Hi guys, I am looking for a very low leakage capacitor. Value flexible, say between 1uF and 10uF, voltage rating >10V. It should have a leakage less than 10nA at DC.
Does such a beast exist? Probably film, teflon?
Jan
Does such a beast exist? Probably film, teflon?
Jan
I've seen 10 uf 63 v film caps in geophysical products. Texcap was the vendor but that was 35 years ago. They were huge. These are not normally stocked at distributors. 1 hz response matters on geophones in oil exploration. Lots of oil exploration systems were built in west Houston where I worked. Look at distributors around where Total buys their exploration equipment. Else you'd have to make a 10000 unit buy.
AVX lists a 10 uf 50 v multilayer ceramic cap. I found some CPO's for $7 at Newark (farnell) 10 years ago, but have never seen them since. Put them in my H182 organ instead of defective tantalum electrolytic caps.
AVX lists a 10 uf 50 v multilayer ceramic cap. I found some CPO's for $7 at Newark (farnell) 10 years ago, but have never seen them since. Put them in my H182 organ instead of defective tantalum electrolytic caps.
Last edited:
Did you measure normally available high quality caps?
Maybe some meet that spec, out of the shelf.
It does not look outrageous, at all.
Get a , say, 15V DC supply, cap in question, a cheap 1M input multimeter, cap in series with input , what DCV does it show?
If 100mV or less, you´re done.
Even without doing the (easy) test, I leave that to you he he, suspect any good quality film cap, specially Teflon as you mention, will pass the test.
Obviously 1uF will be far easier than 10uF .
No I haven't done any tests, just wondering if there are some known brand/types for low leakage applications.
I have some film WIMAs I should check out.
Jan
I have some film WIMAs I should check out.
Jan
Any film cap will easily fit the bill, including the cheapest mylars.
As they are probably not specified/tested (see the typical time-constant spec) for that level/purpose, you would be safer making individual tests before soldering the part into circuit, but I expect that 99.9% of the samples will pass with flying colors.
You can probably buy special caps with a sub-nA guaranteed leakage (for special SH or electrometer applications), but they are going to be tear-drawing expensive.
A 40V to 63V mylar is not going to be bulky, and if you can locate old MKL 25V parts, they might be even smaller.
Ceramics could be OK -or not-, but I would stay away from them: they will be high-k types, with various other flaws, like hysteresis, tempco, voltage dependency, etc.
Some E-caps might do, including tantalum, but that would need to be tested.
Film caps are a safer bet, if you can bear their (relatively) large size
As they are probably not specified/tested (see the typical time-constant spec) for that level/purpose, you would be safer making individual tests before soldering the part into circuit, but I expect that 99.9% of the samples will pass with flying colors.
You can probably buy special caps with a sub-nA guaranteed leakage (for special SH or electrometer applications), but they are going to be tear-drawing expensive.
A 40V to 63V mylar is not going to be bulky, and if you can locate old MKL 25V parts, they might be even smaller.
Ceramics could be OK -or not-, but I would stay away from them: they will be high-k types, with various other flaws, like hysteresis, tempco, voltage dependency, etc.
Some E-caps might do, including tantalum, but that would need to be tested.
Film caps are a safer bet, if you can bear their (relatively) large size
Physically bigger is better because at that current, board and component surface leakage becomes the limitation. Actually getting conductive flux off a PCB is hard these days with the limited choice of board cleaners that are left
You can use several identical low leakage types in series if necessary.
https://www.mouser.com/datasheet/2/315/ABD0000C179-947645.pdf
https://www.mouser.com/datasheet/2/315/ABD0000C179-947645.pdf
When I look at your datasheet, I see the value of the isolation resistor, 100's or 1000's of megohms at specified voltages.
I think this is the spec for leakage current. We often look at the dissipation factor / dielectric absorption but that is only valid for AC, isn't it.
Using a cap in a timing circuit with charge / discharge times of seconds, that's basically DC and DF / DA does not come into play. So it is the isolation resistance.
Jan
I think this is the spec for leakage current. We often look at the dissipation factor / dielectric absorption but that is only valid for AC, isn't it.
Using a cap in a timing circuit with charge / discharge times of seconds, that's basically DC and DF / DA does not come into play. So it is the isolation resistance.
Jan
Tested a few caps along the idea of JMFahey: 20VDC and cap and a DVM with 1M input impedance (Keithley U3401A) in series.
With 1uA leakage, the meter would indicate 1V.
Not a nice picture, I check a couple of WIMA foils, 3.3uF 250V and 1uF 50V. At switch on the level is several volts meaning several uA leakage.
After 5 minutes or so it drops to below 10nA, keeps on dropping but slowly.
To see if this is one-time forming/break-in, I put them away for half an hour, then tested again. Again, started out at several volts, then started dropping.
I'll keep them under power for a few hours, then look again tomorrow morning.
Comments?
Edit: jackinnj, maybe you can do a similar test with your super-duper polystyrenes?
Edit: just looked again, 50uV across 1Meg is 5*10^-12 A = 5pA. But does it hold?
Jan
With 1uA leakage, the meter would indicate 1V.
Not a nice picture, I check a couple of WIMA foils, 3.3uF 250V and 1uF 50V. At switch on the level is several volts meaning several uA leakage.
After 5 minutes or so it drops to below 10nA, keeps on dropping but slowly.
To see if this is one-time forming/break-in, I put them away for half an hour, then tested again. Again, started out at several volts, then started dropping.
I'll keep them under power for a few hours, then look again tomorrow morning.
Comments?
Edit: jackinnj, maybe you can do a similar test with your super-duper polystyrenes?
Edit: just looked again, 50uV across 1Meg is 5*10^-12 A = 5pA. But does it hold?
Jan
Last edited:
Yes it holds: I just tested randomly picked caps at 10V, 2~3 minutes settling time: a 10µF/100V mylar by Epcos: <0.2nA
10µF/63V PC by Ero: 1.3nA
10µF/100V mylar by Ero: 0.3nA.
There is no forming or breaking-in at work with these dielectrics: what you see is an illustration of dielectric absorption.
Each time the cap sees a significant voltage change, it will need some time to settle.
The farthest components of the capacitance are connected through (notional) very high values resistors, in the gigaohm or even teraohm range, and it takes some time for the assembly to settle ~completely (it never really does of course).
Polypropylene, polystyrene, teflon have a low DA. See the DA-boy thread for more detail.
Leakage current is a pure DC parameter, measured after an ~infinite settling time, but in practice waiting for an infinite time is rarely convenient.
In the dielectric, all the previous voltage history is recorded, even days after the cap has been kept shorted. The longer the time, the higher the voltage, the stronger the imprint.
I have seen mylar caps having been constantly operated at high voltages for years behaving almost like electrets afterwards
10µF/63V PC by Ero: 1.3nA
10µF/100V mylar by Ero: 0.3nA.
There is no forming or breaking-in at work with these dielectrics: what you see is an illustration of dielectric absorption.
Each time the cap sees a significant voltage change, it will need some time to settle.
The farthest components of the capacitance are connected through (notional) very high values resistors, in the gigaohm or even teraohm range, and it takes some time for the assembly to settle ~completely (it never really does of course).
Polypropylene, polystyrene, teflon have a low DA. See the DA-boy thread for more detail.
Leakage current is a pure DC parameter, measured after an ~infinite settling time, but in practice waiting for an infinite time is rarely convenient.
In the dielectric, all the previous voltage history is recorded, even days after the cap has been kept shorted. The longer the time, the higher the voltage, the stronger the imprint.
I have seen mylar caps having been constantly operated at high voltages for years behaving almost like electrets afterwards
No. At switch-on the capacitor is empty so all the voltage drops across the resistor.
I would not call the charging of the capacitor leakage.
50 pA actually. The first few dozen seconds you just see charging current, then dielectric absorption (and reforming current when it's an electrolytic cap, but you shouldn't use those when leakage is critical anyway).
I did some work on low current electronics in the mid 1990's, with currents in the pA range and a 1.3 V supply voltage. I vaguely remember that I had an issue with a ceramic disc capacitor that I solved by using an antique ceramic tubular capacitor instead. Film capacitors were also good.
I built circuits with 10 pA bias currents on perfboard that worked fine, even though I hadn't even removed flux residues. I guess that's because they were dry. The slightest amount of moisture could have changed flux residues into electrolytic conductors.
The capacitors that are used in the high-impedance parts of condenser microphones also need to be low-leakage, as they have to work with 1 Gohm to 10 Gohm bias resistors. I think those are usually plain old MKT capacitors, but it's been a while since I last opened up a condenser mike.
Last edited:
I was going to say -how clean and separated are all the critical terminals, and surfaces of pcb, caps etc; even, how dry is the pcb..?
To clean all with Isopropyl alcohol, is surely a necessary check, too; and isn't this kind of problem is why 'guarding' on inputs was a thing ..?
To clean all with Isopropyl alcohol, is surely a necessary check, too; and isn't this kind of problem is why 'guarding' on inputs was a thing ..?
AKG used (and maybe still uses) a kind of red lacquer to improve the insulation of the very high impedance parts of their microphones. No idea what exactly they used, I only know it was red.
Of course (is seen hitting himself on the head)!
Just did a test: took the one cap that has been on the circuit (which was therefor charged up to 20V), reversed it and what do you know: 40uA, rapidly dropping.
QED Gerhard!
Jan
Last edited: