What is a digital signal processor data width?

Digital signal processor (DSP), all floating-point DSP em is a 32-bit fixed-point DSP, and the width is 16 bit, 24 bit and 20-bit DSP, such as Motorola DSP563XX series and Zoran Corporation ZR3800X series. Because EM and DSP external dimensions, the number and the PIN need memory size, and so has a lot to do, so the width of a direct impact on the cost of the device. EM and wider then the size the larger the PIN, the more memory requirements, the greater the cost increases accordingly. In meet the design requirements, try to choose small print width of DSP to reduce costs. In regard to the choice of fixed point and floating point, you can balance the em and the relationship between the development complexity. For example, through a combination of the directive, a 16-bit word wide DSP devices can also implement a 32-bit word wide double-precision arithmetic (course double precision algorithm than the single-precision arithmetic much slower). If a single-precision can satisfy the vast majority of requests, but only a small amount of code needed to double-precision, it is also possible, but if the majority of the computing requirements of precision is high, you need to use a big wide processor. Please note that the vast majority of DSP instruction word and the width of the data words, there are also some, such as ADI (analog devices) of the series of data characters ADSP-21XX to 16-bit and 24-bit word to the directive.