Low-Power and testable circuit synthesis using Shannon decomposition

Swaroop Ghosh, Swarup Bhunia, Kaushik RoyACM Trans. Design Autom. Electr. Syst 2007 · DOI

TL;DR

This paper introduces a novel circuit synthesis approach based on Shannon decomposition and dynamic supply gating. This method significantly enhances circuit testability and reduces power consumption by creating a more structured, inherently testable design, offering a promising solution for the challenges of modern nanoscale integrated circuits.

What Problem Does This Paper Address?

Modern integrated circuits are becoming increasingly complex due to aggressive scaling of transistor geometries. This miniaturization, while boosting performance and integration density, introduces significant challenges in terms of testability and power consumption. New failure mechanisms emerge, and traditional testing methods struggle to keep up. Specifically, detecting faults becomes harder due to increased leakage currents, and test power itself becomes a critical concern for battery-powered devices. Furthermore, ensuring adequate test coverage in the face of these complexities is a major hurdle. This paper addresses the need for a design methodology that inherently improves testability and reduces power consumption, rather than relying solely on post-design testability enhancement techniques.

Methods

The core of the proposed method is Shannon Decomposition-Based Supply-gating synthesis (SBS). Shannon expansion is a technique used in logic synthesis to break down a Boolean expression into simpler, disjoint subexpressions. The paper leverages this by expressing a function $f(x_1, ..., x_n)$ as $x_i \cdot CF_1 + \bar{x}_i \cdot CF_2$, where $CF_1$ and $CF_2$ are cofactors. The key insight is that at any given time, only one cofactor is actively computing, while the other is redundant.

The SBS approach then applies dynamic supply gating to the idle cofactor. This means the power supply to the inactive part of the circuit is temporarily switched off, significantly reducing both dynamic (switching) and static (leakage) power consumption. The selection of the "control variable" ($x_i$) for expansion is crucial and is done to balance the resulting cofactors, ensuring similar static current for any test vector. This structured approach inherently leads to a more testable circuit.

Datasets

The authors evaluated their approach on a set of MCNC benchmark circuits. These are standard, publicly available benchmark suites commonly used in the electronic design automation community for evaluating synthesis and testing methodologies. No new benchmark circuits were introduced in this paper.

Key Insights and Ideas

  • Inherent Testability: The tree-like structure resulting from Shannon decomposition naturally makes circuits more testable. This is because control variables can be isolated and connected to select lines of multiplexers near primary outputs, improving the observability of internal nodes.
  • Power Reduction Through Gating: Dynamic supply gating of idle cofactors is a highly effective way to reduce both active and leakage power. This is a direct consequence of the structured decomposition.
  • Coupling Noise Immunity: The partitioning of the circuit into cofactors, with idle ones being powered down, effectively screens the active cofactor from a significant portion of interwire coupling noise, leading to improved noise immunity.
  • Parametric Yield Improvement: The structured design and reduced leakage contribute to an improvement in parametric yield, making the circuit more robust to process variations.

Key Results

The SBS technique demonstrated significant improvements across several key metrics:

  • IDDQ Sensitivity: An average improvement of 94.4% was observed. When combined with reverse body biasing, this improved to 37%.
  • Test Power: An average improvement of 50.5% in test power was achieved.
  • Test Length: An average reduction of 19.7% for deterministic test patterns and 21.1% for random patterns was reported.
  • Noise Immunity: An average improvement of 50% in coupling noise immunity was achieved.
  • Parametric Yield: An average improvement of 6.25% in parametric yield was observed.

Limitations and Future Work

The authors acknowledge several limitations:

  • Synthesis Flow Modification: The Shannon expansion approach requires modifications to the standard synthesis flow.
  • Area and Delay Overhead: For multi-output circuits, the expansion might lead to increased area due to added multiplexers, and potentially increased delay.
  • Careful Selection: Careful selection of control variables and the number of expansion levels is necessary to minimize area and delay overhead.

Future work could involve combining SBS with existing Design-for-Testability (DFT) techniques for even greater improvements in test length and power. The authors also suggest further exploration into optimizing the control variable selection and expansion strategies.

Summary updated

BibTeX
@article{ghosh2007low,
  author = {Swaroop Ghosh and Swarup Bhunia and Kaushik Roy},
  title = {{Low-Power and testable circuit synthesis using Shannon decomposition}},
  journal = {ACM Trans. Design Autom. Electr. Syst.},
  year = {2007},
  doi = {10.1145/1278349.1278360},
}

Loading related content...

Figure