Linear Analog Calibration

[Neutone]

Sometimes when you post an idea someone else comes along and shows you something you missed. Is their a faster way to perform this? I'm planning to use pre solved values for m and b.

Y=mX+b
(Fractional Interger Compensation)=2^16 Used to avoid floating point math

For analog input
X=Input value(from ADC)
m=(Fractional Interger Compensation)*(Maximum output-Minimum output)/(Maximum input-Minimum input)
b=-(Fractional Interger Compensation)*Minimum input
Y=Output Value (to memory)

For analog output
X=Input value(from memory)
m=(Fractional Interger Compensation)*(Maximum output-Minimum output)/(Maximum input-Minimum input)
b=-(Fractional Interger Compensation)*Minimum input
Y=Output value (to DAC)

The Math
RAW_working=Input value
RAW_working=RAW_working*m
RAW_working=RAW_working+b
Result=RAW_working/(2^16) No processing required as result is located in upper half of RAW_working

16 bit Input_Reading Located within RAW_working (Low bytes)
32 bit RAW_working
16 bit Result Located within RAW_working (high bytes)
32 bit signed m
32 bit signed b
___________________________
This message was ported from CCS's old forum
Original Post ID: 13507

[Tomi]

Yes, it is fast and simple. I use a same method with some differences:
1. I use Y = (1+m)*X + b so if the linearity error is zero (m=0) then no math necessary but an addition.
2. "m" is defined as N/8192 (I use 16-bit arithmetics because the original code was developed for 12CE519 / CCS C V2.61? , no int32 ).
So to divide the value by 2: N=-4096, to double it: N=8192, to keep unchanged: N=0, etc.

But with V3.XXX and int32 you are right, N/65536 is much simpler (and I use it, too ).

:=Sometimes when you post an idea someone else comes along and shows you something you missed. Is their a faster way to perform this? I'm planning to use pre solved values for m and b.
:=
:=Y=mX+b
:=(Fractional Interger Compensation)=2^16 Used to avoid floating point math
:=
:=For analog input
:=X=Input value(from ADC)
:=m=(Fractional Interger Compensation)*(Maximum output-Minimum output)/(Maximum input-Minimum input)
:=b=-(Fractional Interger Compensation)*Minimum input
:=Y=Output Value (to memory)
:=
:=For analog output
:=X=Input value(from memory)
:=m=(Fractional Interger Compensation)*(Maximum output-Minimum output)/(Maximum input-Minimum input)
:=b=-(Fractional Interger Compensation)*Minimum input
:=Y=Output value (to DAC)
:=
:=The Math
:=RAW_working=Input value
:=RAW_working=RAW_working*m
:=RAW_working=RAW_working+b
:=Result=RAW_working/(2^16) No processing required as result is located in upper half of RAW_working
:=
:=16 bit Input_Reading Located within RAW_working (Low bytes)
:=32 bit RAW_working
:=16 bit Result Located within RAW_working (high bytes)
:=32 bit signed m
:=32 bit signed b
___________________________
This message was ported from CCS's old forum
Original Post ID: 13526