Electrolytic Caps: Low ESR vs. "Regular" ESR

[Paul Marossy]

When does it matter which kind of electrolytic cap you use? I know for 99% of stompboxes it doesn't matter, but what about power supplies and stuff like that?

[Hides-His-Eyes]

Draw a 0.1R resistor in your schematic in series with your cap, and see what you see?

It's a big problem in switched mode power supplies; I think it's proportional to frequency?

[R.G.]

HHE is correct, it's an internal resistance that "spoils" the capacitor a little. It's a consequence of the fact that neither the leads outside nor inside the cap, nor the plates of the capacitor are ever truly zero resistance.

It's a big deal when there's any substantial ripple voltage across the cap, because the ESR makes heat inside the cap. In some applications, a higher ESR cap can literally cook itself with ripple current, which goes up with the square of the ripple current. It's not exactly proportional to frequency, but it does change a bit with frequency, especially since the internal inductance keeps current from spreading all over the plate of the capacitor at very high frequencies.

It can bite you even in audio circuits. I recently ...er, tested to destruction  some radial lead capacitors I was trying to cheap out on. They were the main filter caps on an audio power amp power supply. At 120Hz, you'd think a 4700uF cap would have an impedance of X = 1/(2*pi*120*.0047F) = 3.5 ohms, and that all capacitive. If you have 5V ripple on that cap, the current is 5/3.5 = 1.4A. True enough, but the dissipation is 1.4*1.4*ESR, and an ESR of 0.1 ohms is 200mW. Not a lot, but enough to heat up the capacitor internally, especially since the capacitor slug is insulated from the outer case by tar or pitch. It can literally overheat. I had to change them out for some radial lead caps with a much smaller ESR and they stayed cool.

Cap makers don't specify ESR, usually. They specify maximum ripple current, which is kinda the same thing, but it forces back on you to figure out what your ripple current is. That's not always simple.

[Paul Marossy]

It's a big deal when there's any substantial ripple voltage across the cap, because the ESR makes heat inside the cap. In some applications, a higher ESR cap can literally cook itself with ripple current, which goes up with the square of the ripple current. It's not exactly proportional to frequency, but it does change a bit with frequency, especially since the internal inductance keeps current from spreading all over the plate of the capacitor at very high frequencies.

Very interesting. So maybe a way to get around that, to a degree, is to use capacitors with a higher temperature rating? Not that we can totally ignore ESR, but it seems like it would help keep the cap from cooking itself.

It can bite you even in audio circuits. I recently ...er, tested to destruction   some radial lead capacitors I was trying to cheap out on. They were the main filter caps on an audio power amp power supply. At 120Hz, you'd think a 4700uF cap would have an impedance of X = 1/(2*pi*120*.0047F) = 3.5 ohms, and that all capacitive. If you have 5V ripple on that cap, the current is 5/3.5 = 1.4A. True enough, but the dissipation is 1.4*1.4*ESR, and an ESR of 0.1 ohms is 200mW. Not a lot, but enough to heat up the capacitor internally, especially since the capacitor slug is insulated from the outer case by tar or pitch. It can literally overheat. I had to change them out for some radial lead caps with a much smaller ESR and they stayed cool.

Cap makers don't specify ESR, usually. They specify maximum ripple current, which is kinda the same thing, but it forces back on you to figure out what your ripple current is. That's not always simple.

So when a cap overheats because of excess ripple current, does it bulge and start leaking? Or does it blow up?  

I imagine it would be very tough on an electrolytic cap to have a lot of ripple current at a high frequency. What if you used say 50% more capacitance than needed, does that help at all?

[R.G.]

Very interesting. So maybe a way to get around that, to a degree, is to use capacitors with a higher temperature rating? Not that we can totally ignore ESR, but it seems like it would help keep the cap from cooking itself.

For effects, you can generally ignore it. In general it's a power supply and high frequency issue.

A capacitor is supposed to have an impedance that steadily declines with frequency. The higher the frequency, the lower the impedance. This only holds true for so long before the parasitic resistances (ESR) and inductances make the impedance bottom out and start rising again. This is the grain of truth that the hifi tweakos use to hang their mojo supercap theories on. In practice, you have to decide whether you really need forever-decreasing impedance. Sometimes you do - like for RF decoupling of fast ICs. Where the cap is an electro in series with a 1K resistor, the point becomes moot long before the rising inductive impedance gets you.

But to answer your direct question; it's not direct. Higher temp rated caps have different electrolytes and chemical composition that can stand the heat better. That doesn't mean they have lower ESR, just that they're tougher to damage. Higher temp rating does generally mean more careful construction, but it's not just an ESR issue.

So when a cap overheats because of excess ripple current, does it bulge and start leaking? Or does it blow up?

It would if it weren't vented. That is why all electros have rubber plugs or pre-creased tops, some  kind of weak spot in the outer case. If you overheat them by any means, the internal pressure builds up and they would explode, except that they are designed to vent safely without an actual explosion. Bulging is a distinct possibility. My spouse's last computer started getting flaky and then just quit coming up. I replaced it, got her data transferred over, and some months later looked at the motherboard. They used 1000uF/6.3V caps for power bypassing around the CPU. Every one of them had a bulging top and one had vented electrolyte onto the board. Out of curiousity, I replaced the electros. It came right back up and has been working ever since. I run some of my design software on it. Electro caps that are good should NEVER bulge. That's a sign of internal pressure at one time or another, and it is a visual indicator that it's going to die if it hasn't yet.

I imagine it would be very tough on an electrolytic cap to have a lot of ripple current at a high frequency. What if you used say 50% more capacitance than needed, does that help at all?

Yep, high frequency switching power supply use is very hard on them, and has been the driver for many advances in electros. In a high frequency supply, generally you care so much what capacitance is there, as it hardly matters. The ESR and ESL are what matters, and generally it's cheaper to use several smaller ones in parallel to divide the ESR down by paralleling than trying to find magic supercaps. At very high frequencies, the capacitive impedance has become irrelevant compared to the ESR. By the time the ESR is low enough, the capacitance is way low enough.

[Paul Marossy]

Thanks for taking the time to explain that RG. Very informative.

I have a non-operative motherboard from a junked computer at work and it also had every 6.3V cap bulging as well. I had a thought to try and replace them just for kicks to see if it would fire up, but I think you probably answered my question by your own experience. If it had better specs I might have actually tried that.

[12Bass]

Sorry for the sidetrack....

Starting in the 1999, there was a rash of capacitors using bad electrolyte.  This caused numerous failures in the electronics industry, particularly with computers and power supplies.  Many, if not most, of these can be fixed with new capacitors.  I've brought several motherboards and PSUs back to life myself. 

http://en.wikipedia.org/wiki/Capacitor_plague

[Paul Marossy]

Sorry for the sidetrack....

Starting in the 1999, there was a rash of capacitors using bad electrolyte.  This caused numerous failures in the electronics industry, particularly with computers and power supplies.  Many, if not most, of these can be fixed with new capacitors.  I've brought several motherboards and PSUs back to life myself. 

http://en.wikipedia.org/wiki/Capacitor_plague

Huh, interesting!

[boogietone]

So, what would be considered a "Low ESR" value? R.G.'s post a few up talks about problems with a 4700uF with 0.1 Ohms solved by "... chang[ing] them out for some radial lead caps with a much smaller ESR ...". What was the smaller ESR? Is there a general rule of thumb for selecting electros?

Thanks.

[R.G.]

So, what would be considered a "Low ESR" value? R.G.'s post a few up talks about problems with a 4700uF with 0.1 Ohms solved by "... chang[ing] them out for some radial lead caps with a much smaller ESR ...". What was the smaller ESR? Is there a general rule of thumb for selecting electros?

Good question, but far beyond what can be handled here in any detail.

The general rule is "select an ESR that does not interfere with the application's operation, either in terms of its electronic function or its thermal stability and life".

In the case of the low ESR caps, the application was filtering a couple of hundred watts of full wave rectified power to DC. The current spikes in the filter caps were probably in the 10A-50A range This both causes buzz and hum in the audio path if you don't ground it correctly, or accidentally run these wires near signal wires, but the sheer heating caused by the power pulses heated the caps themselves. 

ESR is not fixed, and not even proportional to the capacitance or package size. It's a factor of how the maker actually ran the connection tabs out of the aluminum foils to the leads on the package. The bigger the capacitor, and the more AC current going through the capacitor, the more you need to worry. It really gets critical in output capacitors for switching power supplies where the actual capacitance of the capacitor may be immaterial. The frequency is so high that ripple caused by the capacitance value is much smaller than the ripple caused by the current going into and out of the ESR.

I guess an even simpler rule would be to figure out what size capacitor can will fit and get the biggest ripple current rating capacitor that will go in there.

[Paul Marossy]

I guess an even simpler rule would be to figure out what size capacitor can will fit and get the biggest ripple current rating capacitor that will go in there.

Probably so. That is IF you can find any ESR data on them...

[boogietone]

So, assuming that we use the max ripple current rating as a surrogate for ESR, what is a minimum ripple current rating to look for?

A rough search through the Digikey catalog gives these ranges

1 uF  -> 10mA - 40mA
100 uF -> 61mA - 2A
1000 uF -> 380mA - 4A

Generally, the larger the ripple current rating the bigger the cap and the greater the cost so, as R.G. mentions, that has to be factored in as well if you have limited real estate and/or budget.

Further question:
Do the same ripple current issues apply to electros used that audio signal path? Or, is this discussion only of real interest in the smoothing filter of the power supply?

[R.G.]

Further question:
Do the same ripple current issues apply to electros used that audio signal path? Or, is this discussion only of real interest in the smoothing filter of the power supply?

If you're doing audio signal path, chances are that the signal current will be quite small. You can estimate these issues if not calculate them exactly.

For instance, in a 9V powered effect where you're moving signal from one section to another, and you only know that the input impedance of the next stage is 10K ohms, then the highest peak current that can flow is limited by that 10K input impedance. Since it's AC, chances are that the signal can only swing by half the 9V, so we can estimate the max current will be less than 4.5V/10K = 450uA. So a 1uF cap with a 10ma rating for max ripple current looks pretty safe. In fact, it's OK for 20 times that much current. This is a very conservative estimate, as we've assumed the peak instantaneous current will flow all the time, and that's impossible. The ripple current spec is actually based on RMS current, and that's notably smaller than peak for sines.

The places where ESR bites you are at power filter caps, output caps, especially speaker driving capacitors, and high frequency switching power supplies.