作者: D. Menard , R. Rocher , O. Sentieys
关键词:
摘要: One of the most important stages floating-point to fixed-point conversion, is evaluation specification accuracy. This required optimize data word-length according accuracy constraints. Classical methods for are based on simulations but they lead very long optimization times.A new method an analytical approach presented case linear-time-invariant (LTI) systems. The use this in minimization processes reduces significantly time compared simulation methods. Our allows automatic determination signal-to-quantization-noise-ratio (SQNR) expression at system output format. valid recursive and non-recursive LTI systems takes account quantization modes (truncation or rounding). theoretical concepts different methodology explained. Then, ability efficiently evaluate demonstrated through examples.