Flyback converter - output capacitor ripple current rating

[benmack]

I've been evaluating a few off-the-shelf flyback converters, and they all appear to run the output capacitor at well over its max rated ripple current - am I missing something, or will that give poor lifetime?

For example a mains to 5V 3A DC supply I'm looking at uses a 330uF 25V capacitor which is rated 840mA RMS max at 100kHz. However the actual RMS ripple it sees must be well in excess of 3A! Running thermal tests into full load, in an ambient of 20 deg C the capacitor can reaches 88 deg C, so in a 40 ambient it will exceed its 105 deg C absolute max. And this PSU is from a reputable North American manufacturer, with a 3 year warranty

They all avoid the voltage ripple problem by using an LC filter on the output, thus allowing use of a smaller output capacitor.

We've had a batch of PSUs fail in the field in a matter of months due to the output capacitors cooking, hence my concern

Any thoughts?

 

[andre_luis]

Are you using some low ESR capacitor ?

 

[RCinFLA]

The ripple current causes IR heating within the cap due to the internal resistance of the cap (ESR). You either lower the ripple current or lower the ESR to reduce the internal heating.

There are other degrading effects but the IR heating is usually the most damaging effect on capacitor over time.

 

[andre_luis]

I noticed some SMPS with an array of electrolitic capacitors (with same uF value) in a parallel set.
And I never considered at that sight; the reason could not to be electrical but thermal.

Maybe it was designed to increase total dissipation surface.

+++

 

[Orson Cart]

We've had a batch of PSUs fail in the field in a matter of months due to the output capacitors cooking, hence my concern
Any thoughts?

You are quite correct - these units are simply not designed correctly with regard to true rms ripple current in the o/p caps,
Regards, Orson Cart.

 

[benmack]

Thanks for all the replies :grin:

Yes these designs all use low ESR parts, and yes the ripple limit is due to internal self-heating. And yes two smaller caps will have greater surface area than one larger cap, so running cooler.

I know how to fix the problem (bigger or more caps), I'm just trying to work out why most off-the-shelf designs run their output caps at well over their rated max ripple current. Will these designs be reliable? Are they simply relying on the fact that the rated max is at 105 deg C, so if they can keep the cap cooler than that, they can run at higher ripple current?

Or is it (as I suspect) that they hope that in most applications the PSU will not be used constantly at max load and max ambient temperature for years on end?

I'm looking for a PSU that will typically last 10 years at full load. I have a horrible feeling that most of the PSUs I'm evaluating will barely make it out of warranty period at full load :-(

---------- Post added at 07:34 ---------- Previous post was at 07:31 ----------

You are quite correct - these units are simply not designed correctly with regard to true rms ripple current in the o/p caps,
Regards, Orson Cart.

Mmmm thanks Horse, I'm glad someone concurs and it's not just me going mad :wink:

 

[Orson Cart]

Hello benmack, we are one of the few companies that design and make long life / high reliability power supplies, ours are generally designed to run at rated power for 10 years in a 35 degC ambient. We have been given a great number of asian produced psu's by customers who wanted us to comment on their longevity. Our replies have generally been that if they were to derate by 50% or more then there is a chance they would get 5 years out of it. I suggest the same to you, find a power level where the temp rise on the critical components is moderate, and derate the psu to that level. Or, more quickly, if you have a 50W requirement, use a 100W psu. Regards, Orson Cart.

 

[andre_luis]

...they hope that in most applications the PSU will not be used constantly at max load and max ambient temperature for years on end?

Despite some electronic components are actually being designed to operate at extended temperature ( like an uC at 210oC) their Life-cicle reduces dramaticlly ( the MTBF of mentioned componet is decays to ~10x ).

The classification of Hight temperature means that that components can operates under that agressive enviroments without fails.
However life reduction is the price to pay.

Certainly the overtemperature along the time affetcs the component in an accumulative effect.

+++

 

[benmack]

Thanks Orson, that makes sense given what I'm seeing

Andre, whilst that makes sense for semiconductors and other passives (increase in failure rate), as I understand it, electrolytic capacitors follow a definable ageing process as they dry out, so they have a definite lifetime limit, rather than a random failure rate. The rule of thumb we use with electrolytics is that the expected lifetime doubles for each 10 deg C drop in operating temperature. So start with the endurance figure at max temperature, and reduce the operating temperature to get the lifetime you want. E.g. for a cap rated at 5000 hours at 105 deg C, at 75 deg C you get 40,000hrs (4.5 yrs) and at 65 deg C 80,000hrs (9 yrs). Not an exact science but it has always worked for us (we generally aim for 10 years)

I would also never exceed the ripple rating in my own design - as RCinFLA said above, it isn't just temperature that degrades capacitors

I stumbled across this Badcaps.net - Badcaps Home - gives an idea of how widespread this problem is :-(

 

[som_raparty]

connecting an array of capacitors in parallel will be like reducing the ESR of the capacitors (similar to resistors with ESR in parallel) as a result it helps in better thermal performance