The fixed-point FIR filter model can be created in the Ptolemy II schematic editor called Vergil, as shown below:
An impulse is sent through three version of the FIR
filter. The middle filter uses only double-precision floating point.
The bottom filter operates on a stream of
doubles but uses taps that have been quantized using the
quantize()
function in the expression language. This
quantize
function fits the coefficients of the FIR filter into an
8 bit representation, of which 2 bits are used for the integer
part. The upper filter operates on fix point tokens. This
instance of the FIR filter uses fixed-point versions of the
coefficients using the fix()
function in the expression
language. This function converts each tap value into a fix point
representation using 6 bits of which 2 bits are used to represent the
integer part.