7

I need to measure or sense capacitance from 0 pF to 5pF, with an accuracy of 0.1pF or better. I know capacitance to digital convert chips claim to do that, but is there a simpler/easier way to breadboard a circuit to measure these ultra-low capacitances?

I have a small concentric cylinder where the outer cylinder is physically fixed (and tied to some potential) but the inner one is moving in and out. It's this movement I need to track. The change in overlapping area between the two cylinders produces a change in capacitance. In that way, I keep track of its position at all times by monitoring the change in capacitance.

W5VO
  • 18,738
  • 7
  • 64
  • 96
David M
  • 101
  • 1
  • 7
  • 3
    I think the issue is the parasitic capacitance introduced by your measurement system. Can you explain more about your scenario? Are you trying to test individual SMT capacitors without installing them? Or are you trying to measure capacitance of SMT pads on a PCB? Or what? – user57037 Jan 10 '18 at 06:28
  • Are you trying to measure capacitance dynamically (as per a capacitance probe looking at a rotating metal gear wheel for instance) or some static measurement. Either way you need to provide more detail so an appropriate answer can be targeted. – Andy aka Jan 10 '18 at 09:01
  • 1
    I have a small concentric cylinder where the outer cylinder is physically fixed (and tied to some potential) but the inner one is moving in and out. It's this movement I need to track. The change in overlapping area between the two cylinders produces a change in capacitance. In that way, I keep track of its position at all times by monitoring the change in capacitance. – David M Jan 11 '18 at 02:33
  • Would you be so kind as to add that information to the question by clicking on "edit?" I should have mentioned that in my first comment but I forgot. One more question, also. Would it be acceptable to have a calibration step to establish the min and max displacement, and store the calibration result, somehow? – user57037 Jan 11 '18 at 02:36
  • Yes, calibration is necessary, especially to eliminate initial lead and stray capacitance, which could be seen as some sort of initial offset that would have to be removed from the real reading. – David M Jan 11 '18 at 02:43
  • Possible duplicate of Measure small capacitance value – Dmitry Grigoryev Sep 17 '18 at 07:02

1 Answers1

9

It's quite straightforward to detect a change of capacitance of 0.1pF, as a ratio. The simplest is perhaps to build a relaxation oscillator and measure the frequency and change of frequency digitally, as the test capacitor is connected.

It's very difficult to know exactly how much effective capacitance there is in the rest of the circuit, and any connection jigs, strays, terminals, leads that the ratio is measured with respect to.

The advantage of a relaxation oscillator is that one capacitor terminal is grounded, so the strays are relatively stable. The disadvantage is the strays can be large, quite easily large compared to 5pF.

The alternative is a 3 terminal guarded measurement, which is immune to stray capacitance on either terminal of the capacitor, and only susceptible to strays across it. The third terminal is ground. The method is as follows.

1) Apply a sinusoidal voltage with respect to ground to one terminal of the test capacitor from a known voltage. The strays from this terminal to ground are driven to exactly the same voltage, we are not interested in how much current is required to charge them, the voltage measurement is sufficient.

2) Hold the second terminal at ground, and measure the current required to do that. The most common way to do that is to use a virtual ground op-amp. The strays from the second terminal to ground are held at 0v, so no current flows into them, so the current measurement is accurate.

3) We now know the current through the capacitor for a given voltage across it. Compute the capacitance from impedance and frequency. A capacitive feedback rather than resistive on the virtual ground op-amp allows you to eliminate the frequency from the equation.

Even though the guarded measurement removes the effect of the strays to ground, any strays across the capacitor that are enhanced by your test jig, perhaps a plastic pressure pad that holds an SMD component down onto a footprint, will change the measurement compared to what it would be in circuit without that pad.

Neil_UK
  • 166,079
  • 3
  • 185
  • 408
  • Thanks Neil ... but I've already looked at this type of oscillator and based of the simulations I've run and calculations I've done, the output frequency can be rather high for a 1 - 5pf capacitance, in the range of tens of MHz. (1 / (2.2RC). Unfortunately, I have a lowly MCU and it can't read such high speeds. Plus this is a very low power and space constrained application. The device is a medicine pen (size of a standard ballpoint pen) and I need to track the amount of expelled fluid.There's a whole bunch of mechanical gadgetry in there, so space is really constrained. – David M Jan 11 '18 at 02:48
  • @DavidM use a larger R? – user253751 Jan 11 '18 at 04:38
  • 1
    as immibis suggests, use a larger resistor. With an HC14 oscillator, 10Meg is perfectly usable. An HC393 prescaler will give you up to /256 for very low power, cheap and small, to bring your frequency into Arduino territory. – Neil_UK Jan 11 '18 at 06:20
  • Thanks, immibis and Neil_UK ... but I thought using large R would make the design noisy and/or unstable. Is this a valid concern? – David M Jan 12 '18 at 06:13
  • Using a tiny C and low currents, which means a low frequency, makes the design noisy, there's not a lot you can do after that. The only solution is to increase the currents, increase the frequency, and sense with more power. – Neil_UK Jan 12 '18 at 07:18