We use the ADS8505 to sense current in one of our products. We are looking at modifying that to a high accuracy current sensor, interfacing a fluxgate sensor to the A/D. The fluxgate device outputs a current which we detect with foil burden resistors and an amplifier using a low drift resistor network. I’m doing a Monte Carlo analysis in Spice to estimate the temperature effects. The ADS8505 lists the A/D specs as
Full scale error drift: +/- 7ppm/C typ
Bipolar zero error drift: +/- 2 ppm/C typ
Internal Reference drift: +/- 8 ppm/C typ
I’m modelling the A/D as a variable voltage source (the zero drift) in series with a variable gain (the full scale error and Internal reference drifts). I’m looking at using an external reference to improve drift performance. My questions are:
1.) How do I combine the A/D full scale error and the reference error? It would seem that they both affect the effective “gain” of the A/D. One approach would be to assume that they are the sum of two Gaussian random variables and make the resultant standard deviation equal to the sqrt of the sum of the squares of the individual std devs. Or, I could model them as cascaded gains. Or ???
Note: I did MC runs using both approaches, and the results are similar.
2.) I am modelling most of these drifts as Gaussian random variable, and assigning a GRV to each component, then doing 1000 pass Monte Carlo. What standard deviation should I use for the parameters mentioned above? 3-sigma? More? Less?
3.) Another question relates to the 2.5V reference used for +/-10V input range. Does the drift of the reference directly affect the accuracy, or is there a scale factor? In other words, since the input range is +/- 4 times the reference, is the drift related error 4x or 8x that of the 2.5V reference?
