A Cross-Layer Single-Cycle Approximate Fixed-Point Square Root Unit
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
This work presents a cross-layer modeling and validation methodology for a non-iterative, single-cycle, approximate fixed-point square root unit derived from a first-order binomial series approximation. The proposed unit is a fully combinational datapath comprising a multiplier, a leading-one detector, a onehot encoder, and a linear correction block, entirely described in structural VHDL. The design targets a 65 nm standardcell technology at 1.25 V, characterized in silicon area, power consumption, critical path delay, and power-delay product for 8-, 16-, and 32-bit configurations. A high-level Python reference model enables exhaustive functional verification and provides a comprehensive statistical analysis of approximation errors over the complete input space. The 16 -bit implementation achieves deterministic single-cycle latency with a mean relative error of 9.68%, yielding a latency reduction exceeding 91% compared to sequential non-restoring algorithms and a 71% reduction in critical-path delay relative to exact combinational SD2 arrays. These results demonstrate that the proposed architecture offers a compelling trade-off between accuracy and performance, establishing a structurally energy-efficient and straightforward solution for error-resilient applications.
Fields of Study & Subjects
Low-power high-performance VLSI design
Numerical Methods and Algorithms
VLSI and FPGA Design Techniques
Publication details
- Journal / source
- IEEE Latin American Symposium on Circuits and Systems (LASCAS)
- Pages
- 1–5
- Publication date
- 24 February 2026
- Language
- EN