Hi, Although the lack of floating point can be worked around, it is at times somewhat of a headache to do so. For example, some functions have to be implemented using lookup tables (I'm thinking of atan2). You are right that in the representation of a signal the noise can make the fractional part irrelevant, but the big challenge is to adapt mathematical formulas for fixed-point only.
Mark On Fri, Jun 12, 2009 at 12:03 PM, Eric Brombaugh <ebrombau...@cox.net>wrote: > Mark Porter wrote: > > The only unfortunate limitation on the Cyclone is the absence of floating >> point, which might be a (big) issue sometimes. >> > > I'm curious if there is any FPGA which has native floating point support. > My understanding was that the nature of an FPGA was to provide low level > logic resources - if you need floating point, build it out of the > gates/registers/etc that are provided. > > That said, I've found very few cases where the dynamic range of full > floating point is required in RF/communications front-end processing. I've > seen block floating point used in hardware FFTs in the past, but seldom in > other applications. Most demods are dealing with so much noise and > distortion to begin with that floating point would be, well, pointless. > > Eric >
