As semiconductor manufacturing pushes into the deep submicron (DSM) regime, specifically at the 32nm node, traditional circuit architectures face significant physical hurdles. This paper, originally published in IEICE Electronics Express, introduces a specialized full adder design optimized for these nanoscale constraints, focusing on power efficiency and signal integrity.
The Challenges of Nanoscale Scaling
Shrinking transistors to the nanoscale introduces two primary "demons" that degrade performance in digital arithmetic:
- Leakage Current: In 32nm transistors, current often "leaks" through the gate even when the switch is in the off state. This leads to wasted power and can cause logic errors where a circuit fails to maintain a distinct "0" or "1" state.
- Voltage Swing Degradation: Low-power designs often utilize pass-transistor logic. In older technologies, this was efficient, but at the nanoscale, it frequently fails to provide a full voltage swing. If an output only reaches 0.8V instead of the intended 1.0V, the signal can become unreadable as it propagates through complex logic networks.
The researchers specifically addressed the failure of Shannon-based adders, which, while effective in larger process nodes, suffer from these integrity issues at 32nm.
Proposed Hybrid Architecture
To combat these issues, the authors developed a hybrid architecture that treats the "Sum" and "Carry" operations with distinct techniques:
- Sum Operation: Utilizes Multiplexing Control Input Techniques (MCIT). To ensure signal robustness, two CMOS inverters were added in series at the output. This buffering stage restores the signal to a full logic high or low, preventing the compounding losses seen in previous designs.
- Carry Operation: A novel transistor arrangement was introduced. By ensuring every NMOS branch is connected to ground and every PMOS branch to $V_{DD}$, the design eliminates the electrical fields that drive problematic leakage currents. This ensures a clean logic "0" and "1" without the errors typical of DSM pass-transistor logic.
Methodology and Validation
The design was simulated using the Microwind31 VLSI CAD tool and BSIM4 Spice models, which are the industry standard for modeling 32nm transistor behavior. To demonstrate real-world applicability, the researchers integrated these cells into an 8×8 bit Braun-array multiplier.
Key Results and Performance
The simulations confirmed that the proposed adder cell maintains full signal integrity even at ultra-low supply voltages (0.35V).
Key performance metrics include:
- Ultra-Low Power Consumption: The 8×8 multiplier consumed only 84 μW.
- Superior Area Efficiency: Despite adding inverters for signal restoration, the design is robust enough to function at 32nm where others fail. The resulting multiplier occupied only 828 μm², representing a 73.9% area reduction compared to MCIS-based architectures and a 50.1% reduction compared to Shannon-based designs.
Conclusion
This research demonstrates that in nanoscale VLSI design, ensuring a complete voltage swing is more critical than simply minimizing transistor count. By prioritizing signal integrity, the authors achieved a design that is both smaller and more power-efficient than scaled versions of previous state-of-the-art architectures. This makes the proposed adder cell an ideal candidate for the next generation of ultra-low-power portable devices and digital signal processors.
Summary updated