4
\$\begingroup\$

How should one measure and compensate (If necessary) a wire's stray capacitance in low-level measurements?

Is Guarding applicable here? If so, does it mean having a wire with same signal voltage twisted around signal wire?

In the example below, DUT is a capacitor whose leakage current must be measured. The black rectangle represents a metal shield. Metal shield is connected to LO terminal of source. (Sorry for childish drawing - I am not Da Vinci :D)

Stray Capacitance

\$\endgroup\$
  • 1
    \$\begingroup\$ As Olin points out it's probably not relevant to your task, but you can measure stray capacitance surprisingly well by the RC time constant method using a resistor of a few megaohms and a microcontroller. \$\endgroup\$ – Chris Stratton Jan 6 '12 at 13:21
4
\$\begingroup\$

Several points of confusion here.

  1. To measure capacitor leakage, stray capacitance doesn't matter. You're looking for the current the cap draws in DC steady state.

  2. Guarding (surrounding a trace with actively driven guard traces) is a means of dealing with non-perfect insulators and nulling out the leakage currents that would otherwise run thru them. You recognize that the material has some finite resistance. You arrange for a signal that is actively driven to the same voltage as the sensitive signal to be on the other side of that resistance. Since the voltage accross that resistance is now zero, there will be no current flowing thru it, therefore making it not matter.

  3. Guarding is not accomplished by using a twisted pair. That is for a whole other set of issues. Basically a twisted pair tries to make each signal couple equally to the environment, thereby making any noise picked up from that environment common mode. This is not relevant to your problem of measuring the leakage of a capacitor.

  4. Stray capacitance is very hard to predict. Sometimes you use a shield to trade of capacitance to some unknown source with a higher capacitance to something that doesn't carry noise, like ground.

\$\endgroup\$
  • \$\begingroup\$ Re #3 - Horowitz & Hill suggest guarding can be effective against capacitive effects in addition to leakage currents. A proximate conductor driven by a follower would seem to be an example - ie, a signal twisted with a guard probably would work. But the point about irrelevance to measurement of leakage current remains valid. \$\endgroup\$ – Chris Stratton Jan 6 '12 at 15:46
  • \$\begingroup\$ @Chris: Sounds like you're really making the point that a guard can also act like a shield sometimes, which is true. If you only want to eliminate capacitive coupling from some noise source, the "guard" trace might as well just be connected to ground, making it a shield and not really a guard anymore. Grounding one wire of a twisted pair will reduce capacitive coupling to elsewhere a bit, but if that's what you want you should use shielded cable. Twisted pair has a totally different purpose. \$\endgroup\$ – Olin Lathrop Jan 6 '12 at 15:56
  • \$\begingroup\$ the potential utility you are overlooking is to combat capacitance relative to something fixed, such ground. In the case of a high-impedance AC signal this would be important, and guarding with a co-varying driven signal would be effective. It might not be what we usually think of as the purpose of a twisted pair, but that could be a practical and available physical configuration for the driven guard. I agree that the whole issue of stray capacitance is irrelevant to this question, but your answer contains a mistaken generalization. \$\endgroup\$ – Chris Stratton Jan 6 '12 at 16:19
  • \$\begingroup\$ @Chris: I don't think it's mistaken at all. A twisted pair does not form a good shield. Half the time the signal wire will be exposed to the capacitive source instead of the guard wire. That's better than all the time, but not all that effective. If you're saying to deliberately inject a signal opposite the noise source, you don't need a twisted pair to do that. The opposite signal can be injected in a single place elsewhere, for example. Guarding, shielding, and common mode rejection are three different topics. \$\endgroup\$ – Olin Lathrop Jan 6 '12 at 16:38
2
\$\begingroup\$

One point is to configure the circuit so that the stray cap. is a constant. Long loose wires change their cap. when moved. Using twisted pairs or guarding my actually increase the cap. but will help make it a constant. Measure it by removing only the cap that you want to measure, subtract it out with the circuit intact.

\$\endgroup\$
  • \$\begingroup\$ You seemed to have missed that the OP is trying to measure leakage current of a cap. Extra capacitance in parallel with the capacitor under test is irrelevant. \$\endgroup\$ – Olin Lathrop Jan 6 '12 at 15:40
  • \$\begingroup\$ Good point but I answered the first part of the question, not the second, on the theory that with his confusion he may have meant what he said in the first part, and others seeing the title may want some info related to it. \$\endgroup\$ – russ_hensel Jan 6 '12 at 15:48

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.