Electrolyic Caps - Leakage and Reforming

[Rob Strand]

FYI.

If you didn't know, electrolytic caps age even when sitting in your drawer.
The same problem can occur in the field when you leave equipment unpowered for long periods.
What happens is the leakage increases.  The leakage can get high enough to blow-up the cap,
especially when the circuit can supply high current like on a power rail.

This video demonstrates leakage from some new/unused caps which are 8 y/o,

There's some very significant leakage here.   Enough to blow-up the cap.

It might not be fair blaming this on cheap caps as all electrolytic caps do it.

The video doesn't explain when it going on.
What's going on is explained in the comment by  "Will Thecat".
The idea isn't new, reforming caps goes back to the tube radio days.

Do the experiment yourself:
- Take an old cap (perhaps around 100uF):
- Feed it with say 5V to 10V DC  supply via a 1k resistor.
- Measure the voltage across the 1k resistor with a multimeter. 
   The current through the cap is Icap = Vmeasured/1k.
- When you first power up the circuit current flows to charge the cap.  After that the current shelves off to the leakage value.    The charge time might take 10 to 20 RC time constants, so for a 1000uF cap you will need wait say 20 seconds or more
- A leaky cap might start with a current above 1mA.
- If reforming works you will see the leakage drops to 100uA, 10uA, 1uA etc.  At some point reforming is complete and the leakage will level off to the true leakage value.  It can take anything from 1hr to 2 days.  8hrs is a good rough guide.
  For a good cap the leakage might drop off below 10nA.
  [For higher current resolution up the series resistor to 10k, then Icap = Vmeasured/10k
   Don't forget that will increase the normal charge time].

Some interesting questions come up:
a) Do better quality caps take longer to reach a high leakage in storage?
b) Do they limit how high the leakage gets?  maybe not!
    (if (a) is long then in practice (b) holds by default.)

Here's an interesting case where things go wrong,
https://blog.stuffedcow.net/2022/02/capacitor-plague/

[anotherjim]

Capacitor Plague. Affected even the higher quality brands. The back story I heard is it was industrial espionage gone wrong. Someone took the wrong improved electrolyte formula with them to a new job.

Back in the 1980s, the place I worked at did goods-inward leakage tests on large PSU caps. Someone made a rig to reform the failures. Then I learned that you can reform an aluminium electrolytic in reverse polarity if for some unfathomable reason you wanted to. It's only the original forming that gives it the marked polarity.

[antonis]

Few years ago, I used a bank of aged (over 40 years old) electro caps (20 X 3,300μF/50V) for a power supply reservoir.. PS iddle current was calculated at about 55mA.. When measured, it revealed over double..!!
Didn't put the blame on caps leakage till all particular stages measurements made me to do so..

P.S.
Didn't have the time to observe leakage variation (if any) in relation with duty cycle / load current..

[amptramp]

I had one home made power supply that I bought from another member of the London vintage Radio Club that had an interesting defect.  It had back-to-back filament transformers so you had the voltage stepped down then back up again with two identical transformers.  There was a rectifier and filter bank on the high voltage side and the voltage should have been about 125 VDC under load and higher than that with no load.

I tested it and the voltage started at about 65 VDC and every three seconds, it rose by another volt.  This was a case of reforming the capacitors and it was like a constant current source charging the capacitors - the voltage always rose at the same rate of a third of a volt per second.  This was the only time I have seen and measured the effect of capacitor reformation.  I switched it off after it got to 88 VDC since I had no immediate use for it but it would be interesting to see if it could reform without damaging anything and return to full functionality or if I should just swap out the capacitors.

I don't know if the capacitors that show this defect actually have higher capacitance in the unformed state since the effect may include thinning of the dielectric which would increase capacitance.

[Phend]

Perfect timing, measuring some eletro caps yesterday I came up with a question.
What is ESR and Vloss as shown in the picture.?
The cap is 10uF.
How important are these values in DIY Stompboxes.?

[antonis]

https://www.avnet.com/wps/portal/abacus/resources/article/understanding-esr-in-electrolytic-capacitors/

https://passive-components.eu/capacitors-losses-esrimpdfq/

In breef:
ESR is considered an AC frequency dependent resistance in series with "ideal" cap.. Its main effect is similar with that of non-ideal voltage source internal resistance (aka voltage drop)..
Vloss is a measure of self-discharge voltage drop over time which happens due to leakage current (euivalent parallel resistance)..
IMHO, without knowing the voltage applied or, more importantly, the time over which it was measured, it's kind of useless..

Importance of the above in DIY stompboxes clearly pertains personal criterions..

[pinkjimiphoton]

these days, you can get a variac pretty cheap on like, ali express, banggood.com etc etc

i've encountered many times caps blowing on first power up. not usually with stompboxes, but with amps in particular.

i tend to use the variac to bring stuff up slowly if there's ever any doubt. often, that can save ancient electros, at least for a little while.

its always a surprise just how much of that foil crap fits inside when they explode lol

[Rob Strand]

Then I learned that you can reform an aluminium electrolytic in reverse polarity if for some unfathomable reason you wanted to. It's only the original forming that gives it the marked polarity.

Interesting idea.   Sometimes you get caps which are manufactured incorrectly and have the polarity reversed.  I suspect the problem is the heatshrink labeling more than the plating.

What is ESR and Vloss as shown in the picture.?

ESR is a good indicator when caps dry out.

Reforming won't help decrease the ESR on a dried out cap.

You can use it to diagnose failed or failing PSU filter caps.   These days it's mostly used for filters on switch mode PSUs.   As the cap dries out the ESR goes up, that causes more I^2*ESR losses in the cap, the cap get hotter and hotter and the failure accelerates.  At some point the PSU produces a lot of ripple.  Eventually the cap erupts or explodes.  The PSU might even sing or oscillate (bad).   ESR can be used for 50/60Hz supplies as well. For example the caps in old tube amps sometimes fail.

You need to know what the ESR should be in order to judge the reading is bad.   You expect lower "good" ESR readings from low ESR caps than you do from standard caps.   Typically tables are given of ESR vs cap voltage and Capacitance.  However the tables aren't clear if they apply to low ESR caps or standard caps.   You also expect low values when new, a higher value which is still OK, then finally a high value when it's clearly no good.  The transitions are very grey.  See for example,

https://bkpmedia.s3.amazonaws.com/photos/881_front_lrg.jpg
bigger pic
https://cfnewsads.thomasnet.com/images/large/024/24783.jpg

Vloss is a measure of self-discharge voltage drop over time which happens due to leakage current (euivalent parallel resistance)..
IMHO, without knowing the voltage applied or, more importantly, the time over which it was measured, it's kind of useless..

For audio, leakage can upset DC biasing it usually has to be quite bad and it's not a common problem.  Cap leakage can show up as crackly pots.

That Vloss measurement is non-standard and makes no sense to me unless I read up on the meaning; as per antonis's post.

There's standard ways to measure cap leakage.  The leakage values in the spec sheets are often *much* higher than what you see on good caps.   The specs are measured after a period 1 min, 2min or 5 min.   After that time the leakage will decrease.


I tested two old 100uF caps, standard type not low ESR. 
The test jig was 5V and 1k ohm.
Leakage normally measured at the rated voltage not 5V.

100uF 35V, Sam Hwa, 1977,  (2 units)
- Post charge-up 100uA then quickly drops to 1uA.

- Leakage during reforming

  Time    Leakage @ 5V (resolution 10nA)
  3hr    50nA and 70nA
  4hr    40nA and 60nA
  6hr    10nA and 25nA
  7hr    10nA and 20nA
  8hr   < 5nA and approx 5nA      ;confirming very low leakage after 8hr
  9hr   both measure 0nA, below resolution

- R=10k for more resolution
   Now get 50nA, slowly moving down to 20nA, to less than <20nA.
   The second unit slowly drops to 20nA.

Check: 1 day later (1k)
   Post charge-up 100uA, 250nA, then 80nA @ 1min, 50nA @ 2min
   Post charge-up 100uA, 300nA, then 100nA @ 1min, 60nA @ 2min

   Still very low after 1 day.  We do see a transition from 1min to 2min.

   116uF @ 100Hz, ESR 0.43 ohm @ 10kHz, 0.41 ohm @ 100kHz
   119uF @ 100Hz, ESR 0.41 ohm @ 10kHz, 0.38 ohm @ 100kHz

After reforming the currents in the 1min/2min period were lower than before reforming.
Considering the age the caps are in good condition.

Caveat:

The aim of the test above was to show how the leakage drops off during reforming and that the leakage is very low.  I just used a 5V supply on my bench.

While the 5V supply is OK for demonstration, and will reform caps enough to prevent the catastrophic fails in the video, 5V is not enough to properly reform caps.   

As mentioned reforming is normally done at the rated voltage of the cap.   (Manufacturers actually use forming voltages higher than the rating but that occurs before sealing - I don't want to get into the details.)

If you reform a HV cap at 5V then it could still high leakage at the rated voltage, or worse, it could breakdown and short because the oxide thickness is too thin.