Static Timing Analysis of MAC Units Using Approximate Multipliers for Energy-Efficient Computing
Summary
No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.
Summarise with VIRAT Free to try with your Science Catalogs account.Abstract
Objectives: Multiply–Accumulate (MAC) units form the computational backbone of modern low-power digital signal processing (DSP) and machine learning accelerators. However, conventional exact MAC architectures suffer from high power consumption, increased delay, and large silicon area. The primary objective of this work is to analyze and quantify the timing, power, and area benefits of an approximate MAC architecture using static timing analysis (STA), while evaluating its impact on computational accuracy for error-tolerant applications. Method: An approximate MAC unit incorporating an approximate multiplier is designed and synthesized at the gate level. A conventional exact MAC architecture is implemented as a reference design. Both designs undergo static timing analysis (STA) using synthesized gate-level netlists. Key design metrics evaluated include critical path delay, setup and hold times, timing slack, maximum operating frequency, power consumption, silicon area, and error metrics. Computational accuracy is assessed using error rate and Mean Relative Error (MRE), enabling a comprehensive trade-off analysis between performance and accuracy. Findings: Experimental results demonstrate that the approximate MAC reduces critical path delay from 4.82 ns to 2.10 ns, achieving approximately 56% delay reduction and 129% improvement in maximum operating frequency. Power consumption decreases by about 45%, and silicon area is reduced by approximately 46.7%, resulting in significant PDP improvement. Despite these gains, the design maintains a computational accuracy of 93.8%, making it suitable for applications such as image processing, neural network inference, and multimedia systems.Timing slack improves and indicating superior timing closure and increased robustness against process variations. Despite these improvements, the approximate MAC maintains a high computational accuracy, achieving an accuracy score of 93.8%, which is suitable for error-tolerant workloads such as image processing, neural networks, and multimedia applications. Novelty: The novelty of this work lies in the comprehensive static timing analysis of an approximate MAC architecture, highlighting not only power and area benefits but also substantial improvements in timing slack and frequency scalability. Keywords: Static Timing Analysis (STA), Multiply–Accumulate (MAC) Unit, Approximate Computing, Approximate Multiplier, Digital Signal Processing (DSP)
Fields of Study & Subjects
Low-power high-performance VLSI design
VLSI and FPGA Design Techniques
Numerical Methods and Algorithms
Publication details
- Journal / source
- Indian Journal of Science and Technology
- Volume
- 19
- Issue
- 10
- Pages
- 725
- Publication date
- 26 March 2026
- Language
- EN