Capacitor Life expectancy

[davmax]

Electrolytic Capacitors Temperatures and Life Expectancy

A Badcaps thread covering the replacement of electrolytic capacitors in power supplies developed into a discussion about capacitor sizes, temperatures and life expectancy. This has stimulated the attached contribution.

[davmax]

Re: Capacitor Life expectancy

If there is a need to link into the Elna document please use this link http://www.elna-america.com/products..._technotes.pdf

[gonzo0815]

Re: Capacitor Life expectancy

Great sumary Dave.


To make this complete, it should be mentioned, that the biger sized caps of most series (except the ultrea low esr caps like MCZ or equivalent series) like Panasonic FC or FM have right from the start a loger endurance.
So for Panasonic FM theyr stated with 7000h (12,5) vs 5000 and 3000h for the smaller caps.

If you do the derating wich means doubling per 10°C under spec temp. you will gain a very long endurance even at elevated temperatures.
Considering the fact, that the bigger diameter caps (thus more voltage or higher capazitance) do have in most cases the better (lower) esr too, i think the picture will be mutch clearer abouth the benefits of large diameters.

I duno, why this is not aplicable to the ultra low esr caps, but i assume, that the vendor would not like to assure that a long life, as the low esr types are problably a tight fit an not completely under control.....
So they rated all diameters for the same lowest practical value e.g. 2000h

[davmax]

Re: Capacitor Life expectancy

Ganzo,

The Pany's you mention are not the same as aqueous based caps that have shorter life and lower ESR.

I gave Ultra low ESR cap examples but the Elna paper does not distinguish between aqueous and non aqueous. The principles are the same eg the core temp is higher in a larger diameter cap at a given case temp. My comparisons are within a given type of capacitor. Sure you can get long life caps, no problem. That was not part of the issues being covered.

[PeteS in CA]

Re: Capacitor Life expectancy

The FM series has an aqueous electrolyte.

[davmax]

Re: Capacitor Life expectancy

Thanks Pete. I stand corrected. It appeared to me that as they had about twice the ESR of the Ultra low caps that they not. Apparently there is some difference in electrolyte.

[linuxguru]

Re: Capacitor Life expectancy

Yup - Panasonic FC/FK were non-aqueous, FM aqueous. I'm not sure about the Malaysian FF/FJ/FL, but I suspect that FF is non-aqueous and FJ is aqueous (the latter is much smaller for a given rating).

I still trust the FC and FK more than the others. FF is OK, but the silver stripe suggests that it is not low ESR in comparison to the gold-striped FC and FK.

[gonzo0815]

Re: Capacitor Life expectancy

May be there is not that mutch water in the FM ones, as they have very long life compared with the other water based ultra low esr caps.

But why have the non water based FC`s a lower endurance than FM?
May be Panasonic was a little more on the safer side with the datasheet on them?

[PeteS in CA]

Re: Capacitor Life expectancy

As I understand it from talking to "major manufacturers'" technoids, series such as KY, HE and YXH are lower in water content, series such as KZE, HE and ZL in the midrange, and series such as KZG, HM/HN/HV and MCZ are high water content. Exactly what percentages is proprietary and they don't give out that info. Keep in mind that "water" in this context is probably ultra-pure, far removed from what you get from your faucets at home.

[davmax]

Re: Capacitor Life expectancy

Linuxguru and Pete, again interesting. Something I have not really researched. I guess most times we just have to go by the specs given.

[davmax]

Re: Capacitor Life expectancy

The original write up has ben updated to clarify a few details, particularly in the conclusions.

[999999999]

Re: Capacitor Life expectancy

There is something a bit (ambiguous?) in the document. The measurement of the cap top is not what is applied to the temp figure, it is only used as the only practical (non-destructive) method to gather data to input to the table to arrive at the working core temp. Once the temp of the top is taken and applied to the chart, the chart figure is the only one applicable towards the other cap specs and for purposes of derating or max 85, 105, 125C, etc.

Thus, the larger cap does have a larger conversion factor to approximate the core temp by the table, but it is only as a result of the expectation that the lager the cap, the larger the surface area is.

I mean, the statement in this new document that reads "This would imply shorter cap life for the larger diameters" could be misconstrued, since in an actual cap implementation, the temp measured on the larger cap top would already be lower (if the same ESR, not even considering the expected lower ESR), as much as the difference between respective conversion factors indicated in the chart.

It doesn't in any way change the conclusion that the case temp is an important indicator, but it does effect the summary conclusions at the end. #3 - The manufacturer does provide the core, not top, temp. #4 - Increasing diameter has a more substantial gain, because we know the heat is lower and the dissipation rate is higher. It's granted that the top of the larger cap won't be as hot, all else as equal as possible (same cap series and only voltage or capacitance accounting for the increased cap physical size). Even this is just a ballpark figure, as with larger diameter, increased height also has benefits but to a lesser extent. #5 - It isn't contrary to the Elna document, their doc says the same thing but the message is poorly conveyed. We can assume the larger cap is already cooler running, not that the case would be the same and the core hotter still.

Until a cap that is too large is used such that it's shell is in very close phsyical proximity to something very hot (like literally touching a load resistor in a power supply), the increased size also improves dissipation of heat conducted through the leads from the nearby power circuitry FETs, etc., sunk to the board copper.

[davmax]

Re: Capacitor Life expectancy

There is something a bit (ambiguous?) in the document. The measurement of the cap top is not what is applied to the temp figure, it is only used as the only practical (non-destructive) method to gather data to input to the table to arrive at the working core temp. Once the temp of the top is taken and applied to the chart, the chart figure is the only one applicable towards the other cap specs and for purposes of derating or max 85, 105, 125C, etc.

99999999 you do not seem to have read or inderstood the Elna paper or my contribution.

Thus, the larger cap does have a larger conversion factor to approximate the core temp by the table, but it is only as a result of the expectation that the lager the cap, the larger the surface area is.

The Elna paper clearly states that the core temp of larger diameter cases is higher. Also read my discussion. It is clear that if the case temp is the same as I stated then the core temp is higher in a larger diameter cap.

I mean, the statement in this new document that reads "This would imply shorter cap life for the larger diameters" could be misconstrued, since in an actual cap implementation, the temp measured on the larger cap top would already be lower (if the same ESR, not even considering the expected lower ESR), as much as the difference between respective conversion factors indicated in the chart.

Again you have not read and understood. Yes the case of a larger cap can loose more heat, it just depends what the environment achieves in heating the whole cap. And I thought I had made it clear that the ESR contribution to heat is minor. Quoting Elna it is desirable for deltaT to be less than 5C. My calcs also indicate the very small heat loss in ESR at full rating.

It doesn't in any way change the conclusion that the case temp is an important indicator, but it does effect the summary conclusions at the end. #3 - The manufacturer does provide the core, not top, temp. You need to look at Elna and other specs they do not quote core temp.#4 - Increasing diameter has a more substantial gain, because we know the heat ( look at the calcs I presented and the relatively minor gains there is no guessing or broad statement, they are factual calculations) is lower and the dissipation rate is higher. It's granted that the top of the larger cap won't be as hot, all else as equal as possible (same cap series and only voltage or capacitance accounting for the increased cap physical size). Even this is just a ballpark figure, as with larger diameter, increased height also has benefits but to a lesser extent. #5 - It isn't contrary to the Elna document, their doc says the same thing but the message is poorly conveyed. We can assume the larger cap is already cooler running, not that the case would be the same and the core hotter still. Remember environmental temp is the major contributor and this comes back to case temp. I have even given two scenario examples in the contribution to help with understanding about heat injection that is most often far greater than the core heat.

Until a cap that is too large is used such that it's shell is in very close phsyical proximity to something very hot (like literally touching a load resistor in a power supply), the increased size also improves dissipation of heat conducted through the leads from the nearby power circuitry FETs, etc., sunk to the board copper. This statement does not make a lot of sense. It sounds logical but needs to be more factual. It can be saying that a larger cap absorbs more heat from power circuitry through the leads.

[999999999]

Re: Capacitor Life expectancy

There is something a bit (ambiguous?) in the document. The measurement of the cap top is not what is applied to the temp figure, it is only used as the only practical (non-destructive) method to gather data to input to the table to arrive at the working core temp. Once the temp of the top is taken and applied to the chart, the chart figure is the only one applicable towards the other cap specs and for purposes of derating or max 85, 105, 125C, etc.

99999999 you do not seem to have read or inderstood the Elna paper or my contribution.


Thus, the larger cap does have a larger conversion factor to approximate the core temp by the table, but it is only as a result of the expectation that the lager the cap, the larger the surface area is.

Yes that much was obvious, but you are focusing on the wrong thing, because there is never a case where one would have both caps in any given circuit and then think they need to be aware of a higher temp in the larger cap, in any case it will always be a lower temp measured to begin with.

The Elna paper clearly states that the core temp of larger diameter cases is higher. Also read my discussion. It is clear that if the case temp is the same as I stated then the core temp is higher in a larger diameter cap.

Yes, but the moment you mention "if the case temp is the same" you are potentially misleading the reader because the case temp would not ever be the same, unless either cap is running so cool that the difference is indiscriminable by the measurement device.

[color=blue]Again you have not read and understood. Yes the case of a larger cap can loose more heat, it just depends what the environment achieves in heating the whole cap. And I thought I had made it clear that the ESR contribution to heat is minor. Quoting Elna it is desirable for deltaT to be less than 5C. My calcs also indicate the very small heat loss in ESR at full rating.

I read and understand fine, you still stuck on a backwards interpretation of the information. For our purposes, the issue is of implementation of the cap and measurement. The circuit already exists, that variable is fixed.

What you may know and what was conveyed in the document are not necessarily the same. Information can be technically correct and still presented in a fashion that can mislead, thus the document can be reworded to be more effective towards our purposes.

It doesn't in any way change the conclusion that the case temp is an important indicator, but it does effect the summary conclusions at the end. #3 - The manufacturer does provide the core, not top, temp. You need to look at Elna and other specs they do not quote core temp.

Yes they do. The whole point of having the conversion table is to arrive at a rough approximation of the core temp!
It is possible, but techincally wrong, for another manufacturer to specify case temp. It is more likely the author of the cap spec sheet got it wrong if they specify can temp because the relevant factor is what the type of cap construction can withstand without degradation, otherwise the larger the cap the higher the internal temp would be beyond that 105C (or whatever) rating.

The Elna doc itself mentioned it multiple times "Where deltaT: Temperature increase in the capacitor core", "deltaTo : Temperature Increase at capacitor core, at the maximum guaranteed temperature".

That maximum guaranteed temp is the 105C for most of the caps we use, NOT 105C shell then converted to a higher core temp.
There is ONLY one reason to measure the top of a cap, and ONLY ONE applicable use of that measurement - to approximate the core temp. The moment the core temp approximation is made, the cap top measurement MUST be abandoned, it is invalid for any further calculation.

The remaining question is how one wants to go about approximating the core temp, whether it be by a table (which adds a margin of error, particularly the larger the diameter gets, the more difference in height of the cap, and the greater the deviation in ambient temps from an expected range), OR by calculations of ESR losses and (particularly for our purposes) 'sinking by the cap leads.

#4 - Increasing diameter has a more substantial gain, because we know the heat ( look at the calcs I presented and the relatively minor gains there is no guessing or broad statement, they are factual calculations)

They are calculations based on the simplest equation feasible without knowing the other variables in implementation. They cannot know you are installing these right next to a FET or load resistor, for example. The factual state of the calculation is only as good as the information quantity, assuming everything else is a constant. That is the only way they could present it, but already it does show 4C by your calc even ignoring the other reasons for higher heat.

... is lower and the dissipation rate is higher. It's granted that the top of the larger cap won't be as hot, all else as equal as possible (same cap series and only voltage or capacitance accounting for the increased cap physical size). Even this is just a ballpark figure, as with larger diameter, increased height also has benefits but to a lesser extent. #5 - It isn't contrary to the Elna document, their doc says the same thing but the message is poorly conveyed. We can assume the larger cap is already cooler running, not that the case would be the same and the core hotter still. Remember environmental temp is the major contributor and this comes back to case temp. I have even given two scenario examples in the contribution to help with understanding about heat injection that is most often far greater than the core heat.

It does not come back to case temp! Any time a reference is made to case temp it is wrong automatically, unless mentioned only as a way to derive the core temp, and then using core temp for any further presumptions or calculations.

Until a cap that is too large is used such that it's shell is in very close phsyical proximity to something very hot (like literally touching a load resistor in a power supply), the increased size also improves dissipation of heat conducted through the leads from the nearby power circuitry FETs, etc., sunk to the board copper. This statement does not make a lot of sense. It sounds logical but needs to be more factual. It can be saying that a larger cap absorbs more heat from power circuitry through the leads.

The statement makes sense, you're just stuck in argument mode. No it cannot be saying a larger cap absorbs more heat through the leads, with the exception being that beyond a certain size of cap (typically at 16mm size) the leads do increase in diameter which could increase heat 'sink in from the copper, but with the larger sized cap it also means that by virtue of the larger radius it is practically always further away from the (other) heat source so a lower thermal density at the leads, and then with this larger diameter part the cap shell surface area is still larger.

The difference is more than you will concede so I suggest you put it to the test, which I have. Recall I've always advocated larger caps, I've used them for replacements whenever possible but sometimes you do what you can - I have installed both 12.5mm and 10mm caps on many boards, in parallel, when there was not enough room to replace every prior 10mm cap with a 12.5mm, every other cap had to be smaller than 12.5mm to allow clearance. In cases were the ripple or 'sunk heat was at a high enough level (which is not always the case, certainly in some uses it won't matter for longevity purposes) the larger sized cap was noticably cooler, often around 15C case (top) which is approx 12C core difference. At 12C the lifespan difference can be significant, about halved.

[davmax]

Re: Capacitor Life expectancy

999999999. I am quite sure that you are absolutely correct that if you insert larger diameter caps you will get better surface area cooling to remove the injected energy typically coming from the motherboard. This is a very good point explaining your results /experience. I have been trying to point out it is not wise to make generalised statements without backup evidence because capacitor series vary and if long life is really required it is best to use those specified for long life. When you first made this statement and supplied the Elna paper I checked the evidence to find out what gain could be expected because your information did not provide an answer to "how much life is increased". You have now provided a practical measure and that is good thanks.

Remember your first claims that there was no injected heat into the cap and that the heat generated by ripple was a major reason for cap failure and this heat was better removed with larger caps. Niether is correct. I believe you now acknowledge injected heat and hopefully understand that ripple heat is a minor player compared to environmental heat. My calculations reveal the later. It would seem that max ripple rating is limited by three things 1. ESR 2. Internal thermal resistance 3. The ability to reabsorb hydrogen generated by the ripple current.

Experience like yours is good but we cannot neglect the data or formula (both basic and from manufacturers). For example you state that manufacturers specify core temp. The Elna paper makes it very clear that the manufacturers maximum guaranteed temperature is To and that the core temp is To + delta T. Therefore it is clear that the core temp is above that of that specificied by the manufacturers and this is defined as being 10% higher in 8mm caps and 20% higher in 12.5 to 16mm caps. Why can you not understand this?

Everything I have stated in the my write up is correctly based on manufacturers information. It is correct to state that for the SAME case temp a larger diameter cap will have a higher core temp, but this has to be understood in context. I may have made a miscalculation or mistake and I welcome any error correction that is based on sound fact. I have worked at creating some understanding of the relationship of core dissipation and temperature in relationship the cap life and I have certainly discovered some points that informed me.

The main unanswered question for me is the fact that some series of caps do not have a specified life time increase with diameter whilst other long life caps do have an increase. I can only speculate that this is due to the elecrolyte differences and/or manufacturing method.

To conclude you are correct in that larger caps can and do dissipate more heat. It is important to know what improvement it brings. It is also important to get the surrounding issues correct (such as environment heat) and well understood, particularly the role of case temperature as the major contributor and measure of capacitor lifetime.

[davmax]

Re: Capacitor Life expectancy

99999999 I do agree that I have made a mistake in conclusion 4. My summary there was based on equal case temps and overlooks what you have experienced that for a given heat injection a larger diameter cap will dissipate more of the energy and therefore case will run at a lower temperature above ambient air than a smaller cap. Having said that, all it is not that simple. Take the example of high temp injection from MOSFETs at a 60C temp. Heat will be conducted through both cap leads, one to the core and one to the case. The energy going to the case will be dissipated by the case and the temperature is reduced, however the injection to the core is different the core to case thermal resistance is high so the core temp is likely approximate 60C plus delta T. So this raises the prospect that even though the case is cooler by changing the cap diameter the saving in core temp will be less. So it still seems better to use caps rated by the manufacturers for long life.

[davmax]

Re: Capacitor Life expectancy

Since I raised the prospect of the internal thermal resistance I thought I would take a closer look. I find that I mistakenly applied the Elna table 2-1 as a factor applied to case temp to determine core at a measured temperature. This is in fact not the case , their factor applies to the temp increase at the case due to core dissipation alone. To measure this the case temp must be taken without any ripple ie ambient temp, then ripple applied and the temp rise noted. eg if this temp rise is 5C then the core temp for a 12.5mm cap is +20% or 6C. Not much. So this indicates that core temp is close to case temp. I checked this by applying the Elna equation (3) to a 1800ufd Samxon GC cap ESR 12mohm and max ripple 2.2amps. This revealed an internal dissipation of 58mW at max ripple and an ambient to core temp rise of 4.6C (within the desired limit). Have been concerned for some time that my previous thermal resistance values were too high and it was hoped that this new calculation would improve the figure. Well a 4.6C rise for 58mW means a thermal resistance of 12.61mW/C , better than my original 7mW/C for a 10mm cap but still quite a high thermal resistance when compared 1W/C etc.

If you find a mistake in the above I would be glad to receive your input.

You stated that my calculations were simple and did not relate to the practical world taking account of MOSFETs etc. Sorry the calculation of cap internal heat from ESR loss, then the resultant core temp rise and thermal resistance calcs are all completely independent of external influence. By working on these calcs it became clear what heat is generated by ESR loss and its whole relevance to cap life when compared with heat input from external sources. Hot systems kill caps not ESR loss.

[Topcat]

Re: Capacitor Life expectancy

Excellent read and very informative!! They say you learn something new every day, and when threads like these come up, that saying holds true!

[999999999]

Re: Capacitor Life expectancy

There are two other points that can be significant.

1) We have not established ripple currents to be as low as those used to rate the caps. Certainly this is a large variable, particularly with overclocking or some kinds of odd upgrades (running a Tualatin CPU on a 440BX board for example) or very crude VRM subcircuits like PCChips & Co. like to use on occasion.

2) When considering ultra low ESR caps, their spec methodology, significant digits and rounding is starting to create significant error. Suppose 0.018 Ohm impedance (@ 100KHz, etc), is it really 0.0175 rounded up, or 0.01849 rounded down, or something closer to 0.018? Already a 5% margin for error and this only considering worst case, that the spec is a max guaranteed threshold. Similarly so for lifespan ratings, they round off to 1000 hrs, but if you look at the parts per family for many caps and begin to graph them, you will see they have to be rounding quite a bit off, 1000 hr. increments are far too large for caps rated (typically) up to 5000 hrs.

While there is conduction of heat through leads, not necessarily so much through the case, but we could conclude the immediate area has a higher ambient temp as a result, thus lower deltaT to the cap casing, lowering heat conduction. If we consider total surface area, a 12.5mm cap may still have 3:2 surface area advantage (this ratio also assumes that a larger diameter cap is also available in larger height).

When calculation attempt to attain a number independent of factors beyond ESR losses, it can ignore aging of the cap, that we're seeing quite a few non-defective caps fail from the heat so it can't be ignored as a factor in additive core temp increase. It is true that ESR alone won't typically kill any good cap, but it certainly does kill some of the crap ones. I recall an MSI Tualatin something-or-other mATX board that had caps I'd expected to fail because they were so hot even when new, and it ran fine for over 2 years with plenty of ventilation, but the day came when their failure was determined and replacement with lower ESR Panasonic FM caps resulted in a large drop in temp. Unfortunately I didn't take temp measurements with the prior caps, no solid data on that but what would someone conclude about this market difference in temp except it was the ESR difference? However, it was not a comparison of same family of cap, so it may be that ESR difference do matter, just not enough if the same cap family and the size increase isn't much. Towards this end, I typically fit the largest caps that will fit, and if that seems to excessively raise the uF values too much, at that point I might prefer moving up in voltage (rating).

[davmax]

Re: Capacitor Life expectancy

99999999 you raised some good points.
2) The rounding effect turns out to not be significant because of the manufacturing tolerances and the way the specs are quoted. Caps vary within a batch and from batch to batch so the specs cover this
a) Capacitance is +-20%
b) 100Khz impedance may well have a +-10% tolerance however the spec sheet specifies the maximum value ie you may well get many with a lower impedance.
c) Lifetime is specified as coming to an end when capacitance is +-25 to 30% of initial value or the dissipation factor has increase 200%. Looking at the lifetime charts in the Elna paper the worst case shows a 5% decline in capacitance over the last 1000hrs. In summary with all the tolerances a rounding to the nearest 1000hrs is not a problem.

All manufacturers data is carefully provided to enable designers to create products with sufficient margin to cover the worst case. Some products that cut costs may not be that good.

You comments re ESR and cap failure need attention. Yes indeed crap caps fail due toa faulty electrolyte that fails to re-absorb hydrogen and degrades the Al foil reducing capacitance and raising ESR. Process builds over operational time with the hydrogen excess growing, often enough pressure is created to blow the vent. This is not the case for high quality caps and it is likely that ESR will double by end of life, a doubling from the usual about 5C to a 10C temp rise this will halve the life at that time. Therefore with quality caps ESR is not really an issue, operating temp is.