Mantracourt Electronics Limited
DCell & DSC MantraCAN User Manual
46
How to Set Up Linearity Compensation
A linearity correction can be set up either from sensor specification/calibration data, or more commonly from in-
system testing results.
Assuming we do not have any prior information on linearity errors, the usual approach is to do a series of controlled
tests with accurately known test loads.
Just as with temperature compensation, it is
possible
to obtain a detailed graph of linearity error and then choose a
‘best-fit’ piecewise linear curve for the compensation table.
However, it is generally good enough, and much simpler, to simply test at several different points and then apply an
exact correction at those points. If the error curve is reasonably smooth, this should give exact results at the test
points, and reasonably accurate values in between.
NOTES:
Linearisation tests should only be done
after
the cell calibration is set, because the correction values are dependent
on the cell calibration.
Similarly, linearisation testing should only be done at the calibration ‘reference’ temperature, or after temperature
compensation is installed, to avoid temperature effects from distorting the results.
The linearisation tests should not reveal any significant remaining linear trend in the errors.
If errors do appear to lie on a definite line, this could drastically reduce the accuracy of the correction.
If this does happen, it shows that the cell calibration is wrong and should be redone.
The table points must always cover more-or-less the whole range of output values to be used, because corrections
are extrapolated outward beyond the first and last points.
It is always worthwhile including more test-points than will be used in the correction table, because this gives
confidence that no regions of rapidly changing error have been missed.
Tests should be done both with steadily increasing
and
decreasing load values, as hysteresis effects (for load cells)
are often of a similar size to non-linearities.
Parameter Calculations and Example
Based on the simple method outlined above, we suppose that we have obtained test results for a series of precisely
known load values –
test loads Xi give readings of CRAW = Ci, for (i = 1..n)
Then calculate the errors that need to be removed at these points –
Ei = Xi
–
Ci
Now just enter these values into the correction table, remembering to scale the errors –
CLN = n
CLX
i =
X
i
CLKi = 1000
•
Ei
Example
Suppose we have a load cell and Cell calibration giving a result in the range 0-500 KgF.
The following test results were obtained using a series of known test loads –
For test load of x1 = 0Kg
:
CELL reading c1 = 0.0010
For test load of x2 = 100.13Kg :
CELL reading c2 = 100.44
For test load of x3 = 199.72Kg :
CELL reading c3 = 200.57
For test load of x4 = 349.97Kg :
CELL reading c4 = 349.75
For test load of x5 = 450.03Kg :
CELL reading c5 = 449.98
We choose these precise test points as our linearisation reference points, so
CLN = 5
CLX1 = 0.0010
CLX2 = 100.44
CLX3 = 200.57
