A low-power FFT processor with multi-level switching-activity reduction and quarter-wave twiddle compression in SKY130 CMOS

This paper presents the design, implementation, and evaluation of a low-power 256-point Radix-2 Decimation-in-Frequency (DIF) Fast Fourier Transform (FFT) processor for edge digital signal processing (DSP) applications. The proposed processor is implemented using the SkyWater 130 nm open-source CMOS process with the sky130_fd_sc_hd standard-cell library and the OpenLane v1 RTL-to-GDSII physical design flow. The architecture targets energy-constrained systems, such as Internet of Things (IoT) sensor nodes, embedded spectral analyzers, biomedical monitoring devices, and low-frequency edge DSP accelerators. The processor performs the 256-point FFT computation using a 16-bit unscaled datapath across eight butterfly stages. It employs a single time-multiplexed Processing Element (PE) featuring a 5-cycle pipeline, which reuses a single complex multiplier across all 1024 butterfly operations. A 12-bit quarter-wave twiddle ROM exploits cosine/sine symmetry to derive all 128 twiddle angles from only 65 stored coefficients, reducing the ROM area by 75%. The 16×12-bit complex multiplier features sign-aware rounding and trivial twiddle bypass for the W0 and W64 rotations. Seven RTL-level low-power techniques are co-integrated: adder/subtractor operand isolation, trivial twiddle bypass, complex multiplier operand isolation, Register Array clock-enable gating on dual independent memory banks, integrated clock gating, twiddle ROM latch gating, and Gray-code FSM encoding. These techniques effectively reduce redundant signal transitions without modifying the underlying FFT algorithm. VCD-annotated switching analysis using OpenSTA reports a true global switching activity factor of $$\alpha$$ = 0.0551, yielding a post-layout estimated power of 11.20 mW. The complete design is physically implemented and achieves zero DRC and LVS violations, with all antenna warnings fully mitigated via automated diode insertion. Timing closure is achieved at 20 MHz with a setup slack of + 30.08 ns and a hold slack of + 0.44 ns (WNS = 0.0 ns). The physical layout occupies 0.646 mm2 and contains 37,778 standard cells. Functional verification against a NumPy double-precision reference confirms a Signal-to-Noise Ratio (SNR) of 58.1 dB and an Effective Number of Bits (ENOB) of 9.38 bits. The total energy per 256-point FFT frame is 3.155 µJ, resulting in a highly competitive Figure of Merit (FoM) of 140.1 pJ/point2.

Authors

Institutions

Publication Details

Journal
Discover Informatics
Published
2026-11-19
DOI
https://doi.org/10.1007/s44564-026-00030-1
Primary Topic
Low-power high-performance VLSI design
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

A low-power FFT processor with multi-level switching-activity reduction and quarter-wave twiddle compression in SKY130 CMOS

Vinod Babu Pusuluri, Shyam Peraka, Srinivasa Reddy Badam, Tejesh Varma Vulchu
Discover Informatics
Low-power high-performance VLSI design
article

A low-power FFT processor with multi-level switching-activity reduction and quarter-wave twiddle compression in SKY130 CMOS

Vinod Babu Pusuluri, Shyam Peraka, Srinivasa Reddy Badam, Tejesh Varma Vulchu
article en

Abstract

This paper presents the design, implementation, and evaluation of a low-power 256-point Radix-2 Decimation-in-Frequency (DIF) Fast Fourier Transform (FFT) processor for edge digital signal processing (DSP) applications. The proposed processor is implemented using the SkyWater 130 nm open-source CMOS process with the sky130_fd_sc_hd standard-cell library and the OpenLane v1 RTL-to-GDSII physical design flow. The architecture targets energy-constrained systems, such as Internet of Things (IoT) sensor nodes, embedded spectral analyzers, biomedical monitoring devices, and low-frequency edge DSP accelerators. The processor performs the 256-point FFT computation using a 16-bit unscaled datapath across eight butterfly stages. It employs a single time-multiplexed Processing Element (PE) featuring a 5-cycle pipeline, which reuses a single complex multiplier across all 1024 butterfly operations. A 12-bit quarter-wave twiddle ROM exploits cosine/sine symmetry to derive all 128 twiddle angles from only 65 stored coefficients, reducing the ROM area by 75%. The 16×12-bit complex multiplier features sign-aware rounding and trivial twiddle bypass for the W0 and W64 rotations. Seven RTL-level low-power techniques are co-integrated: adder/subtractor operand isolation, trivial twiddle bypass, complex multiplier operand isolation, Register Array clock-enable gating on dual independent memory banks, integrated clock gating, twiddle ROM latch gating, and Gray-code FSM encoding. These techniques effectively reduce redundant signal transitions without modifying the underlying FFT algorithm. VCD-annotated switching analysis using OpenSTA reports a true global switching activity factor of $$\alpha$$ = 0.0551, yielding a post-layout estimated power of 11.20 mW. The complete design is physically implemented and achieves zero DRC and LVS violations, with all antenna warnings fully mitigated via automated diode insertion. Timing closure is achieved at 20 MHz with a setup slack of + 30.08 ns and a hold slack of + 0.44 ns (WNS = 0.0 ns). The physical layout occupies 0.646 mm2 and contains 37,778 standard cells. Functional verification against a NumPy double-precision reference confirms a Signal-to-Noise Ratio (SNR) of 58.1 dB and an Effective Number of Bits (ENOB) of 9.38 bits. The total energy per 256-point FFT frame is 3.155 µJ, resulting in a highly competitive Figure of Merit (FoM) of 140.1 pJ/point2.

Discover InformaticsVol. 1(1)
National Institute of Technology, Puducherry (IN), Rajiv Gandhi University of Knowledge Technologies (IN)
Industry, innovation and infrastructure
Low-power high-performance VLSI design
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon