Access Official Technical PDF

Share Summary:LinkedInEmail

Full adder power optimization is a useful design consideration wherever arithmetic logic is repeated across a digital system. A full adder processes three one-bit inputs—A, B, and carry-in—and produces a sum and carry-out. Those functions appear in arithmetic units and in operations such as address calculation and multiply-accumulate processing, so the cell’s power and area can influence the larger design.

A 45nm technology study simulated three implementations: 10-transistor, 14-transistor, and conventional 28-transistor full adders. The reported comparisons show reductions in both power and area for the proposed designs relative to earlier versions. The results illustrate the potential value of compact circuit structures, while also highlighting why cell-level comparisons need to be interpreted in the context of their simulation conditions and the requirements of the target design.

Why full adder design affects power and area

Full adders are core logic elements in many arithmetic circuits. A change to an adder cell can therefore matter beyond a single operation, particularly when the function is instantiated repeatedly. Designers typically weigh power and physical area alongside correct logic behavior and the ability to drive connected circuitry.

In CMOS circuits, power includes dynamic and static components. The study discusses leakage current as a contributor to leakage power and notes that leakage from individual transistors accumulates at the circuit level. Reducing transistor count may reduce the number of contributing devices, but transistor count alone does not establish the overall power or performance of a design. Circuit topology, operating conditions, and implementation details also matter.

Three transistor-count approaches

10T full adder

The 10T implementation uses ten MOSFETs and is presented as an area- and power-oriented alternative to designs with more transistors. Its schematic and layout were evaluated in a 45nm process setup at a 1V supply. The study describes PMOS and NMOS widths of 240nm and 120nm, respectively, and provides output, leakage-current, and DC power waveforms.

14T full adder

The 14T design uses pass-transistor-based XOR and XNOR circuits. The paper describes earlier 14T work as offering improvements in delay and power compared with preceding full-adder designs, while identifying threshold-voltage loss and remaining power consumption as concerns. Its simulation setup also uses a 1V supply and the same stated transistor widths. An output waveform is shown at 100MHz.

28T conventional full adder

The 28T implementation uses a conventional static CMOS structure, with pull-up and pull-down networks intended to provide full-swing outputs and driving capability. The reported simulation uses a 10fF load capacitance and examines supply voltages from 0.7V to 1.5V. The paper also presents an output waveform at 1V and 50MHz.

Reported power and area reductions

The paper compares its proposed implementations with earlier designs and reports the following reductions:

  • 10T full adder: power reduced by 59.31% and area by 40.68%.
  • 14T full adder: power reduced by 62.06% and area by 37.93%.
  • 28T full adder: power reduced by 41.49% and area by 58.50%.

The reported power values are 8.561nW for the proposed 10T design, 29.67nW for the proposed 14T design, and 205.4nW for the proposed 28T design. The comparison values listed for the earlier designs are 21.04nW, 78.21nW, and 351.1nW, respectively. These figures are simulation results, not a universal ranking: the study compares each implementation against its corresponding earlier design, and the reported setups differ in places.

How to interpret the results in a design workflow

For engineers assessing a full-adder cell, the study offers a starting point for comparing compact and conventional structures. However, the reported percentage reductions do not by themselves determine which topology is best for a processor datapath, controller, or other arithmetic block. Selection depends on the intended operating conditions and the surrounding circuit.

A practical evaluation should confirm that the candidate cell:

  • Produces correct sum and carry outputs across the required input combinations.
  • Meets the design’s supply-voltage, load, and operating-frequency requirements.
  • Has power and area measured under comparable simulation conditions.
  • Provides suitable output behavior when integrated into the next level of logic.

The paper describes Cadence SPECTRE simulations and presents schematics, layouts, waveforms, and comparative figures. It does not provide a detailed timing comparison for all three designs in the summarized results. Engineers should therefore treat the power and area findings as specific to the reported evaluation, and verify timing and other integration requirements independently before choosing a cell.

Conclusion

The 45nm comparison indicates that transistor-level redesign can reduce simulated power and area across 10T, 14T, and 28T full-adder implementations. The largest reported power reduction is for the 14T design, while the largest reported area reduction is for the 28T design. For practical use, those results are most valuable as a reminder to evaluate power, area, output behavior, and operating conditions together. A cell that performs well in an isolated simulation still needs validation within the target arithmetic path and its actual design constraints.

Access Official Technical PDF

About The Author