Post Go back to editing

AD7175 - average value changes

Category: Hardware
Product Number: AD7175-8

Hi,

I'm measuring 4 differential analog input with an AD7175-8 at a frequency of 100Hz.

To do so I set the AD7175 in single conversion mode, configure and activate the 4 analog channels and start a new measurement with a frequency of 100Hz.

Everything work fine but I saw a strange behaviour when changing the configuration of the AD converter. Indeed, although my analog signals are constants and there is a small offset difference when changing the filter type (sync5 + sync1 / sync3), the filter size or passing from the single conversion mode to the continuous one. Those offsets are not big (2000 - 3000 ticks out of the 24 bits range), but I don't understand them. I thought those filter where completely digital and it should not change the offset of the signals measured.

Thanks for your help

  • Hi,

    May I know the data you are measuring using the sinc5+1 and sinc3 filters? Please also include the register settings and the process/steps you did in measuring the data. Thank you.

    Regards,
    Rod

  • I'm measuring pressure and temperature data.

    Here is the schematic that I use:

    And here is my register configuration:

     - 4 Channels enabled, all other disabled
     - Configuration bipolar, external reference, AIN+ buffer enabled, AIN- buffer enabled, REF+ buffer enabled
     - sync5+1 filter, ODR0 = 9
     - single conversion mode, delay 40 us

  • Hi,

    Can you complete the table below for the offset values you obtained when changing the filter type, filter size, or conversion mode. For the inputs, kindly measure the actual input values from the input pin. Thank you. 

    Regards,

    Rod

  • Ho Rod,

    It's difficult to make such table with meaningful data, because the temperature that I measuring is never really constant. Moreover the error is so small that I don't think the measurement I made with my multimeter is really relevant. However, here is my table:

    Filter type Filter size Conversion mode AIN2 [V] AIN3 [V] REF+ [V] ADC output average value
    SINC5 + 1 9 → 2.5kHz single 4.082 0.937 5.013 10'537'800
    SINC5 + 1 11 → 500Hz single 4.082 0.938 5.014 10'534'400
    SINC3 9 → 1667 Hz single 4.082 0.938 5.014 10'535'400
    SINC5 + 1 9 → 2.5kHz continuous 4.082 0.938 5.014 10'534'300

    But we can see a clear offset of about ~2'000 ticks (~120 ppm) when changing from sync5+1 to the sync3 filter. This offset is repeatable and is easily seen when plotting the conversions output over time.

  • Good day,

    I have tried changing the filter size and filter type (from sync5+1 at 2.5k SPS to sync3 at 5k SPS, SING_CYC enabled) given AIN2 and AIN3 with 4.082V and 0.937V respectively, a REF+ of 5.013V, and the same configurations you provided; however, I am only measuring a code difference of around 100-200 codes (from the Mean) while the noise values changes. Changing the filter size, filter type, and conversion mode in my setup only shows a code difference of around 100-200 codes.

    Did you perform any calibrations on your channels?

    Are the values on your table represent each of your 4 channels, or is it a single channel with 4 trials of different configurations?

    Also, can you provide a step-by-step procedure on how you measure your data? 

    Regards,

    Rod

  • Hi Rod,

    Thank you for taking the time helping me.

    No, I did not perform any calibration. The chip is straight as received from the the factory. However I've seen a similar behaviour on multiple chips.

    The values I give in the table represent a single channel with 4 trials of different configuration. A few minutes separate every trial.

    I'm not sure what you want to know about the step-by-step procedure. I have an AD7175 converter connected to microcontroller. At startup the microcontroller configure the A/D and then goes read one conversion of each channels (4 used) every 10 ms. The measured data are then transmitted to a PC where I have a small house-made tool to plot the data.

    One thing that I noticed in your simulation is that the configuration seems to be BIPOLAR but I'm measuring in UNIPOLAR, that explains the difference of average value.

    There is something that I can't understand, from where does the difference (very small I agree) of average comes from in your simulation? If I understood correctly those filters are only digital and of type FIR, I don't see how they could change the average value of a DC signal

  • Good day,

    The noise performance changes per filter and odr configurations which is described in the datasheet. With this, we would expect that the average from my simulation may slightly differ due to the noise differences, however, the difference should only be minimal unlike as what you are seeing. In your measurements it does not look like noise; it really looks like an offset difference.

    It is possible, for example, to perform ADC noise test in your system. Isolate the input source by simply shorting the inputs to the GND or to a common mode (midsupply) and measure 100 samples. Then, get the pk-pk noise and rms noise and compared it with the datasheet. You can also check if the plot/waveform also shift the shorted input measurement. Kindly plot your waveform per configuration. Thank you.

    Regards,

    Rod