AD5933 evaluation board measuring low impedance liquid

Undoubtedly the easiest way would be to feed the chip with a lower-frequency clock, there is a connector and some jumpers on the board to configure it for that. For measuring RC networks the accuracy of the external clock does not need to be of quartz oscillator quality to provide reasonably accurate impedance measurements. Inaccuracy in the frequency is going to show a bit in the resulting data, but is not going to scramble the results beyond recognition as in case of measuring high-Q resonant circuits.

If the external clock is not an option, tfsoft provided a reference to a paper describing a method of extracting reasonably accurate data out of the chip a frequencies down to 100Hz while running on the onboard clock, but it involves quite a lot of tedious math.

Probably the most difficult would be to redesign your measurement cell in a way that the values in the equivalent RRC network would produce a characteristic response at higher frequencies, preferably above 10KHz.

Thank you for your answer Snorlax. I have tried using a function generator, at 100 kHz as a clock for the AD5933 evaluation board and use 100 ohm resistors as Rf and as calibrated resistor.

I read in the datasheet that the lowest clock I can get from this setup is 30 Hz, but what is the maximum frequency that I can achieve with 100 kHz clock?

My result is that if I run the calibration several times, the results will be different every time. Is there any way I can improve the accuracy of my experiment?

First things first: does your board has the AFE? Is this the board you are using? Not sure where the 30Hz came from, could you please provide a reference?

The chip can synthesize the sine waves at frequencies between MCLK/2^29 and MCLK/16, but cannot really measure accurately below 5-10KHz at MCLK=16MHz due to some internal limitations you can read about in the paper found by tfsoft. By lowering MCLK to 100KHz, you can go proportionately (160 times) lower in frequency and get reasonably accurate measurements at, respectively,  ~30-60Hz and higher. If your board has AFE, it might need some modification to pass frequency this low across the AFE.

I used the board that you refer to (UG-364), and there is a reference in the documentation that said so, in the measuring lower frequencies part.

There is no additional AFE added to the UG-364 for the time being (I haven't really looked into the board manuals).

The measurements seem okay, but the accuracy is a bit off. Any suggestions?

Thanks for the reference, not quite clear how exactly those results were experimentally obtained.

If I got it right, you did connect your board to external generator at 100KHz ( by presumably connecting the cable to CLK1, opening the jumper LK3 and closing the jumper LK4)?

There are two things that likely affect the accuracy:

1. there is a high-pass filter in the on-board AFE produced by R7, R8 and C7 with a cutoff frequency of about 135 Hz, which is suppressing the excitation voyage coming to you resistor under measurement. Perhaps increasing C7 to 470nF should help

2. the demo software is not aware of the external clock coming to the board from generator and still thinking that MCLK is 16MHz. It is programming into chip the frequency codes calculated based on 16MHz. To get your 30 Hz on 100KHz clock you need to trick the software to produce the correct frequency codes for the chip by supplying 30Hz*16MHz/0.1MHz = 4800Hz for the "Start Frequency" and 10Hz*16MHz/0.1MHz = 1600Hz for the "Delta Frequency." 

Hi Snorlax,

Snorlax said:

If I got it right, you did connect your board to external generator at 100KHz ( by presumably connecting the cable to CLK1, opening the jumper LK3 and closing the jumper LK4)?

That is correct.

As you said, adding a 470 nF in parallel to the C7 helps and the measurement is now quite reliable. As for the software, you just need to set the external clock to 100 kHz and the frequency as needed (in my case from 30 Hz - 200 Hz), and the reading is acceptable.

Another question that I have in mind is how low can the measurement go? Can the system measure down to 10 ohms of impedance? Thanks.

Good point re the demo software, it does have the functionality to read the value for MCLK. I used a different one that I had to trick and rushed to assume that you may have bumped into the same issue.

Good to hear your measurements seem to be on the right track. 10 Ohms might be a bit tricky: at your excitation voltage of 2V p-p the current across your impedance under test should be 2V / 10Ohm = 0.2A p-p, while the maximum output current your A1 OPAMP  AD8606 can produce and A2 OPAMP can consume is only 80 mA. If the OPAMS hypothetically could handle the current, they would be quite warm and so the Rfb as (2V)^2 / 10 Ohm is 0.4 Wt p-p. Brute force solution would be to use high-power OPAMPs and beef up the power supply.

If your device under test is linear, you can run your measurements at 0.2V excitation voltage setting and set the PGA gain at 5x while your Rfb should be equal to the lowest impedance in the range of your measurements. Or keep the PGA gain at 1, but increase the Rfb to 5 times the lowest impedance you expect to see in the course of your measurements. 

Good point re the demo software, it does have the functionality to read the value for MCLK. I used a different one that I had to trick and rushed to assume that you may have bumped into the same issue.

Good to hear your measurements seem to be on the right track. 10 Ohms might be a bit tricky: at your excitation voltage of 2V p-p the current across your impedance under test should be 2V / 10Ohm = 0.2A p-p, while the maximum output current your A1 OPAMP  AD8606 can produce and A2 OPAMP can consume is only 80 mA. If the OPAMS hypothetically could handle the current, they would be quite warm and so the Rfb as (2V)^2 / 10 Ohm is 0.4 Wt p-p. Brute force solution would be to use high-power OPAMPs and beef up the power supply.

If your device under test is linear, you can run your measurements at 0.2V excitation voltage setting and set the PGA gain at 5x while your Rfb should be equal to the lowest impedance in the range of your measurements. Or keep the PGA gain at 1, but increase the Rfb to 5 times the lowest impedance you expect to see in the course of your measurements.