A New Complex Floating-Point Representation Using Auto- Correlation Algorithm for DSP Processors
E. Thenmozhi Ramyah, S. Pavithra, Vijayan Prabhakaran · 2015
Complex number computation is a big deal in mathematical calculations and it is well practiced in all modern processors, as we all familiar with real and imaginary parts in complex numbers, computation against the real and imaginary parts becomes a difficult task over a specific inputs. Here a DSP processors is taken for signal manipulation and a new method is introduced which reduces the bits size for better output. In this paper, a new complex floating point representation for complex numbers is introduced and is compared with IEEE 754 standard and a common DSP fixed point. The resulting system will use fewer bits than IEEE 754, which keeps the dynamic range and precision. It is also proposed that it has good quantization noise analysis using auto-correlation algorithm. This new algorithm reduces the former IEEE 754 Single precision representation total bit by 1 bit and also retains variables with a better exponent and mantissa parts.