span and offset and equating error with offset is not entirely implausible. I will assume that is what is going on.
The idea is based on the linear transform
Where C is the offset and M the span.
I normally proceed as follows
- Inject a zero signal
- Adjust offset for zero indicated
- Inject a full-scale signal
- Adjust span for full scale indicated
Repeat the above at least 3 times, until within expected tolerance.
Sometimes you can’t inject exactly zero and 100%, but you will need two points as widely saved as possible. And some arithmetic.