Post Go back to editing

problem calibrating eval-AD5933

Category: Software
Software Version: 1.0

Hello Everyone,

I recently bought a eval-AD5933 board to do some cure monitoring of a resin system. For this, i want to calibrate my system with a 1MΩ resistor. However, i keep running into issues.

When i connect a 1MΩ resistor to both the Rfb and Z pins on the board and calculate the gain factor, it gives a value of 1,01191066098752E-10. When i then start a frequency sweep, the resulting impedance is very low, around 0,0001 Ω. I have tried many combinations of connecting other resistors, leaving out the Rfb resistor, leaving out the Z resistor, connecting to the T2&T3 pins but nothing seems to be working. Has anyone else had this problem or does someone have a solution?

Thanks in advance,

Roy

Parents
  • Just to make sure that this is what you described, your 1M resistor is supposed to be connected between the two "Z" female socket pins (very close to each other), marked as such on the board. At the same time the Rfb resistor should be placed between the two female socket pins (farther apart) where the "RFB" resistor symbol is drawn between them. When the board arrives there is usually a 200k resistor connected there for the Rfb. T2 and T3 are not connected to anything as C1, C2 and C3 are not populated.

    If you expect that the unknown impedance you are going to be measuring is larger than 1M, with your settings it would make sense to put 1M for the Rfb and, for starters, calibrate with 1M resistor as well. Also, you might consider using the "Download Impedance Data" button to store the raw data form the chip in CSV format - very useful for debugging and, when in doubt, posting it here. For example, a dataset with no Z is connected, Rcal connected and some Zunknown connected would pretty much tell whether the system functions properly.  

  • As recommended, i connected the 1M resistor between the Z female pins and a 1M resistor between the Rfb female pins. The resulting frequency sweep shows very low impedance values, shouldn't this show 1M as well then? XLSX

  • It most certainly should show 1M.
    From the raw Re, Im and Magnitude data everything else seems to be working as advertised. Assuming that the software implements the gain calibration procedure as described in the datasheet (p. 17, GAIN FACTOR CALCULATION) and calculating the gain ourselves from the Magnitude column as Gain = 1 / (1M *Magnitude) we get it to be around 1.01E-10 (hopefully this is something that you see in the "Calculated Gain Factor" field of the software UI). Now, following the datasheet, the impedance should be calculated as Impedance = 1 / (Gain * Magnitude), which, with the Gain we calculated and the Magnitude data, yields values close to 1M as expected, but this is not what the software reports in the Impedance column. Somehow it appears to be exactly 1M * Gain. Hopefully somebody from ADI could comment on this behavior.
    The good news is that with the raw Re and Im data available you can perform your own calculations not relying on the demo software.

Reply
  • It most certainly should show 1M.
    From the raw Re, Im and Magnitude data everything else seems to be working as advertised. Assuming that the software implements the gain calibration procedure as described in the datasheet (p. 17, GAIN FACTOR CALCULATION) and calculating the gain ourselves from the Magnitude column as Gain = 1 / (1M *Magnitude) we get it to be around 1.01E-10 (hopefully this is something that you see in the "Calculated Gain Factor" field of the software UI). Now, following the datasheet, the impedance should be calculated as Impedance = 1 / (Gain * Magnitude), which, with the Gain we calculated and the Magnitude data, yields values close to 1M as expected, but this is not what the software reports in the Impedance column. Somehow it appears to be exactly 1M * Gain. Hopefully somebody from ADI could comment on this behavior.
    The good news is that with the raw Re and Im data available you can perform your own calculations not relying on the demo software.

Children
  • Yes i see, You're right in saying that calculating the impedance manually shows the right values. I guess you should never assume software uses the right calculations! good life lesson. very weird that the software delivers the wrong impedences in the impedence column though, but thanks for the solution!

  • BTW, you seem to sweep a rather narrow frequency range. In case you want to stay on the same frequency throughout the sweep, you can program in the Frequency Increment equal 0. If you eventually need to sweep a wide frequency range you might consider calculating Gain at every single point of your sweep and storing it for the subsequent calculations of the unknown Zs - gives much better accuracy than the calibration procedures implemented in the demo software.
    If the impedance of the curing epoxy samples are expected to give you interesting behaviour, you might consider treating the Gain as a complex number, which is calibrated with a frequency-independent real impedance Rcal + * 0 and from the Re + * Im data from the output registers calculate the complex impedance of your sample.