conversion of bipolar input signals to unipolar differential signals using resistance network in ADS 8284I 18 channel ADC

I am using ADS8284IB   which is a 18 bit +- 2.5 LSB  integral linearity ADC.  With this ADC I am not getting +- 2.5 LSB accuracy.  I have used some hardware to convert bipolar single ended signal in to true bipolar differential signal. This hardware uses two op-amps (OP07).  To convert bipolar input signals to uni-polar differential signals I have used resistance network as suggested in the data sheet of ADS8284 ADC. I have used 1.2K resistance with 1% tolerance in the resistance network.  The error I am  getting is  about 400 uV on +-4 volt range. The schematic diagram is attached.

Please suggest alternate solution for

(1)        Bipolar single ended signal in to true bipolar differential signal. If some standard  IC’s are available.

(2)        Bipolar input signals to uni-polar differential signals in place of resistance network.

 Please suggest solution to reduce the error in conversion.

Please suggest solution,  I am still waiting for the reply

Hello Seema,

These are two possible solutions:

1) Apply signal offset at the op-amps positive inputs:

2) A single fully-differential amplifier (THS4521) setting the common mode to 2.20V.

Note that to meet the +-2.5LSB accuracy the ADC offset and gain error need to be corrected. Thus, every time the potentiometers are adjusted the ADC codes must be read to correct for the overall gain and offset of the complete signal path. This can be very time consuming with four potentiometers.

Therefore, it is easier and faster to simply do a three-point calibration applying three voltage inputs (-10V, 0V, 10V) to extract the offset and gain error of the complete signal path. Then, the software can apply this correction on the data and the accuracy of the measurements should be +-2.5LSB or better.

For proper calibration by either potentiometers or software, it is critical to measure any voltage with a voltmeter that has better accuracy than the ADC. For 16bit or higher resolution, we typically calibrate converters measuring with the Agilent 3458A.

What voltmeter are you using for calibration? If the meter is not better than the ADC, the error measured can simply be voltmeter error.

Besides of meter accuracy, another source of error is the input capacitor type. For high resolution ADCs, input capacitor must be C0G or NPO type because it keeps capacitance and leakage very constant for any applied input. X7R and X5R capacitors change their capacitance and leakage for different voltages and they must be avoided at high-resolution ADC inputs.

Best regards,
Rafael