HCS12 datasheet(369/452 Pages) FREESCALE

369 / 452 page

S12CPUV2 Reference Manual, Rev. 4.0

Freescale Semiconductor

369

used to divide the sum-of-products by the sum-of-weights. TMP1 : TMP2 is

transferred into Y : D and TMP3 is transferred into X.

9.7 Custom Fuzzy Logic Programming

The basic fuzzy logic inference techniques described earlier are suitable for

a broad range of applications, but some systems may require

customization. The built-in fuzzy instructions use 8-bit resolution and some

systems may require finer resolution. The rule evaluation instructions only

support variations of MIN-MAX rule evaluation and other methods have

been discussed in fuzzy logic literature. The weighted average of singletons

is not the only defuzzification technique. The CPU12 has several

instructions and addressing modes that can be helpful when developing

custom fuzzy logic systems.

9.7.1 Fuzzification Variations

The MEM instruction supports trapezoidal membership functions and

several other varieties, including membership functions with vertical sides

(infinite slope sides). Triangular membership functions are a subset of

trapezoidal functions. Some practitioners refer to s-, z-, and

π-shaped

membership functions. These refer to a trapezoid butted against the right

end of the x-axis, a trapezoid butted against the left end of the x-axis, and a

trapezoidal membership function that isn’t butted against either end of the

x-axis, respectively. Many other membership function shapes are possible,

if memory space and processing bandwidth are sufficient.

Tabular membership functions offer complete flexibility in shape and very

fast evaluation time. However, tables take a very large amount of memory

space (as many as 256 bytes per label of one system input). The excessive

size to specify tabular membership functions makes them impractical for

most microcontroller-based fuzzy systems. The CPU12 instruction set

includes two instructions (TBL and ETBL) for lookup and interpolation of

compressed tables.

The TBL instruction uses 8-bit table entries (y-values) and returns an 8-bit

result. The ETBL instruction uses 16-bit table entries (y-values) and returns

a 16-bit result. A flexible indexed addressing mode is used to identify the

effective address of the data point at the beginning of the line segment, and

the data value for the end point of the line segment is the next consecutive

memory location (byte for TBL and word for ETBL). In both cases, the B

HCS12 Datasheet(PDF) 369 Page - Freescale Semiconductor, Inc