AD5933 impedance drop when frequency increase

Hi All,

I have a AD5933 impedance measurement question, when I measured frequency 1k-100kHz for only resistor, and calculated the impedance by the formula which datasheet provided, I got decreased impedance with increase frequency.

My circuit only have AD5933 chip with 1uF and 10uF capacitor connect between VDD and GND.

I used 51kohm resistor for feedback resistor and calibrated by 200kohm resistor, which I followed this article(https://www.analog.com/media/en/technical-documentation/application-notes/AN-1252.pdf) to calculate my feedback resistor and calibration resistor. My VDD = 3.3V, Gain = x1, Frequency increment = 1kHz, clock = internal clock(16M). The resistor range I want to test is 100kohm to 510kohm.

First picture tested on the calibration resistor, and this seems okay.

Second picture tested on 100kohm resistor.

Third tested on 330kohm resistor.

Last tested on 510kohm resistor.

And I had find someone said the testing range of different calibration resistor and feedback resistor, like if I want to test 100k-1Mohm resistor, I need to use 100kohm resistor for calibrating and 100kohm resistor for feedback, but the result looked similar like above.

Thank if anyone help me!!!!

  • One suggestion would be to calibrate using 100k resistor instead of 200k - one should always use the lowest calibration resistor value that yet does not cause the saturation of the input OPAMP(s) to fully utilize the ADC dynamic range. The formula for Rcal on page 5 of your reference does not pass a simple physical test: if nothing is connected to the chip, Zmax is theoretically infinity and this formula does not yield any usable value. Always calibrate using Zmin and, if you do not have the AFE, the Rfb calculated by the previous formula on the same page.

    What type of resistors do you use in your measurements? The resistors should better be some film resistors with minimal parasitic reactance. There is always some parasitic capacitance between VIN and VOUT pins of the chip and parasitic capacitance of the circuits connected to those pins, which is in parallel to whatever you are measuring. At 100KHz a mere 2pF of parasitic capacitance would have an equivalent impedance of  ~800k, which is in parallel to your resistors, but not taken into account in your calculations. With a very crude assumption that this parasitic capacitance did not affect your calibration with 200k resistor (yet another reason to use the lowest possible Rcal resistor), you can even roughly estimate this parasitic capacitance from, say, the 510k sweep: at 100KHz it appeared as 300k due to a parasitic impedance of 730k in parallel to the resistor, which corresponds to ~2.18pF capacitance.

    Not that important, but the internal clock is 16.776MHz: worth taking it into account if excitation frequency has to be accurate.

  • Thanks for your reply!!

    One suggestion would be to calibrate using 100k resistor instead of 200k - one should always use the lowest calibration resistor value that yet does not cause the saturation of the input OPAMP(s) to fully utilize the ADC dynamic range. The formula for Rcal on page 5 of your reference does not pass a simple physical test: if nothing is connected to the chip, Zmax is theoretically infinity and this formula does not yield any usable value. Always calibrate using Zmin and, if you do not have the AFE, the Rfb calculated by the previous formula on the same page.

    I'll try using 100k resistor for calibration, and still use 51k resistor for feedback.

    What type of resistors do you use in your measurements?

    I use carbon film(all of the resistors I use besides 200k) and metal film resistor(only 200k) in my measurement.

    There is always some parasitic capacitance between VIN and VOUT pins of the chip and parasitic capacitance of the circuits connected to those pins, which is in parallel to whatever you are measuring. At 100KHz a mere 2pF of parasitic capacitance would have an equivalent impedance of  ~800k, which is in parallel to your resistors, but not taken into account in your calculations. With a very crude assumption that this parasitic capacitance did not affect your calibration with 200k resistor (yet another reason to use the lowest possible Rcal resistor), you can even roughly estimate this parasitic capacitance from, say, the 510k sweep: at 100KHz it appeared as 300k due to a parasitic impedance of 730k in parallel to the resistor, which corresponds to ~2.18pF capacitance.

    I want to ask how to know the parallel reactance on my board? If I use different board I need to calculate from the result I receive from chip, right? And how to calculate correct phase shift on my board? The phase seems affect by the parasitic capacitance. And if there is any solution to avoid the parasitic capacitance?

    Not that important, but the internal clock is 16.776MHz: worth taking it into account if excitation frequency has to be accurate.

    Okay, Thanks for your remind!!

    Sorry for I'm not the student from EE, so I have lots of questions. And thanks for your help very much!!

  • I want to ask how to know the parallel reactance on my board?

    It is tricky to perform AC resistance measurements in hundreds of kOhm range exactly due to parasitic capacitance that is not known, but surely affects the measurements. The phase is certainly affected by parasitic capacitance as what you are measuring is a parallel RC network instead of pure resistor. It is impossible to get rid of the parasitic capacitance completely, only reduce the influence of it by using resistors with very short leads, or better yet chip resistors and making sure that no sizeable wires are connected to Vout and Vin pins of the chip. You are using carbon film ones, those should be reasonably good in terms of parasitics. You may try taking the parasitic capacitance into account when calibrating your system, but you would need to get through some cumbersome math.
    One step would be to try estimating the capacitance between Vout and Vin pins of the chip without any resistor attached in between. First you can try calibrating your system with lower value resistors, say Rfb = 5.1k and Rcal = 10k where effect of parasitic capacitance is negligible and then connecting some very large resistor, like 510k or 5.1M, in place of Rfb, taking a sweep without anything connected between Vout and Vin pins and calculating the capacitance from that data.
    Hard to recommend anything else without knowing what the project goal is.

  • Thanks for your help!!

    It is tricky to perform AC resistance measurements in hundreds of kOhm range exactly due to parasitic capacitance that is not known, but surely affects the measurements. The phase is certainly affected by parasitic capacitance as what you are measuring is a parallel RC network instead of pure resistor. It is impossible to get rid of the parasitic capacitance completely, only reduce the influence of it by using resistors with very short leads, or better yet chip resistors and making sure that no sizeable wires are connected to Vout and Vin pins of the chip. You are using carbon film ones, those should be reasonably good in terms of parasitics. You may try taking the parasitic capacitance into account when calibrating your system, but you would need to get through some cumbersome math.

    I'll do this.

    One step would be to try estimating the capacitance between Vout and Vin pins of the chip without any resistor attached in between. First you can try calibrating your system with lower value resistors, say Rfb = 5.1k and Rcal = 10k where effect of parasitic capacitance is negligible and then connecting some very large resistor, like 510k or 5.1M, in place of Rfb, taking a sweep without anything connected between Vout and Vin pins and calculating the capacitance from that data.

    Ok I'll try this. But why not fix Rfb resistor and change the resistor between Vin and Vout to calculate the RC parallel? So, leave the Vin and Vout connect without anything, and read the data from chip will only get the parasitic capacitor?

    Hard to recommend anything else without knowing what the project goal is.

    I'm going to use this chip to calculate randles circuit for electrochemistry impedance measurement.

  • But why not fix Rfb resistor and change the resistor between Vin and Vout to calculate the RC parallel?

    You can do it with the assumption that different resistors between Vin and Vout have the same parasitic capacitance, then you can both calibrate your system and estimate the parallel capacitance.

    So, leave the Vin and Vout connect without anything, and read the data from chip will only get the parasitic capacitor?

    This will get you the part of parasitic capacitance between Vin and Vout pins and the traces connected to those two pins, but would not give you the parasitic capacitance of the resistors when you connect those to your circuit. This additional capacitance is in parallel to whatever exists between the Vin and Vout when nothing is connected.  

    I'm going to use this chip to calculate randles circuit for electrochemistry impedance measurement

    Those measurements are normally done in some kind of electrochemical sell. Why do you think your cell will have equivalent resistance in hundreds of kOhms? Also those measurements are hardly ever done without AFE, which controls DC bias applied across the cell. If you just connect your Vout and Vin to the cell electrodes, the voltage difference between DC bias from table 1 in your reference and  VDD/2 will be applied across the cell. Excitation voltage of 2V p-p is likely to trigger some electrochemical reactions at the electrode interface so you may get a nonlinear response from the cell.

    It looks like your measurement system and software are working, so maybe it would be worth reviewing the construction of your electrochemical cell and figuring out what is needed to be between the cell and the AD5933 chip to get the results. You might need to find someone with EE background to collaborate on this project as you will likely need some kind of AFE.