Fixpoint Conversion

The following model converts a sequence of floating-point numbers to fixed point, and then further quantizes the fixed point numbers.

The precision of the first fixed-point conversion is controlled by the parameters of the DoubleToFix actor. The default precision is [4,2], which means that a total of 4 bits are used, 2 of which are integer bits. By default, the input is rounded to the nearest representable number. You can change the strategy by changing the quantization parameter of the DoubleToFix actor.

The precision of the second second quantization is controlled by the FixToFix actor, which has a default precision of [3,2]. Thus, the second quantizer reduces by one the number of fractional bits. In addition, it uses an overflow model called "overflow_to_zero". You can modify these properties by right clicking on the actor icons above and selecting "Edit Parameters".

The fixed-point numbers are represented in two's-complement notation. Thus, the set of numbers that can be represented with a [3, 2] precision is:

  { -2.0, -1.5, -1.0, -0.5, 0.0, 0.5, 1.0, 1.5 }