Thursday, November 19, 2026
Science
No Result
View All Result
  • Login
  • HOME
  • SCIENCE NEWS
  • CONTACT US
  • HOME
  • SCIENCE NEWS
  • CONTACT US
No Result
View All Result
Scienmag
No Result
View All Result
Home Science News Technology and Engineering

Open-Source 130-Nanometer CMOS Processor Cuts Fast Fourier Transform Power Consumption

November 19, 2026
in Technology and Engineering
Denise Maddox
By Denise Maddox Scienmag Editorial Profile - Mechanical Engineering
Reading Time: 5 mins read
0
Open-Source 130-Nanometer CMOS Processor Cuts Fast Fourier Transform Power Consumption

Open-Source 130-Nanometer CMOS Processor Cuts Fast Fourier Transform Power Consumption

65
SHARES
587
VIEWS
Share on FacebookShare on Twitter
ADVERTISEMENT

Engineers have designed, physically implemented, and verified a low-power 256-point Radix-2 Decimation-in-Frequency Fast Fourier Transform processor using the open-source SkyWater 130-nanometer CMOS fabrication process. Targeting energy-constrained edge digital signal processing workloads such as Internet of Things sensor nodes, embedded spectral analyzers, and biomedical monitoring devices, the hardware accomplishes the complete 256-point computation through an unscaled 16-bit datapath across eight butterfly execution stages. Post-layout power analysis annotated with value change dump switching data indicates that the integrated architecture operates at an estimated 11.20 milliwatts when running at a target clock frequency of 20 megahertz under a 1.8-volt supply. The full physical layout was completed through automated place-and-route tools, yielding zero design rule and layout-versus-schematic violations on a 0.646-square-millimeter silicon die.

The mathematical operation of the Fast Fourier Transform serves as an essential building block across wireless communications, audio processing, biomedical sensor filtering, and Orthogonal Frequency Division Multiplexing transceivers. In battery-driven and continuous-monitoring embedded platforms, the computational hardware must preserve signal fidelity while minimizing dynamic power and physical silicon area. Many conventional implementations rely on streaming pipelines such as single-path or multi-path delay commutator topologies to maximize data throughput. While effective for broadband communications, continuously active butterfly processing units and pipelined memory stages toggle internal signal nodes constantly, creating substantial dynamic power dissipation even when the application requires only moderate throughput. To resolve this imbalance, the investigators selected a sequential, time-multiplexed single processing element architecture that minimizes active silicon while exposing opportunities to eliminate unnecessary internal switching activity.

Dynamic power consumption in CMOS digital logic is fundamentally governed by the product of the circuit switching activity factor, the effective load capacitance, the operating frequency, and the square of the supply voltage. Because operating frequency and voltage rails remain constant for a given operating mode, reducing the effective switching activity factor represents the primary mechanism for lowering operational energy. The researchers incorporated seven register-transfer-level low-power mechanisms to suppress unneeded transitions across arithmetic datapaths, memory interfaces, coefficient lookup tables, and clock networks. These mechanisms consist of adder and subtractor operand isolation, trivial twiddle factor bypassing, complex multiplier operand isolation, dual-bank register array clock-enable gating, integrated clock gating, twiddle read-only memory latch gating, and Gray-code state machine encoding, implemented without altering the functional computation of the transform.

Within the arithmetic datapath, the butterfly execution unit incorporates 17-bit adder and subtractor circuits to calculate sum and difference terms. These arithmetic structures are actively needed only during specific computation and write-back sub-cycles, meaning that permitting input signals to toggle during inactive cycles propagates redundant transitions through carry chains and internal logic gates. By introducing enable-controlled masking logic at the adder and subtractor inputs, operands are forced to zero whenever the control state machine is not performing active butterfly computations. Similarly, twiddle multiplications involving zero-degree and ninety-degree rotations represent trivial mathematical operations. The control logic detects these specific rotational indices, routing inputs directly to outputs through multiplexers, sign inversion, or coordinate swapping. This direct bypass circumvents the full 16-by-12-bit complex multiplier, while complementary multiplier operand isolation forces multiplier inputs to zero during idle intervals and trivial rotations.

Memory and clock network switching typically account for the dominant portion of sequential dynamic power dissipation in digital signal processors. In this architecture, internal storage is structured as two independent 256-word by 16-bit flip-flop-based register arrays that separate real and imaginary signal components. High-density flip-flop cells were selected instead of rigid foundry static random-access memory hard macros, avoiding macro-padding overhead, inflexible power-ring routing constraints, and floorplan congestion in a modest 256-word footprint. To limit dynamic toggling, the memory control logic asserts a chip-enable signal strictly during valid idle synchronization, sample loading, butterfly computation, and output unloading cycles. When this chip enable de-asserts, the register-file banks remain inactive, latching prior values to prevent downstream switching. Integrated clock-gating cells featuring latch-based enable paths disable clock distributions across inactive pipeline registers, twiddle registers, and multiplier write-back stages, while Gray-code encoding ensures only single-bit Hamming distance changes during primary control state transitions.

The physical processing core executes 1024 total butterfly operations across eight computation stages using an iterative, in-place architecture. To eliminate dual-port memory read-write access collisions and bypass complex data-forwarding networks, the processing element operates under a deterministic, non-overlapped five-cycle pipeline schedule. During the first cycle, the processing element reads the initial operand from the register array, followed by the second operand read in cycle two. Cycle three executes twiddle factor acquisition and complex multiplication, cycle four carries out 17-bit butterfly addition and subtraction, and cycle five writes the updated results back into the register storage. This strictly ordered schedule enforces a predictable execution latency of 5,634 clock cycles per complete 256-point frame, which corresponds to 281.7 microseconds at 20 megahertz, yielding a throughput of approximately 3,550 transforms per second.

To compress the silicon area required for coefficient generation, the design incorporates a quarter-wave twiddle read-only memory. Rather than storing the full set of 128 unique complex coefficient values, the design exploits cosine and sine trigonometric quadrant symmetries. Only 65 discrete cosine values, spanning angular indices from 0 to 64, are stored in a 12-bit Q0.11 fixed-point numerical format, which compresses the required storage area by 75 percent. Compact quadrant-selection logic comprising basic standard-cell exclusive-OR gates and inverters extracts real and imaginary twiddle factors across all 128 angles. Following complex multiplication, the 28-bit full-precision products are scaled back to the 16-bit format using sign-aware rounding with an arithmetic shift of 11 bits. To prevent catastrophic integer rollover and numerical wrapping during high-amplitude transient peaks, hard saturation logic bounds the intermediate 17-bit butterfly results to maximum and minimum 16-bit signed integer limits.

Physical implementation of the processor was completed using the OpenLane version 1 open-source register-transfer-level-to-GDSII flow and the SkyWater 130-nanometer high-density cell library. Logic synthesis was executed using Yosys targeting an initial 50-nanosecond clock period constraint, producing a gate-level representation of 37,778 standard cells, including 8,396 sequential flip-flops, 11,121 clock and signal buffers, and 10,456 multiplexers. Floorplanning established a core utilization of 52 percent across an 823.8-micrometer by 823.8-micrometer die footprint, resulting in a post-placement-and-routing design area of 0.646 square millimeters. Global and detailed placement and routing were handled by Replace, OpenDP, FastRoute, and TritonRoute engines, while twenty detected antenna rule warnings were fully eliminated through automated antenna diode insertion. Physical sign-off verification using Magic and Netgen confirmed zero design rule and layout-versus-schematic violations.

Static timing analysis conducted with the OpenSTA engine at the typical-typical process corner of 1.8 volts and 25 degrees Celsius established positive timing closure at 20 megahertz. The worst-case data path arrival time measured 19.34 nanoseconds against the 50.00-nanosecond clock period, which combines with a library flip-flop setup requirement of 0.58 nanoseconds to produce a positive setup slack of 30.08 nanoseconds and a hold slack of 0.44 nanoseconds. From this constrained critical path, the maximum calculated theoretical frequency reaches 50.2 megahertz. An ablation analysis evaluated the quantitative impact of progressively enabling the low-power techniques under identical 20-megahertz and 1.8-volt operating conditions, revealing that sequential clock gating and memory-enable controls yielded the largest individual reductions, while the complete seven-technique suite reduced overall power consumption by 39.46 percent relative to the unoptimized baseline.

Functional correctness and signal fidelity were verified through gate-level simulation with standard delay format back-annotation against a double-precision NumPy software model requantized to fixed-point format. Testing with multitone sinusoidal inputs yielded a signal-to-noise ratio of 58.1 decibels and an effective number of bits of 9.38 bits. Gate-level activity extraction generated a measured global switching activity factor of 0.0551, representing roughly 11 percent of the theoretical worst-case toggle rate. Post-layout power analysis estimated total energy dissipation at 3.155 microjoules per 256-point transform, translating to 12,324 picojoules per transform point and yielding a normalized figure of merit of 140.1 picojoules per point squared. The authors noted that these metrics reflect post-routing electronic design automation estimations rather than physically probed silicon test chips, and while the sequential single-element architecture intentionally trades raw execution throughput for minimal footprint and low power, it delivers verified layout closure suited for open-source integrated circuit workflows.

Subject of Research: Low-power digital signal processor design in open-source 130-nanometer CMOS

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

Article References: Peraka, S., Pusuluri, V. B., Vulchu, T. V., Badam, S. R., & Veeranki, Y. R. (2026). A low-power FFT processor with multi-level switching-activity reduction and quarter-wave twiddle compression in SKY130 CMOS. Discover Informatics, 1(1), Article 41. https://doi.org/10.1007/s44564-026-00030-1

Image Credits: AI Generated

DOI: 10.1007/s44564-026-00030-1

Keywords: Fast Fourier Transform, SkyWater 130 nm, OpenLane, CMOS processor, low power electronics, digital signal processing, switching activity reduction, decimation in frequency, pipelined architecture, quarter wave twiddle, integrated clock gating, operand isolation, static timing analysis, VLSI design, edge computing, application specific integrated circuit

Cite Scienmag News

Denise Maddox. (November 19, 2026). Open-Source 130-Nanometer CMOS Processor Cuts Fast Fourier Transform Power Consumption. Scienmag. https://scienmag.com/open-source-130-nanometer-cmos-processor-cuts-fast-fourier-transform-power-consumption/

Denise Maddox. "Open-Source 130-Nanometer CMOS Processor Cuts Fast Fourier Transform Power Consumption." Scienmag, 19 November 2026, https://scienmag.com/open-source-130-nanometer-cmos-processor-cuts-fast-fourier-transform-power-consumption/. Accessed 19 November 2026.

Denise Maddox. "Open-Source 130-Nanometer CMOS Processor Cuts Fast Fourier Transform Power Consumption." Scienmag. November 19, 2026. https://scienmag.com/open-source-130-nanometer-cmos-processor-cuts-fast-fourier-transform-power-consumption/

Tags: application specific integrated circuitCMOS processordecimation in frequencydigital signal processingedge computingFast Fourier Transformintegrated clock gatinglow-power electronicsOpenLaneoperand isolationpipelined architecturequarter wave twiddleSkyWater 130 nmstatic timing analysisswitching activity reductionVLSI design
Share26Tweet16
Previous Post

Snake Fruit Peel and Banana Corm Produce High-Energy Biopellets Through Carbonization

Next Post

Long-Read Sequencing Reveals Dynamic mRNA Splicing Shifts During Retinal Organoid Development

Related Posts

Hilfer Fractional Stochastic Systems Achieve Approximate Controllability via Multivalued Control Maps
Technology and Engineering

Hilfer Fractional Stochastic Systems Achieve Approximate Controllability via Multivalued Control Maps

November 19, 2026
Survey Highlights Institutional and Attitudinal Obstacles to University AI Adoption in Jordan
Technology and Engineering

Survey Highlights Institutional and Attitudinal Obstacles to University AI Adoption in Jordan

November 19, 2026
Fabric Structure and Mechanical Rigidity Drive Microplastic Fiber Shedding During Laundering
Technology and Engineering

Fabric Structure and Mechanical Rigidity Drive Microplastic Fiber Shedding During Laundering

November 19, 2026
Glucosamine Coated Silica Nanoparticles Show Multifunctional Delivery Potential for Motor Neuron Disease
Technology and Engineering

Glucosamine Coated Silica Nanoparticles Show Multifunctional Delivery Potential for Motor Neuron Disease

November 19, 2026
Diffusion Model Remodels Aging DNA Methylation Signatures While Preserving Individual Identity
Biotechnology

Diffusion Model Remodels Aging DNA Methylation Signatures While Preserving Individual Identity

November 19, 2026
Most Converted Swedish Buildings Keep Ventilation Types While Airflow Frequently Drops
Technology and Engineering

Most Converted Swedish Buildings Keep Ventilation Types While Airflow Frequently Drops

November 19, 2026
Next Post
Long-Read Sequencing Reveals Dynamic mRNA Splicing Shifts During Retinal Organoid Development

Long-Read Sequencing Reveals Dynamic mRNA Splicing Shifts During Retinal Organoid Development

  • Mothers who receive childcare support from maternal grandparents show more optimized

    Mothers who receive childcare support from maternal grandparents show more parental warmth, finds NTU Singapore study

    27656 shares
    Share 11059 Tweet 6912
  • University of Seville Breaks 120-Year-Old Mystery, Revises a Key Einstein Concept

    1061 shares
    Share 424 Tweet 265
  • Bee body mass, pathogens and local climate influence heat tolerance

    682 shares
    Share 273 Tweet 171
  • Researchers record first-ever images and data of a shark experiencing a boat strike

    546 shares
    Share 218 Tweet 137
  • Groundbreaking Clinical Trial Reveals Lubiprostone Enhances Kidney Function

    531 shares
    Share 212 Tweet 133
Science

Embark on a thrilling journey of discovery with Scienmag.com—your ultimate source for cutting-edge breakthroughs. Immerse yourself in a world where curiosity knows no limits and tomorrow’s possibilities become today’s reality!

RECENT NEWS

  • CRISPR Screening Identifies Essential RNA-Binding Proteins and Prognostic Signature in AML
  • Review Documents Twenty-One Native Mosquito Species in Ireland Amid European Disease Shifts
  • Survey Shows Public Awareness Remains Low as U.S. Mpox Cases Persist
  • Optimized Dung Beetle Clustering and Adaptive Fusion Extend Sensor Network Lifespan

Categories

  • Agriculture
  • Anthropology
  • Archaeology
  • Athmospheric
  • Biology
  • Biotechnology
  • Blog
  • Bussines
  • Cancer
  • Chemistry
  • Climate
  • Earth Science
  • Editorial Policy
  • Marine
  • Mathematics
  • Medicine
  • Pediatry
  • Policy
  • Psychology & Psychiatry
  • Science Education
  • Science News
  • Social Science
  • Space
  • Technology and Engineering

Subscribe to Blog via Email

Enter your email address to subscribe to this blog and receive notifications of new posts by email.

Join 5,151 other subscribers

© 2026 Scienmag - Science Magazine

Welcome Back!

Login to your account below

Forgotten Password?

Retrieve your password

Please enter your username or email address to reset your password.

Log In
No Result
View All Result
  • HOME
  • SCIENCE NEWS
  • CONTACT US

© 2026 Scienmag - Science Magazine

Discover more from Science

Subscribe now to keep reading and get access to the full archive.

Continue reading