DC leakage in coupling caps (theoretical question)

[Raoul Duke]

From what I’ve picked up in my short time learning about amp building and tweaking - in a perfect world; coupling caps should leak “0” current when in circuit at operating voltage and temperature. I’ve researched this for the past few days and read a lot of conflicting opinion - so I wanted to ask here hoping that the more experienced folks might weigh in.

My theoretical question is:

In practical applications, is there an “acceptable” allowance given the variables that age, run time, temperature, wall voltage etc. where some leakage is possible (expected)?

Example: in a typical 400v cap, less than .003uA is acceptable?

My understanding is the DC leakage affects the following stage(s) bias - but mathematically there must be a margin that these successive stages operate in given the previously stated variables given that we don’t live in that perfect world?

Or am I way out there with the “new guy” type questions?

[martin manning]

Ideally for a common cathode preamp stage you would like to see 0 V on the grid. If the preceding coupling cap is a little leaky, you might measure 2 or 3 mV. That would be acceptable to most people. So what does that mean about the leakage current? Consider the cap as being the upper resistor in a voltage divider. Assume there is 400V on the opposite end of the cap, you measure 2 mV on the grid (the output of the divider), and that the grid leak (lower resistor) is 1Meg. Ignore the meter’s input impedance (10-20 Meg) and the tube’s DC input impedance. That means the cap looks like a 499 Meg resistor, and it’s leaking 0.8 uA at 400V.

[Raoul Duke]

Thank you Martin,
That’s kind of what I’m getting at (I think). Depending on the specific circuit (and I know that can vary a lot) and the specific “ear” (more variation there too) - there must be a range where it becomes somewhat inconsequential/indescernible - or not.

Holistically/practically speaking WRT a given circuit, is monitoring the preamp plate voltages the most obvious tell-tale that too much DC is getting by or are there more indicators? Fender board noise comes to mind because I’ve read that it’s sometimes conductive paths between eyelets and sometimes leaky caps. So if the preamp plates are higher than spec then that would point to a preceding cap leaking enough to be the likely cause of the problem?

Last week’s discussion in my other thread (AC30) sent me down a rabbit hole of sorts and I’ve been reading a bunch about leakage and it’s causes/effects etc. I’ve seen everything from “absolutely not” to “some (unquantifiable amount of) leakage is part of the ‘mojo ” and quite a bit in between - so I kept reading until I reached saturation and then decided to ask my trusted advisors here at TAG for some context and perspective.

I always appreciate the lessons learned here.

[martin manning]

In either case (a leaky cap or a conductive board) the grid voltage will go up a bit, and that will cause the tube to draw more current, lowering the plate voltage and increasing the cathode voltage.You can also measure the voltage across the plate resistor or the cathode voltage and calculate the current to see if it’s out of line. Checking the grid voltage to see if it’s positive is the most direct way to detect leakage.

[trobbins]

Holistically/practically speaking WRT a given circuit, is monitoring the preamp plate voltages the most obvious tell-tale that too much DC is getting by or are there more indicators?

To add to Martin's comments, your above comment about monitoring a plate voltage is sort of correct, with the proviso that imho it is a 'check' not a 'monitor', as its benefit is to check the plate voltage shift with the coupling cap connected and when disconnected. Checking for tag board insulation leakage is more onerous, as the grid link would need to be lifted from the board and a temporary grid leak resistor connected. I find measuring anode voltage as a 'check' is less sensitive to other issues like meter input resistance and meter measurement resolution, and relates better to what voltage shift is doing to the operating loadline of that stage.

The decision to replace a coupling cap is more vague, as it requires an awareness of what leakage is typically observed with a particular cap model, and whether that leakage could degrade over time. It can also require an awareness of collateral damage, such as the demise of a coupling cap to the output stage can cause damage, but a preamp stage is likely just going to cause signal colourisation or demise. Also keep in mind that tube-swapping can give some context as to how much the idle anode voltage can change, but still be 'nominal' for the application.

[Raoul Duke]

Understood and thank you both,
This gives me plenty to ponder and definitely connects some dots.

Martin - your grid voltage pointer makes more sense to me. Seems obvious now after the fact - but that’s me slowly piecing this together.

Trobbins - your measuring anode voltage is a great tip as well; now that I understand the difference (WRT monitoring the plates) thanks to your explanation.

Great info, thanks again!

[B Ingram]

From what I’ve picked up in my short time learning about amp building and tweaking - in a perfect world; coupling caps should leak “0” current when in circuit at operating voltage and temperature. ...
My theoretical question is:

In practical applications, is there an “acceptable” allowance given the variables that age, run time, temperature, wall voltage etc. where some leakage is possible (expected)?

The "theoretical question" is about "practical application"?

The Shortest Practical Answer is that if a coupling cap has any noticeable or (easily) measurable leakage, then you probably don't want it around anymore. Things are never going to "get better."

The Perfect Answer (for this, and for issues around tube testing) is "The capacitor's condition is acceptable if it continues to deliver satisfactory performance."
- This answer is frustrating for novices (and myself 15 years ago), because a lack of experience means it's hard to know what constitutes "satisfactory performance."
- This answer is perfect because the goalposts move depending on the requirements of the circuit using the part/tube.
- This answer is also perfect because no matter the test result, the part/tube is acceptable as long as the circuit doesn't exhibit an "obvious malfunction."

_______________________________________________________________

Martin gave an example of a cap leaking 0.8 microamperes, and behaving like a resistor of 400v/0.8µA = 500MΩ (really 499MΩ because of the 1MΩ grid-leak).

That's a very small leakage current, but consider what happens if it increases to a very small leakage current of 2µA.
- 1MΩ x 2µA = 2 volts
- Most 12AX7 stages have only 1v of bias, so this extra input throws things off considerably.

What if that 2µA of leakage happens at a stage where the next-tube's grid-leak is only 220kΩ? And what if there's a negative bias voltage of -45v present?
- 220kΩ x 2µA = 0.44 volts
- An extra "0.44 volts" on a bias of -45v makes the bias -44.56 volts which is "basically unchanged"

And what about the resistance this implies?
- 400v / 2µA = 200MΩ ---> 199MΩ after we subtract the 1MΩ grid-leak
- If we had a 500MΩ resistor, fingerprints on the body might be enough make it behave like a 200MΩ resistor.

_______________________________________________________________

Back to the Practical Answer.
- For the most part, modern caps with plastic dielectrics don't leak DC volts.
- If you have encountered acid-free paper for archival stuff, you know they don't make that kind of paper by accident.
- Before the use of plastic film in caps in the mid-20th century, a significant % of caps used a paper dielectric.
- Paper dielectric works fine at first, but over decades breaks down and becomes increasingly leaky.
- A paper dielectric cap that has started leaking is never going to get less-leaky; the process is moving in one direction.

Some paper caps are worse than other paper caps about breaking down & developing leakage. But if they do, when is the leakage too much? It's when the leakage affects circuit-performance "enough to notice" that the cap needs replacement. But some communities rip-and-replace any paper-dielectric cap because they want the item fixed for the next couple decades (not going back & forth for repair as caps fail one-by-one).

Since the answer is "it depends" and varies with the application, some folks will default to the simplest concrete answer: the only acceptable leakage is zero (easily measurable) leakage.

[Raoul Duke]

Thank you, that makes a lot of sense.

Your “It depends” explanation is what I was driving at. I just wasn’t sure if that more pragmatic approach was even a thing in this situation given the seemingly concrete “0 leakage” tenet that seems pretty common from what I’ve been reading.

The “theoretical” part is my question coming from the perspective of “is “0” leakage the only correct answer?” - which stems from my lack of technical knowledge here. I wasn’t even sure it was a question that would merit response from folks with a solid technical background; so I couched it as such. Looking at the way I worded it - it is a bit contradictory/funny. I guess I meant “in theory, is there an acceptable margin; and if so is it a percentage based on the application in circuit or is it subjective/performance based?”
Hadn’t had my coffee yet…

This is all great learning for me and I really appreciate everyone patiently sharing their experience and knowledge. Thanks again all!