Numerical issues
Naim Dahnoun · 2017
The majority of digital signal processing (DSP) algorithms are based on the dot product. Digital signal processors which process digitalised data use fixed-point and/or floating-point arithmetic. These processors can be divided into two categories: fixed-point and floating-point processors. This chapter shows typical fixed- and floating-point applications. Different applications use different numerical formats. It is important to note at this stage that the high data rate and the high precision required for the new communication standards like Long-Term Evolution (LTE) are now easily supported by the KeyStone floating-point formats. The chapter discusses the various formats supported by the KeyStone processors. Floating-point processors are ideally suited for handling operations on numbers with a large dynamic range. However, due to the fact that numbers in any format can only be represented with a limited number of bits, the number's accuracy will be affected. Fixed-point numbers have a lower accuracy than floating-point ones.