摘要
Adaptive noise data filtering in real-time requires dedicated hardware to meet demanding time requirements. Both DSP processors and FPGAs were studied with respect to their performance in power consumption, hardware architecture, and speed for real time applications. For testing purposes, real time adaptive noise filters have been implemented and simulated on two different platforms, Motorola DSP56303 EVM and Xilinx Spartan III boards. This study has shown that in high speed applications, FPGAs are advantageous over DSPs with respect of their speed and noise reduction because of their parallel architecture. FPGAs can handle more processes at the same time when compared to DSPs, while the later can only handle a limited number of parallel instructions at a time. The speed in both processors impacts the noise reduction in real time. As the DSP core gets slower, the noise removal in real time gets harder to achieve. With respect to power, DSPs are advantageous over FPGAs. FPGAs have reconfigurable gate structure which consumes more power. In case of DSPs, the hardware has been already configured, which requires less power consumption? FPGAs are built for general purposes, and their silicon area in the core is bigger than that of DSPs. This is another factor that affects power consumption. As a result, in high frequency applications, FPGAs are advantageous as compared to DSPs. In low frequency applications, DSPs and FPGAs both satisfy the requirements for noise cancelling. For low frequency applications, DSPs are advantageous in their power consumption and applications for the battery power devices. Software utilizing Matlab, VHDL code run on Xilinix system, and assembly running on Motorola development systems, have been used for the demonstration of this study.
Adaptive noise data filtering in real-time requires dedicated hardware to meet demanding time requirements. Both DSP processors and FPGAs were studied with respect to their performance in power consumption, hardware architecture, and speed for real time applications. For testing purposes, real time adaptive noise filters have been implemented and simulated on two different platforms, Motorola DSP56303 EVM and Xilinx Spartan III boards. This study has shown that in high speed applications, FPGAs are advantageous over DSPs with respect of their speed and noise reduction because of their parallel architecture. FPGAs can handle more processes at the same time when compared to DSPs, while the later can only handle a limited number of parallel instructions at a time. The speed in both processors impacts the noise reduction in real time. As the DSP core gets slower, the noise removal in real time gets harder to achieve. With respect to power, DSPs are advantageous over FPGAs. FPGAs have reconfigurable gate structure which consumes more power. In case of DSPs, the hardware has been already configured, which requires less power consumption? FPGAs are built for general purposes, and their silicon area in the core is bigger than that of DSPs. This is another factor that affects power consumption. As a result, in high frequency applications, FPGAs are advantageous as compared to DSPs. In low frequency applications, DSPs and FPGAs both satisfy the requirements for noise cancelling. For low frequency applications, DSPs are advantageous in their power consumption and applications for the battery power devices. Software utilizing Matlab, VHDL code run on Xilinix system, and assembly running on Motorola development systems, have been used for the demonstration of this study.
