Chapter 5: Floating-point arithmetic

Abstract

Fixed point numbers suffer from limited range and accuracy. For a given word length both fixed point and floating point representations give equal distinct numbers. The difference is that in fixed point representation the spacing between the numbers is equal, so smaller numbers when truncated or rounded give a much larger error than the larger numbers. However floating point representation gives different spacing between numbers. We get denser distances between numbers when the number is small and sparser distance for larger numbers. So the absolute representation error increases with larger numbers.