leakage of caps...

[Martin Czech]

:::tgd for metalized polycarbonate film caps is (lower than) 30 *10^-4 at 1
:::KHz.
:::
:::Solving the equation for R, and using these values gives:
:::
:::R = 1 / 30 * 10^-4 * 2pi * 10^3 * 22 * 10^-9 = 2,4 MOhm
:::
:::This is the what my teacher told me. Is this be right? Seems quite a low
:::resistance...

I guess this is most probably nonsense. The problem is that
several effects are mixed up:

1.: there is dielectricum leakage. From my understanding this should
    lead to long time discharge. It will depend on temperature and
    of course on cap quality, as well as on history, ie. overvolatge
    damage etc.
    
2.: In order to raise capacitance in a given volume, special dielectrica
    with higher epsilon are used (instead of dry air). This is polycarbonate
    or polystyrol or paper and so on. These are compounds with large molecules.
    They are very good isolators, but the large molekules may interact with
    the coulomb forces the electric filed imposes on them. In other words:
    the molekules twist and turn such that a minimum of field energy is
    obtained. This means: the capacitance will change, and so will the voltage.
    This looks like leakage, but it isn't. It takes some time for the molecules
    to turn, so I'd apply the same voltage for a long time (say an hour)
    to the cap, and then look at the discharge curve. In contrast to this
    you can discharge a cap (also wait very long time in order to allow
    turn back). Now charge it very shortly. The discharge curve should
    be different. Leakage should totally discharge the cap (well, dirty leakge
    has some rectifying effect, so some charge may stay), whereas dielectric
    absorbtion (this is the proper name for molekule twisting) should
    only lower the voltage a bit. A good description is given in Analog trouble
    shooting by Mr. Bob Pease. Recommended reading!
    You see that both effects could get into the tan(delta) measurement
    which is only an overall figure to see what the cap is like.
    
We are talking about nA here. And grease, dirt, input leakage will make a joke
out of your measurments. Use air insulation. Use very high impedance voltaghe
followers with low drift.
    
So I disagree with your teacher.

Saying this it is clear to me that a usual s&h will suffer from all kinds of 
leakage that are much larger then the good cap leakage effects.

m.c.