pyfixp

A numpy based Python library for fast fixpoint arithmetics.

This library was originally created for the Python Filter Design and Analysis pyfda project but can also be used standalone.

(Re-)quantizers read and write numpy scalars and arrays, allowing easy interfacing to e. g. float stimuli and matplotlib plotting. Quantizers are instances of the Fixed() class, their quantization and saturation behaviour and the number of integer and fractional bits is controlled with a dictionary.

Most routines operate on scalars and arrays alike, currently the following operations are supported:

(Re)Quantization ('floor', 'round', 'fix') with two's complement wrap around ('wrap') or saturation behaviour ('sat' )
Conversion of binary, hex, decimal, Canonical-Signed-Digit (CSD) format strings to float and vice versa
A basic direct form FIR filter in the submodule pyfixp.filters

Example

In the following example a quantizer is defined with an output format of 0 integer bits and 3 fractional bits, overflows are wrapped around in two's complement style ("wrap") and additional fractional bits are simply truncated ("floor"):

import pyfixp as fx
q_dict = {'WI':0, 'WF': 3,               # number of integer / fractional bits
      'quant':'floor', 'ovfl': 'wrap'}   # quantization / overflow behaviour
Q = fx.Fixed(q_dict)                     # instance of fixpoint class Fixed()
third_q = Q.fixp(1/3)
print(f"q<1/3> = {third_q}\n")
for i in np.arange(12)/10:               # i = 0, 0.1, 0.2, ...
    print("q<{0:>3.2f}> = {1:>5.3f}".format(i, Q.fixp(i))) # quantize i

The options are shown by entering fx.Fixed? in the notebook.

More examples can be found in the the Jupyter notebooks of the doc subdirectory.

intro_pyfixp.ipynb: An introduction to using pyfixp for quantizing and saturating signals
fixpoint_filters.ipynb: An introduction to fixpoint filters and how to implement them with pyfixp (work in progress)

You can use the notebooks interactively using the Binder service:

chipmuenk/pyfixp

pyfixp

Example