Reforming long unused electrolytic capacitors

[RoGeorge]

The legend says long unused capacitors might pop when powered up, because when left unpowered, the oxide layer slowly dissolve back in the electrolyte.  A too thinner oxide layer leads to increased DC leakage, so when powered again at nominal voltage, the capacitor may fail.

That never happened to me, but I was curious about the leakage level before and after reforming a decades long unused capacitors.

Measured the leakage while stepping up the voltage volt by volt, from zero up to a few volts above the nominal voltage.  Each voltage step was kept constant for 10 minutes.  During all this time, the current passing through the capacitor is measured every 5 seconds.

This is the current for a 200uF/12V electrolytic from 1981, brand IPRS, EG6133, at first test (capacitor not yet reformed).

When measured with the MTester, it was showing ESR=0.92ohms, Vloss=7.4%, and 754uF instead of its nominal 200uF.

Then reformed it, this is the current during reforming it for 5 hours at constant 15V.

The reforming was done at 15V, 3V more than its 12V nominal, because I've read that at manufacturing the forming process is made using 135-200% the nominal voltage.  After reforming, the MTester measured ESR=1.5ohms, Vloss=3.2% C=317uF, much closer to 200uF, but still far from 200uF.  Maybe the reforming voltage or the reforming duration were too low.

Then, did another run with voltage steps from 0 to 15V, and this is the new leakage current after reforming for 5 hours:

The leakage is much lower now.  Before reforming, the current was about 100..1000 times bigger! 

Compared at the same scale, the current before (red) and after (blue) forming looks like this:


Attached is the CSV data and the .gp scripts, in case anybody wants to closer inspect the data with Gnuplot.



Later edit:
========
Added data for another capacitor, brand OST, Low ESR, 2200uF/16V, 105C, from 2003:
https://www.eevblog.com/forum/projects/reforming-long-unused-electrolytic-capacitors/msg5680727/#msg5680727

[SteveThackery]

Very interesting! Thank you for posting it.

I've decided to adopt the policy of replacing all electrolytic capacitors in kit which is more than 20 years old (give or take, I'm not religious about it).  As far as I know, electrolytics are the only electronic components which have an ageing mechanism which operates even when the device is unpowered.

Do you think this is wasteful and the wrong thing to do?

[trobbins]

Perhaps it would have been better, and more practically enlightening, to limit at 12V - the rated voltage.  Taking it above 12V is moving outside of what anyone else would do for a reform, and is going beyond datasheet allowance, even though there is a 110% style surge capability.  My view is that factory forming is a once off process.

Certainly there is a need for reforming after years of inactivity, and I've found that good quality 1960's e-caps typically reform to very low leakage current levels at rated voltage after decades of inactivity.  Turning any equipment on after more than a decade of inactivity is a big risk, unless via a variac or by assessment knowing cap recharge current is limited.

[Conrad Hoffman]

It's great to see somebody post some data on this. I've seen manufacturers recommend reforming of new parts that have been on the shelf for a while, before measuring them. My experience with very old caps is less successful. They may reform OK, and measure OK, but then they quickly deteriorate if not kept under voltage. It might take hours to weeks, but if left unused, the device can't be relied on to power up without damage. I don't have a rule for age and some of my test equipment is running fine with 50+ year old electrolytics. If the value, dissipation factor and DC leakage are good, the part is good. Further, if those numbers are good, the likelihood of the cap failing in the near future is near zero. Aging, fortunately, is a slow process. A lot depends on the initial construction. You might find this booklet, given to new employees, interesting. (It reads backwards and sideways because it was intended for printing.)

[Gyro]

...
I've decided to adopt the policy of replacing all electrolytic capacitors in kit which is more than 20 years old (give or take, I'm not religious about it).  As far as I know, electrolytics are the only electronic components which have an ageing mechanism which operates even when the device is unpowered.

Do you think this is wasteful and the wrong thing to do?

You are right not to be religious about it. I think there is some danger of replacing older caps that will last for ever. In particular, I feel uncomfortable when replacing capacitors with ones with a significantly smaller body size. I often significantly over-rate the voltage just to get the size up. If a big cap still reads as good without re-forming, my instinct is to leave it alone, especially if I will mechanically stress things getting it out. I do tend to judge these things on a case by case basis though.

[RoGeorge]

Do you think this is wasteful and the wrong thing to do?

Don't know, it depends.  The dissolving of the oxide layer (and the need of reforming the electrolytic capacitors) happens only when they are left unused for many months, or many years.

When voltage is applied, the oxide layer grows back naturally.  The only problem is when the oxide layer is too thin and the voltage is applied suddenly and without limiting the current.  If all these happen, then the current become high enough to get permanent damage by shorting/venting/exploding the capacitor.

[edpalmer42]

I've also heard that electrolytics can dry out over time.  If true, this would be a fatal, non-recoverable fault.  I believe that some high-end electrolytics are hermetically sealed and so would be immume to this.  I think I've heard of Tantalum capacitors like that.

[Haenk]

I've also heard that electrolytics can dry out over time.

That certainly can happen; last significant case was an old 1972(?) cap, pretty large one even, that was completely dry - and therefor unusable. Likely the machine was last used like 50 years ago, I had it for about 30 years (and didn't ever turn it on).

[MisterHeadache]

Interesting experiment, thanks for sharing this data.

It's fair to say that one could perform a current monitored reform process like this on any restoration project, if the goal were to avoid replacing an original part if it wasn't necessary.  I haven't had that experience yet - for most of the 40+ year old equipment I've restored to functionality it just wasn't worth my extra time to test the caps.  Most items had less than a dozen electrolytics and film caps anyway, so replacing them with modern equivalents (or in some cases superiors) was much faster and at negligible costs (especially compared to the costs of other items that did need replacing).  But each to his own...

[trobbins]

Imho a restoration helps if you have experience with the make/model of capacitor being assessed for rejuvenation.  For sure I'll immediately not waste time on certain e-caps.  But for some I will go through the reformation process and take the time - typically for me that relates to high-end metrology bench equipment from the 60's and 70's, where it is rare for a main e-cap not to pass acceptance.

With respect to risk/hazard, obviously the first filter cap in a power supply requires the most assessment, as it's supply resistance is the lowest and its initial peak current could well be very high, and it would experience the worst operating conditions for ripple current and higher voltage.  In contrast, e-caps further along an RC distribution ladder are often protected by significant series resistance.

Reforming does take time, so this is certainly diy territory.

[Hiemal]

I kinda wonder how well polymer style caps fair in this regard. Solid polymer caps shouldn't dry out, but there are hybrid types that do still have some form of liquid inside. Would those too need reform in a few decades as well? 

[peter-h]

Very interesting data - thanks for doing it!

I have a load of 1970s/80s test gear (HP3314A kind of stuff) and my feeling is that

- there have been good and bad electrolytics manufactured (and this is years before we started getting the chinese crap) and HP/Tek used the good ones

- if the kit is turned on periodically, say every few months at least, it should have a very long life

- stuff that has not been turned on for say 20 years is often dead, for various reasons, perhaps mostly something in the PSU blowing up

- the Japanese stuff was good (I have an Iwatsu SL-4602, 40 years old now, which works perfectly)

[RoGeorge]

I'm no expert in recovering capacitors, so can't advice much about this.  Some are desoldering and reform the capacitors outside the circuit, others are just powering the entire circuit through a variac, while slightly raising the mains voltage, it depends of the appliance.


I don't really need that 200uF/12V electrolytic from 1981.  I've reformed it only only out of curiosity, because when measured was with a MTester it was showing 800uF, way more than it's nominal 200uF.

Increased capacitance is a sign of less aluminium oxide.  A thinner layer of oxides also means the max voltage supported become lower than the nominal value printed on the electrolytic capacitor.


Dissolving the oxide layer back into the electrolyte is specific to aluminium electrolytics, I don't think reforming will work for polymer or other type of capacitors, or at least I didn't encounter such reforming other than for Al electrolytics.


As for why wasting time measuring that, no particular reason.  I was digging through the scrap boxes, and found an old adjustable supply 0-12V/1A that I've built during the 80's, and was left unused since then.  The capacitor after the bridge (another one, a 2200uF/25V from 1982) was hanging in one terminal, so I've desoldered it and measured its capacitance, out of curiosity.

Instead of finding it dried, it was showing slightly more than nominal, and not much leakage, kind of funny.  So I've posted that in another thread:  https://www.eevblog.com/forum/chat/rate-my-42-years-old-2200uf25v-electrolitic-capacitor/ 


From one to another, tried a few more long unused 2200uF, and found them slightly different, so I thought it might worth posting the data in this thread instead.


For example, in another box found a 2200uF/16V (the 3rd from the photo, low ESR, 105C, brand OST) with a very strange behavior.  Except for the first in the photo, these were scraped from a former AT power supply.  Fabricated in 2003, and in use probably until 2005..2010 or so, then scavenged and left unused.  This capacitor was showing some "jumps" in current. 

First test for 2200uF/16V from 2003, starting with 10V to 20V (maybe starting from 10V was brutal, IDK), 10 minutes each step.  It shows a lot of current jumps while the voltage was perfectly steady and constant (and good wires/contacts, I've double checked the jumps were not caused by faulty contacts):


Then, tried again, in the hope that previous measurement also reformed it a little:


Better, but still strange.  BTW, looking at the DMM display, those jumps are in fact slow variations, many seconds long.  The logging was made at 5 seconds, which make them appear as spikes, but they were more like 1/F noise in reality:

Then I've become curious if the jumps were happening at a lower voltage, too, tried again, this time starting from 0V and going up to 20V:


Turned out the jumps happens at lower voltage, too. 

At this point I've decided to let it reform for 5 hours at 16V (its nominal voltage), and this was the measured current during reforming:


And a last run after reforming, still jumpy, but much better then at the first test:


Note that the scale is not the same along the different plots.

For this capacitor I didn't aggregate all the plots together on a single scale, because the big variations will make the other plots look like a flat line.  For a closer inspection please find the attached .zip, with the CSV measurements and the .gp Gnuplot scripts to visualize/explore the data at different zooms.

[trobbins]

Afaik some capacitance meters can be 'fooled' into measuring a higher value of capacitance if leakage is higher.  Perhaps worth cross-checking against an impedance bridge style instrument where leakage and D-Q are separately determined from C.

[KerimF]

More than two decades ago, I bought many hundreds of 470u/50V capacitor then I forgot about them (the project for which I bought them was dropped).
About 2 years ago, I needed them, but I found out that their leakage became rather high (low Rpar) which is natural in general.

I used to refresh such an old capacitor manually, by charging it to its maximum voltage then discharging it while limiting the currents. In each cycle, I monitor its leakage current by the brightness of an LED (connected in series) at the end of charge. But since their number was rather big, I built the basic circuit below.



Here, the voltage applied on the capacitor under test (Ctst) is close to 50V. When its leakage current (represented by Rpar) becomes relatively low, the LED turns on, almost continuously. R5 limits the discharging current. The resistance of R9 determines the lowest leakage current that turns on the LED (the higher the resistance, the lower the leakage current to turn on the LED). There are three DC supplies which are connected in series. The first/bottom one (here 15V) supplies the opamp (LM358 or LM324 could be used). The third/top one (here 10V) supplies the timer NE555. The second/mid one (here 25V) is added to let the voltage on the capacitor under test be close to the voltage of interest (here 50V).

I agree that this refresh circuit is somehow complex, if not very complex, for this job. But the refresh time happened to be 45 sec in the average. I assembled the circuit above on a breadboard. Then while I was working on my computer desk, I placed the board on it by me. Every one minute, or two, I replaced the refreshed capacitor with an old one. If I remember well, about 300 capacitors were refreshed in one day.

For instance, I presented here what I did while I don’t expect this fast complex method will be used by someone instead of the very simple slow one (a voltage source and a series resistor).

Kerim

[RoGeorge]

Interesting method, have you tried keeping them at constant DC, without pulses, then compare the results?

Asking because so far everybody seem to be using constant voltage, not pulses, including during the manufacturing process.
https://www.rellpower.com/wp/wp-content/uploads/2018/09/AEappGUIDE1.pdf
I don't see why reversing the current would grow the oxide layer faster than just keeping a voltage applied.  Maybe "shaking" the oxide crystals by reversing the current makes them align and grow faster.    Will give your method a try, if I'll find a batch of unused capacitors of the same type.  I'm curious if reversing the current (pulsed DC) makes the reforming faster.

Thanks for sharing. 

[iMo]

There is not such thing like "hermetic" unless the capacitor is put in a metal or glass package with some glass-metal sealed feedthroughts. Would be interesting to know what is the chemical composition of the electrolyte..

[peter-h]

You can buy "milspec" sealed caps; mostly tantalum. They come up on Ebay under "milspec"

[KerimF]

Interesting method, have you tried keeping them at constant DC, without pulses, then compare the results?

Good question.
But I couldn't do it two years ago and can't do it now.
We used to have 24/24 mains voltage (220V, 50Hz). Unfortunately, after the world started in year 2011 to save the people among whom I was born and live, we had no more electricity for many years (we had to rely on local AC generators and DC to AC inverters when real necessary). Lately, we have electricity 4/24 at best (it is much better than 0/24).

Since long and by accident, I noticed that by charging/discharging an electrolytic capacitor helps in reducing its leakage current. Only now and thanks to you, I heard of the simple method.
For instance, I have some of the 470uF/50V which I reformed two years ago. Their leakage current is still low as it was.