Testing capacitors for electrical leakage

[David G4EBT]

Just a few musing about testing caps for leakage/replace on sight?

From time to time on the forum, the topic of caps is mentioned and debates ensue about the wisdom or otherwise of routinely changing capacitors on sight, notably waxy paper caps (DAC90A etc) and Hunt's Mouldseal caps, simply because they have a bad name, rather than on the basis of a proper diagnosis.

It isn't unusual to see comments such as:

'I've checked all the caps and they look OK', without outlining what those checks were. (IE, just a visual inspection?). For waxy paper or Mouldseal caps the two concerns are: Is the capacitance within spec and is it 'leaky' in the electrical sense at it's rated Voltage? For electrolytic capacitors, are they within their stated tolerance, and are do they have an acceptable ESR (effective series resistance), or have they lost capacitance and morphed into becoming a low valve resistor, which is often the case in physically small electrolytics in transistor radios?

We know of course that reservoir & smoothing caps hum us a tune if they're not on their best behaviour.

I tend to change Mouldseal caps on sight because if I test them, they're invariably duff, and as often as not, waxy caps are too, but not always, though many will take the view that if they're not duds now, they will be err too long, and it can take as long to test them properly as it does to replace them. Likewise, in my experience, cathode bypass caps on valve radios/record player amps (mostly 25uF/25V), are invariably duff, and the yellow/red Plessey and blue Philips electrolytics in transistor radios call for close attention.

Checking the capacitance (and in the case of electrolytics, the ESR), is easy enough with a Chinese Multi-tester costing under a tenner, but Mouldseals and waxy paper caps must be tested at their rated voltage - it's not a bit of use checking them on the high Ohms range of a multi-meter, putting maybe just a Volts or so across the cap.

To illustrate this, I've just replaced a 0.05uF 250V Mouldseal cap in a Heathkit signal generator, though admittedly in circuit, it wouldn't be subjected to 250V. It's capacitance was 0.1UF, so yes, it still was functioning as a cap, and I doubt that being double the stated capacitance would have mattered all that much. As many might do, I checked it on the 20Meg Ohm on a multi-meter and it showed no leakage. However, when I tested it on a Robin 3131 insulation tester at its rated 250V, the leakage was 250k Ohms. So in effect, what was masquerading as a 0.05uF capacitor was 0.1uF with a 250k resistor in parallel.

In contrast, the 0.047uF 1,000V polypropylene cap I replaced it with showed no leakage at all when tested at 1,000 Volts.

It this particular instance it wouldn't have mattered had the cap been left in place, but it certainly would had it been an audio coupling cap, or an aerial/earth isolation cap on a live chassis radio. So whether or not to routinely replace Mouldseals and waxy paper caps, it seems to me that the two options are:

A) Lift one end, test its capacitance and check it's leakage if any, at its rated Voltage and if passes those tests, refit it and leave it in place. (Not everyone will have the test equipment to do that).

B) Accept that after maybe 60 or 70 years, routine replacement isn't so much a lack of diagnosis, but sensible practice based on experience over time. Just don't change them all in one session, but one at a time, testing as we go or it will be another tale of woe along the lines of: "the set was working after a fashion - I changed all the caps (all?) and can't get a sound out of it now".

Pic 1 below shows the 0.05uF Mouldseal showing up as 0.1uF. (Physically, it looks perfectly sound).
Pic 2 shows no leakage on the 20 meg-Ohm range of a multi-meter.
Pic 3 shows the leakage as 250k Ohms at 250Volts. (Top scale reads 0.25 meg-Ohms).
Pic 4 sketch of how the cap is (mis)behaving.
Pic 5, a new .047uF 1,000V rated cap showing infinite resistance when tested on the 1,000V range.

Whether or not Mouldseals or leaky waxy caps left in circuit will cause a problem (we know that mains filter caps and audio coupling caps in the output stage will, hence countless open circuit output transformers on DAC90As for example), the cumulative effect will surely degrade a radio's performance, if not to a noticeable degree.

I guess it depends on whether we just want to 'mend/do up/get going' or restore a radio to how it performed when it left the factory. That depends on whether the radio deserves such attention and is a 'keeper'. Personally, if I'm not going to restore a radio and keep it, I don't bother with it at all.

Am I the only one who has to bite his tongue when well meaning friends and neighbours say: Hey, you mess around with old radios don't you - is this any good to you before I throw it out". How very dare they - I no more 'mess around' with radios than do dentists mess around with their patients' teeth!

Rambled and dribbled David.

[G6Tanuki]

The pragmatist in me says "it takes as long to unhook one end of a capacitor and test it (probably finding it's leaky or out of spec) as it does to unhook both ends, replace with a known-good modern-part, and move on".

With experience you learn to spot those parts that have a greater-than-50% probability of being defective - so why bother testing when there's a 50%+ chance you'll be replacing anyway?

There again, my repair/restoration mindset is aimed at producing a piece of gear whose current performance and future reliability is at least as good if not better than the original was.

OC17x/AF11x-series and Lockfit transistors are another replace-on-sight thing for me. Even though they may be working now, what will they be like in five years?

(Some people may want to retain aged parts for cosmetic reasons; to me that's like not repairing dents or rusted chassiswork on your car)

[ms660]

Used to use a neon, a resistor or two and a DC supply (BY100 diode straight off the mains) to test capacitors when I was in the Radio and TV servicing trade, after a while you could get a good idea of capacity by the duration of the glow and the degree of leakage by the intensity, after a while it got consigned to the "no further use but shame to throw it away drawer" because the usual suspects were changed anyways whatever their electrical status might have been.

Here's an ancient article from a condenser manufacture aimed at the Serviceman regarding condenser leakage:

https://www.americanradiohistory.com...ox-1935-03.pdf

Lawrence.

[kalee20]

I tend to regard ALL paper capacitors with suspicion.

If I bought an old car, barn stored for 30 years, I'd change the oil as a matter of course, for a modern equivalent. I regard paper capacitors as essentially shelf-life-limited components.

[Station X]

I like to test capacitors, mainly to prove to myself that I'm not a mere capacitor changer! Having got the set or W.H.Y working I then replace all waxies ONE AT A TIME because I know they'll likely fail at some future date.

There seems to be a misconception among newbies that changing all the capacitors in a set will get it working. In my experience this is rarely the case and there are invariably other faults. Where newbies go wrong is in replacing all the capacitors in one go, often introducing new faults. They also tend to replace perfectly good low value (pf) caps in the RF and IF sections spoiling the alignment.

[M0FYA Andy]

I'm not convinced that when an old capacitor is tested and has apparently increased in value that you're actually measuring capacitance. Most capacitance testers simply measure the current flowing when AC is applied. This is fine for a good, non-leaky capacitor, but with a leaky capacitor the current is a function of that due to capacitance plus that due to leakage. So the apparent 0.1uF in the equivalent circuit above is probably fictitious.

Andy

[Diabolical Artificer]

I only test at FWV for leakage if other tests indicate a problem like high ESR or excessive ripple on a rail or as I did yesterday with NOS/caps wots been in the shed for years.

Yesterday I tested some TCC 2.5u 400v PIO cans using a cap leakage tester, EG variac, doubler, few R's and neon. I plug my DMM into this on microamps then crank the voltage up, if the meter has the same figures ish as the voltmeter, EG 30uA/300v (dialetric being a linear constant resistance) and the neon doesn't light, I know its a good un. Rarely test for actual capacitance.

Andy.

[GrimJosef]

As well as testing at full working voltage it's also worth checking at full working temperature. I've seen Hunts paper-in-oils measure OK at switch-on but be leaking after half an hour when they're a bit more than warm to the touch.

To be honest, while test gear is OK there's something to be said for testing in situ where possible. A suspect coupling cap feeding the grid of a known good (i.e. negligible grid current) output valve will reveal any leakage as a positive voltage on that grid. Watch it for the first half hour and you'll know whether there's a problem with it.

I've also come across cases where the 'tests' amount to tests to destruction. A 350V capacitor which fails after it's been Meggered might have been fine up to that point. It's ruined now though. That hardly feels like the spirit of restoration.

Cheers,

GJ

[HamishBoxer]

I agree, find the fault first and then do a one at a time change.

['LIVEWIRE?']

I'm glad I'm not the only one who regards OC17x/AfF11x transistors as 'replace on sight' items. Ditto with Hunts 'Modlseals' and waxies ( and those red & Yellow Plessey Electrilytics. All of these, plus no doubt a few other devices will have by now long exceeded their predicted service life, so why faff around testing them??

[Hartley118]

My own interest in vintage radio emphasises the 'Repair' rather than the 'Restoration' element. For me, it's very much a hobby rather than a business, so I have the privilege of regarding faultfinding as an interesting puzzle to be solved rather than adopting the more time-efficient commercial approach of 'replace on sight'.

I agree that a normal testmeter is pretty useless in measuring capacitor leakage. If it's a coupling capacitor, then I'll first check for any trace of positive volts on the valveholder grid socket: no need even to remove the chassis for that test. Sometimes even ancient wax capacitors can prove surprisingly good; on the other hand, other apparently solidly encapsulated more recent capacitors can prove disastrous. That all adds to the interest. Incidentally, why is capacitor leakage apparently non-linear, only becoming obvious at the higher voltages? What's going on?

Testing one's own capacitor stock is another issue: many of us have been involved in the hobby so long that we can't assume much about those components that have been lurking in the drawer for a few decades. Some years ago, I acquired a battered but still accurate 250V Megger which I find very handy for non-destructive testing of leakage in paper capacitors.

Sometimes, we can suffer from the 'too much information' syndrome. Back in the valve days, I don't remember ever questioning the ESR of an electrolytic. The assumption was that it was sufficiently low not to be of concern. Now those clever Chinese testers are giving us another parameter to worry about. If I were designing a switch-mode power supply, I'd be very interested, but I'm not.

But of course, if the capacitor is one of those old colourful red black & yellow Plesseys, the ESR answer will inevitably be infinity!

Martin

[ricard]

Originally Posted by David G4EBT

To illustrate this, I've just replaced a 0.05uF 250V Mouldseal cap in a Heathkit signal generator, though admittedly in circuit, it wouldn't be subjected to 250V. It's capacitance was 0.1UF, so yes, it still was functioning as a cap, and I doubt that being double the stated capacitance would have mattered all that much. As many might do, I checked it on the 20Meg Ohm on a multi-meter and it showed no leakage. However, when I tested it on a Robin 3131 insulation tester at its rated 250V, the leakage was 250k Ohms. So in effect, what was masquerading as a 0.05uF capacitor was 0.1uF with a 250k resistor in parallel.

Not wanting to doubt you here, but I'm honestly curious: if indeed the capacitor behaves as 0.1µF in parallel with 250k, how come it doesn't show up as 250k on the (low voltage) multimeter? That would indicate that the shunt resistor which models the leakage current has a value which drops the higher the applied voltage is. Does the leakage current follow some sort of squared relationship to the applied voltage (if it were a perfect resistor the current would be proportional to the voltage), so that at, say, 1 volt the equivalent resistor would be much higher than 250k?

[Station X]

I always think of the resistance of a capacitor as being "insulation resistance" requiring a high voltage to break it down.

By way of analogy, you wouldn't test a steam boiler rated at 200 PSI at 10 PSI would you?

I've tested leaky capacitors with slowly increasing voltages and they don't follow any kind of law. some break down gradually and others suddenly. Testing at the full rated voltage is good enough for me.

[1100 man]

Hmm.. Over the last few years I've drifted down from the 'replace on sight' end of the spectrum to the 'replace only when proven guilty' end.

My latest project (Thorn 2000 colour TV) is a case in point. It has 400 capacitors! Most people would consider all the Callins electrolytics (of which there are lots!) to be replace on sight. As it turns out, so far only ONE of these has been faulty and I certainly have no plans to change all of them. By faulty, I mean actually causing degradation of sound or picture.

Lockfit transistors also have a bad reputation. I've never had any problems with them. My Pye CT205 has an all transistor colour decoder with almost all the transistors being Lockfits. The only failure was a non Lockfit!

Plessey red & yellow caps again get bad press. My GEC 2000DS TV has several which I tested for leakage (at working voltage) and capacitance. Much to my surprise only one was faulty enough to need changing. The rest are working quite happily!

I was fully expecting to have to rebuild the capacitor block in my Volksempfanger radio which dates from about 1939. All 5 capacitors, tested at working voltage, displayed very low leakage and the capacitance was within acceptable limits.

When I start a new project, it depends very much on condition as to what my approach will be. A basket case which has been in a wet shed for 50 years, tends to need more replacement components before power is applied than something that has been dry stored.

So my approach now tends to be testing rather than replacement. The object for me is enjoyment which does not mean getting the thing working in the shortest possible time! I enjoy the process of fault finding and locating components that are faulty enough to cause a problem.

All the best
Nick

[Herald1360]

Originally Posted by G6Tanuki

(Some people may want to retain aged parts for cosmetic reasons; to me that's like not repairing dents or rusted chassiswork on your car)

Not really. Dents may be cosmetic (patina) but rusty chassis may be structurally unsafe (functional).

An RF cathode bypass could well be left alone in its low Z circuit but grid coupling, screen bypass, AGC filter etc should be replaced (repaired) or restuffed (restored).

[GrimJosef]

Originally Posted by G6Tanuki

(Some people may want to retain aged parts for cosmetic reasons; to me that's like not repairing dents or rusted chassiswork on your car)

Another 'not really': I usually want to retain aged parts for the same reason I want to retain the whole aged wireless (or hifi, or whatever) - which is precisely because they and it are aged. They are survivors from a previous age. They are valuable to me because they are old and rare. A penny black is useless as a stamp. But as a mark of what a society achieved at a certain point in time it has value. If I only cared about the functionality then I'd chuck all these old wireless and listen to broadcasts over the internet on my phone. To me that's like what we do when we chuck away all the old components and replace them with shiny modern ones.

Cheers,

GJ

[stevehertz]

My modus operandi is:

1. In the case of vintage radios and TVs I change all waxies on sight. Similarly with other recognised caps that tend to be faulty such as the blue lozenge caps in Grundig sets. I cut one end of electrolytics (normally much easier than trying to desolder a multi-wire junction joint) and check using a Chinese tester for capacitance and ESR. I make good the cut wire by inserting both ends into a short 'tube' of bare tinned copper wire and soldering. So no burnt insulation, excessively heated components, swearing etc.

2. For 70s and later hifi gear I LOOK carefully for signs of leakage or bursting of all caps. Otherwise I employ recognised testing methods to point me in the direction of faulty caps. To test caps I desolder them and remove from the pcb. Again the Chinese tester is employed. I most certainly DO NOT perform blanket and unnecessary total recaps as is so popular in the US and fast making its way to a town near you.

So, two quite different MOs for me based on hands on experience of working on these two quite different types of apparatus.

[G8HQP Dave]

Originally Posted by M0FYA Andy

I'm not convinced that when an old capacitor is tested and has apparently increased in value that you're actually measuring capacitance. Most capacitance testers simply measure the current flowing when AC is applied. This is fine for a good, non-leaky capacitor, but with a leaky capacitor the current is a function of that due to capacitance plus that due to leakage. So the apparent 0.1uF in the equivalent circuit above is probably fictitious.

Perhaps. However, one of the main causes of leakage in an old capacitor is moisture getting in. Moisture is a conductor, although a poor one. Moisture is also a dielectric, with a permittivity around 80. Hence it is to be expected that moisture will cause leakage and an increase in capacitance value.

[mark2collection]

Spurred on by this thread & comments thus far, I've brought an electrolytic capacitor into work & tested the component on a (calibrated) HP LCR Meter.

The capacitor comes from an early 50's set, last used in the mid-60's. 25uf at 25 volts, Cathode bypass.

L- Inductance read as :- 0.00 mH (short circuit).
C- Capacitance read as:- 32.9 uF with a dissipation factor of 0.229
R- Resistance read as:- 3.7 Ω @ 120Hz, and 1131 mΩ @ 1KHz.

I've been winding the volts up slowly on a bench power supply, by the end of the week I hope to re-test the capacitor (once at full volts) on the LCR meter again, and see how the readings compare.

Could be interesting.

Mark

[GMB]

However, one of the main causes of leakage in an old capacitor is moisture getting in.

This is a common belief but my research suggests the water is usually a result of decomposition of the paper dielectric. The capacitance does indeed rise a bit, but after driving the water out under high vacuum the capacitor ends up with no leakage but a reduced capacity.