HV capacitor leakiness test?

[Al (astral highway)]

Hello, especially amateur licence holders with experience of EHT B+ voltages.

I have a doubler circuit with paralleled 0.85uF capacitors, one or more of which is leaky. I know this as the transformers were saturating and overheating under no load, and the diode chain also failed, from passing too much current. I've been running the Xformers for 10 mins disconnected from the failed cap doubler and they are within their normal heating range.

I will redesign with better overcurrent detection and protection but I'm having difficulty isolating the 'guilty' capacitor. They're £10 a shot! (Those I've used so far were culled from street finds in discarded microwave ovens, but these are getting scarcer as people move to Inverter based ovens, the latest generation).

The caps are rated 2100VAC. Obviously I understand leakage current but I'm wondering how people would tackle this test for a definitive and trustworthy result, specifically for paper in oil capacitors (pretty sure, aren't they?) given the cost of getting it wrong.

Cheers,

[GrimJosef]

My approach (usually limited to a few hundred volts, to be fair) is based on having a supply of suitably high voltage. It doesn't need to be able to deliver significant current though. I ground one side of the supply's output and connect the other side to one terminal of the capacitor under test, via a resistor capable of protecting the supply if I fear the cap will fail to a hard short circuit. I connect the other terminal of the cap via a cheapish digital multimeter to ground and I set the meter on its maximum volts range. If I can find one I wire a switch across the meter and close it.

Then I turn the supply on (or wind the volts up gently if that's an option). Once I'm convinced that the capacitor is DC charged I open the switch which has been shorting out the meter and I check for any voltage reading. These meters are typically 10Mohm input resistance so on their voltage range even 1uA of leakage through the cap will register a clear 10V on the meter. By working down through the voltage ranges in principle I could measure less than 1nA leakage current.

Things worth noting are

1. That any noise on the supply's output will appear in full across the meter which might mislead us into thinking that the cap is intermittently/noisily leaky,

2. If we really don't want to risk blowing the meter up in the event of a sudden cap failure it might be wise to wire some overvoltage protection across it (e.g. HV Zeners),

3. Capacitor failure can be very nonlinear - they can go from testing fine to failing badly over just a few percent increase in voltage,

4. If they run in a warm environment then failure may occur only when they've been at full operating temperature for a while,

5. Sometimes even mechanical details can matter - clamping too tightly can bring internal metalwork into contact.

Cheers,

GJ

[Al (astral highway)]

Hi GJ,

That's a very clear and useful protocol you've got there. I like it. Thanks for sharing that and undoubtedly useful advice for anyone testing most caps.

It is, however, going to be very difficult to scale up by a factor of ten, I have no DMM (as you say, necessary for the high input resistance) that can tolerate 2-3KV (leaving out the doubler - the xformers have 2100V on the high side).

Hmm...

[Craig Sawyers]

It is worth looking out a Megohmmeter. There seems to quite a choice out there from China-land which will go to 1kV at 2G-ohm for low tens of quid.

Craig

[Richard]

Hi Al

Unless your caps are early they will have a built in discharge resistor, could this be your problem?
Also some have a built in HT diode, from one terminal to case.

I worked for a company dealing with commercial and domestic microwaves for 30 years, I will pop in sometime and see what caps I can find, there were boxes of them.

There was a box of 1uF 2500V, with no discharge resistors.
Also a pile of HT transformers, 2500V, 1000W, with both ends of the HT winding coming out, fully insulated, unlike most which have the inner end connected to the frame.

Richard

[GrimJosef]

Originally Posted by astral highway

It is, however, going to be very difficult to scale up by a factor of ten, I have no DMM (as you say, necessary for the high input resistance) that can tolerate 2-3KV ...

I just use an everyday DMM. One terminal is grounded and the voltage on the other terminal is set by the cap's leakage current. If I suspect that the leakage current might be as much as a milliamp then I wire a 10kohm resistor across the meter.

Having checked your numbers again in this case it really will be necessary to have a shorting switch across the meter while the cap charges up. 0.85uF and 10Mohm would put the bulk of the supply voltage across the meter for many seconds.

Cheers,

GJ

[russell_w_b]

Originally Posted by GrimJosef

My approach (usually limited to a few hundred volts, to be fair) is based on having a supply of suitably high voltage. It doesn't need to be able to deliver significant current though.

If you can get hold of an old insect-zapper (or one of those air ioniser thingies), they have a 230V - 2kV transformer in that has a suitably high-impedance, and hence low current secondary. But you'll still need a variac or other means of winding up the voltage to that required.

[Al (astral highway)]

Ahah! I've answered my own question.

There are five caps in total, so by applying the same test to each, I should get the same kind of result.

Here's what I did, using only a DMM. It is a very safe test indeed, which is another plus.

1) Short each cap to insure no residual charge (there is an internal 10 MOhm resistor, but still, not worth any risk).

2) Set DMM to max ohms range (20MOhms on my DMM)

3) The capacitor and body resistor form a charging circuit that will eventually reach 5 time constants, when any capacitor reaches a steady state of charge. The capacitors in the circuit are between 0.85 and 1 uF. Because of the value of the safety resistor internally in parallel with the cap, 1 time constant is given by 0.85 to 1 uF * 10 MOhms, 0r 0.85 to 1e6 * 10e-6.

Since the exponents are the same, negative six and six, and cancel each other out, it's actually a very easy bit of mental maths.
---------


Let's take cap number one, 0.95uf .

1 time constant is therefore 0.95 *10 =9.5 seconds.

So 5 time constants is 9.5 *5 = 47.5, call it 48 seconds. The tolerance is around plus or minus 3%. So we're looking at near enough 50 seconds or 46 seconds, keeping things simple.

Cap one charges to a steady state around the value of the body resistor in 51 seconds, in spec. Reversing the leads, it climbs back to the same state in the same time.

I stop the clock but keep the DMM charging. The cap holds its charge and continues to read the value of the body resistor. This cap is ok,.

Caps two, three and four climb faster to the value of the body resistor but continue to charge to 14 or 15MOhms, then climb to read infinity. These caps are leaky.

Cap number five also reaches a steady state in 53 seconds, in line with the 10MOhm value of the body resistor. I reverse the leads and apply the same test. The cap holds its charge and the DMM maintains the reading .

By carrying out this test, I work out that I can keep two caps but throw three away.

In the rebuild, I will put two caps in series for longevity, because when the power supply is not under load, the caps are stressed beyond tolerance.

The test is very simple because there's already a resistance hard-wired into the caps. I hope others can benefit from it. It's crude but reliable!

[Al (astral highway)]

Originally Posted by Richard

Hi Al

I worked for a company dealing with commercial and domestic microwaves for 30 years, I will pop in sometime and see what caps I can find, there were boxes of them.

There was a box of 1uF 2500V, with no discharge resistors.
Also a pile of HT transformers, 2500V, 1000W, with both ends of the HT winding coming out, fully insulated, unlike most which have the inner end connected to the frame.

Hi Richard,

I'm sending you a PM...

That would do me a huge favour, thank you! I'm now on the lookout for 2 x
the high voltage xformers and six caps rated as you describe...

The bigger the xformer (wattage) the better, so very interested in the ex-commercial microwave oven ones. The ones you describe, without one winding returned to the E-core, I think saturate less than the domestic ones, which draw 2-3A current on no load!

Please could you let me know what you find, I can contribute something for your trouble!

Thank you!!

[Ed_Dinning]

Hi Al, a lot of the commercial microwave oven transformers use a TIG welded lam stack.
This weld bead can easily be removed with an angle grinder as it is not deep and the lams re-stacked after and rewinds.

Ed

[Andrew B]

If you look carefully, you can see the gnd/body secondary connection coming out of the bottom of the winding, an easy job to disconnect it and join another piece of wire to the end of it.

[Richard]

Hi

Yes you can, but the end of the winding has very little insulation relative to the core, which can be an issue dependant on what you are trying to do.

[Al (astral highway)]

Originally Posted by Andrew B

...an easy job to disconnect it and join another piece of wire to the end of it.

I know, but there's a difference between MOTS, which as you say are typically wired with the secondary grounded to the E-core (I've seen tens of them) and EHT transformers designed to have isolated secondary windings on both sides. There is a safety reason for the grounding of one side, and if we lift it, the whole thing is under a lot more dialectric stress..!

[Al (astral highway)]

Originally Posted by Ed_Dinning

...a lot of the commercial microwave oven transformers use a TIG welded lam stack.
This weld bead can easily be removed with an angle grinder as it is not deep and the lams re-stacked...

Good call, Ed, I didn't realise that. I am going to practice removing a shunt on a burnt-out domestic Microwave oven transformer, it looks a bit fiddly to do without damaging the primary, although perhaps easier if I remove the red magnetron filament winding first??

Thanks Ed!

[Argus25]

Astral Highway,

If you had a very high voltage probe with high resistance input, you could charge your caps up to the full operating voltage and simply wait and measure to see how much charge (&voltage) they keep with time.

I think with your high voltage work going the way it is, what you ideally need to have on hand is a 1000:1 frequency compensated EHT probe.

The whole issue of high voltage probes is a somewhat vexed one. There are the types of probes that are good for 30kv for TV EHT etc, but these probes act as a very effective LPF because of the high series resistance and stray capacitance, they are useless for waveform information. There is a Tek oil filled frequency compensated probe, that can make it to 30kV to 40kV with the oil, but its very expensive.

I was faced with the dilemma of measuring high frequency waveforms that would go at times to 50kV. The only way to do this was to make my own frequency compensated high voltage probe. The compensation capacitor takes the form of a rod parallel to the oil filled chamber with the resistor in it, the story of it is here:

http://nebula.wsimg.com/1b62eb4a5f8e...&alloworigin=1


I scavenged the oil from a Lucas ignition coil.


With this probe I was able to measure some very brief off load ignition coil secondary voltages on a Thyratron based CDI unit; go to this article and check out the -40kV recording of the secondary voltage of an unloaded ferrite cored ign coil scroll to page 27 of this article. A recording like this would be impossible without the probe.


http://www.worldphaco.com/uploads/TH...OTION_Inc..pdf

In the end I use this probe all the time now, for any high voltage recordings/measurement of CRT anodes, focus chains, waveforms, for example I can connect it to the anode of the H output tube in a TV set and look at the voltages/waveforms there no problems at all, so its turned out to have multiple uses, even though it was designed initially for ignition system investigations.

PS: if you are building an EHT probe at some stage, don't be tempted to use a series resistor chain for the main dropper resistor. While it seems like a sensible idea, if there are high frequency components in your measured waveform, there is more stress on the initial resistors in the chain at the probe input due to stray capacitance and they usually fail here. It is better to use a dedicated single high value resistor made for the job and preferably in oil.

Hugo.

[Phil G4SPZ]

Originally Posted by russell_w_b

If you can get hold of an old insect-zapper (or one of those air ioniser thingies), they have a 230V - 2kV transformer in that has a suitably high-impedance, and hence low current secondary. But you'll still need a variac or other means of winding up the voltage to that required.

Even cheaper and easier are those inexpensive electronic fly-swatters that look like a tennis racquet, obtainable for a few pounds at household discount shops. I measured the output of mine at 1.8kV, generated from two AA cells by some form of inverter circuit. Powered from a variable low-voltage PSU, it should provide a variable EHT source relatively safely.

I've not used one for this purpose myself, but there's no reason why it shouldn't work.

[Al (astral highway)]

Hey Hugo,

Thank you for posting details of the EHT probe you designed, I like it a lot and as you say, very useful to measure what is really going on with waveforms including complex waveforms. The second article you researched, wrote posted does illustrate this complexity very neatly. A lot of food for thought here!

I do ‘get’ your interest in these ignition systems - I designed and built a transistorised automotive ignition about 17 years ago now, lasted five years!

Phil, good idea there, appreciate it.

[russell_w_b]

For what it's worth, I've always checked capacitor leakage the 'ugly' way by putting a 1kV Megger in series with an Avometer on the highest current range, then switching down as it charges up, until I can make it - if it makes it - to the 50uA setting.

After a minute or so, I open the circuit (or else the cap back-feeds through the Megger and stands to damage the Avo) then, after a few minutes, I check the voltage across the cap with a DMM, then again, a few minutes later, before discharging the cap via a lamp.

It might not be the most elegant method but I've never had a problem with it. The wax-paper 6uF 1.6kV in my Beethoven exploder capacitor didn't show leakage until over 500V were applied.

It ought to be easy enough to knock up a little HV source if you can't get hold of a ready-made one as Phil suggests, with his tennis racquet.

[Al (astral highway)]

Hi Russell,

That’s a great idea and I do understand the importance of trying to test leakage at the operating voltage.

Thing is, though, these caps always come with an internal parallel body resistor of around 10 MOhms. It means no tests are possible that are looking for how well the capacitor holds its charge, only whether it will charge at all.

Thanks for your idea , will keep it in my box of tricks for other caps

[russell_w_b]

Originally Posted by astral highway

Thing is, though, these caps always come with an internal parallel body resistor of around 10 MOhms.

Then you're looking for a maximum resistance of 10 Megohms. Any significant leakage would show up as a lower value resistance, being that of your leakage resistance shunted with 10 Meg.

If you use a 1kV Megger you'll draw no less than 100uA at steady-state. Anything more than 100uA will indicate a leakage current.