repair? - No, not by my standards. But sometimes that's all you can do in a given situation.
That said, my repair was still at least 2x cheaper than what a fly-by repair tech would have charged to install the same or worse PSU as the original crap.

[momaka]

Re: Annoying heavy load-failure in Tecnoware Free Silent 500W

Momaka, as you suggested me, I also have been desoldering the resistors (see the picture), actually 470 ohm, since most of them have been baked enough for the colour bands to appear all as dark-brown ones :-) Anyway, the dissipating heat is 17x17/470 = 0.61 W which is 20% too much for a halfwatt resistor. Therefore I removed thm all (V = -11.5 to -12 V). I measured every voltage under a 21 W car stop bulb (see picture) connected to the 12 V rail only. Not so much indeed, but better than nothing.

Good riddance with that load resistor! I think the heavy load on the 12V rail will keep the -12V rail voltage is spec too.

Even if that 470 Ohm resistor was rated for 1 Watt, 0.61 Watts is still enough to make it run very hot. A good rule of thumb for resistors is to dissipate no more than half of the rated power (so for 1 Watt, that would be 0.5 Watts). I guess the PSU manufacturers here weren't too concerned with that, though .

If the voltage were in spec with that 21 Watt car bulb on the 12V rail, chances are, they will be in spec when the system is loaded as well (provided the CPU isn't too much of a load on the 12V rail). Speaking of which...

...and the other has been stressed for 10 days of intensive computation and holds super.

What did the voltages measure in that PC when it was idle and under load? Sometimes this can be a decent indication if the PSU is actually capable of running the PC in the long term. Generally, if many of the voltages are too close to being out of specs, that indicates you've likely reached the PSU's limits or are getting very close.

The other pictures are concerning the previous release of the ATX posted in this thread to recap. Here, there is the usual failure under intense load (ATAPI error #11 and #9).

So that's a different revision of the PSU than the one is post #14 below bu same model number?
https://www.badcaps.net/forum/showpo...4&postcount=14
If yes, I say this revision (in post #42) is even more terrible. One cap per rail means you will have to use HUGE caps on the output to keep ripple and noise low.
For the 12V rail, you will want at least 3300 uF. Same with 3.3V and 5V rails. I see you only have a single 1800 uF cap in there. I'm not surprised at all you are getting unstable PC / HDD errors here again.

1. why did they put a 2200 uF 6.3 V cap in the 5VSB rail after the wattage in the specification is even lower with respect to the newer release (10 W vs. 12.5 W)?

If the PSU had other 2200 uF 6.3V caps elsewhere originally, they probably did this because it was cheaper to buy many caps of the same size and capacitance (to reduce cost per cap). Either that, or perhaps because they knew they used low quality capacitors and wanted the 5 VSB cap to last a bit longer.

In general, most older 5VSB designs need to have at least 2000 uF worth capacitance. Whether you do it with two 1000 uF caps or one 2200 uF cap is irrelevant. And having two 2200 uF caps is even better in terms of filtering ripple on the 5VSB, but not necessary. 5VSB topology is usually Flyback type, so low ESR and high ripple-current -rated caps are recommended.

2. is there any concern about the presence of a central, small electrolytic cap between the fast diode and the two small transistors (I had to also replace a 0.25 W 10 ohm resistor - red ellipse - which was rather baked together the PCB)?

If it looks like that cap got overheated and you have a spare cap, then replace it. Looks like that electrolytic cap is for smoothing voltage for the BJT drive circuitry on the secondary side. Not sure how important it is for this design, though.

Also, if that 10 Ohm resistor was really burned, then you should probably also replace ceramic cap C19 as well (ahh, more parts to change, I know! ). The 10 Ohm resistor and C19 form a snubber network, which limits voltage spikes going to the secondary rectifiers. Since C19 appears to be connected to that F16C20C rectifier, then this snubber network is for the 12V rail (which makes sense, because your PCs were/are stressing the 12V rail the most, and those ceramic caps were likely not coping well with the high loads.)

I recommend replacing the ceramic cap in spot C19 with a polypropylene (PP) / film capacitor. Reason for that is because ceramic caps tend to have a terrible thermal coefficient (their capacitance greatly varies with temperature), whereas PP/film caps are much more stable. Use same capacitance as original. Voltage rating of 50V should suffice, but 100V will be even better. If the resistor burns again, you might want to consider replacing with a 1W watt rather than 1/2 Watt.

3. what may the green spots on the ChengX cap black envelopes near the diode heatsink mean? I can see no exploded component, however.

I was just looking at the pictures myself before I read your questions and was wondering the same thing too. :\
Aluminum corrosion is usually green or dark gray, but I think it would be weird for that to go through the paint like this. Hopefully someone else can chime in with answers or ideas.

4. last but not least: how can one MEASURE the actual (instant) load of a PC to an ATX? I mean experimental readings, not average computing in some web sites. I begin to suspect that the PSUs simply can't afford the intense load of my computations!

Easy, but less accurate way:
Use a Watt meter (like P3 Kill-A-Watt or similar) to measure the power drawn from the wall. These cheap PSUs have 65-75% typical efficiency, so the DC power drawn by the computer will be only 65-75% of what you measure with the Watt meter - and most of that from the 12V rail.
Example:
PC draws 170 Watts from the wall at full CPU load. Then that means the PC is drawing between 170 x 0.75 = 127.5 and 110.5 Watts of DC power (of which, 85-90% is likely from the 12V rail).

Harder, but much more accurate way:
You will need a meter that can measure at least 20 Amps of DC current continuously (most multimeters say they can measure 10 or 20 Amps only for a few seconds). Splice it between the motherboard and PSU on the CPU 4/8-pin 12V connector to measure the current. This will present the majority of the load on your PSU if there is no medium/high power dedicated video card. In your case, you said you had a GeForce 6200, so add 10-20 Watts for that.

P.S. I am inspecting the Corsair VS450, which we had speaking about in a previous post. It's indeed full of CapXon and JunFu caps! I will be posting some pics since it's not yet clear to me which cap belongs to which rail.

Probably better to start a new thread about that one so we can keep better track of the discussion. Feel free to link the thread here, though (once it's created.)

[goodpsusearch]

Re: Annoying heavy load-failure in Tecnoware Free Silent 500W

What if I put bigger capacitors in the one-cap-per-rail PSU? Let's say I would double capacitance. If I remember, two same caps in parallel is equivalent to a double capacitance cap.

2x2200uF or maybe 2x3300uF is the maximum you can put. If the caps esr is too low for the psu design then you may have oscillations.

[stj]

Re: Annoying heavy load-failure in Tecnoware Free Silent 500W

2 caps in parallel also lowers the esr

[omega]

Re: Annoying heavy load-failure in Tecnoware Free Silent 500W

2 caps in parallel without a coil in series among them is definitely better than just 1 cap per voltage output.

What if I put bigger capacitors in the one-cap-per-rail PSU? Let's say I would double capacitance. If I remember, two same caps in parallel is equivalent to a double capacitance cap.

[omega]

Re: Annoying heavy load-failure in Tecnoware Free Silent 500W

You can try improving these power supplies but no one here can guarantee to you that they are going to be safe to use on a power hungry and expensive system, such as an Intel quad core computer. I wouldn't.

Thanks for the warning. Beside recapping them, I also replaced the 16A Fast Recovery 12V rail with a Schottky 20A. Hopefully it will increase the MTBFof these 20 PSUs