4.3.1. Bit Growth
Using fixed-point types preserves the extra information of binary point position through hardware blocks, so that it is easy to perform rounding and shifting operations without having to manually track the interpretation of an integer value. A fixed-point type change propagates through your design, with all downstream calculations automatically adjusted.
In a typical mathematical algorithm involving multiplication and addition, data width grows as signals travel through the arithmetic blocks. A large data width implies better accuracy generally, but more hardware resources and potentially lower fMAX (such as in large adders).