Quantization Noise: Simulating Arbitrary–Precision Fixed–Point and Floating–Point Roundoff in Matlab

The basic idea of the first two possibilities is to evaluate each operation in Matlab, and follow each by a roundoff operation. For example, if a and b are two finite wordlength numbers which are to be multiplied, the result is evaluated as c = Q(a ∗ b). This is a fast and simple approach, and generally provides good results. There are, however, some limitations which will be discussed below. The last two possibilities are based on objects. The methods provided by the object libraries allow precise simulation of virtually any finite wordlength operation occurring in practice, at the cost of slower execution.