Replacing Caps in Power supplies

[davmax]

Most power supplies that use crap caps also place higher than necessary voltage rating caps, often 16Volt on the 5 volt output and some on the 3.3 volt .

If replacing caps with high quality caps check where the cap is and the colour of the output wires to which it is connected. This can save you getting unnecessary wide diameter caps.

With Orange wires use 6.3V
Red wires use 6.3V
Yellow use 16V
Purple use 6.3V
White use 6.3V careful this is a -5V output
Blue 16V careful -12V

[japlytic]

Re: Replacing Caps in Power supplies

Typically, I would recommend 10V units for 5V lines (red, purple, white).

[davmax]

Re: Replacing Caps in Power supplies

That is interesting. Why 10Volt? Is peak ripple a concern?

Take purple line. The cap on output is fed by a 5 volt linear regulator and no ripple.

[davmax]

Re: Replacing Caps in Power supplies

Japlytic. I agree that 10V is worthwhile away from the output connection of the main +5V rail ie the cap before the filter choke is best 10V. In this respect my previous listing could mislead. Thanks for raising the point.

[999999999]

Re: Replacing Caps in Power supplies

Most power supplies that use crap caps also place higher than necessary voltage rating caps, often 16Volt on the 5 volt output and some on the 3.3 volt .

If replacing caps with high quality caps check where the cap is and the colour of the output wires to which it is connected. This can save you getting unnecessary wide diameter caps.

With Orange wires use 6.3V
Red wires use 6.3V
Yellow use 16V
Purple use 6.3V
White use 6.3V careful this is a -5V output
Blue 16V careful -12V

You must be in a different geographical region, the vast majority of poorer PSU in midwest US don't use 16V on 5V or 3V, but 10V is common. This is not necessarily a problem, it may be a good thing as the lower quality caps seem to be rather aggressively rated for voltage as evidenced by their smaller can size per claimed ESR and uF values. Either that, or they're outright lying about the uF spec.

If an "unnecessarily wide diameter cap" will fit, it can be the better choice. There is seldom a preference to fit a smaller diameter part where a larger will fit, if there is nothing it's too close to such as a load resistor or hanging over PCB jumpers that could theoretically short. Larger diameter cap has longer life and lower ESR given same height, or better with larger diameter parts typically coming in taller can sizes as well.

To put it more simply, suppose a PSU had room on the PCB for a 12.5mm diameter cap, but originally had a 10mm, 2200uF, 10V cap installed on the 5V rail. If you replaced it with a 12.5mm, 2200uF, 16V, that is an upgrade and superior even if the exact same family of cap.

Given a choice in same cap size, it is generally preferred to use higher mfd value before higher voltage - up to a point, past which the inrush current gets exceedingly high but it's not typically an issue on a PC PSU. Even so, larger can size in same cap family from increased voltage rating is still a positive thing so long as it isn't also resulting in significant reduction in uF value. Even some reduction in uF value is anticipated on a few poor PSU as they used poor quality caps with thin layers in an attempt to get more uF from any given size can and the replacement higher quality cap will not have as high a uF value in a can size that can fit in the space provided.

[davmax]

Re: Replacing Caps in Power supplies

999999999 you have made some good points and are very correct in that a larger diameter cap is usually better. Whilst placing a 16v cap say instead of a 6.3V is an upgrade in voltage and possible ripple rating it is a fact that electrolytics should be operated not too far below their voltage rating so that they form correctly with the specified capacitance. This is why caps are created for different voltages. 16volt is a bit too high for a 5 volt rail.

[999999999]

Re: Replacing Caps in Power supplies

I haven't ever had a cap performance issue from the slight difference in 6.3 to 16V. I suspect it would be more of a problem if the voltage rating were far higher

[davmax]

Re: Replacing Caps in Power supplies

No doubt you are right. If using high quality caps I cannot see any real benefit in placing larger diameter caps with excess rating. If you are happy that is all that matters.

[999999999]

Re: Replacing Caps in Power supplies

The benefit is as already mentioned.

1) Larger cap has lower ESR, (all else being equal except the higher voltage rating)

2) Larger cap rated for longer life (all else equal again)

3) Larger cap has lower thermal density (")

Even if we ignored #1, the lower thermal density and larger physical size (larger surface area means higher rate of heat transfer) means lower operating temp IF it had the same ESR. Now combine this benefit with the lower ESR and the longer life rating, and you will typically get a substantially longer life from the cap.

That doesn't mean I'd automatically grab for the higher voltage rated cap, the higher uF / larger sized cap alternative has similar benefits and ripple reduction too.

[gonzo0815]

Re: Replacing Caps in Power supplies

I absolutely agree wit 999999, loock at the spec sheet, and youl will find that an Panasonic FM with 12,5 diameter will have up to 7000h lifetime at 105°c. I would say this is awsome, but shure, this would probably be overkill in a good ventilated PSU.
The high voltage cap will probably develop a thining of his oxid layers, but this is no problem.

[davmax]

Re: Replacing Caps in Power supplies

I read you the first time. I simply said I (personally) could not see the benefit. Most PS I have worked on have the caps fitted tightly with very little space, therefore if good quality caps of correct diameter and voltage will do the job well why change. I have agreed with your proposal otherwise if there is space but feel that the gain of excess diameter and voltage is insufficient.

Since you have raised the heat issue, yes in general what you say is true in that some caps types do have longer life with wider diameter. But we should not generalise. If you look the specs for Rubycon MCZ and Samxon GC (ultra low ESR caps) you will find the life spec does not change with diameter.

Also the caps lifetime specs often state "After applying rated voltage and ripple current at 105C ". I take this to mean that the maximum ratings can be applied at 105C which seems to state that the externally generated temperature rather than the internal temperature is the major factor in capacitor life. In a mobo example the external temperature comes from both conduction (mobo planes and hot airflow from the CPU) plus radiated heat.
It is therefore difficult to always state how effective the cap surface area is at removing heat, when in some cases it can aid heat absorbtion.

Hope this is helpful

[PeteS in CA]

Re: Replacing Caps in Power supplies

I take this to mean that the maximum ratings can be applied at 105C which seems to state that the externally generated temperature rather than the internal temperature is the major factor in capacitor life.

It's the capacitor core temperature that matters, and heat is heat whether it comes from the ambient, the ripple current or both. That's the word I've heard from both UCC and Rubycon technical people.

[999999999]

Re: Replacing Caps in Power supplies

I read you the first time. I simply said I (personally) could not see the benefit. Most PS I have worked on have the caps fitted tightly with very little space, therefore if good quality caps of correct diameter and voltage will do the job well why change. I have agreed with your proposal otherwise if there is space but feel that the gain of excess diameter and voltage is insufficient.

Unfortunately it's true, far too often a cap is landlocked and a larger size just won't reasonably fit. I think we are considering when it would fit and in that case, it's not exess diameter but rather your part has real benefits instead of as much of a compromise the manufacturer made to control costs.

Since you have raised the heat issue, yes in general what you say is true in that some caps types do have longer life with wider diameter. But we should not generalise. If you look the specs for Rubycon MCZ and Samxon GC (ultra low ESR caps) you will find the life spec does not change with diameter.

Remember that a lifespan spec is till the point at which the other parameteres fall below a threshold, that a larger diameter cap may fall below it's own (better) parameter thresholds but still remain superior to what a smaller diameter (but still as tall or taller as we're considering same or very similar uF, same family, just higher voltage rating) would've degraded to at that point.

It will remain closer to the functional requirements of the circuit, the remaining question being whether the (PSU or whatever) had need to last that long but since the caps in LC SMPS filtering do tend to be among the shorter lived parts, and since the PSU may be exhibiting a tendency to put these caps under fair stress (else they wouldn't have failed and need replacement), it is better to use the higher voltage part when possible as we can't appreciate a few cents cost difference replacing a single-digit # of parts like a PSU manufacturer could over their entire parts selection process.

Also the caps lifetime specs often state "After applying rated voltage and ripple current at 105C ". I take this to mean that the maximum ratings can be applied at 105C which seems to state that the externally generated temperature rather than the internal temperature is the major factor in capacitor life. In a mobo example the external temperature comes from both conduction (mobo planes and hot airflow from the CPU) plus radiated heat.
It is therefore difficult to always state how effective the cap surface area is at removing heat, when in some cases it can aid heat absorbtion.

No, it will not aid in absorbtion, if the ambient air is X degrees the cap would be at that temp no matter the diameter plus it generates heat too for further rise over that (anytime a cap of low ESR is called for). If the heat is sunk to it by the copper, the primary path is through the leads not the jacketed shell or plug so again we have a larger surface area on the cap to radiate away the heat and lower thermal density.

Combine that with the typically lower ESR (or even ignoring it) the cap will always be cooler unless the size interferes with airflow around it or places it practically against a tall, hotter part. These latter two things can be determined per-PSU instead of generalized.

There is a very real benefit from the larger cap (all else equal but next step (sometimes 2 steps) up in voltage), quite possibly as much or moreso than choosing one of the premium grade caps versus a more typical low-ESR rather than lowest-ESR type. You will also find this happening in better, more expensive equipment PSU rather than space and cost constrained PC PSU, though it is done within limits as the larger size allows higher uF too or instead.

In short, the benefit is very significant, and the only good reason not to use slightly higher voltage rated cap is if you can instead use higher uF, or due to space constraints or parts availability. It is as significant as most choices relating to cap lifespan to get this larger size / similar otherwise. The only remaining question is which parameters to change in same family of cap, voltage or uF, with uF typically being the more desirable.

If I were going to generalize, you might see double the life with this one change alone. Sometimes more, sometimes less (than double), but this is considering a typical PSU use where one will fit. I don't mean to be an advocate for it though, I'd as soon go with higher uF (same cap family) if I had the two options and uF wasn't getting excessive already.

[davmax]

Re: Replacing Caps in Power supplies

Pete. You are absolutely right. That the core temp is what matters. I am trying to get across the point that hot systems kill caps. Take the manuafacturer's 105C max, that temp does not all come from internal heat the major heat component is external. Obviously the core is at a higher temp at rated ripple.

999999999 I agree with you on most points. I believe your idea to use larger , longer life caps is ideal for 24/7 systems. We both agree that it must be possible to fit them. Using 7000hr caps is certainly going to stretch life by 3.5 compared with a 2000hr cap.

Increasing capacitance will help particularly those output voltages that may not have huge cap filtering on the mobo.

Oh and the caps do absorb heat through the leads and case. Hot system kill caps. I am sure you have experienced this.

[PeteS in CA]

Re: Replacing Caps in Power supplies

You can place unconnected caps in a hot environment and they will swell (hydrogen gas). With aqueous electrolyte caps, that seems to happen around 120C, and with non-aqueous electrolyte caps, around 140C-150C. Obviously (hopefully?) the ambient around a cap in a VRM circuit won't be anywhere near hot, but if the ripple current and thermal resistance from core to ambient cause the core to be 40C above ambient, an 80C ambient around the cap (hypothetical numbers) will kill the cap pretty quickly. There is considerable variation vendor to vendor, series to series and value/voltage to value/voltage in how fully the can is filled and (I believe) in the thermal resistance from core to ambient. Thus, a part that skimps on the amount of foil by using more highly etched foil (greater surface area, yielding higher capacitance per square cm.) may have a higher thermal resistance core to ambient. Such a cap would have a higher core temp than a better cap with otherwise identical conditions - I(ripple), ESR and ambient temperature.

[PeteS in CA]

Re: Replacing Caps in Power supplies

Oh and the caps do absorb heat through the leads and case.

Leads, e.g. when connected to a voltage or ground plane that is also serving as a "heatsink" to a hot MOSFET or Schottky rectifier and close to that device.

[davmax]

Re: Replacing Caps in Power supplies

Thanks for you input Pete. Yes it is all a very hazy area because we do not know the indivual characteristics, chemical , foil and thermal resistance. We only have the 105C reference and tolerance to full ripple to work with. Hopefully no design approaches or exceeds the cap maximum ripple value. We have to assume 105C represents case temp, although I suspect this is the ambient used in testing and the rated ripple actually raises the case still further, we are not told, so taking it as case temp is on the safe side. Certainly case temp is easy to measure and useful.

I guess noboby is telling us the critical core temp, but your testing gives us a good clue.

Thanks again.

[999999999]

Re: Replacing Caps in Power supplies

105C is the cap (core) temp, the lifespan rating is not valid if your cap gets hotter than 105C. Ambient temp can be anything, they don't try to resolve that variable but rather the only known constant which is their part.

[davmax]

Re: Replacing Caps in Power supplies

999999999. Are you sure 105C is core temp? I would like a link to a reference comfirming that. Core temp rating would make it very difficult to rate a capacitor because a core temp rating would require knowing thermal resistance etc where case temp is easy to measure and I believe used to estimate life. Shock news for me, I would like to know more.

[999999999]

Re: Replacing Caps in Power supplies

999999999. Are you sure 105C is core temp? I would like a link to a reference comfirming that. Core temp rating would make it very difficult to rate a capacitor because a core temp rating would require knowing thermal resistance etc where case temp is easy to measure and I believe used to estimate life. Shock news for me, I would like to know more.

There are a couple ways of looking at it. One is to consider the core and the other is to use an equation factoring for the ripple current, ESR, ambient temp, radiation (as effected by surface area and deltaT, etc towards the goal of estimating the core based on a case measurement or (less accurately) only presuming generated heat vs radiation rate. I'm sure you can imagine that in a highly dynamic PC load scenario, it could be very difficult to mathematically solve to any degree of precision. If you would rather call 105C the ambient you essentially do the reverse of the above mathematically, still having to factor for the further thermal rise of the cap, deviations from spec'd ripple, etc. So the short answer is, neither is a complete, fixed temp spec, rather only in context of all the other specs as constants.

Here's a related PDF, probably the shortest path is to just do as it mentions measuring the temp on the top then using their table to estimate core.
https://cdn.badcaps-static.com/pdfs/...0dc91bca69.pdf