Three electrode measurement for AD5933?

Hi everybody,


we're using the AD5933 for bioelectrical electrode impedance measurements. Usually, the electrodes have to be immersed in a saline solution for measuring, and a 2 point measurement (Electrode to Reference electrode) adds unwanted potential offsets to the measurement. For a more accurate measurement, a technique using 3 electrodes is employed:


(Fig. 4 from this link)

Ralph Group : projects : Fabrication and characterization of graphene electrodes for electrochemistry experiments

Figure 4

Is there any way to connect the AD5933 in such an arrangement to measure impedance? Do you have any additional suggestions for performing such a measurement?


Thanks and Kindest Regards,

Oscar Cota

  • 0
    •  Analog Employees 
    on Mar 20, 2017 3:27 PM over 3 years ago


    Im sorry but I don't know exactly what you did, I need more details.

    What is the RFB value used?

    Could you provide a schematic about the load connected? Or sensor in the measurements..



  • I'm currently using this AFE:

    The first stage takes the AD5933 200mV peak-peak signal to 0.18Vdc, 28mV p-p, the second stage is the buffer like you have above with the three electrodes (I'm not using a transistor for current), the third stage is for some gain, and the fourth stage moves the DC bias back to Vdd/2 for the AD5933 internal op amp. The gain is 1 for the internal op amp. 

    I've tested the circuit on the scope and verified the frequencies, DC and AC voltages, and ensured that the signal wasn't clipping. I'm sweeping from 100 kHz - 1 Hz, the PGA gain is 1, my calibration resistor is 747 ohms (interested range is roughly 200~2000 ohms), and I'm interested in the Nyquist plot (real and imaginary values). I've measured different values of "unknown" resistors just fine. I'm expecting graphs like the one below, where as more of the biomarker is added, the semi-circle should be getting larger. 

    Example from online:

    Here's an example of data I collected. The black lines are the attempt at circle curve fitting for the value of interest.

    Removed the last two data points (1.5 Hz and 1 Hz) so it's easier to see

    The circle shape starts out alright but then gets kinda wild. 

    This is the professional EIS device's results. They aren't the best as a whole (the circles should be getting larger as more biomarker is added), but the individual curves are nicely shaped.

    Do you have any further suggestions for making the AD5933 work like a potentiostat or work in a three electrode configuration? 

  • The RFB is 20k ohms. I'm using the AFE below and in practice the Reference and Counter are not shorted together, but connected to electrodes in the biomarker solution like the examples Oscar gave above. 


    The red boxes which I tried to add in Paint are multiplexers and the red thing between them in the calibration resistor.

    For measuring the impedance, I start at 100 kHz and measure the calibration resistor and then the biomarker solution, then I manually step the frequency down to 63 kHz and measure the calibration resistor and then the biomarker solution, and I keep decreasing the frequency until 1 Hz. Then at the end of the sweep, the code calculates the gain factor and the actual real and imaginary values for the solution.

  • AD5933 is not quite suitable for measurements under 10 KHz - there are artifacts from the way the part performs "DFT". 100 Hz is achievable by working through some cumbersome math, but single Hz range will require reducing frequency of the external clock. 

  • I'm using a 16 MHz - 4.1 kHz programmable clock for the AD5933 MCLK and I use the 4.1 kHz for the 1 Hz and I have to slow the I2C clock.