Capacitor Life expectancy

[davmax]

Re: Capacitor Life expectancy

Quick note on previous post. End of life dissipation factor of no more than 200% of specified value. Better wording.

[999999999]

Re: Capacitor Life expectancy

It seems I entirely disagree with your last posts davmax, as we cannot claim a variability from part to part somehow offsets a known reduction in all of them. It is a very substantial different rounding off lifetime ratings, and an end of life dissipation factor is only applicable to that rounded off (thus inaccurate) rating. These shifts in values as a cap ages, make the other seemingly linear degradations more rapid, a snowball effect even on caps we see vent.

It is untrue that "All manufacturers data is carefully provided to enable designers to create products with sufficient margin to cover the worst case". Almost all cap replacements we do (original product caps as well as our replacement intentions) violates that, the data only cites a few thousand hours and calculations to arrive at our expectations incur ever larger margin of error. In short, product manufacturers are painting their products in the best light, which we can't blame them for doing, marketing is done in several ways to make the sales.

In the end the alternatives come down to one thing, whether one wants to spend a great deal of time calculating out a bare minimal acceptible lifespan projection or to spend a slight bit more money per part to use the larger caps. I can't say I have ever regretted using the larger caps, except that one time I put one too tall next to an AGP slot and the !@#$% video card wouldn't fit anymore.

I for one vote to increase the amount of time to edit posts beyond 10 minutes. Some things I wrote could be reworded better too.

[davmax]

Re: Capacitor Life expectancy

999999999 I am amazed that you have completely misunderstood and misread the information I gave you. Not once did I claim that production variability offsets or compensates for anything. As an negineer I am very comfortable iwth the data sheet figures. Particularly the ESR value is stated as the MAXIMUM value, you just cannot get better than that it covers all variability, rounding is no issue.

[999999999]

Re: Capacitor Life expectancy

I am comfortable partially agreeing and disagreeing, but I must warn that (other) engineers feeling they could just apply the specs to dissimilar situations is one of the causes of premature capacitor failure. Any engineer know there has to be a little fudge, and more fudge is usually better until the beancounters get involved.

[davmax]

Re: Capacitor Life expectancy

999999999 you have expressed concern at a 5% rounding error for ESR. Lets look at the significance of the 5%. The Elna paper states that manufacturers try to keep the ambient to core temperature rise down to 5C with max ripple. Using their equation 3) and table 2-1 most of the 5C rise occurs as case temperature and about 1C or less as core to case. Even if we take the whole 5C a 5% error represents 0.25C, case to core about 0.05C niether is significant. (Temperature rise is proportional to ESR value). The manufacturers specifications are not misleading in any significant way.

[davmax]

Re: Capacitor Life expectancy

I have long wanted to establish the relationship between core temp and case. I have seen many estimates. I am thankful to 999999999 for stimulating me into an investigation to solve this issue. My first attempt at using the Elna paper at the beginning of the thread came unstuck by misapplying Table 2-1 covering the constants for different case diameters. This lead to core to case thermal resistance levels that I found unsettling, they did not exactly fit what might be expected. So I had another crack and I have broken down the Elna code, so to speak, by applying thermal principles and dissecting the Elna equation 3).

The result has turned out to be amazing simple and comfortably lines up with a range of facts.
Some interesting facts were revealed:
1. Elna states that manufacturers try to keep the ambient to core down to 5C and table 2-1 defines a corresponding core to case down to about 1C.
2. That the major contribution of core temp rise is the case to ambient thermal resistance (case temp contribution).
3. The fact that case and core are so close makes it entirely feasible to measure operating case temp and determine life. Something already done intuitively for some time.
It has been an education for me and I hope that I can pass this on by attaching the documented process in future. But first I am referring the document back to Elna to ensure that if by chance I have made a monumental blunder interpreting their omissions and anomalies that correction can be made. TC we certainly learn each day, you can now discard some errors I made earlier.

[999999999]

Re: Capacitor Life expectancy

I didn't necessarily express concern as much as recognizing that it is one minor contributing factor. You want to dismiss each thing in turn as small, instead of seeing the additive effects of several small things.

[davmax]

Re: Capacitor Life expectancy

Really, I am giving you factual information as best I can by trying to place the facts in proportion and reality. You seem unwilling to accept facts referenced to manufcaturers data. How do these things add? ESR determines the max ripple current and life time is the dependent on temperature. Where do they add? Please explain.

[davmax]

Re: Capacitor Life expectancy

I have decided to stick my neck out and post my latest investigation into capacitor core the case relationship. I have sent a copy to Elna for evaluation and comment. Everything fits so well and must be very close to the real world I thought it well worth posting now.

Please see attached.

I will be away for two weeks from next Tuesday.

[999999999]

Re: Capacitor Life expectancy

Life is quite effected by temp, and per specific model of cap, yet the expectations for specific model of cap are rounded off in increments of 1000 when only 2000-3000 hr. life being typical. That's a substantial margin for error.

Suppose a cap rated for 2000 was really rounded up from 1500, or one at 3000 hr is actually 3499. Now consider we aren't looking for 1500 hours of service but closer to 10 years total (maybe not that long, but if you're going to replace caps then shouldn't they no longer be a first failure point if possible?).

10 years, 24/7 = 87,600 hours.

Supposing a cap degrades 5% aforementioned period, we're potentially looking at several times that 5% change in value.

On the other hand looking at ESR, suppose it's only slightly different, suppose that and size change only results in 4C change in temp (though I refer back to the issue of the case size also radiating at higher rate and leads conducting into the core), that temp difference will accelerate aging, and at the end of one period the ESR value is now rising faster. During the next period there is a larger difference in ESR between the two caps, the one that started out with higher ESR now has an even larger change in ESR for each subsequent period, having progressively higher rise in temp and wearing out faster, drying up sooner.

As briefly mentioned previously, it may not matter in some implementations. Depends on the circuit it's placed in, how much margin was afforded by the particular cap(s) used. We could hope it never mattered, but since we're typically looking at either a failure repair or preventative measure for anything but blatantly defective cap scenarios, we have situations where there is expectation these factors are significant.

[davmax]

Re: Capacitor Life expectancy

I understand what you are saying about life being out by 500hrs. Numerically it is large but we need to refer to curves published by the manufacturers that indicate the rate of decline over time. Typically in the last 1000 hrs the worst change I have seen is 5%, it will be less for 500hrs. So a 500 hr rounding may result in a error in capacitance of around 2.5% this really is not significant when compared with an initial capacitance error of +- 20%.

Your description of ESR change is right. But this is a part of the normal aging and accounted for in the manufacturers life rating. ESR changes steadily over life and may reach 200% of initial value at end of life. In some cases as you might expect degrading occurs more rapidly near end of life. Take a look at the life curves for capacitance and tan0 in the Elna paper, here you can see what type of error 500hrs would make.

[999999999]

Re: Capacitor Life expectancy

Yes, rate of decline over time, that the decline over time is not linear. A cap that starts out with certain better specs, ages more gracefully in any given (particularly for our uses) circuit. It could be only a 2.5% "error" in the initial period, ~ 2K-3K hr. typical spec'd lifespan, but since the error rate change is non-linear our several years intended lifespan has the potential to substantially exceed the error seen during the first period, during the life of the product the cap is used in.

In the following graphs we can see it is not linear but a rapdily changing curve from degradation. http://www.elna-america.com/tech_al_reliability.php

It really is significant. This is adhering to the manufacturers specs, making a conclusion that the first period is the only one to be considered is going against specs.

[gonzo0815]

Re: Capacitor Life expectancy

Well, i have read the summary, and i have no problems to accept, what Davmax has carefully put toghether, it seems to be very accurate.

But fore shure, if the conlusion drawn from this data should be, that there is no difference (or may be a marginal) between a 10mm sized cap and a 12,5.mm than i fore shure won`t agrree completely.

IMHO, there are a lot of things we and probably the cap manufakturers do not know abouth caps. There are simply to mutch unknown facts especially abouth the Ultra low esr water containing types, to consider a cap as a exact device.

But we all know for most caps, that the bigger the size, the more elekrolyt content, the higher the endurance.

So even if the datasheet does not guarantee this for the ultra low esr type, i am shure that it is true, but that someone want to be on the safe side.
This is understandable as even Rubycon can`t look into the future, if their MCZ can hold up for about 10 years in reality. And if they state an endurance of more than 2k they exactly do this.

So in the end those 500h will add to the unknown but probably substantial higher endurance of the bigger case sizes.
Compared with the substantial lower ESR, wich can for a given circuit easily be doubled wihtout any significant side efect, compared with the higher esr of the smaller case size.

Shure, if there ia a comparable series of caps with even higher endurance, but same case size and same ESR that shoul be fore shure the first choice, but this is not very common for the ultra low esr types not even for the Panasonic FM`s.

If i would desing a circuit by myself, wich should last more than five years, shure, then i would prefer to use more caps but not those ultra low esr caps.
That would be the easiest way to avoid the unknown long therm behaivior of those water based elektrolyt and it is obviousely, that i could use even those very high encurance types with very high ESR values.
I then could calculate the lifetime very accurate, as there is enough knwowledge availiable for the old types of caps.
Then there would obviousely no point in adding bigger caps, btu may be one of my customers want tjhis circuit to last another 10years, thus he uses again the biggest cap he could get.

[davmax]

Re: Capacitor Life expectancy

Gonzo

Thanks for your input. I have a feeling that you may not have read my latest input attached added recently. In the first I did come to a point where it appeared that there was little gain in larger diameter caps (based on same case temp). This was not correct, I had made an error. My second attachment clearly shows a benefit in larger diameter caps and all calcs seem to hold up to real life. Please check out the second write up. I still have preference for caps specified for long life.

[davmax]

Re: Capacitor Life expectancy

Yes, rate of decline over time, that the decline over time is not linear. A cap that starts out with certain better specs, ages more gracefully in any given (particularly for our uses) circuit. It could be only a 2.5% "error" in the initial period, ~ 2K-3K hr. typical spec'd lifespan, but since the error rate change is non-linear our several years intended lifespan has the potential to substantially exceed the error seen during the first period, during the life of the product the cap is used in.
I agree that the curves you refer to are non linear and might be expected, however they give a general picture with no values in time or parmeter decline. You have been very keen to be practical but here you have generalised and additionally claimed that the 2.5% error in 500 hrs can only be in the initial stages. You have not even bothered to look at the curves I referenced as evidence of what I quoted. These curves are plotted for actual and real cap types that indicate the rate of decline with real values. Go back to the link and check, incidentally it is the link you sent me previously. We must stick to manufacturer produced information based on measured values from life testing.
No argument for arguments sake just refer to well defined facts please.

[gonzo0815]

Re: Capacitor Life expectancy

Don`t worry Dave, i have read your latest attachment, but i have had the freedom to coun`t the +500h as negligible ;-)
But may be i was to generouse this time.

Any way, good work Dave.

[davmax]

Re: Capacitor Life expectancy

Thanks Gonzo. I have just re-read my last write up and as often happens when reading after some time it is found that some explanations are not as clear as they should be. So I have edited a critical part of the comparisom of 10mm and 12.5mm caps. It is shown that there is only a 2.23C temperature improvement in reducing the core temp and there could be no improvement at all. Better case temps of 12.5mm have been measured in the field but the measurement conditions and methodology must be carefully checked. Here we are only considering core loss contribution. See revision below (Are the Elna equations and tables correct? We have to assume they are.)



Real values applied to Elna equation 3)



Example1. Samxon GC 1800uF 6.3V, ESR 12 mohm, Ripple max. 2.2amps, dimensions 10mm dia x 20mm high.



Applying the values to equation 3)

I2R = 58mW at maximum ripple

Area = 7.07 cm2

Radiation coefficient =1.5mW/cm2 (the worst case in a 1.5 to 2 range)

The result = 5.47 oC case temp rise

This represents a thermal resistance of 0.094 oC/mW



Referring to Table 2-1 Elna do not make it clear how this is applied but it is required to find the missing link ie the core to case thermal resistance affecting temperature rise. This table seems to be the only clue.

Table 2-1 gives a core/case temperature ratio of 1.1 for a 10mm diameter cap. Applying this factor to 5.47 oC gives a core temp of 6.02 oC a small increment of 0.55 oC above case and indicates a lowish thermal resistance of 0.0095 oC/mW (9.5 oC/W)



Example 2. Samxon GC 1800uF 16V, ESR 9mohm, Ripple max 3.19A

Dimensions 12.5mm dia x 20mm high.



To ascertain a gain due to the increased diameter a ripple current of 2.2amps will be taken from example 1.



Applying the values equation 3)

I2R = 43.56mW

Area = 9.19 cm2 (approx 30% greater area than 10mm)

Radiation coefficient = 1.5mW/cm2

The result = 3.16 oC case temp rise

A thermal resistance of 0.0725 oC/mW a 22.8% improvement when compared with a 10mm cap.



Core temp = 3.16 x 1.2 = 3.79 oC or an increment of 0.63 oC above case. An internal thermal resistance of 0.0145 oC/mW or 14.5 oC/W somewhat higher than the 10mm cap.



10mm to 12.5mm comparison

Here it has been shown that for the same ripple current (2.2A) the temperature rise is reduced by 2.23 oC (6.02-3.79) when using the larger 12.5mm capacitor. The gain is found in the reduced core loss and the 22.8% improvement in case to air thermal resistance. This will lower both the case and core temperatures, life gains are 43% for 5 oC and 100% for 10 oC therefore a 2.23oC improvement will not gain much in terms of life. Again it is clear that external temperatures are the dominant determinant of capacitor life time.



The above does not consider the fact that when using a capacitor of lower ESR the ripple current will increase because of the constant voltage nature of the ripple source. For the 12.5mm example the ripple rise could cause a heat loss rise from 43.56mW to 77.44mW. If this does in fact happen there will be an improvement in ripple removal but the core temperature rise will be 6.7oC and there will no life gain over the 10mm at all.



The calculations taken from the Elna definitions indicate that the core to case thermal resistance is sufficiently low that there will be a minimal impact on the core to case temperature when heat is injected into the connecting leads.

[999999999]

Re: Capacitor Life expectancy

The ripple source is not a constant voltage for practical purposes, if it were the ripple current would be negligible. While it is true that is the goal, it is these caps themselves which are what makes that happen, they can't be simultaneously described as being in a constant voltage scenario when in the typical LC circuits, right after induction.

While we might say a lower ESR part sees more ripple current if paralleled with other inferior caps than with more of it's kind, seldom is this the capacitor arry employed except for space constraints where the one with the lower ESR was still more of a benefit than putting in the higher ESR part instead.

[davmax]

Re: Capacitor Life expectancy

Sorry, 999999999 perhaps I misled you and I should not used the words "constant voltage". Ripple is in fact seen as an AC voltage source that is loaded by the CPU and the capacitor ESR in parallel with one another. The ripple the CPU receives is directly related to the ripple bypassed by the capacitor/s. Therefore as cap ESR decreases more ripple flows through the cap and less through the CPU. Thanks for pointing out the incorrect naming, but the principle of increased ripple with lower ESR remains correct as I am sure you will agree.
The term "Constant AC voltage source" can be used but it could be confusing.

[davmax]

Re: Capacitor Life expectancy

My last post acknowledged that confusion could occur over the words "constant voltage" However some clarification as to what this means is appropriate.
When calculating circuit conditions it is essential to use a specific a fixed value such as in the example cap calculations a ripple current of 2.2 amps was used. This is fixed and therefore "constant". To obtain this current value the ripple voltage must also be "constant". I think it will be understood that calculation would be very difficult if values were not held "constant" ie fluctuating. So the term Constant Voltage source is appropriate, although it can be misunderstood, and it can apply to either a DC or AC voltage source. Clearly in the cap examples ripple is an AC component.