Design and realization of a high speed 64 x 64 - bit multiplier for low power applications

Official URL: http://risc01.sabanciuniv.edu/record=b1083690 (Table of Contents)

Wireless communication systems, including third generation cellular radio systems and wireless LANs, have become tremendously popular in recent years. These systems can be implemented using various platforms, like digital signal processors, ASICs and FPGAs. Most digital signal processing systems incorporate a multiplication unit to implement algorithms such as correlations, convolution, filtering and frequency analysis. These algorithms are used in applications such as finite impulse filters (FIR), infinite impulse filters (IIR), discrete cosine transforms (DCT) and fast Fourier transforms (FFT). Moreover, there has been a rapid increase in the popularity of portable and wireless electronic devices, like laptop computers, portable video players and cellular phones, which rely on embedded digital signal processors. Since the desire is to design digital systems for communication applications at best performance without power sacrifices, the need for high performance and low power multipliers is inevitable. Since multiplication is one of the most critical operations in many computational systems, there have been many algorithm proposals in the literature to perform multiplication, each offering different advantages and having tradeoffs in terms of speed, circuit complexity, area and power consumption. This thesis focuses on an ASIC implementation of a multiplexer-based multiplication method, an efficient algorithm which is applicable to low power applications. Recently, it has been proved that the multiplexer-based multiplier outperforms the modified Booth multiplier both in speed and power dissipation by 13% to 26%, due to small internal capacitance. After analyzing the performance characteristics of conventional multiplier types, it is observed that the one designed using multiplexer-based multiplication algorithm is more advantageous, especially when the size of the multiplied numbers is small. In order to verify the superiorities of this algorithm, we performed an implementation, in which the bit size of the multiplicand and the multiplier is comparably large. Thus, realization of a 64 x 64-bit multiplier block has been done in 0.35 M [micron] CMOS technology using Cadence Design Framework tools. The final multiplier structure operates at 12.8 ns with an approximate dynamic power consumption of 1mW. Also, using the same algorithm, another block of 32-bit x 32-bit multiplier is designed and is sent for fabrication.