HomeProjectsPeoplePublicatons
Search:
   
 

Paramererizable Multiplier Library

Hungchi Lee, M.S. 1999

Multiplication is an important part of real-time digital signal processing (DSP) applications ranging from digital filtering to image processing. The multipliers used in such applications require many different operand size. An efficient way to design multipliers with different sizes is through the design of the parameterized cells. A parameterized cell is a graphic, programmable cell that lets you create a customized instance each time you place it by allowing the designer to specify certain parameters. The parameter is a setting that can control the size, shape, or contents of the cell instance. The big advantage in using parameterized cell is that you can speed up the time of chip design by eliminating the need to create a lot of chips with the same function but different sizes. Also, parameterized cells can save disk space by creating a library of cells for similar parts that are all used the same cells.(for example in my design the regular multiplier, complex multiplier and multiplier-accumulator are almost use the same cells).

However, parameterized design is not the only considerations; low power dissipation and small chip area are also needed because of the dense packing of transistors in today’s DSP chips. The main objective of this thesis is to design a parameterized cell library by using Cadence CAD tool, called "parameterized cell (Pcell)", to automatically generate a 16x16-bit multiplier, a 12x12-bit complex number multiplier and a 12x12-bit complex number multiplier-accumulator(MAC) for the second-generation digital backend receiver.

The three main considerations for the design are a high multiplication speed, low power dissipation, and a small rectangular chip area. All the designs are using SGS-THOMPSON MICROELECTRONICS’s HCMOS7 0.25um technology. All layout and logic testing are done with Cadence CAD tools with Unicad design kit, and EPIC. All delay and power simulation are done with Meta Software’s HSpice and PowerMill.