Specifications

Limitations
As stated earlier, there are limitations to using this custom function technique.
These limitations are directly proportional to the non-linearity of the desired
waveform. For example, suppose the function X*X*X (or X
3
) is to be represented
over a range of ±1000. The resulting binary range would be ±1024 and the
segments would be partitioned at 1024/64 intervals. This means that every 16 units
would yield an Mx+B calculation over that segment. As long as numbers are
inputted that are VERY close the cardinal points, good values will result. Strictly
speaking, perfect results will be received only when calculated at the cardinal
points, which may be reasonable for an application if the input values are limited to
exactly those 128 points.
The waveform may also be shifted anywhere along the X-axis and Build_table()
will provide the necessary offset calculations to generate the proper table. Be
aware, too, that shifting the table out to greater magnitudes of X may also impact
the precision of the results depending upon the linearity of the waveform. Suffice it
to say, the best results will be produced and it will be easiest to comprehend what is
being done if the waveform stays near theX=0point since most of the
measurement results will have 1e-6..16 values for volts.
One final note. Truncation errors may be seen in the fourth digit of the results. This
is because only 15 bits of the input value is sent to the function. This occurs
because the same technique used for Custom EU conversion is used here and the
method assumes input values are from the 16 bit A/D (15 bits = sign bit). This is
evident in Table 1 where the first and last entries return ±0.9999 rather than ±1. For
most applications this accuracy should be more than adequate.
Generating User Defined Functions
380 Appendix E
Artisan Technology Group - Quality Instrumentation ... Guaranteed | (888) 88-SOURCE | www.artisantg.com