FSP 5VSB Question

[mockingbird]

Hello,

I've got 4 350 Watt FSP PSUs (FSP350-PN/B), all of them have visibly blown Teapo caps on the 5vsb. All the other caps look OK. I've repaired this exact same PSU model before, but now that I have the opportunity to repair four more of them, I would like some further advice on what to use for the 5vsb.

Here are some photos from other threads that match the layout of my PSU:



Now those are what I'm assuming are the two 5VSB caps that are part of the pi filter that make up the circuit.

Now let's look at the specs for the Teapos, and the Nichicon HE (UCC KY equivalent) I used to replace them:
Teapo "SC" 10V 1000uF 8x20mm 1080 0.065
Nichicon "HE" 10V 1000uF 8x20mm 1050 0.069

Close enough, I thought at the time, and I still think this would work fine. Fact is, I have sold many of these PSUs with small failure rates. It just so happens to be that these PSUs came from an extraordinarily warm environment, had a constant high load on them (1st gen Phenom X4), and were left on 24x7 in a room with little to no ventilation. A testament to Teapos quality when they are used in PSUs... I'm starting to think that PCBONEZ is right and that Teapo SC is actually a decent series because none of the big caps are bulged whatsoever, though I don't have a meter to actually test them.

Now, I don't see any reason why not to use something like this:
Nichicon "HM" 10V 1000uF 8x20mm 0.018 1870

but I have some questions:
1) Is it excessive heat or excessive ripple which kills these 5vsb caps?
2) I know you've posted this a thousand times over PCBONEZ, regarding how to decipher actual cap life from manufacturers' datasheets, but maybe you can explaint o me in laymens terms if caps with higher ripple suppression actually have a better heat tolerance? Will a higher ripple cap last longer than a lower equivalent in the same heat conditions or does ripple rating apply strictly to ripple alone? In which case, will a higher ripple cap help me here if the cause of 5vsb cap failure is heat in the first place?
3) If the cause is heat, should I use some of those automotive caps (Though I doubt I could get an automotive cap with the same ripple level as an ordinary cap - which leads me to believe that ripple ratings have nothing to do after all with heat tolerance.)
4) Do you think the Nichicon HE I used to replace the SC is OK eventhough it has slightly inferior specifications (I know Nichicon rips its specs from UCC - almost everyone rips their specs from UCC - did Teapo just rip the spec from KY series and embellish it a tiny bit)?

[999999999]

Re: FSP 5VSB Question

1) Usually it's the same thing. Their specs are to be taken as complimentary to each other. They are spec'd to last x # hours with Y amount of ripple at up to Z temperature. Higher ripple significantly reduces working hours, sometimes even enough to cause very fast failure. I would never use a cap rated for lower ripple than the original.

2) Caps with higher ripple spec and same temperature spec AND hours lifetime spec AND same size, exhibit less self heating from lower ESR, so their operating temperature is lower in the same ambient (chassis) temperature, contributing to longer life. Yes a higher ripple value will help, though between different brands with variations in electrolyte formula, some may exhibit signs of heat stress sooner than others. This is a good reason to avoid off brand capacitors even if their spec sheet claims the same or similar initial ripple spec, ESPECIALLY if all specs seem similar but they magically manage to make the capacitor physically smaller.

3) Ripple ratings do have to do with heat tolerance but the key is they are not rated to the same ripple values so with less ripple and other design considerations that don't have to include handling high ripple, lifetime can be extended.

4) The nichicon will probably last as long as the Teapo, but realize the Teapo was picked to cut costs. Ideally you would fit the largest capacitor you can fit in the available space without touching anything else, with the lowest ESR reasonably possible and 105C temperature rating. If a larger diameter will not fit (single sided board so you can drill a new hole to increase lead spacing if necessary), seek a taller capacitor though your choices are limited in 8mm diameter. With that in mind, you could go a bit higher in capacitance but better to go up to 16V instead of 10V.

The other factor is whether the extra work sourcing and extra cost shipping parts you may not have, is worth the extension in lifespan or not. While the other capacitors are not venting yet, running continuously at moderate to high load in a room with little ventilation will take its toll eventually.

[PCBONEZ]

Re: FSP 5VSB Question

exhibit less self heating from lower ESR

No, lower ESR is not what does it.

Assuming the 12v max for the ATX spec is the actual Ripple Voltage present Ripple Volts = 120mV [or 0.120v]

Now two caps with different ESR:

Cap 1:
If
Ripple Voltage = 0.120v
ESR: .040 ohms
Then
Ripple Current = 3 amps
Power = 0.36 Watts

Cap 2:
If
Ripple Voltage = 0.120v
ESR: .020 ohms
Then
Ripple Current = 6 amps
Power = 0.72 Watts

Thus lowering the ESR causes the cap to heat MORE, not less.
- The lower ESR results in more Ripple Current -through- the cap.
.

[999999999]

Re: FSP 5VSB Question

Hardly.

If your ripple voltage in-circuit is the same then your ripple current is too (average), both caps were doing the same amount of filtering but the one with half the ESR was heating up half as much.

Lowering ESR does cause higher peak current through the cap but for the purposes of heating the concern is RMS current. That peak ripple value is more quickly reduced to a smaller voltage difference with the lower ESR capacitor.

This is easily demonstrated when someone replaces a crap quality capacitor with a very low ESR version. ESR may be half as much and it's the same circuit it's in, but the temperature is certainly not double the rise over ambient temp, it's lower than the original had - assuming same physical size, metal can electrolytic (all else equal) capacitor.

[everell]

Re: FSP 5VSB Question

Looks like folks agree that there is correlation between ripple current and heat, and that this is the culprit that causes capacitor failure.

I want to add a little more mud to the water. I am of the opinion that an electrolytic capacitor is NOT the best type capacitor to be using in a circuit running at 60 to 130 Khz. A film and foil capacitor would be a much better choice. But just try buying a 2200 ufd film and foil capacitor....and due to the huge size, try getting one into a power supply case!

I have noticed that the 5vsb output filter capacitors are a common failure, especially the first capacitor in the filter. Considering that the 5vsb is always on when the computer is plugged into the wall, whether the computer is turned on or not........and when the computer is OFF but the 5vsb is on, the fan is NOT running, the 5vsb capacitors take a lot of punishment.

One thing I am trying.......is to add a .1 ufd film and foil surface mount capacitor just ahead of the input filter capacitor. There is usually plenty of room to add it to the two traces that are usually close together (5vsb and ground). My thinking is that the .1 ufd film and foil capacitor can off load some of that high frequency current from the electrolytic capacitor and thereby extend its life. What I am using is Digikey part number 399-1805-1-ND

[PCBONEZ]

Re: FSP 5VSB Question

1) Is it excessive heat or excessive ripple which kills these 5vsb caps?

Excessive Ripple Current essentially -is- Excessive Heat.
There are two sources of heat.
- Internally created Heat from Ripple Current passing through the cap.
- Heat from outside the cap.

The Ripple Current rating is the max Current that won't overheat the cap Internally.

2) I know you've posted this a thousand times over PCBONEZ, regarding how to decipher actual cap life from manufacturers' datasheets, but maybe you can explaint o me in laymens terms if caps with higher ripple suppression actually have a better heat tolerance? Will a higher ripple cap last longer than a lower equivalent in the same heat conditions or does ripple rating apply strictly to ripple alone? In which case, will a higher ripple cap help me here if the cause of 5vsb cap failure is heat in the first place?

"Will a cap rated for more Ripple Current be less affected by ambient Heat."
- No. - 105C is 105C regardless of where it comes from.
If ambient Heat is what is heating the cap to 105C then what it's Ripple Current rating is won't matter at all.
-
In reality a cap's Internal heat comes from -both- ambient and Ripple Current.
Amount from ambient plus X amount from Ripple Current = Total Internal Heat.
[Presumably] A cap with a higher Ripple Current rating 'should' heat less for a given amount of Ripple so the total heat input should be less and it should level off at a lower Internal temperature.

3) If the cause is heat, should I use some of those automotive caps (Though I doubt I could get an automotive cap with the same ripple level as an ordinary cap - which leads me to believe that ripple ratings have nothing to do after all with heat tolerance.)

If you can find them with appropriate Ripple and ESR that isn't a bad idea.
I've looked into doing that a few times and never found suitable high temps caps.
That said:
Several brands have come out with new high temp series in the last year or two and some have respectable ESR. I didn't look into it in great detail because when I saw them I was updating my cache of data sheets and I didn't want to get side tracked.

4) Do you think the Nichicon HE I used to replace the SC is OK eventhough it has slightly inferior specifications (I know Nichicon rips its specs from UCC - almost everyone rips their specs from UCC - did Teapo just rip the spec from KY series and embellish it a tiny bit)?

For a 1000uF I'd probably spilt the difference and use a ZL or a KZE. [Easy to find.]
.

[999999999]

Re: FSP 5VSB Question

Looks like folks agree that there is correlation between ripple current and heat, and that this is the culprit that causes capacitor failure.

I want to add a little more mud to the water. I am of the opinion that an electrolytic capacitor is NOT the best type capacitor to be using in a circuit running at 60 to 130 Khz. A film and foil capacitor would be a much better choice. But just try buying a 2200 ufd film and foil capacitor....and due to the huge size, try getting one into a power supply case!

I have noticed that the 5vsb output filter capacitors are a common failure, especially the first capacitor in the filter. Considering that the 5vsb is always on when the computer is plugged into the wall, whether the computer is turned on or not........and when the computer is OFF but the 5vsb is on, the fan is NOT running, the 5vsb capacitors take a lot of punishment.

One thing I am trying.......is to add a .1 ufd film and foil surface mount capacitor just ahead of the input filter capacitor. There is usually plenty of room to add it to the two traces that are usually close together (5vsb and ground). My thinking is that the .1 ufd film and foil capacitor can off load some of that high frequency current from the electrolytic capacitor and thereby extend its life. What I am using is Digikey part number 399-1805-1-ND

I agree but why add only 0.1uF? Instead of $0.40 each for those Digikey 399-1805-1-ND, some electronics surplus sites have much less expensive caps like the following 0.56uF/50V for 3 cents each. They may exhibit more thermal drift than some but in this type of application it wouldn't matter:

http://www.bgmicro.com/56uf50vmonoradialcap.aspx

... or of course the more clear solution in this case is a solid/polymer capacitor instead of the 'lytic... these things just cost more but lifespan at elevated temp is MUCH better.

[mockingbird]

Re: FSP 5VSB Question

PCBONEZ... Thanks for the clear and concise answers... And I would also like to thank you "999999999", eventhough your answers are contradictory to PCBONEZ...

Yes, I can get polymer caps, but their ESR is really low, and according to PCBONEZ, lowering the ESR of the cap in this application defeats its purpose in the first place... With that low an ESR, it will perform on an inferior level than the Teapo SC that was there in the first place... (Though I admit I have zero idea of the math necessary to compute the higher heat to lower ESR ratio).

For a 1000uF I'd probably spilt the difference and use a ZL or a KZE

Which would you pick from these:
10V 1200uF 8x20mm Panasonic "FK" 0.044 1220
6.3V 1200uF 8x20 UCC "KZE" 0.041 1250
6.3V 1200uF 8x20 Panasonic "FM" 0.030 1560

Thanks.

[shovenose]

Re: FSP 5VSB Question

the 10V ones, definitely!

[999999999]

Re: FSP 5VSB Question

PCBONEZ... Thanks for the clear and concise answers... And I would also like to thank you "999999999", eventhough your answers are contradictory to PCBONEZ...

Yes, I can get polymer caps, but their ESR is really low, and according to PCBONEZ, lowering the ESR of the cap in this application defeats its purpose in the first place... With that low an ESR, it will perform on an inferior level than the Teapo SC that was there in the first place... (Though I admit I have zero idea of the math necessary to compute the higher heat to lower ESR ratio).

False. With all else equal, lower ESR = lower temperature rise in the same circuit.

Understand that if your ripple voltage is the same with either capacitor you choose, your capacitor is charging to that voltage level so long as it is a type rated for the frequency used -- it is specified to have the ESR it does at a certain frequency for comparison purposes.

With both charging to the same voltage (and discharging to the same low level, so we have the same peak ripple voltage) and having the same capacitance, you have more power loss in the higher ESR capacitor AND the duty cycle of the switching circuit has to rise slightly higher to make up for that loss while still keeping same regulated output level. Lower ESR capacitors not only run cooloer themselves but help other components operate at a little lower temperature too.

[mockingbird]

Re: FSP 5VSB Question

Why? I only need a 5V cap here... The only reason they used 10V in the first place was because of the tendency of higher voltage caps to have better ripple ratings... Why not use a lower voltage cap that still meets the voltage requirement and has a -better- ripple rating... And Newark has a great deal right now on the KZE...

Only 15 cents

[999999999]

Re: FSP 5VSB Question

^ because you can do both, use a cap series with a better rating and improve it even more by using one with a higher voltage rating, which lowers ESR more without increasing the inrush surge current through the circuit as much by not having a large increase in capacitance.

They did the equivalent in the first place when they picked a 10V part, it was just cheaper for them to do that than to pick the higher quality cap rated for lower voltage, it was NOT because the ripple rating is some special value you need to try to duplicate in this particular circuit.

[PCBONEZ]

Re: FSP 5VSB Question

Hardly.

If your ripple voltage in-circuit is the same then your ripple current is too (average), both caps were doing the same amount of filtering but the one with half the ESR was heating up half as much.

Lowering ESR does cause higher peak current through the cap but for the purposes of heating the concern is RMS current. That peak ripple value is more quickly reduced to a smaller voltage difference with the lower ESR capacitor.

This is easily demonstrated when someone replaces a crap quality capacitor with a very low ESR version. ESR may be half as much and it's the same circuit it's in, but the temperature is certainly not double the rise over ambient temp, it's lower than the original had - assuming same physical size, metal can electrolytic (all else equal) capacitor.

Wrong.

You can't get around the law of conservation of energy like that.
It doesn't work.
.
.
The switcher section is going to create the same energy [watts] in the form of Ripple regardless which cap is there or even if there is no cap at all.
[I used volts because it's the simplest way to represent that power without getting into heavier/lengthy math.]
.
The Ripple Current for that amount of Ripple energy has two choices.
- Go through the load.
- Go through the cap.

Lower ESR means more Ripple Current goes through the cap - and less through the load.
- Which means the load gets less heating from Ripple.
- Which means the cap gets more heating from Ripple.
(The total Heat dissipated doesn't change, only -where- it is dissipated changes.)

There is NO WAY to reduce the amount of Ripple energy that goes to the load without sending that energy (which becomes heat) somewhere else.
(That's IAW the Law of Conservation of Energy.)

Somewhere else is the capacitor.
.

Thus when you take a cap in a given circuit and switch it with a lower ESR cap the new cap DOES dissipate MORE heat.
There is no way to get around it. - IT DOES. - And the ONLY thing running cooler is the load.
.

[PCBONEZ]

Re: FSP 5VSB Question

Also:
.
Anytime you are dissipating less heat from Ripple in the capacitor you are dissipating MORE heat from Ripple in the load.
.
If it actually worked that way you would want to raise ESR, not lower it.
.

[999999999]

Re: FSP 5VSB Question

^ With all due respect, this is very well established fact. The switcher is not creating the same energy because it has a feedback loop. Further, high voltage spikes would cause downstream parts to consume more current.

If you have no cap there or insufficient capacitance or too high an ESR your power supply is less efficient.

If the ripple current goes through the load at a higher value because of your higher ESR capacitor, you used more power, but the capacitor STILL charges up to the voltage, just not as quickly, but it does charge to it so it took the same amount of current to bring a given uF capacitance value up to that voltage but with higher loss due to the higher ESR. If the cap hadn't charged up to that voltage, the average sensed voltage in the circuit would be lower and the switching duty increases but actually BOTH things happen.

Lower ESR does not mean more ripple current goes through the capacitor, it means the capacitor can handle more ripple current at a given frequency but that at any particular ripple the cap sees in use, it heats up less. Regulated power supplies do not regulate based on what the load is doing, they regulate based on the sensing circuit near the capacitor, deciding to increase or decrease switching duty based on the sensed voltage the capacitor is holding that portion of the circuit to.

It is very clear scientifically proven fact that lowering ESR lowers capacitor temperature in switching supply circuits. The whole point of specifying the very low ESR for capacitors is so you know they will heat up less in switching supply circuits.

Meh, we (or I and someone) argued about this a few years back so it seems not worth the time to rehash again... just think about it though, if low(ering) ESR was not important then they would not use low ESR capacitors at all. Why is it important? What effect is there? Reduced heat.

Well I'm done, not going to keep arguing this anymore...

[PCBONEZ]

Re: FSP 5VSB Question

^ With all due respect, this is very well established fact. The switcher is not creating the same energy because it has a feedback loop. Further, high voltage spikes would cause downstream parts to consume more current.

Bogus and/or irrelevant.
.
There is no feedback loop that has to do with or is fast enough to even detect Ripple. It is for the DC component ONLY.
(If there was then we wouldn't need caps to start with.)
.
Further you don't compare what happens with two caps and an ESR change when the rest of the circuit is in a transient.
(Thus the DC load would be steady state and feedback would be at null during any valid comparison.)
.
To do a PROPER comparison you assume the DC load is the same.
If the DC load is the same then so is the amount of Ripple being created.
.

[PCBONEZ]

Re: FSP 5VSB Question

It is very clear scientifically proven fact that lowering ESR lowers capacitor temperature in switching supply circuits.

Wrong.
No such thing have ever been shown because it isn't true.
~Thousands~ have assumed/rationalized that is true and repeated it over and over but that doesn't make it true.

The whole point of specifying the very low ESR for capacitors is so you know they will heat up less in switching supply circuits.

Wrong again.
The point in specifying low ESR caps is to reduce heat of the OTHER components.
The caps do that by rerouting the Ripple Current [and thus the Heat from Ripple] to themselves.
.

[999999999]

Re: FSP 5VSB Question

^ I want to know what direction you pray towards and which god it is because what you claim is not possible in a rational world using science. Using science the whole point of lowering ESR as much as possible is better performance and cap life in a SMPS.

To argue the opposite is a very strange madness and not one worthy of my time. I should not have even come back for this reply, your ego/trolling crazy post just made me give it one more try to promote sane intelligent science into a topic when it doesn't seem like that will make a difference.

I urge you to discuss this with capacitor manufacturers since you seem to think you know more than they or the SMPS industry does.

[PCBONEZ]

Re: FSP 5VSB Question

The rational world that uses science adheres to it's laws.
One of those Laws is Conservation of Energy which you either don't understand or ignore.

Unlike yourself I don't believe in gods. - I can also do math.

Cap manufacturers don't say anything that supports your theory -except- to someone that doesn't comprehend what they're reading.

And apparently you have -no idea- why low ESR caps are used.
- It is absolutely NOT to make the SMPS live longer.
The SMPS would live just as long with good quality GP caps.
The difference is lacking low ESR caps it would be sending more Ripple to the load.
- The quality of the power to the load is the -why- in using using low ESR caps.
.

[999999999]

Re: FSP 5VSB Question

The purpose of the
bypass capacitor is to store energy and to supply the high
instantaneous currents during switching and to prevent
the supply from drooping. Any losses within the capacitor
limit its ability to supply this current. The losses also
cause heat to be dissipated within the capacitor
contributing to reduced system efficiency. In portable
electronic devices, lower system efficiency reduces
battery life or requires the use of physically larger
batteries. In order to provide adequate supply decoupling
and maintain low loss, the capacitor must have low
impedance at high frequencies. This requires a capacitor
which has low ESR.

Capacitors with too
high an ESR will self-heat too much and not regulate the
current properly. Obviously the self-heating will also
reduce the operating life of the capacitors, and the switch-ing power supply performance and life will be reduced
accordingly.

http://www.epn-online.com/page/new18...apacitors.html

These lower ESR capacitors can help engineers to achieve further improvements in the goals of efficiency, reliability and system cost.

Lowering ESR reduces self heating, resulting in lower power losses and increased efficiency. It also allows the capacitor to handle larger ripple currents without exceeding the target temperature.

http://books.google.com/books?id=Z1b...d=0CKMBEOgBMBA

However, foil capacitors are generally described by their manufacturer as being more suitable for high frequency pulse applications because their lower ESR reduces self-heating.

https://cdn.badcaps-static.com/pdfs/...3127f28bdd.pdf

A general formula for ripple current capability is derived
from elementary physics. The power generated (Pgen)or rate
of heat generation in a capacitor is:

Pgen = I^2rms x ESR

where Irms is the rms (root mean square) value of the ripple
current (I - in amperes), and ESR is the equivalent series re-sistance of the capacitor (in ohms).

How many more quotes do you need? These came from only the first page of Google search results. Show accredited references that lowering ESR causes more heat. It's mind boggling that anyone would think it does.