My suspicion is that the capacitance between the connecting wires is changing as you change the geometry. This is normal behaviour, and you need to fix the cable geometry in order to get consistent results. To confirm this you could measure the connecting wires without the sensor (just fasten them to an insulator with the same spacing as the sensor and see whether the change in capacitance is somparable to that with the sensor (allowing for the parallel connection of course).

In my opinion you do not use the right test fixture. I would suggest Kelvin probes, like this:

http://www.datatec.de/HZ184-Kelvin-Messadapter.htm#detail_text

(sorry for German link)

If you want to use your test fixture you have to connect your capacitor directly to the fixture connections without any wires inbetween (or as the forspeaker said wires have to be geometrically fixed and calibrated).

I am really appreciate your advises .....

but I am wondering something 

To :  Mr. Robert Alan Cottis and Mr.Иван Трајић 

I have tried the way you suggested me, my measurement is quite stable and the wire does not have significant effect to my device(Cylindrical actuator), let say i get 40+30j ohm,  Can I conclude that the actual impedance is 40+30j ohm ? 

how can I make sure that my current setting give me the actual impedance ? 

To: Anna Schröter 

I can open in English form. Have you ever used that kind of test fixture ?

Does it give you a stable value ? 

I will check with the Keysight Tech. for that kind of test fixtures 

Thanks for your time to answer my question..... 

Best Regards,

Yannapol S.

Hi, Yannapol

The actual impedance of your sample under test (SUT) depends on frequency. It seems that you SUT has two resonances near 100 kHz.  It is probably caused by two parallel LCR circuits with some coupling factor.

As previous speakers already noted, the overall impedance depends on connecting wires. You may use shortest wires with fixed geometry to minimize this influence. However, the impedance analyzer is calibrated for its test fixture, and large objects may still significantly influence the measurement results. This is caused mainly by stray capacitances between ground and SUT. In your case (relatively small dimensions and low frequency) these errors are probably small.

 Another possibility is to measure the parameters of your fixed connection wires (inductances, capacitances, resistances) and add corresponding elements to the electrical model of SUT. However, as already noted, long connections will not match well with the calibration parameters of the standard test fixture of impedance analyser. Use of longer probe (calibrated for that analyzer) proposed by Ms Shröter solves the problem. However, these probes can be used only in the limited frequency range.  100 kHz?  It is possible to design test fixtures for large SUT and wider frequency range, but this is more complex task.

BR