Let's talk about testing capacitors

[TomCarlos]

Without using the dedicated power supply or at least the 12k current limiting resistor the whole system makes no sense at all.
Rather you will damage your current meter.

Gotcha.... I watched the Part 3 video and at the 26:00 minute mark, he explains the 12K resistor. In Part 4, he adds a diode after that resistor.

I will experiment with my tester and see if I can incorporate those two additions.

Thank you.

[Helmholtz]

Leakage current is the DC current flowing through the cap when connected to a DCV source.
The current must be limited by a series resistor. In cap literature I found series resistor values ranging from 1k and 100k.
For most practical purposes we're interested in the final sustained current at rated voltage.
With ecaps the leakage current drops for a while after charge up.
Reason is an electrochemical process called re-forming, which repairs tiny voids in the insulating Al2O3 layer.
There is no quick way to reliably measure leakage. Minimum measuring time should be 2 minutes.
Major ecap manufacturers recommend to measure leakage after a re-forming time of at least 1 hour when a cap hasn't been used for a while.

My own method is simple:

Cap with 10k series resistor connected to HV source. Voltmeter (DMM) connected across resistor.
Leakage current is found from the voltage drop across the resistor divided by resistor value.
Let sit until the resistor voltage stabilizes.

With HV ecaps 100µA is good, 1mA is borderline.
With film/foil caps even 1µA means bad leakage.

[Helmholtz]

To add:
Acceptable leakage current depends on the application, i.e. the circuit position of the cap.
E.g., a leakage current of 1mA with a reservoir cap will go unnoticed, but if the filter cap after a 10k dropper has a leakage current of 1mA, it will increase the voltage drop by 10V which might not be acceptable.

[Tom Phillips]

Also note that the "acceptable" specified leakage of aluminum electrolytics varies (i.e. rises) with increasing rated capacitance and voltage.
Attached is a snip from a Kemet data sheet that I chose at random. The full data sheet is available at https://www.mouser.com/datasheet/2/4...SG-3316600.pdf

[Helmholtz]

According to the Kemet formula above a 32µ/500V cap would have a leakage current of 0.97mA @20°C.
Not really a good value for a tube amp, considering that the leakage current will be 2 to 4 times larger at an ambient temperature of 50°C (as can be expected inside an amp).
I have a datasheet of an F&T (dual) 32µ/500V cap specifying a leakage current of 128µA @20°C.
I found even lower spec values for EPCOS ecaps
This example shows that there is considerable difference between brands and cap types.

[TomCarlos]

Thank you Tom P. and Helmholtz.

I have started to dive into spec sheets for caps and looking at various leakage calculation formulas. I see what you mean!! There are definitely differences from one cap to another.

The table below comes from an F&T 500V Radial Cap (that Helmholtz is referring to). Plugging in the 32uf and 500v, that gives us the 128uA. So I guess the leakage is dependent on the Manufacturer, Cap Voltage Rating, etc.

= .008 * C(uf) * U(v) + 4uA
= .008*32uf*500v
=.000128
=128uA plus the 4uA for good measure

Always learning something new!!

[Helmholtz]

Note how much the leakage drops after 1h re-forming.

[g1]

Multiplying capacitance by .008 rather than .06 is a huge variation. I'm quite surprised at the range.
Here is a bunch more values.

[tedmich]

Im reminded of the scene where Sean Young has to be questioned for HOURS in order to tell if she's a replicant in Bladerunner.

Since we're not talking people here maybe just buy new caps?

[Helmholtz]

Since we're not talking people here maybe just buy new caps?

Maybe because not all new caps are good enough for the circuit?
Look at the published data.
Values are given for new caps as they left the factory but after a storage time of 1 or 2 years leakage will be worse to unknown degree.
It is useful to know how much leakage current the circuit can tolerate.

Given good capacitance and ESR values, I'd prefer a 10 year old cap to a new one when both have the same leakage.current after re-forming.

[The Dude]

..........Given good capacitance and ESR values, I'd prefer a 10 year old cap to a new one when both have the same leakage.current after re-forming.

That's one of the reasons I don't ascribe to the idea of wholesale replacing caps simply because they're old.

[R.G.]

There's an issue lurking there. Industrial chemists are working hard to increase yield, improve specs, and reduce manufacturing cost. The tradeoffs are not obvious, as witness the capacitor plague from the early 2000s. The issue is that a few years is enough time to get significant changes in the internal chemistry and manufacture of electro caps. Ten year old caps very likely come from a different era of chemistry and maybe manufacturing process. The chemists and industrial engineers work with reams of historical data and large, statistically significant sample batches to make finer guesses from. They have data you don't have.

It's entirely possible that ten-year-old caps with same leakage ARE in fact more trustworthy than new ones. And that nine-year-old ones are dramatically worse. And that five-year-old ones are the best of all. We -sadly- don't have enough data to know or to plot trends. Those super-reliable ten-year-olds were once untrustworthy one-year-olds.

Then there's survivor bias. In ten years, the "weak" caps of that vintage have failed and been replaced. The survivors, however few, look pretty good. Then there's failure rate. Caps follow the usual bathtub curve of reliability, best we can tell. Infant mortality wipes out a fraction early, reliability is a fairly stead % per 1000hours for the bottom of the bathtub, then wear-out starts, and the failure rate goes up a lot. Where is -this- capacitor in the bathtub? Of this batch of ten-year-olds, are one or two of them about to give up the ghost?

We don't know. All we have are the statistics. A remarkably few electro caps will survive to 20, 30, 40 years. A relatively huge number of new ones will survive to five years. We can believe in ancestor's goodness, but in reality, we're playing the statistics and making our bets.

[Helmholtz]

We don't know. All we have are the statistics. A remarkably few electro caps will survive to 20, 30, 40 years. A relatively huge number of new ones will survive to five years. We can believe in ancestor's goodness, but in reality, we're playing the statistics and making our bets.

We can measure leakage, ESR and capacitance.

BTW, I have a lot of 32 years old Siemens Long Life ecaps (LL series, 450/600V, 22µ and 47µ), axial) that still measure better than many new caps. (That was a special order version for my company, where we needed a peak voltage of 600V and an operational life of 50,000 hours. We bought in millions.)
Not saying that older caps were generally better, but I'm happy to leave an original cap in a vintage amp alone if it still measures ok.
Ecaps don't die abruptly.

[The Dude]

I see this often here: Should I replace output transistors? Should I replace such and such capacitors? Are my input jacks bad? Etc. I'm not advocating leaving old caps without testing them- same as any other part. I see even seasoned techs sometimes forget they have measurement tools right in front of them.

[R.G.]

We can measure leakage, ESR and capacitance.

Yes, we can. Those things are indicators that are more or less correlated to failure.

BTW, I have a lot of 32 years old Siemens Long Life ecaps (LL series, 450/600V, 22µ and 47µ), axial) that still measure better than many new caps. (That was a special order version for my company, where we needed a peak voltage of 600V and an operational life of 50,000 hours. We bought in millions.)
Not saying that older caps were generally better, but I'm happy to leave an original cap in a vintage amp alone if it still measures ok.

Wow - those must have been very expensive. It certainly is possible to tweak the chemistry and process to get very long life. Those things must have been from before etched foil construction, even.

My point was not that any vintage or age is better, just that they vary in expected lifetime, the part we haven't seen yet; and the decision to leave a cap in or replace it amounts to a bet on its remaining lifetime. If you get good results from measuring and leaving it in, great.

Ecaps don't die abruptly.

I've had a few die abruptly. In my previous job I designed power supplies for a while, so I tended to be who got yelled at when something melted down in a customer's office, for whatever reason. It's unusual, but it does happen. Design for reliability is a demanding task, and all you have to go on is the statistics.