Biasing ADC input

Easiest and most accurate solution is using the "voltage divider" with two resistors connected to the reference. Use 0.1% resistors.

#1 requires an extra reference, including its tolerances, thus requiring calibration (AD has its own internal reference, thus compensating its own measurments) #2 would be the solution (if ADC doesn't sink enough, take for example order of 10k) Any arising voltage drops due to the R- network can be incalculated in software #3 has the disadvantage of different leakage currents for every device, thus requiring calibration

Specs: Using a TC7660 or TCM829 for the negative supply to the AD594. I'm using the internal 2.56V Vref of the AVR ADC. Vref is exposed on a pin. The AD594 can source 5ma on its output. The ADC has 100M input resistance.

Googling came up with the following suggestions: #1)Add a low voltage reference in series to temperature sensor.

#2) Try connecting the output of the temp sensor to the ADC through a 1k resistor. Take another 1k resistor and connect it between the reference voltage and the ADC input. This may not work if your temp sensor will not sink enough current.

#3) Many ADC's have voltage reference outputs that can be used to bias the analog input(s) of the ADC. This can be done by connecting a resistor from the ADC reference output to the analog input. The Voltage reference output can be bypassed to analog ground with a small capacitor to improve the ripple rejection. The bias resistor value can be selected based upon the ADC input leakage current. Consider a resistor value such that the maximum ADC input leakage current alone causes less than a 1 LSB voltage across the bias resistor. In this way the "offset code" of the ADC will not be overly input bias current dependent.

#1 seems straight forward enough, however, #2 only requires two resistors and #3 only one.

Questions:

- Is one or is two resistors the better solution?

- #3 gives specifics on selecting the resitor value. Any help with this, I'm lost?

thx, t