Access Official Technical PDF

Share Summary:LinkedInEmail

As transistor dimensions and supply voltages shrink, process variation can make otherwise identical circuits behave differently. For digital logic, those differences can affect delay and energy as well as the ability to meet timing requirements. An evaluation of Schmitt trigger variability mitigation in 7 nm FinFET full-adder layouts shows both the potential value of the technique and the trade-offs engineers need to consider.

The approach replaces selected internal inverters in a full adder with Schmitt Trigger (ST) inverters. Because a Schmitt Trigger uses different switching thresholds for rising and falling inputs, its hysteresis can help reject input disturbances. But the modified circuit also changes the cell’s electrical and physical characteristics. The right question is therefore not simply whether variability improves, but whether the improvement justifies changes to delay, energy and area.

Why process variability matters in FinFET circuits

Manufacturing variation changes device characteristics, which can shift circuit performance and power consumption. In FinFET technology, relevant physical variations include work function, gate length, fin height and fin thickness, as well as parasitic resistance. Work function fluctuation can contribute to threshold-voltage variation, which in turn affects transistor behavior.

FinFETs provide stronger gate control than planar devices and can reduce some short-channel effects. That does not remove the effects of manufacturing imprecision. As dimensions become smaller, variations can still influence a logic cell’s delay, energy use and operating margins.

A full adder is a useful circuit for examining these effects. It produces a Sum and a Carry Out from two input bits and a carry input, and full-adder cells can be arranged in arithmetic datapaths. Their topology influences signal paths, loading and circuit behavior under variation.

How Schmitt Trigger variability mitigation works

A conventional inverter switches at a threshold as its input rises or falls. A Schmitt Trigger has separate switching thresholds for those transitions, creating a hysteresis window. This characteristic is commonly used to make circuits less sensitive to noisy or slowly changing inputs.

In the evaluated technique, ST inverters replace internal inverters in four full-adder architectures: Mirror CMOS, Transmission Gate Adder (TGA), Transmission Function Adder (TFA) and Hybrid. The change is intended to improve robustness to process variation. However, the Schmitt Trigger’s own switching behavior is affected by device variation, and its transistor arrangement can add delay. Improvements in one metric are not guaranteed to appear in every output or topology.

What the 7 nm layout evaluation measured

The designs were implemented using the ASAP7 7 nm FinFET academic process design kit. Both the original and modified versions of each adder were laid out and passed through physical verification. Parasitics were extracted from the layouts before electrical simulation, allowing the comparison to account for layout-related effects rather than relying only on idealized circuit schematics.

Simulations considered two supply conditions: nominal operation at 0.7 V and near-threshold operation at 0.4 V. Process variation was evaluated with 2,000 Monte Carlo simulations, varying PMOS and NMOS threshold voltages according to a Gaussian distribution based on work function fluctuation. Delay and energy were assessed for both Sum and Carry Out. The normalized standard deviation (standard deviation divided by mean) was used as a measure of sensitivity to variation: a lower value indicates less relative spread in the measured results.

Results depend on topology and operating voltage

The results show that the ST technique can reduce variability, particularly for energy, but does not provide a uniform benefit across all cells. At nominal voltage, energy variability decreased for all four adder designs. Reported improvements varied by topology and output, with the Hybrid adder showing a substantial improvement in energy robustness. Delay results were mixed: some output paths became less sensitive to variation, while others worsened or showed little change.

Near-threshold operation presents a different design picture. Lower supply voltage reduced mean energy for the evaluated circuits compared with nominal operation, but delay became more vulnerable to variation in some cases. The TGA and TFA showed high delay deviations and maximum delays, consistent with the signal degradation associated with their pass-transistor logic. The Schmitt Trigger technique improved energy robustness across the evaluated designs, with reported gains reaching approximately 65% for an energy measure. Delay robustness, however, did not improve in every case.

These findings underline why “robustness” should be evaluated by metric and output path. A cell may become more consistent in energy while gaining delay variability, or improve one output while worsening another. Results from one adder architecture should not be assumed to apply to a different topology.

Account for implementation penalties

The modified designs also carried costs. In the evaluation, adding Schmitt Triggers increased mean delay and energy in several comparisons. The size and direction of those penalties varied with the adder and operating condition. The additional circuitry also affects layout area, so area must be included in any design comparison rather than treated as a secondary concern.

For engineers and system integrators considering this approach, a practical assessment should compare the original and modified cells using the constraints that matter to the target design:

  • Measure delay and energy variability separately for Sum and Carry Out.
  • Evaluate nominal and near-threshold operation when both are relevant to the design.
  • Include extracted parasitics and layout area in the comparison.
  • Check whether improvements in variability outweigh changes to average performance and energy.

Conclusion: choose the technique by design context

Replacing internal full-adder inverters with Schmitt Triggers can improve resistance to process variability in 7 nm FinFET layouts, especially for energy-related measures. The benefit is topology- and condition-dependent, and it can come with higher delay, energy consumption or area. For industrial and computing-system designers, the key implication is to evaluate variability mitigation at the layout level and across relevant operating conditions. The technique is a potential design option—not a universal replacement for careful circuit-level trade-off analysis.

Access Official Technical PDF

About The Author