Linearity Error AD7767

Hi

I'm using the AD7767 for a high precision measuring device. There is an input stage that reduces my input voltage from +/-12V to +/-2,4V and a Diff Amp to drive the ADC. My ADC input filter consists of two 15 Ohm resistors and one 10nF capacitor. Input stage works very well but I recognize an ADC linearity error six times higher than the promised error of 7,6ppm.

It seems to me, that the input resistance of the AD7767 is not constant and causes the error together with the resistor of the input filter. If I reduce the filter resistors to 5 ohms the linearity error decreases from +/-44ppm to +/-14ppm.

All your application notes show an input filter with 15 ohms resistor and even the eval board uses this values.

Is it possible to get more information about the AD7767 input resistance and its behaviour?

Please can you tell me how I can get the typical 3ppm performance without eliminating the input filter?

Thanks Stefan

AD7767LinError.pdf
  • 0
    •  Analog Employees 
    on Jul 29, 2013 9:50 AM

    Hi, Stefan.

    The analog input has an input capacitance of 22 pF as shown on page 3 of the datasheet. You can find the analog input structure on Page 16 of the datasheet. This will make the THD performance of the device greatly affected by the values of the input filter. The recommended values for differential configuration is shown on Page 20 of the datasheet. AD7767 eval board uses combination of 15 ohms resistor and 2.2 nF capacitor as the input filter. These are empirically tested values to have an optimal performance versus its datasheet spec.

    By the way, what input frequency your application is operating? The linearity of the eval board that gives +/-3ppm is tested with 324.25 Hz sine wave input. The input frequency should not be an integer division of the ADC sampling rate which was set to 1.024 MHz in this case.

     

    Regards,

    Johnny

  • Hi Johnny

    My MCLK is 1.000MHz. Voltage is applied very slow. After applying a Voltage I wait for 1second than I average the Voltage from 256 samples bevor I increace the voltage to the next step. There was only one ramp over the whole ADC range in 4096 steps.

    I'm confused about that steps of approx. 20ppm in my lin measurement. they will be only one third if I reduce my filter resistances from 15 to 5 ohms. This looks like the ADC input resistance is not constant.

    Did you use an input filter for your lin measurements?

    Regards

    Stefan

  • 0
    •  Analog Employees 
    on Aug 5, 2013 3:46 PM

    Hi, Stefan.

    Sorry for the delay.

    It is important to point out that R and C in the input filter forms the time constant for the input signal to settle prior to acquisition. Their values are selected to provide enough current to charge the sampling capacitor. Input capacitance is typically 22 pF but its value is dominated by the sampling capacitor. However, ADC inputs are not easily characterized by impedance since the sample switch continuously opens and closes depending on the sampling rate.

    You mentioned that you used one 10 nF capacitor in the input filter. Would you mind sharing how did you connect it together with the 2 resistors? It would make us understand better if you also impart some portion of your schematic. The AD7767 Eval Board uses 15 Ohms and 2.2 nF input filter values for each differential input and an X7R capacitor is selected to provide temperature stability on its capacitance value.

    Regards,

    Johnny

  • Hi Johnny

    Encolsed my schematics of the ADC and the comparison of the different filters. Further I monitored my reference voltage, but it is stable within 1ppm.

    Regards

    Stefan

    AD7767LinError.pdf
  • 0
    •  Analog Employees 
    on Aug 8, 2013 12:00 PM

    Hi, Stefan.

    The RC combination of 15 Ohms and 2.2 nF of the evaluation board gives a time constant of 33 ns. Take note that the external RC network must settle within the acquisition time of the device. You notice the capacitors (C99 & C100 in the eval board) between the lines were not populated.

    If we're going to consider the 10 nF in your schematic considering that the line is differential, we can think of it as 5 nF equivalent capacitance for each line to GND . Hence, this gives an effective capacitance of 5 nF in parrallel to 1 nF of each line so the time constant is 15 Ohm * 6 nF = 90 ns. This is the time required for the op-amp to charge the capacitor which is 3 times more than the eval board. When you drop the resistor value to 5 Ohms, the time constant is 5 Ohms * 6 nF = 30 ns which is almost the same as the evaluation board.

    The combination of R & C is used to determine the stability of the op-amp. For the evaluation board, 15 Ohms and 2.2 nF were chosen to allow the ADA4841 to settle in enough time for the ADC to complete the signal acquisition.

    With R set in 15 Ohm, you can try lowering the value of the 10 nF in your filter to have an effective capacitance of 2.2 nF. Better yet, remove the 10 nF and replace the 1 nF with 2.2 nF.

    Regards,

    Johnny