Aluminum electrolityc capacitors operated close to rated voltage

[Conrad Hoffman]

Well, in practice, 80% sounds about right. Not for any reason than that's how most people seem to pick 'em. I think the temperature rule is aimed more at the higher temperatures. Above I talked about falling temperatures. That was more about equipment running in a cold environment, or powering things up right off the UPS truck in the winter. Not a good thing to do in a marginal design case.

[mclute0]

As per my post above, it is up to the engineer to apply the suggestions of the capacitor manufacturer.

The linked article lists the manufacturers estimate cold effect as well as derating for cap aging and a link to that manufacturer's derating calculators.

It could be that if a designer were cheap, they might only estimate the cap failure to be the length of whatever warranty they are offering. IF they offer no warranty, they could just as well not derate as suggested by the cap manufacturer and let their device fail relatively sooner than later.


Not a good way to make repeat customers, that is for sure.

[mclute0]

Well, in practice, 80% sounds about right. Not for any reason than that's how most people seem to pick 'em. I think the temperature rule is aimed more at the higher temperatures. Above I talked about falling temperatures. That was more about equipment running in a cold environment, or powering things up right off the UPS truck in the winter. Not a good thing to do in a marginal design case.

The US Navy requires derating Aluminum electrolytic caps by 70% as of 2007.

https://www.navsea.navy.mil/Home/Warfare-Centers/NSWC-Crane/Resources/SD-18/Products/Capacitors/Derating/

[Zero999]

IF they offer no warranty, they could just as well not derate as suggested by the cap manufacturer and let their device fail relatively sooner than later.


Not a good way to make repeat customers, that is for sure.

And not an option in the EU, where you have to provide a minimum warranty.

This life-temperature dependence actually impacts how you should derate the voltage on the capacitor.

In practice, aluminum electrolytic capacitors typically are used at about 80% of their rated voltage.

I don't believe either of these statements are true..

You are certainly entitled to an opinion. I would suggest you take it up with Cornell Dubilier and ask them to change their capacitor usage guide.

from https://www.cde.com/resources/technical-papers/AEappGuide.pdf

Estimating Lifetime for Capacitors without an Online Calculator
We offer online calculators for many of our capacitor series such as our screw-terminal, snapmount and flatpack capacitors. For our capacitor series without a calculator, you will find on its datasheet a “load life rating” with an ambient test temperature and duration at the rated ripple current, which is tabulated for each capacitor within that series. To estimate the minimum lifetime, you may select a capacitor whose tabulated ripple current is at least equal to your application’s ripple current, and apply the “doubles every 10 °C” rule between the load life test ambient temperature and your application’s ambient temperature. If your DC voltage is derated, then you may multiply the lifetime by the voltage multiplier Mv given in equation (6). This will be an estimate of the minimum expected life

...

Operating Lifetime Model

Onset of wear-out is determined mainly by the capacitor’s average operating temperature. Operating voltage has some effect. For capacitors operating at moderate temperatures the operating life doubles for each 10 °C that operating temperature is reduced. Our online lifetime calculators available at  http://www.cde.com/technical-support/life-temperature-calculators our best estimates of the useful lifetime for many of our more popular capacitor series. The expected operating lifetime is approximately

Lop = Mv × Lb × 2((Tm – Tc)/10[°C]) (5)

The above equation is the simple, classic “doubles every 10 °C” rule used for many decades, and Mv is a DC voltage deratingmultiplier equal to

Mv = 4.3-3.3 VA / VR (6)

Thus for example when the DC voltage is derated by 10%, this multiplier is equal to Mv = 1.33. As in the failure rate equation discussed in the previous section, VA is the applied DC voltage,
VR is the rated DC voltage, Tc is the core temperature, Tm is the maximum rated core temperature and Lb is the base lifetime.

Thanks for posting that. It also specifies the life at the rated ripple current, so the original poster can expect it to last for much longer, as the ripple current will be much lower, probably zero most of the time.

Well, in practice, 80% sounds about right. Not for any reason than that's how most people seem to pick 'em. I think the temperature rule is aimed more at the higher temperatures. Above I talked about falling temperatures. That was more about equipment running in a cold environment, or powering things up right off the UPS truck in the winter. Not a good thing to do in a marginal design case.

The US Navy requires derating Aluminum electrolytic caps by 70% as of 2007.

https://www.navsea.navy.mil/Home/Warfare-Centers/NSWC-Crane/Resources/SD-18/Products/Capacitors/Derating/

Military has much more stringent requirements, compared to commercial/consumer electronics, so that's probably irrelevant.

[Siwastaja]

You are certainly entitled to an opinion. I would suggest you take it up with Cornell Dubilier and ask them to change their capacitor usage guide.

You demonstrate well the problem of "authority argument".

While everything they say is true, it has a context, and the problem is, you don't seem to understand it.

Whether the application is of high ripple current or low ripple current type totally changes everything. Because the Cornell Dubilier text does not emphasize this directly, you as a designer must be experienced and knowledgeable enough to see it. "core temperature" being a parameter is a red flag!

In high ripple current rating application, the "hour rating" dominates, and all the classical derating formulae apply.

In low ripple current rating application, the aging is more similar to "shelf storage" aging. Except that certain amount of bias voltage will actually help maintain the capacitor. At some point, too high voltage starts doing damage instead. But it's totally different mechanism! It's not related to heating, like the hourly ratings in SMPS applications.

To me, it seems OP's case is a nearly zero ripple current one, although I would like to have a confirmation of that.

So instead of appealing to authority, you need to build understanding, to be able to give correct advice.

[mclute0]

I apologize if you take offence to my statements.

Perhaps some meaning is lost in translation. I am not a scientist that researches and designs capacitors. I freely admit I may be wrong. I like so many other people only follow what we have been instructed by manufacturers.

I stated my understanding of the general rules of derating capacitors used by a number of sources and was told that it was not to be believed. No reason given.

I then quoted 'A' manufacturers directives on derating. If that is bad, I am again sorry.

I work in a field where things change. Rules that were true before change. The internet exacerbates this problem by retaining information that is used for refence effectively forever.  When old material on the internet has wrong information, it makes people like me think and say the wrong things.

If I am not permitted to use references to illustrate my understanding of general principals of implementation without being accused of "an argument from authority" because I cited my sources, I don't know what to say.

[Siwastaja]

No need to apologize, you are the only one taking offense.

You can go on and blah blah  until cows come home, but that does not address the actual issue that the text you linked to is actually about thermal aging of capacitors due to ripple current; the usual lifetime/derating consideration mostly relevant for SMPS design. It's not the best resource for OP's case which probably deals with small enough ripple current that it does not cause capacitor aging.

So the aging mechanisms OP will see are different.

Sadly, I don't have a good source of information off-hand. Shelf-life aging and low-ripple aging under DC bias are much, much less discussed in appnotes, because expected lifetime will normally be very long, and manufacturers don't want to give guarantees that long.

[mclute0]

blah blah? OK, have a nice life!

[bdunham7]

Sadly, I don't have a good source of information off-hand. Shelf-life aging and low-ripple aging under DC bias are much, much less discussed in appnotes, because expected lifetime will normally be very long, and manufacturers don't want to give guarantees that long.

It would take a long time to accumulate that information as there probably isn't a good way to do accelerated testing and manage to correctly mix in all of the factors.  Even moderate ripple and temperature applications like linear PSUs with very large capacitors where the capacitor temp might be 30C and the ripple a miniscule percentage of the rating result in units that still work after hundreds of thousands of hours.  So after 30 years of testing they can tell you what the spec is was...

I the OPs case the capacitor will typically see 6 to 9V, so unless there is something unusual going on I don't see a problem.  However, I have to wonder if something unusual is going on. 

[uer166]

I apologize if you take offence to my statements.

Perhaps some meaning is lost in translation. I am not a scientist that researches and designs capacitors. I freely admit I may be wrong. I like so many other people only follow what we have been instructed by manufacturers.

I stated my understanding of the general rules of derating capacitors used by a number of sources and was told that it was not to be believed. No reason given.

I then quoted 'A' manufacturers directives on derating. If that is bad, I am again sorry.

I work in a field where things change. Rules that were true before change. The internet exacerbates this problem by retaining information that is used for refence effectively forever.  When old material on the internet has wrong information, it makes people like me think and say the wrong things.

If I am not permitted to use references to illustrate my understanding of general principals of implementation without being accused of "an argument from authority" because I cited my sources, I don't know what to say.

There is a cool presentation that I need to dig out somewhere, sorry for the blunt "I don't believe that's true" without further explanation. But the gist of it is, most manufacturers' aging calculators don't include the voltage stress in the equation whatsoever, it's not even a variable. That tells us that the voltage stress on a elcap is at the very least a minor stressor, if not totally negligible compared to thermal stress. Most literature has well established that it is temperature alone (no matter the cause, although of course ripple current is the major one, but could be anything), is the major stressor that affects aging, not the DC bias. Different manufacturers have secret sauce data and don't seem to even agree about whether voltage stress is a major factor or not between themselves.

You happened to chose one of the manufacturers that does indeed include it (C-D), but that is not the whole story and IMO misleading. It's misleading because it might lead people into believing that reducing DC bias will extend lifetime, which isn't true in 99% of cases in the field, but reducing the temps absolutely, and dramatically would.

[uer166]

E.g. from Nippon Chemi-Con which has been already posted somewhere:

Where a capacitor is used at lower than the rated voltage, the lifetime may not be adversely affected, which means that the effect of the applying voltage is negligibly small, while the effect of the ambient temperature and heat generation due to ripple current is significant.

However, for capacitors of larger size and higher rated voltage contain a larger volume of electrolyte, difference in applying voltages can affect degradation of the oxide layer, other than the diffusion of electrolyte. Therefore, for screw mount terminal type capacitors with the rated voltage of 350Vdc or higher, the lifetime estimation includes the effect of applying a lower voltage than the rated voltage (derating voltage).

I'm not arguing that blanket de-rating a cap by 80% will hurt anything, instead I'm claiming that it:

• Is not the whole story, and doesn't explain the why
• Is sub-optimal for no good reason
• Is not necessary in most cases with minimal attention. In cases where it is needed, that "minimal attention" will catch them.
• Distracts from the real problems which are often overlooked and what kill products in the field, namely thermal effects (whether due to ripple or external heating)

[Siwastaja]

There is a cool presentation that I need to dig out somewhere, sorry for the blunt "I don't believe that's true" without further explanation. But the gist of it is, most manufacturers' aging calculators don't include the voltage stress in the equation whatsoever, it's not even a variable. That tells us that the voltage stress on a elcap is at the very least a minor stressor, if not totally negligible compared to thermal stress.

The problem is also that it's hard to say if voltage is a direct physical stressor, or an indirect one. By indirect, I mean: when you consider heating, if you pick two capacitors with the otherwise same ratings, the larger voltage version comes in larger package, with more surface area, so it cools better (assuming they still have same ESR). This is the indirect reason why voltage derating is sometimes suggested to remedy thermal aging. I think that ripple current rating should be the primary parameter to derate, though, in such thermal lifetime application. I do both, a 15V switcher with 1A of ripple current gets a 25V (as 20V isn't a typical size) cap with 2A ripple current rating.

OP's application is close to "DC biased shelf life" case, though. Voltage and ambient temperature are the two factors (ripple current isn't since it's near zero; and core temperature simply becomes equal to ambient temperature).

It's well known that when the voltage rating is exceeded by a lot - maybe say 30-50% - DC leakage current starts to increase steeply, resulting in heating, thermal runaway, and destruction. Who knows if there is some certain magical overvoltage % which causes slow damage. In any case, applying rated voltage may actually lengthen the life by keeping the passivation layer formed, and as the capacitors were originally formed at a significantly over the rated voltage (like 30-40% IIRC), 10% overvoltage likely does not do any damage when it is free of transients.

But it's hard to find reliable data. I guess OP's case is fine if they print "don't use lithium cells" in the manual. Those who do will be unlikely to have any problems.

[bdunham7]

OP's application is close to "DC biased shelf life" case, though. Voltage and ambient temperature are the two factors (ripple current isn't since it's near zero; and core temperature simply becomes equal to ambient temperature).

We're assuming this, but without confirmation.  Suppose whatever is going on with their device results in an occasional 15V spike from inductive kick of some sort? 

[Zero999]

I don't see the need for the argument. I found that application note interesting. Yes, as I said, it's not completely relevant to this application, but it's still useful information nonetheless.

OP's application is close to "DC biased shelf life" case, though. Voltage and ambient temperature are the two factors (ripple current isn't since it's near zero; and core temperature simply becomes equal to ambient temperature).

We're assuming this, but without confirmation.  Suppose whatever is going on with their device results in an occasional 15V spike from inductive kick of some sort? 

400V capacitors on switch mode power supplies routinely cope with those sorts of spikes.

[Siwastaja]

We're assuming this, but without confirmation.  Suppose whatever is going on with their device results in an occasional 15V spike from inductive kick of some sort?

That is unlikely because large electrolytic capacitors by their very nature damp inductive peaks very well - lot of capacitance, high enough ESR. For this reason, just slapping a standard elcap is the classical silver bullet to tame large MLCC banks and their hotplug voltage transients.

If the application is just some sort of AA battery holder with short wires into the electrolytic cap, I don't think any significant voltage transient is possible, unless the user connects some external power supply into the battery holder - which would definitely void the warranty.

BTW, OP, is there reverse polarity protection? If not, adding one would be a good measure. Maybe just use a boringly simple series diode for the job, and it also sheds some half of a volt off the voltage, solving the voltage rating issue as well.

[bdunham7]

It's misleading because it might lead people into believing that reducing DC bias will extend lifetime, which isn't true in 99% of cases in the field, but reducing the temps absolutely, and dramatically would.

There is going to be a relationship between lifetime and both DC bias and temperature, but it isn't going to be a simple one that applies at all values.  For example, I don't think lowering the temperature from 1000C to 970C will increase the capacitor life 8X, nor will raising it from -25C to -15C reduce it by half.  It's all very complex and all the rules are just gross oversimplifications that make it possible to make some decisions.

I would point out that failure rate and lifetime, while related to an extent, can also be looked at as different issues.  Derating addresses more than simple models, it also addresses manufacturing variances in both the component and the device in which it is installed, as well as abuse, user error and other random or unknown issues.   

[TimFox]

A long time ago, I posted a question about voltage derating of electrolytic capacitors.
Basically, in my youth I learned that one should not operate an electrolytic "far below" its voltage rating, as well as not above it.
I have sought a more quantitative statement ever since.
None of the replies to my question was quantitative.
In vacuum tube days, the power supply filter capacitors were rated at a few hundred volts, and the cathode resistor bypass capacitors were rated at around 10 V, and they were not confused.

[Conrad Hoffman]

Long ago, in the '60s, Sprague was very free with technical data, unlike today where everything is a state secret. Technology has changed, though the general trends are probably still true. I have several of the old Sprague papers on my site, all of which are worth reading, but you might look at page 16 of this one- http://www.conradhoffman.com/papers_lib/Sprague_62_7.pdf One of the other papers talks about running caps at low voltages and says there's no issue right down to zero volts, or at least not enough to worry about.

[mclute0]

It's misleading because it might lead people into believing that reducing DC bias will extend lifetime, which isn't true in 99% of cases in the field, but reducing the temps absolutely, and dramatically would.

There is going to be a relationship between lifetime and both DC bias and temperature, but it isn't going to be a simple one that applies at all values.  For example, I don't think lowering the temperature from 1000C to 970C will increase the capacitor life 8X, nor will raising it from -25C to -15C reduce it by half.  It's all very complex and all the rules are just gross oversimplifications that make it possible to make some decisions.

I would point out that failure rate and lifetime, while related to an extent, can also be looked at as different issues.  Derating addresses more than simple models, it also addresses manufacturing variances in both the component and the device in which it is installed, as well as abuse, user error and other random or unknown issues.   

A I stated in my orignal post "Always check with the manufacturer since all capacitors may vary."

I was taught the adage that real engineers do the math, psychics and contractors make guesses. Rules of thumb are fine for making estimates, but they are no substitute for the math.

If an engineer designs an automobile device for average temperatures in Florida, it will be invalid if the owner moves to Alaska and takes their automobile with them. The only real way I know to approach that issue is to design not only for the worst-case scenario, but to also add my own personal unknow error factor.

If a bridge has over engineered specs nobody notices, if it is under engineered people could die.

Without knowing the application or the criticality of any device it is easy to say an 8v spec is no different than a 9v spec, but it all depends on how many lives and how much money are in play. The rules of play are set by the manufacturers of the component, they may not always be followed to the letter, but they should always be considered.

[TimFox]

Long ago, in the '60s, Sprague was very free with technical data, unlike today where everything is a state secret. Technology has changed, though the general trends are probably still true. I have several of the old Sprague papers on my site, all of which are worth reading, but you might look at page 16 of this one- http://www.conradhoffman.com/papers_lib/Sprague_62_7.pdf One of the other papers talks about running caps at low voltages and says there's no issue right down to zero volts, or at least not enough to worry about.

Thanks for the reference.  The curves on p. 16 show a monotonic decrease in failure rate with respect to voltage reduction that are truncated below 50% of rated voltage, apparently because of too few failures in the tests.

[mclute0]

Long ago, in the '60s, Sprague was very free with technical data, unlike today where everything is a state secret. Technology has changed, though the general trends are probably still true. I have several of the old Sprague papers on my site, all of which are worth reading, but you might look at page 16 of this one- http://www.conradhoffman.com/papers_lib/Sprague_62_7.pdf One of the other papers talks about running caps at low voltages and says there's no issue right down to zero volts, or at least not enough to worry about.

or https://www.conradhoffman.com/papers_lib/Sprague_62_4.pdf


It is a shame that work like that has not been converted to an internet friendly form.

[bdunham7]

I was taught the adage that real engineers do the math, psychics and contractors make guesses. Rules of thumb are fine for making estimates, but they are no substitute for the math.

Well math doesn't do you much good unless you have accurate data and models.  A manufacturers specification doesn't give you the underlying information and the way those specs are determined can vary.  At least a well-considered rule of thumb can add in some empirical data--the wisdom of experience.

If a bridge has over engineered specs nobody notices, if it is under engineered people could die.

Whether the bridge collapses or not is a matter of the actual materials and construction used, not their listed specifications.  Headline or advertised specifications are especially unhelpful and misleading in many cases.  I suppose modern expectations would be that specs would be exaggerated best-case maximums, but not always.  How much weight can a 1-ton pickup truck haul?

[mclute0]

I was taught the adage that real engineers do the math, psychics and contractors make guesses. Rules of thumb are fine for making estimates, but they are no substitute for the math.

Well math doesn't do you much good unless you have accurate data and models.  A manufacturers specification doesn't give you the underlying information and the way those specs are determined can vary.  At least a well-considered rule of thumb can add in some empirical data--the wisdom of experience.

If a bridge has over engineered specs nobody notices, if it is under engineered people could die.

Whether the bridge collapses or not is a matter of the actual materials and construction used, not their listed specifications.  Headline or advertised specifications are especially unhelpful and misleading in many cases.  I suppose modern expectations would be that specs would be exaggerated best-case maximums, but not always.  How much weight can a 1-ton pickup truck haul?

I don't disagree, models are math.


How much weight can a 1-ton pickup truck haul? What kind tone? A Tonne? Short Ton? Long Ton? Metric Ton? Ton Shortweight? Ton Longweight? They are all different specifications. With options or without? With driver or not?

I can guess from your words and context what you mean, but to find out I need to either check the manufactures specification and use the math to answer or destructively test some number of pickup trucks and math to find a statistical answer.

[brumbarchris]

The sequence to this thread focuses on the broader problem and solution:

https://www.eevblog.com/forum/projects/alkaline-battery-characteristi-effect-on-product/msg4204183/#msg4204183

[oz2cpu]

i just did a test and teardown video of a lab dc powersupply, sold in high volume,
it contain 35V large size caps, and they charge them to 42V DC all day long,
the unit is 20 years old, and nothing bad happened from this..
but it is a VERY bad design style, it reveal they take specifications light,.

some handle this fine, others not at all,,

it is recommended to perform design DERATING,
that means you always run at a voltage that is, the wanted margin, UNDER the max specified voltage,
this way you are sure to create good quality designs,