Testing for leaky caps

[skodajag]

I know this has been discussed before, but I’ve never been successful in using the search facility on this forum. So, eg, searching the title above brings up numerous unrelated posts. I'm sure it’s me. So, can you please direct me to the info on this invaluable forum which tells the best way to rest for leaky non-electrolytic caps both in and out of circuit? Advice on using the search facility would also be appreciated.
Thankyou.

[Silicon]

The simplest method is to connect a known resistor (i.e. 10kOhm) in series with the capacitor and apply a small voltage (i.e. 9V).
If the capacitor leaks, a small current will pass through the resistor and this will generate a small voltage.

This can be measured with a digital voltmeter and the current can then be calculated.

Some capacitors may only break down at higher voltages.
You need a variety of voltage sources and resistors to safely test a 400V capacitor for example.

The capacitor could be a dead short and be capable of passing large currents.

[Silicon]

I click on 'search' then 'advanced search'.

I select the forum I think will narrow down the options i.e. 'components and circuits'.

I type in something like 'test capacitor' in the keyword box and I then click on the 'search now' button.

There are many search options but I rarely alter those for the initial search.

[GMB]

Testing for leakage in-circuit is hard unless you are lucky with the circuit.

Out of circuit the rule is to measure the leakage current with about the working voltage applied (with resistor in series to protect the meter in case of shorts).
For HT caps an insulation tester can be very useful for doing this.
For very low voltage caps you can use an ordinary resistance meter (checking which way it applies its voltage).

The forum search feature is rather useless because adding more words widens the search rather than narrows it. So far I haven't found a syntax that improves that. But you can use a regular search engine to search the site by adding "site:www.vintage-radio.net" in the search so it only looks on this site.

[GMB]

The simplest method is to connect a known resistor (i.e. 10kOhm) in series with the capacitor and apply a small voltage (i.e. 9V).
If the capacitor leaks, a small current will pass through the resistor and this will generate a small voltage.

No, you must use a voltage near its working voltage or the results can be misleading. Breakdown effects are non-linear and might not occur at all at low voltage.

[Station X]

Originally Posted by GMB

But you can use a regular search engine to search the site by adding "site:www.vintage-radio.net" in the search so it only looks on this site.

Or use the "Google Custom Search" box at the top of each forum page.

[stevehertz]

Not saying it's conclusive or the 'best' method, but I simply select ohms x 100 on my Avo 8 and connect its leads directly to each end of the capacitor. After an initial swing of the needle (depending on the cap value) it should make its way to infinite resistance. That's a good cap. Though as has been stated, caps can break down at higher voltages than are applied using an Avometer. But it's a good, quick test.

[skodajag]

Originally Posted by stevehertz

Not saying it's conclusive or the 'best' method, but I simply select ohms x 100 on my Avo 8 and connect its leads directly to each end of the capacitor. After an initial swing of the needle (depending on the cap value) it should make its way to infinite resistance. That's a good cap. Though as has been stated, caps can break down at higher voltages than are applied using an Avometer. But it's a good, quick test.

Thanks. Is that a test done in circuit, or out of circuit, or in circuit with one side of cap disconnected?

[stevehertz]

In general you cannot test caps in circuit due the effect of parallel paths (resistors, inductors, transistors, etc) that would make the cap appear to be leaky. If it's a valve circuit and you can see from the schematic that there are no parallel paths, then that's ok. But in general I tend to avoid testing in circuit. It just a case of lifting one end of the cap.

[jimmc101]

Originally Posted by stevehertz

Not saying it's conclusive or the 'best' method, but I simply select ohms x 100 on my Avo 8 and connect its leads directly to each end of the capacitor. After an initial swing of the needle (depending on the cap value) it should make its way to infinite resistance. That's a good cap. Though as has been stated, caps can break down at higher voltages than are applied using an Avometer. But it's a good, quick test.

Should mention that on an AVO8 (and all? analogue multimeters) the polarity is reversed on the ohms ranges.

Jim

[trobbins]

If this is for a coupling cap in a valve circuit then you may be able to do a test in-situ. For a common circuit with a coupling cap between stage A to stage B, and the cap going to stage B's input grid, with just a grid leak resistor to 0V on that grid, then remove stage B's valve and measure the DCV across the grid leak resistor. Caveats are that stage B's valve can be removed safely and without adverse consequences (eg. the supply rails don't rise beyond safe levels), and that stage A has the coupling cap connected to its supply rail (eg. via a plate load resistor). The leakage current is then a calculation that includes the input resistance of your meter. Typically the measured voltage is 0.0V, but increases with cap leakage current.

[Diabolical Artificer]

Personally if I'm testing power supply caps in a DUT I just check the DC value first if that's lower than expected I flick my meter onto AC volts to check the ripple. If the ripple is over 100mV ish I then unsolder the wire/s from the positive terminal or lift the positive leads from the PCB then test using a power supply with my DMM on DC milliamps, red lead to the supply, black lead to the cap. The power supply's negative goes to chassis/ground. If the result is over 100uA the cap is suspect.

The latter method needs care though, you need to bring the voltage up slowly or put a resistor in series to stop current surge, you also need to disconnect the ammeter when discharging the cap after test, else you'll keep blowing the meter fuse.

Lastly when using a variable supply you need to check the the cap is at the right test voltage, check with a meter, then disconnect the meter otherwise it will effect the test reading. Test at 10% over voltage as a last test, if a cap passes all tests it's good.

Andy.

[Station X]

One way of testing a grid coupling capacitor (that capacitor) in circuit is to disconnect the control grid end of it and measure the voltage between the disconnected lead and chassis. It should measure zero, but I've seen a hundred volts or more when the cap is leaky.

[stevehertz]

Originally Posted by jimmc101

Should mention that on an AVO8 (and all? analogue multimeters) the polarity is reversed on the ohms ranges.

Jim

Yes, but this is of no consequence when testing non polarised caps, for which the OP was asking for advice for testing. Ok, he did say non electrolytic caps, but as tantalums are also polarised (as per electrolytics), then yes, good advice in that case, connect Avo black lead to cap +ve. I have a switching panel on my Avo.

[Station X]

It's good to know that someone wants to know how to test non-electrolytic capacitors.

Reading some threads the approach seems to be change ALL the capacitors including the ones in RF circuits with values in pF. It obviously doesn't always work or we wouldn't have "What do I do next?" threads.

[G6Tanuki]

For non-electrolytic capacitors used in valve circuits, my approach involves a variable-voltage source, a 100KOhm resistor and a neon-bulb.

Lift one end of the capacitor so it's not shunted by any other resistance-paths.

Hook the 100KOhm resistor and the neon in series with the capoacitor then feed it from the variable voltage supply, and wind up the voltage!

The neon will initially light as the capacitor charges, but should then go out.

Very occasional flashes [let's say one every ten seconds] make me suspect the capacitor.

A flash every minute is OK. Fast flashing or the bulb staying on and the capacitor gets condemned.

And I wind the voltage uyp to the "Rated working voltage" as marked on the capacitor's outside - even if it is actually being used in an application where it will be exposed to a much lower voltage. Quite often you will find seemingly-high-voltage capacitors used for things like decoupling a receiver AGC line or the detector circuit or the first-audio stage, where there will be at most only a few tens of volts involved. But these are high-impedance points where you may have several Megohms in series with the capacitor, which if its resistance is a Megohm it can seriously upset the biasing/time-constants of the circuit.

In practice, there are some capacitors I don't bother to test: any Waxies, the black Philips "Tar-bomb" type, and the metal-cased TCC Metalpack/Metalmite types [which despite their shiny aluminium outer case and rubber-sealed ends are just tarted-up Waxies].

Don't bother testing these... just go ahead and replace. Personally I like the appearance of nice shiny yellow MKT/MKP capacitors that show a proper rework of a piece of gear has been done properly.

[stevehertz]

Originally Posted by G6Tanuki

In practice, there are some capacitors I don't bother to test: any Waxies, the black Philips "Tar-bomb" type, and the metal-cased TCC Metalpack/Metalmite types [which despite their shiny aluminium outer case and rubber-sealed ends are just tarted-up Waxies].

Don't bother testing these... just go ahead and replace. Personally I like the appearance of nice shiny yellow MKT/MKP capacitors that show a proper rework of a piece of gear has been done properly.

Totally agree.

[David G4EBT]

Originally Posted by G6Tanuki

And I wind the voltage up to the "Rated working voltage" as marked on the capacitor's outside - even if it is actually being used in an application where it will be exposed to a much lower voltage. Quite often you will find seemingly-high-voltage capacitors used for things like decoupling a receiver AGC line or the detector circuit or the first-audio stage, where there will be at most only a few tens of volts involved. But these are high-impedance points where you may have several Megohms in series with the capacitor, which if its resistance is a Megohm it can seriously upset the biasing/time-constants of the circuit.

In practice, there are some capacitors I don't bother to test: any Waxies, the black Philips "Tar-bomb" type, and the metal-cased TCC Metalpack/Metalmite types [which despite their shiny aluminium outer case and rubber-sealed ends are just tarted-up Waxies].

Don't bother testing these... just go ahead and replace. Personally I like the appearance of nice shiny yellow MKT/MKP capacitors that show a proper rework of a piece of gear has been done properly.

I concur with all of that.

I've retrospectively tested waxy caps, mouldseals and 'tar bombs' that I've replaced, and invariably they've been leaky when tested at their working voltage, and their capacitance out of spec. Given that they'll usually be at least six to seven decades old, we shouldn't be surprised.

Pic 1: 0.05uF Mouldseal capacitor on multi-meter 20 MegOhm range showing no leakage.
Pic 2: Tested at 250V on Robin insulation tester, shows 125K Ohms. (display shows 250K, but on that range, the displayed figure is halved).
Pic 3: Shows what the capacitor should be, and what it actually is. (When tested for capacitance on an MK328 meter, it measured 0.1uF).
Pic 4: Capacitors I replaced on a 1948 Murphy A124, tested after removal.
Pic 5. 'Tar bombs' replaced in a Stella 105U, which behaved more like low value resistors at their rated voltage.

I have a Robin 3131 insulation tester but also use a 'Victor VC60B', which is often sold under other brand names, and seems generally to be well regarded. Typically £35.00 on eBay and Ali-Express from China, or £45.00 on Amazon in UK. Useful for much more than just testing caps for leakage of course. (Often mis-described as 'Meggers').

A couple of YouTube video reviews of VC60B is use:

https://www.youtube.com/watch?v=q0aLf1TKwVo

https://www.youtube.com/watch?v=cBspKnKsk_A

Hope that's of interest.

[Uncle Bulgaria]

Originally Posted by skodajag

I know this has been discussed before, but I’ve never been successful in using the search facility on this forum. So, eg, searching the title above brings up numerous unrelated posts. I'm sure it’s me. So, can you please direct me to the info on this invaluable forum which tells the best way to rest for leaky non-electrolytic caps both in and out of circuit? Advice on using the search facility would also be appreciated.
Thankyou.

I find the search function pretty good. I think it works on basic keywords rather than natural language, so I use a variety of keywords, but the more you have the more you have to search through them. Remember search modifiers: using "quotation marks" searches for the whole phrase within them. Using + before a phrase, i.e. "leaky caps" +test would search for the phrase 'leaky caps' but only include posts with the word 'test' in as well.

For your query, I would have tried two words: 'capacitor testing', which throws up a variety of threads (using proper nouns rather than slang/abbreviations usually gets better results, along with the smallest number of keywords). Try several different searches with different terms and you can often narrow things down, or sometimes I find a thread with a link to a further thread which gives you a more field-specific keyword from one of the real experts.

I made a capacitor tester using a board bought from Everyday Practical Electronics magazine, which is great for measuring leakage at up to 630V. You can set the time to apply the voltage and see if the leakage current decreases to an acceptable value (whether the capacitor 'reforms'). It's proven useful with every project. I've attached the relevant pages - boards are still available.

In general, with my limited experience, I use the information from the vintage-radio.com part of this very website to identify capacitors likely to be troublesome before power is applied (look under the 'components' heading - grid coupling waxed paper types etc.) and replace them. I put them on the tester anyway and am gratified by their failure to hold charge.

[Craig Sawyers]

I use a Megohm meter. These were made identically by different manufacturers. AVO made one, and other manufacturers identical ones. 100, 250, 500 and 1000V. 10^5 to 10^11 ohms switchable.

The AVO one is the RM290, and the Sullivan one is T2900. Darned useful if you can find one. General Radio made one too - the 1862 and 1863.

Taking the example of a dreaded Hunts 350V paper cap. On a DVM it might measure fine. But put on one of the above meters, you get maybe 10^6 ohms at 100V and the meter might peg at 250V. on the 10^5 ohm range. So for those notorious parts the leakage is a strong function of applied DC voltage.

Craig

Craig