FPGAs let you define the structure of the hardware, and implementing fixed point is a lot simpler in terms of logic gages used and usually higher performance. Y

FPGAs let you define the structure of the hardware, and implementing fixed point is a lot simpler in terms of logic gages used and usually higher performance. You absolutely could use floating point, but the HW cost would be high, so most folks convert to fixed and use that. Then the question becomes how many bits, as you can have arbitrary width fixed point, so must choose at the design time depending on the application requirements.