
Originally Posted by
brickeyee
A/D converters have a couple of different properties that can contribute to errors.
Resolution is the first one.
How many bits does the converter use to determine a value.
A 10 bit converter has 1,024 possible values.
Each additional bit doubles the number of possible values.
12 bits produces 4,096 values.
14 is 16,384.
The 'value' of each bit is a function of the range (in whatever unit you carer to use).
The range is set by whatever gain the output of the D/A and its amplifiers are set up
to produce.Many A/D converters have a nominal 1 volt input range.
Over that 1 volt range there are gain errors, linearity, and
offset errors.
In many cases there is also some 'digital curve fitting' coming into play.
This is the reason for multiple points in the setup.
Commonly simple curve fitting will use zero, and a single weight near maximum.
More sophisticated curve fitting might use 0, 1/4 of the range, 1/2 of the range, and then a fill scale setting.
Easily accomplished with a set of weights that are equal.
Generally the more points there are the better.
One simple trick is to use a larger number of equal weights.
This allows the simple use of the same value of weights.
There are high resolution 'successive approximation' converter.
They can get into 20 bits and higher.