How much error in interpolation by Acc-51P will be expected?
We designed the circuitry and algorithms of the Acc-51B and Acc-51E to make the errors they introduced so small that they could not practically be measured. We believe that we have succeeded in this. We have not been able to create a test that could show any errors created by the interpolator and not the input signals. We use 14-bit A/D converters with +/-.5 LSB non-linearity (integral and differential) specs when we really need only 11 or 12-bit resolution to get 4096 states per line. We use very high-quality gain-stabilized op-amps with precision components on the front end. Users who are concerned about accuracy should use the differential inputs we provide to eliminate possible offset concerns. The arc tangent calculation used is fundamentally ratio metric and the analog parts that we use are lot-matched so that they have a strong tendency to vary in the same way; so variations tend to be canceled out in the calculation. The arc tangent calculation is accurate to 14 bits before it