What's the deal with this? I read somewhere that a high ripple current figure for an electrolytic is a bonus; an expensive and sought-after characteristic. However, I'd always thought ripple was anathema to PSUs. Can some expert kindly explain the relationship between ESR, ripple current, capacitor ageing and whatnot?
** Ripple current is simply the RMS value of the current flowing into and o ut of a capacitor as it charges and discharges - something it must do in or der to smooth rectified AC.
The current's magnitude depends on the p-p ripple voltage and the value of the cap and is somewhat greater than the DC load current.
Ripple current heats the cap according to "I squared R" where R is the actu al ESR at the frequency and temperature of operation. So a low ESR cap has a higher ripple current rating.
However, I have not come across a electro cap where operation at 100Hz and with the usual 10% to 15% ripple voltage exceeds its ripple current rating.
When an electro cap is used on the OUTPUT of a SMPS the rating becomes more important as the ripple voltage at high frequencies can be small while the current is large. Luckily the ESR of a electro caps improves at high frequ encies and also when the temperature rises.
Electro cap life is normally a function of operating temperature, halving f or each 10 degrees C rise above room temp. It is purely due to the liquid e lectrolyte escaping from the cap as vapour - a few molecules at a time.
Manufacturers use about 3 time more electrolyte than needed inside electros so the cap will have a long life without change is ESR or capacitance unti l nearly all the electrolyte is gone.
the capacitor is not perfect - it also has internal resistance (and inductance). As ripple current flows in and out of the capacitor it causes heat via the internal resistance. To much heat and the service life is short. Way too much heat and it vents messy crap all over the place.
Using Opera's mail client: http://www.opera.com/mail/
ESR is an apparent series resistance inside the capacitor. Ripple current is the capacitor charging and discharging from the rectifier pulses of current and the load current in between the rectifier current.
The ripple current flows through the apparent series resistance and produces heat. heat accelerates the processes that cause ESR to increase.
Low ESR is good, but its things like SMPSUs that need the more expensive parts.
Very often you can improve matters by putting metalized foil or MLCC capacitors in parallel with electrolytics.
Well, no. The SMPS has an input filter, at a few hundred volts bias, that only sees an amp or so, as well as a 3.3V output filter, that sees near a hundred amps of ripple. If you put the low-ESR cap at the input filter, it just means that your ON/OFF switch makes a fat spark every time you apply power. Switch life, and RF radiated interference, are your new problems.
and, if you DON'T put a low-ESR cap at the output filter, the ten milliohms of ESR, which feeds that hundred amp load, can have up to 1V of effect (i.e. your regulation suffers).
If the ESR of the output caps and capacitance of the output caps are both in the 'good enough' range, go on to the next problem. One 10 milliohm capacitor won't do, but fifteen in parallel is OK. The CPU that takes all that power, has a dozen power input pins, anyhow: there's wiring resistance to consider, so a microohm ESR for the capacitor doesn't help much.
That depends on the exact parameter you are attacking.
It is true that a lytic will have higher ESR at lower temperatures, but whe n exposed over time to higher temperatures, especially cycles, their averag e ESR will go up. While this is still dependant on temperature the basic fa ctor of it is worse, higher ESR.
Leiberman was up here like last year with a chart he made with a pot of wat er, thermometer and ESR meter. I would almost like to see how those same ca ps fare after being in service for a couple of years. If he still has the t ime to burn of course.
I would not buy an ESR meter, I can measure that with a scope or a few othe r means. But after being in service they deteriorate and we all know that. But if they are not leaking fluid out or anything, what is causing it ?
The only explanation is that the electrolyte is losing its electrolytic pro perties. There was no leak, the plates did not change, the application of t he electrodes to the plates did not change, or did they ? You know a bit mo re about the construction of these things so you speculate on that. However I think that a problem with the connections would cause higher ESL, moreso than ESR. The only thing left is the electrolyte. And I remember the almos t wars over their formulae.
The leaky ones were a different story, they leaked all over the boards in a lot of shit. I got to the point where I did not replace them, they already done leaked. I washed the shit out of the board and was usually done with i t. The thing is seven years old and this is fixed now, see you in seven yea rs. The only real problem is when that shit ate the copper right off the bo ard. But that is why I got the big bucks.
Your explanations are very clear I must say. Now, given that v. large value electros with v. low ESR are expensive, would there be any saving from buying lots of smaller values and placing them all in parallel (space permitting) to equal the value of the one large one? And would there be any performance hit to take from doing things this way? Thanks again.
** Doing this has been a common practice for decades when low levels of output noise are required for a SMPS. The total capacitance value is not as important to this task as the final ESR of the combined parts.
Of course, using a DC choke between two sets of electros works even better.