RE: CN0349 Drifting

Hi Brandon, I realize it has been almost a year on this issue but I am having this issue again and I think the first answer was a false positive. I don't know why I am having this issue again or why it seemed fixed the first time.

I am using the same commercial probe (Sensorex CS200TC-K=1.0) in a 49 mS/cm standard and am having significant drifting but also sometimes it goes through periods of stability. This is the case no matter what probe or probe material configuration I seem to have. Sometimes measuring on board resistors Rfb =100 ohms and Rcal =100 ohms  at 100kHz results in a magnitude of 3609.2531 and then completley random may be a magnitude of 9658 instead. I have seen it change con the drop of a hat when reuploading code to my MCU

I think this may be a code problem because if I am correct, the AD5934 does a frequency SWEEP. So it calculates the real and imaginary components for every frequency in the sweep. If I am measuring conductivity of a solution why would I want to do a sweep? I should want to do it at 1 frequency because there is capacitance in water like this:

And I want Rsol as the conductivity. where Cs is the capacitance between probes and Cdl a property of the solution.

Note: the drifting does not occur with a precision resistor! So my thought would be the capacitance in the water running things amuck.

So I should want to choose the frequency to be really high, like 100kHz and have no increments in frequency. However even after adjusting my code to this, I am encountering drifting. It could be drifting upwards or downwards. I would rather it drift up than down because up could indicate that the solution is evaporating and getting more conductive. Additionally, I have noticed that the wires to the probe from the CN0349 are critical. Some wires I have seen attenuate the signal like crazy. In one instance using alligator clips I witnessed a change in data values from .9 to 32 just from the wire. I can't seem to escape this drifting so I think the lowest common denominator must be the software and how I am getting the impedance. I have an oscilloscope I just don't how to troubleshoot this problem. If you are still willing maybe I could run my code by you?

  • 0
    •  Analog Employees 
    on Aug 29, 2019 3:36 PM over 1 year ago

    I split your question out into a new forum, because the other question was "answered".

    The AD5934 has the ability to do a frequency sweep if you want it to, but it doesn't have to, its all in how you program the device.  Please feel free to share your code and we can look at it.  Would you also have a software flow diagram that outlines how you are taking the measurement?  What processor/board are you using to take the data? 

    I would certainly expect the leads of the probe going to the board as being extremely critical.  Remember the leads are going to have inherent real world RLC characteristics, and those are going to change your measurements.  In any of these types of systems proper calibration is going to be required, before you are going to be able to obtain an "accurate" measurement.  

    The resistor test is a great test because that should give you a stable and repeatable result.  Once you have that figured out (which it sounds like you do) it would be interesting to put that resistance in your alligator clip leads and make the same measurement.  How much did that change?  Is that repeatable?  This will tell you the error associated with the alligator clip leads.  Then you can try you conductivity probe again and connect it directly to the board and through the alligator clips to see what this does to the measurement.  You know the value of the alligator clip error, and you should be able to subtract that out from your measurement data.(Make sure you are using calibrated solutions from Hach or other trusted vendor when using the probe)

    There is not simple answer here, before these systems are very complex when you start debugging them.



  • 0
    •  Analog Employees 
    on Sep 24, 2019 2:43 AM over 1 year ago

    Hi jgirgis: from your descriptions I think you could try to use your oscilloscope to check the TP1 signal on the CN0349 PCB board at first, this signal should be a sine wave have same frequency with excitation signal. Because the pin 2 and pin 4 of J1 connected to U2B inverting input directly, so the stray capacitance of probe cable may cause unstable oscillation, please try to keep the probe parasitic reactance as small as possible.

  • Thank you Brandon, I don't think I was notified about this thread merging which I why I am about a year late.  I deeply apologize and I thank you so much for your help. I think the issue has RLC problems correlated with it, yes. But additionally there is 3.3Vpp going into the water. I talked with some electrochemistry experts on this and 3.3Vpp is a high enough excitation value to be unstable in water. Especially in seawater, with many many ions an unsymmetric I versus V curve most likely exists. I think I will start to think about modifications I can make such as a 4-wire measurement instead.

  • Thank you R.L, I am sorry I am about a year late, I wasn't notified that the thread was continuing.  I deeply apologize. Yes the reactance is an issue. But there is also a double layer capacitance in series with the "water resistor" being measured, that has on the order of 10uF/cm^2 of probe area, so the transient time it takes for the signals to settle across the "water resistor" could be upwards of 50 seconds depending on the probe area. I don't believe the AD5933 is capable of allowing its excitation signal to persist that long.

  • AD5933 can produce excitation signal indefinitely long, as long as it stays powered up. The drift you are describing can be a number of things, quite often it is some unaccounted for DC bias that ends up applied across the measurement cell. DL capacitance is notorious for being sensitive to extraneous DC.