Capacitor ESR

[Squarewave]

Hello.

I'm going to ask a bit of a silly question.

Is ESR ever a problem in capacitors which are not electrolytic?

Reason I ask is that I do have some high in ESR electrolytics, but I'm considering maybe replacing the lot, non electrolytics also.

Normal tests on the non electrolytics appear ok, but I'm looking to rule out everything.

Thanks.

[Kosmic]

No, the problem is really with aluminum electrolytic capacitors. Over time, the seal break and they dry out or leak and ESR increase. I think tantalum are mostly shorting and breaking quite drastically. So, I don't think you will see much variation in ESR before they die.

Normally I double check all the electrolytic capacitors and if 50% are bad or suspect I just change them all.

[T3sl4co1l]

Film capacitors can suffer.  The capacitance typically drops a lot in the process, as portions of the metallization become disconnected.  The wear mechanism is self-healing (mainly due to mains surges, EMI capacitors and CFL resonant capacitors being the most vulnerable here), and corrosion of the metallization due to H2O and O2 diffusing into the plastic (the metallization is extremely thin, microns; it's not passivated, it's all processed under vacuum).

Tim

[David Hess]

Some circuits like many low dropout regulators depend on the relatively high ESR of an electrolytic capacitor for stability so blindly replacing them with low ESR capacitors can cause improper operation.

[Kosmic]

Interesting. What's the rationale behind this ?

I rarely see any mention of the bypass capacitor ESR when reading LDO specsheets.

[T3sl4co1l]

If you don't see mention of it, you're reading the wrong datasheets and should run away...

Most modern LDOs are rated for "any cap", or at least very low ESR caps, so that ceramics are acceptable.

The input side cap doesn't seem to matter to them, but the output side cap does.

The other classic LDO gotcha: if you don't see "CMOS" (or sometimes, even if you do), or a plot of GND/ADJ pin current versus input voltage, run away!  Some bipolar LDOs can draw huge GND pin currents in dropout, as they try to saturate the pass device to full throttle, not realizing there's not enough supply voltage to achieve it anyway.  Again, newer ones tend to be good about this, and provide proof of their superior performance.

Tim

[Kosmic]

Most modern LDOs are rated for "any cap", or at least very low ESR caps, so that ceramics are acceptable.

That was my understanding.

Any example of LDO needing high ESR output cap ?

[T3sl4co1l]

I've seen much worse but don't remember what ones.

This is a bit of an example:
http://www.onsemi.com/pub/Collateral/LP2950-D.PDF
Fig.17, note instability for low ESR at 0.1uF load.  (This isn't a good example, because they claim stability for more C, at any reasonable ESR.  0.1uF is pretty small, easily met by the bypasses on a couple of loads!)

Note also, while they don't give GND/ADJ pin current, they do at least give Fig.6, which is on the scale of 100s uA, and you do see a blip where it's struggling, but it's only struggling to the tune of say 20-50uA, not crazy.

So I wouldn't mind this one.  It has risk factors (bipolar, C and ESR limits).  On closer inspection, they are hardly insurmountable.  The documentation seems adequate to support this conclusion.

If you see one that merely claims (and this may be in the headline text, or buried deep in the application section) that it needs C and ESR of such-and-such range, treat it suspiciously.  If you see no documentation whatsoever about the matter, run away!

Peeking in my junkbox, I see https://www.alldatasheet.com/datasheet-pdf/pdf/53200/FAIRCHILD/KA378R12C.html which doesn't make any claims about capacitance (except to suggest >47uF being somehow helpful), and which shows dropout current consumption in dramatic fashion!

Tim

[Kosmic]

Thanks a lot for the info. Really interesting.

I'm in the middle of replacing bunch of 10uF cap on a board and went with some rubicon cap. They have really low ESR.

So I will just do a "due diligence" and double check everything

[David Hess]

Some circuits like many low dropout regulators depend on the relatively high ESR of an electrolytic capacitor for stability so blindly replacing them with low ESR capacitors can cause improper operation.

Interesting. What's the rationale behind this ?

I rarely see any mention of the bypass capacitor ESR when reading LDO specsheets.

Some datasheets just neglect to mention it.

The problem is that if the output resistance of the regulator combined with a "perfect" capacitor produces an additional 90 degrees of phase shift before the loop gain falls below unity, then the feedback loop will oscillate.  The capacitor's ESR serves to add phase lead before this can happen improving stability.

This is especially a problem at low output currents where output resistance rises and with low dropout designs where the higher impedance drain/collector is used for the output rather than the lower impedance source/emitter.  So low dropout micropower regulators are most likely to suffer from this problem although it also crops up with high current regulators at low output currents.

Newer, or at least better since this trick has been around for decades, integrated low dropout regulators use a trick where the output transistor structure is tapped before the output for AC feedback effectively adding series resistance to the transistor output to replace the output capacitor's ESR.  This allows them to use any type of output capacitor or at least a wider variety.  Also, any resistance in series with the output used as a current shunt helps as well if AC feedback is taken before it.

[Tomorokoshi]

ESR isn't the only thing to watch out for. With MLCC capacitors, DC bias may reduce the capacitance significantly. On the switching side this may be somewhat offset by the frequency response of an electrolytic capacitor that otherwise would have been there.

https://www.analog.com/media/en/technical-documentation/application-notes/AN-1099.pdf

[Kosmic]

Some circuits like many low dropout regulators depend on the relatively high ESR of an electrolytic capacitor for stability so blindly replacing them with low ESR capacitors can cause improper operation.

Interesting. What's the rationale behind this ?

I rarely see any mention of the bypass capacitor ESR when reading LDO specsheets.

Some datasheets just neglect to mention it.

The problem is that if the output resistance of the regulator combined with a "perfect" capacitor produces an additional 90 degrees of phase shift before the loop gain falls below unity, then the feedback loop will oscillate.  The capacitor's ESR serves to add phase lead before this can happen improving stability.

This is especially a problem at low output currents where output resistance rises and with low dropout designs where the higher impedance drain/collector is used for the output rather than the lower impedance source/emitter.  So low dropout micropower regulators are most likely to suffer from this problem although it also crops up with high current regulators at low output currents.

Newer, or at least better since this trick has been around for decades, integrated low dropout regulators use a trick where the output transistor structure is tapped before the output for AC feedback effectively adding series resistance to the transistor output to replace the output capacitor's ESR.  This allows them to use any type of output capacitor or at least a wider variety.  Also, any resistance in series with the output used as a current shunt helps as well if AC feedback is taken before it.

Make sense. Thank you for the explanation.