This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

LMP91200: Oscillation on VOCM output

Part Number: LMP91200
Other Parts Discussed in Thread: ADS124S08

Hi

 I have designed pH measuring device. It is 4-20mA loop powered, isolated using flyback converter. It has STM32, ADS124S08 as ADC and LMP91200 as front end. All running on 3.3V. Vref is connected to 3.3V.

I'm trying to have accuracy better than 1mV but experiencing some oscillation:

1. INP and VCM shorted on PCB and no cable connected, gives me pretty acceptable 200uVpp noise (measured by reading ADC)

2. when I connect RG58 cable ~1m shorted at the far end and INP and VCM shorted on PCB I start to have 3mVpp kind of sinusoidal noise on VOCM line. The VOUT is also affected but it's 500uVpp. It looks like below (x axis is number of sample and it's 20sps - ADC set to: single sampling, 400sps, sinc3, chop on.

3. the filter between LMP and ADC is: 1kR on VOCM and VOUT, 10nF on each ADC input to gnd, also 100nF between ADC inputs but currently not fitted. 

How to get rid of that noise/oscillation?

Regards

Krzysztof

  • Krzysztof,

    The sensor output of the LMP91200 is very high impedance (2.5Meg Ohms). If you look closely in the data sheet at layout section in the data sheet you will see that it is necessary to add a guard ring around the traces on the PCB going to the sensor. That being said you are using a coax and have the sensor 1 meter away. Yes you will see a lot of noise. The fact that you can take a measurement under these conditions is very good.

    Any situation in which the source to be measured has a very high output impedance and is vulnerable to leakage current will require a driven shield. The use of a triaxial cable may be used as the outer shield of the triaxial cable can be held at the same potential as the sensor wire. Remember that any noise on the guard may be amplified in the output.

    What you will need to do is use a driven shield as a guard ring in your triaxial cable to protect low-current circuits against leakage current. A driven shield is often referred to as a driven guard ring. It is used in situations where the tiny leakage of current through the insulating surfaces of a wire or PCB board would otherwise cause error in the measurements or functionality of the device. The basic principle is to protect the sensing wire by surrounding it with a guard conductor that is held at the same voltage as the wire so that no current will flow into or from the wire. This is achieved using guard1 and guard2 of the LMP91200 which matches the guard voltage to the sensing wire voltage. The leakage from the shield to other circuit elements is of little concern as it is being sourced from a buffer which has a low output impedance.

    Cleanliness is very important as well. Make sure to remove all flux from the PCB and wire solder connections at both ends. Any flux or moisture will become a leakage path for current and will cause you errors and noise. 

  • Hi Gordon, 

    Thanks for reply. Here are my comments:

    1. datasheet says Zoutvcm is 4Ohms. Unless you mean something else?

    2. I have the guard ring on PCB (on all layers)

    3. I use standard coax rg58 at the moment. Triaxial cable on order.

    4. I clean the pcb as much as possible using either flux cleaner spray or IPA then drying this with hot air.

    But have oscillation on VCM/VOCM which supposed to be low impedance. What is allowed capacitance load for VOCM ? 

    Regards

    Krzysztof

  • Krzysztof,

      I was referring to the guard ring pins as being low impedance and matching the voltage of INP which is high impedance. This is the driven protection needed to eliminate noise.

    Just having the guard ring on the PCB is not enough when you have a 1M cable. Properly using the triaxial cable will help you alot.

    Excellent, on the cleaning this is very important and Ionic contamination on the board can be <1Meg Ohm and the INP has a ~2.5Meg Ohm input impedance. 

    If you don't have a capacitor on the VCMHI pin add a 1000pf cap to this pin. Don't go higher capacitance than that. 

    The VOCM is a mirror of VCM and is used only if you are measuring the VOUT with a differential ADC. The VOCM output is set to 1/2 VREF, which is your zero. If you are measuring with a single ended ADC, then you will have to mathematically account for the zero.

    One last comment, everywhere you have a bypass capacitor or filter on this part, add a parallel 1000pf capacitor across the bypass capacitor.

    Let me know how this works out when you get the triaxial cable. 

  • Gordon,

     I followed your suggestions and added 1nF caps:

    1. I have added 1nF to VCMHI – I think it made tiny improvement.
    2. I also added another cap to VREF pin. I had 100nF only. Now I have 1uF and 100nF
    3. 1nF added to VDD1, VDD2

    This did not make much improvement.

    So now my schematic looks like:

    What actually helped is R82 and C97. It made the noise on PH_VOCM around 100uVpp. Also, PH_VOUT was on <500uVpp level. This was measured with RG58 cable shorted at both ends, or only far end.

    Then I connected my artificial pH probe:

    And I still had good level of PH_VOCM (300uVpp) but PH_VOUT got back to 10mVpp.

    Then I used pH probe from DFROBOT SEN0161 (in jug with tap water, this has ~1m coax cable) and this actually gives me satisfying levels of noise: <200uVpp on PH_VOCM, ~500uV on PH_VOUT.

    So, I think this resolved my problem. Gordon, do you think the RC filter is acceptable, or can cause some issues I can’t see now?

    Ps. Triaxial cable – I have got it but actually not tested. Customer requirement is to use 2-wire squeeze connector so I could not use triaxial cable anyway.

    Regards

    Krzysztof

  • Krzysztof,

      The VCM output is basically the power supply to the probe. This output is raised to create the zero for the probe. The output impedance is about 4 ohm. Adding a series resistor to VCM increases the output impedance. The current in the probe will be changing as the levels change. I believe the currents are very low, however a changing current will create a changing VCM to the probe. You should be able to determine if the voltage drop across the series resistor will effect the accuracy of the measurement. It may be so small that it doesn't make a difference. Be prepared to potentially add a correction factor to your measurement at higher current levels. Otherwise it should not hurt the system.

    The 1nf capacitors are for high frequency noise. Helps with EMI and susceptibility. Due to the high impedance of the INP and probe, It is a good idea to keep the 1nf caps. 

    A good ground plane on your board will also help. If you are getting these numbers, you probably have this.

    Good luck with your design.