Energy-recovery CMOS for highly pipelined DSP designs

We compare the frequency-versus-power dissipation performance of two energy-recovery CMOS implementations to that of a conventional, supply-voltage-scaled design. The application is a small but complete DSP function. All three designs are based on the same high-level organization and conform to the same I/O specification. SPICE simulations indicate that an energy-recovery design which requires only a small degree of modification to the conventional design offers more than a two-fold reduction in power across a wide range of operating frequencies.

[1]  W. Athas.  Energy-Recovery CMOS , 1996 .

[2]  P. T. Yang,et al.  An FIR Filter Generator , 1992 .

[3]  W. Athas,et al.  A sub-CV2 pad driver with 10 ns transition time , 1996, ISLPED '96.

[4]  N. Tzartzanis,et al.  A resonant signal driver for two-phase, almost-non-overlapping clocks , 1996, 1996 IEEE International Symposium on Circuits and Systems. Circuits and Systems Connecting the World. ISCAS 96.

[5]  R. Jain,et al.  Low-power DSP circuit design using bit-level pipelined maximally-parallel architectures , 1993 .

[6]  Sven Mattisson,et al.  Hot Clock nMOS , 1985 .

[7]  Mary Jane Irwin,et al.  Low power considerations in the design of pipelined FIR filters , 1995, 1995 IEEE Symposium on Low Power Electronics. Digest of Technical Papers.

[8]  Nestoras Tzartzanis,et al.  Design and analysis of a low-power energy-recovery adder , 1995, Proceedings. Fifth Great Lakes Symposium on VLSI.