??? 03/25/04 06:57 Read: times |
#67399 - RE: Linearisation using curve fitting Responding to: ???'s previous message |
First i would like to thank Mr. Joseph, Mr. Micheal and Mr. Kai for their invaluable help for me.
I will first explain Mr. Kai doubt about input and output. The input is Actual COUNT it can vary from 0 to 1200000 (signal range-2.5V to +2.5V, ADC AD7710). Please note the actual signal range is equated to 1200000 by 2 point calibration). The sensor is having non linearity errors as given by Error counts in my excel file under column ERROR CT (Y). For example if i keep a weight of 200 gram then the output given by the sensor is 400270 counts. But ideally it should be 400000. so i have error of 270 counts which i want to remove from the sensor counts by subtracting form actual count. The error count is always positive for this sensor. Most of the operations of weighing scale is done on counts instead of calcualted wt (of count) to reduce errors. The point i want stress that each sensor is unique in their nonlinearity behaviour but repeatable. So i have to find the non-linearity for each one and get the coefficents for the behaviours and feed into EEPROM for run time compensation. One thing i am ignoring is the temperature compensation of magnetic field which is very sensitive. I have thermistor mounted close enough to magnet for temperature measurement. Mr. Kai you reply on coefficent i need some time study and give feedback. I hope that explained properly. Once again thank you. V.Nagarajan. |