Capacitor temperatures and lifetime

[davmax]

Another thread in this forum covered my attempt to unravel some of the issues surrounding capacitor temperatures and lifetime.

My effort was focused on the theory presented by manufacturers and to plugin two typical capacitor examples. The objective was to obtain accurate information that would enable a pretty thorough understanding of the isssues.

If you are interested as I was in the following questions then I have attached the researched answers.

• What does the manufacturer's maximum temperature of 105oC represent? Ambient, case or core?
• What is the core to case temperature relationship?
• What is the lifetime impact of maximum rated core loss?
• Why case temperature measurement?
• What is the life time gain when increasing case size from 10mm to 12.5mm?
• Why does case size not change the 105oC lifetime specification?
• How does increasing the case size of a normal lifetime capacitor compare with that of a long life capacitor lifetime?
• Is storage lifetime different?

I really hope this will be found useful. Constructive feedback is welcomed.

[davmax]

Re: Capacitor temperatures and lifetime

The Chemi-con link in the pdf document has changed to:

[gonzo0815]

Re: Capacitor temperatures and lifetime

Nice read as always.

[999999999]

Re: Capacitor temperatures and lifetime

1) That one manufacturer uses ambient temp instead of case or core, is not a proof others do unless explicitly stated. While measuring ambient is a reasonable fixed scenario, it could only be in the context of running the cap at the other peak datasheet spec'd value (such as ripple current), otherwise regardless of fixed ambient temp, the core temp would vary, and then include the variable of different caps having different ripple ratings. To some extent, it makes calculations more difficult until the circuit parameters are known, but we can still see relative relationships to contrast different caps, especially within same manufacturer (presuming fixed test methodology per same manufacturer).

2) Although we can consult a datasheet max current and do a calculation, it fails to address real world implementions, that the product design does not necessarily adhere to this. Consider that in operation of the equipment, many caps do significantly exceed ~5C case rise over ambient. The interpretation of this is problematic, it leads to suspicion that some questionable quality caps don't actually adhere to their specs and this accounts for the increase in case temp, leading to a marked decrease in actual life vs rated life.

3) Ripple current increase with a lower ESR cap has a mitigating factor and it may even decrease. It depends on context, the circuit. The cap has to charge up to some average voltage either way, but when the ESR is higher, the peak voltage in the waveform is higher and this causes more power usage by the powered parts during that period, and thus more current.

They are (for most of our purposes) regulated circuits so the lower threshold (sensed voltage level) for regulation is the same but the on-time for the switching is a higher duty cycle caused by the active parts consuming more power from the higher peak voltage. More power flows through the regulation stage and the higher peak voltage offsets the higher ESR in increasing charge rate, and ESR losses.

Perhaps the last time I wrote something about this I did not make it clear enough, what further way can I clarify this? Envision a picture of a waveform. With higher ESR cap the ripple is larger. That larger ripple, higher voltage, is seen by the powered part. This causes more current by itself, but even more ripple given the inductors remain fixed variables. The powered part never sees as low a depression in voltage from the higher ESR, because the regulation sensing has turned on again sooner or stayed on for longer duration, because such controllers achieve regaultion up to the minimal voltage level not down from the max voltage level - until the integral OVP subcircuit kicks in but this is a higher threshold, a delayed reponse since the powered part consumes power in relation to voltage, not a constant current.

If it were not for the circuit being a regulated switcher, and if the powered parts had constant current consumption, THEN we could assume the higher ESR cap had a ripple current decrease.

4) Larger can size caps are typically starting out with longer lifetime ratings. Some may disagree about the significance of rounding off to 1000 hour increments, but any mathmatician will know, rounding off to 1000 hours it parts rated for a mere 2000-4000 hours typical (or 3000 avg), is already potentially a 17% error rate, and by looking at where the changes in life rating occur we could approximate a more accurate life rating by plotting a curve of size vs rated life, that it would be a curve in many cases instead of a stair-step shape. Hybrid or solid caps are the wildcard here, especially at the more typical ambient temps seen in a computer rather than as high as 105C, their derating curve is much steeper.

Finally, an operating circuit has a minimam threshold for proper performance. Suppose for example a circuit needed caps with (summed) ESR at least as low as 3mOhm. The "lifetime" until the bank of capacitors fails to remain below this value, is not necessarily the same as lifetime until the cap falls below it's own manufacturer spec'd values. Actually it is always different, in that if two caps were aging at the same rate but one starts out with lower ESR, it retains a longer viable lifespan for the circuit in which it is placed as all of their ESRs would rise, but the potential is for the ESR difference to allow one to function acceptibly still in the circuit, and for the one that did not, there is potential for the circuit to use even more power as mentioned in #3 above. A bit more information about life may be found in the following PDF, as well as mention of cap life equations using core, not case or ambient temp. It does not mean that's how all manufacturers do it, just another example of it not being a constant among different manufacturers or engineers. See especially pg. 4-5.