[help] looking for a highly accurate phase detector for low frequency

[soleilsword]

I'm looking for a highly accurate phase detector for my capacitive sensing application. In the application, the capacitance is in the range of 1 ~ 10 pF. I've been able to get the capacitance via analog device's AD7745. However, I was requested to have the dissipation factor measured as well. I've been trying to seek a chip to do the job, unfortunately not yet. I've came up with the idea of using a phase detector, with the information of which, I'll be able to calculate an approximate value of dissipation factor myself later on. So my question is, is there such a phase detector that can detect the phase difference between the voltage and current across the capacitive sensor with a very high accuracy? with accuracy I mean: the capacitor is a nearly ideal capacitor with dissipation factor of less than 0.001, so the most interesting part of phase difference is between 89 ~ 90 degrees...

The interested frequency range is DC ~ 1Mhz.

thanks !

 

[chuckey]

I think that the angular error is .057 degrees - not a lot! I was thinking of an exclusive or gate. If you feed this with two square waves, its output is the difference between them, i.e if they are co-incident there is no output, if they are 180 degrees out of phase there is a constant 5V DC. So with your figure there would be .057/180 X 5 = ~0 (.3 mV), not enough to get excited about. It could easily be swamped by what are normally considered as secondary effects, such as noise and switching transients.
Another technique could be a I and Q demodulator. This demodulates a phase shifted carrier into inphase and quadrature components. If you use your applied voltage as the demodulation reference carrier, then the I and Q channels will give you the inphase and quad components. If you phase invert the reference oscillator at a low frequency rate, then the I and Q channels should be the same amplitude but opposite in sign, so you can use any amplitude differences between the two phase to change the reference frequency's phase to null the out of phase component, the minimum value being the true one. Then the problem is that of accurately measuring the amplitudes.

Frank

 

[soleilsword]

I think that the angular error is .057 degrees - not a lot! I was thinking of an exclusive or gate. If you feed this with two square waves, its output is the difference between them, i.e if they are co-incident there is no output, if they are 180 degrees out of phase there is a constant 5V DC. So with your figure there would be .057/180 X 5 = ~0 (.3 mV), not enough to get excited about. It could easily be swamped by what are normally considered as secondary effects, such as noise and switching transients.
Another technique could be a I and Q demodulator. This demodulates a phase shifted carrier into inphase and quadrature components. If you use your applied voltage as the demodulation reference carrier, then the I and Q channels will give you the inphase and quad components. If you phase invert the reference oscillator at a low frequency rate, then the I and Q channels should be the same amplitude but opposite in sign, so you can use any amplitude differences between the two phase to change the reference frequency's phase to null the out of phase component, the minimum value being the true one. Then the problem is that of accurately measuring the amplitudes.

Frank

Dear Frank,
thx for the reply.
Indeed, in this application an accurate determination of dissipation factor (phase shift between capacitive / resisitive component or current/voltage component) is very crucial, which normally goes down to 0.001, or even 0.0001. I've been using a precision LCR meter for the measurement, and I've been quite satisfied with the measurement, but of course I wanna get away with the benchmarking tool and have my own system.
From what I understood of your post, you're more in favor of the 2nd solution. I don't understand completely the concept though.... Could you elaborate it a bit more, or maybe elaborate it with a few numbers so I have a more concrete idea?

thanks a lot for your help!!!

 

[FvM]

Besides using a dedicated phase sensitive demodulator, the complex signal vector can be extracted in digital signal processing.

But whatever method you use, e.g. pure analog or digital signal processing, the peformance of the analog frontend (e.g. voltage source, I/V converter) and a precise system calibration are essential points. Finding a reliable calibration standard isn't the easiest one. Even an air capacitor will show some dissipation by humidity influence.

 

[soleilsword]

Besides using a dedicated phase sensitive demodulator, the complex signal vector can be extracted in digital signal processing.

But whatever method you use, e.g. pure analog or digital signal processing, the peformance of the analog frontend (e.g. voltage source, I/V converter) and a precise system calibration are essential points. Finding a reliable calibration standard isn't the easiest one. Even an air capacitor will show some dissipation by humidity influence.

Hi FvM, thx for your answer.
I believe the dissipation is caused by the non-perfect nature of the capacitor :-> the existance of resistive component in the capacitor. and that's what I'm trying to find out here now -> dissipation factor (loss tangent) of the capacitor.
I'd like to have a chip that can do the job for me.

But just come back to your answer.the complex signal vector you were talking about, can you give me an example of product that can do the job?

 

[chuckey]

Link to I Q system :- https://www.edaboard.com/threads/85577/
I = inphase ~ resistive loss
Q = quadrature phase ~ capacitive reactance
Frank

 

[FvM]

A straightforward way to process the digitized measurements would be a FPGA or fast signal processor. For signal frequencies up to 1 MHz, we would e.g. use 4 MHz ADC sampling rate.