Low leakage capacitor?

[zahc]

What kind of capacitors hold their charge the best? I'm looking at making an analog integrator that fills up a cap until a certain voltage is reached, but I don't want the charge to leak out of the cap as fast as I put it in.

[charby]

analog integrator

Shocking people for answers?

[Gewehr98]

Acrylic or glass aquarium + metal plates + mineral oil. 

Charge.

Zap.

Repeat.

 

[charby]

Acrylic or glass aquarium + metal plates + mineral oil. 

Charge.

Zap.

Repeat.

 

Tesla type capacitor?

[Gewehr98]

Yup.

You can also use 2-liter soda bottles. 

I'll have to take pics of my dad's Tesla Coil sitting in one part of his inventor's workshop/barn.  250-500K volts, he plays with it like it is some sort of kid's toy. 

"Hold this fluorescent tube right here!"

"Why, Dad?"

"Just do it, and watch!"

 

[Rudy Kohn]

You probably want to go with a polyester, polystyrene, polypropylene or mica capacitor--they are cheap and have low leakage.  Avoid any kind of electrolytic caps, as they are the worst for this.  I hope you can do with nanofarads, as it's tough to get the plastics/mica caps bigger than that.

If you want to get really fancy, try glass or teflon (or vacuum).

Source:  Horowitz and Hill, Art of Electronics.

Hope this helps. 

[zahc]

I have some .1uF polypropylene caps left over from an amp. I'm not sure of the values I need yet, but it needs to fit on a small board. I'm making an integrating auto-exposure circuit for my pinhole camera, because very long exposures are hard to estimate when light keeps changing.

[Headless Thompson Gunner]

How much capacitance do you need?  How long do you need it to hold a charge?

Polypros are the usual answer when you need high quality capacitance, but most all caps hold their charge pretty well and would probably work for you.  Just avoid the electrolytics.

[zahc]

As long as it can hold a charge to within 10% over a couple days it should be fine.

I don't know what value to use cause I haven't decided on the rest of the circuit, any suggestions?

[Rudy Kohn]

I think 10% over a couple of days is a little tough for most capacitors.  I'm not sure, but I think leakage will tend to discharge a cap (even a good one) on the timescale of hours or minutes, rather than days. 

You might consider a leakage compensation circuit--basically set up a current supply that supplies a current proportional to the capacitor voltage (use a good high-input-impedance FET op-amp or something), which, with proper tuning, will compensate for the current lost by leakage.  Figure out the series resistance of the capacitor you intend to use (you probably still want a non-electrolytic one) and build a current supply to throw this current at the capacitor.

You should be able to effectively cancel the leakage with this, even for days, and it will even work passably well with a big ceramic capacitor, even though ceramics are not as good as mica or plastic for leakage.

Good luck!

[Marvin Dao]

I think 10% over a couple of days is a little tough for most capacitors.  I'm not sure, but I think leakage will tend to discharge a cap (even a good one) on the timescale of hours or minutes, rather than days.

10% over a couple of days is actually no problem for polystyrene and polypropylene caps. There's a little experiment detailed here where someone tested the leak rate of some polypropylene caps.

[Rudy Kohn]

Cool.  I didn't know that.  In that case, either one of those should be fine, but I'd test the leak rate myself beforehand, just to avoid any surprises.

Hmmm, I just checked the datasheet for a brand of polystyrene caps on Newark, and got a time constant of 9 hours or so, based on the largest possible capacitance and 2x the minimum stated resistance.  (Newark part no. R=10^12 Ohm, C=33 nF, R*C=33000 seconds)
I dunno, I still think 10% over a couple of days is gonna be tough.

The high-capacitance polypropylenes were huge and expensive... and the datasheets didn't list a resistance as far as I could see.

Marvin, do you know what brand of caps were tried and who tried them?  Those were the best I could find on Newark that would fit on a small board.

[zahc]

Even if it works for up to 2 hours, it will cover 90% of my uses. The 10% figure is just thrown out there too. This is photography where increments of exposure are routinely measured in factors of 2. The leakage compensator circuit is clever, though I don't trust datasheet values very far for things like shunt resistances. Although you could test the leak compensator circuit in isolation and see how well it works.

The high-capacitance polypropylenes were huge and expensive

Yes, they are. I bought a bunch for building an amp because people on the internet said they 'sounded better'. Since then, I built a vacuum tube headphone amplifier that has a capacitively coupled output with some common low-ESR 400uf electrolytic capacitors, which are about the size of a .5uF polypropylene cap. You aren't supposed to use electrolytic capacitors for audio because they 'sound bad' but I'm absolutely ecstatic with the way the amp sounds. Teaches me to pay attention to audiophiles.

[FTA84]

Retracted

[zahc]

I'm not sure I understand. The momentary switch shorting out the capacitor is to reset the circuit.

I don't know how I could implement this with a flip-flop? I entertained the idea of using a microcontroller, but that would be too much work.

[Marvin Dao]

Cool.  I didn't know that.  In that case, either one of those should be fine, but I'd test the leak rate myself beforehand, just to avoid any surprises.

Hmmm, I just checked the datasheet for a brand of polystyrene caps on Newark, and got a time constant of 9 hours or so, based on the largest possible capacitance and 2x the minimum stated resistance.  (Newark part no. R=10^12 Ohm, C=33 nF, R*C=33000 seconds)
I dunno, I still think 10% over a couple of days is gonna be tough.

The high-capacitance polypropylenes were huge and expensive... and the datasheets didn't list a resistance as far as I could see.

Marvin, do you know what brand of caps were tried and who tried them?  Those were the best I could find on Newark that would fit on a small board.

Bob Pease did the testing. Charged cap hooked up to a high impedance unity gain follower to a DMM. The only brand mentioned was the Panasonic polypropylene 1 µF's that leaked at about 2-3 mV/day when charged to 9V. From what I can tell, he tested a few other brands and types that fared similarly but he didn't mention which ones. He did mention that paper and oil caps (audiophile favorites) were particularly poor performers though.

Looking at it again though, input bias current from the op-amp might be much larger than leakage current making the capacitor's leakage current irrelevant.

[FTA84]

I don't want the charge to leak out of the cap as fast as I put it in.

My apologies, I did not realize this was the problem which you were trying to solve.

[Rudy Kohn]

zahc, if that's the case, then you'll almost surely be fine with the polystyrene caps, but the ones I looked at on Newark were all "ship from UK, extra $20," so you might want to try somewhere else (that is, if you generally use Newark--maybe you know a better supplier?).  Good luck with the project.

Marvin, thanks for the information--that's really neat.  I would never have expected the leakage (with input bias current added on, to boot) to be that low.  Very interesting.