CoChat

Direct FFT oversampling without zero-padding

Open access
Donciu, Codrin, Temneanu, Marinel Costel, Serea, Elena · 2025 · Scientific Reports

The DOI matches its Crossref record; 20 of the 20 references it cites were found in a registry or an index.

Abstract

The Fast Fourier Transform (FFT) is the most powerful and widely used method for transforming signals from the time domain to the frequency domain, being employed in cutting-edge research fields. Achieving high-accuracy FFT involves increasing the number of frequency points through frequency oversampling, typically by extending the time-domain signal with zeros (zero-padding) and applying the FFT to the extended signal. While increasing the number of zeros enhances the precision of the results, it also substantially raises the computational load. To address the high computational demands associated with traditional frequency oversampling, this paper proposes a direct oversampling method. By employing a non-integer number of cycles within the FFT kernel, the method achieves spectral analysis results identical to those obtained with zero-padded oversampling but with a significantly lower computational cost. Specifically, it reduces the complexity from O(Mlog 2 M) to O(Mlog 2 N), where N is the original signal length and M is the oversampled frequency resolution. This approach eliminates zero-padding altogether, providing a mathematically equivalent and more efficient alternative for applications requiring high spectral resolution.

Overview

CoChat Summary

TL;DR

The paper proposes FIN-FFT, a direct frequency-oversampling method that evaluates the DFT at non-integer frequency indices, avoiding zero-padding while producing the same spectrum as a zero-padded FFT. It reports reduced complexity from O(M log₂ M) to O(M log₂ N) and demonstrates mathematical equivalence, with initial FPGA synthesis benchmarking.

Key findings

  • FIN-FFT computes M frequency bins directly from an N-point sequence by repeating FFTs with fractional twiddle factors, without appending M−N zeros; the paper proves this is mathematically equivalent to zero-padded FFT (Eq. 11).
  • The computational complexity is reduced from O(M log₂ M) to O(M log₂ N), with the gain growing with the oversampling ratio M/N; for M ≫ N this is described as a significant efficiency advantage.
  • In the worked example, N = 128 and M = 256 (oversampling rate 2, step 0.5): zero-padding would add M−N = 128 zeros, while FIN-FFT applies the FFT twice to the same 128-point sequence, first with r = 0 and then with r = 0.5.
  • Table 1 positions FIN-FFT as full-spectrum, non-integer-bin, preprocessing-free, O(M log N), and real-time ready, unlike zero-padded FFT, IpDFT, Zoom FFT, Chirp-Z, and pruned FFT; Fig. 7 reports percentage reduction in operation count for N = 16, 128, 1024, and 16,384, increasing with oversampling rate.

Methods

  • Theoretical derivation: the DFT is extended to non-integer arguments, and the paper proves equivalence between a zero-padded FFT of length M = s·N and FIN-FFT evaluated at fractional index (k + r).
  • Algorithm design: standard FFT structure is modified with an additional outer loop executed M/N times, replacing the kernel argument k with (k + r), where r starts at 0 and increments by the oversampling step N/M; outputs are interleaved to produce an M-length oversampled spectrum.
  • Evaluation: an illustrative N = 128, M = 256 example, complexity analysis, a qualitative feature comparison table, and mention of initial FPGA synthesis benchmarking; no real-world dataset or quantitative spectral-error experiment is described.

Limitations

  • Stated: the method is formulated for integer oversampling factors, which may limit flexibility for arbitrary frequency resolution; future work is proposed for rational or adaptive oversampling.
  • Stated: a systematic analysis of numerical stability under finite-precision conditions, especially for high-resolution or real-time implementations, remains to be done.
  • Stated: benchmarking has so far been conducted through FPGA synthesis only; evaluation on CPUs, GPUs, and embedded platforms is needed. Evident from the text: no quantitative spectral accuracy or error results are reported, and the comparison with existing methods is qualitative rather than benchmarked.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1038/s41598-025-21270-5

Cites

20 of 20 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Figures

Fig. 1 Analogy illustration of picket fence effect.
Fig. 1 Analogy illustration of picket fence effect.CC BY-NC-ND · PubMed Central
Fig. 2 Analogy illustration of picket fence effect for a zero-padded.
Fig. 2 Analogy illustration of picket fence effect for a zero-padded.CC BY-NC-ND · PubMed Central
Fig. 3 Comparative illustration between classical FFT (a), zero-padding FFT (b) and the proposed method FIN-FFT (c).
Fig. 3 Comparative illustration between classical FFT (a), zero-padding FFT (b) and the proposed method FIN-FFT (c).CC BY-NC-ND · PubMed Central
Fig. 4 Pseudocode and block diagram of the FFT algorithm.
Fig. 4 Pseudocode and block diagram of the FFT algorithm.CC BY-NC-ND · PubMed Central
Fig. 5 FIN-FFT signal flow graph implementation for N = 128, with oversampling factor of 2.
Fig. 5 FIN-FFT signal flow graph implementation for N = 128, with oversampling factor of 2.CC BY-NC-ND · PubMed Central
Fig. 6 Pseudocode and block diagram of the FIN-FFT algorithm, showing repeated FFTs with fractional frequency shifts and interleaved output to achieve oversampled spectral analysis.
Fig. 6 Pseudocode and block diagram of the FIN-FFT algorithm, showing repeated FFTs with fractional frequency shifts and interleaved output to achieve oversampled spectral analysis.CC BY-NC-ND · PubMed Central
Fig. 7 Percentage drop in computation across increasing signal’s size for proposed method against zero-padded.
Fig. 7 Percentage drop in computation across increasing signal’s size for proposed method against zero-padded.CC BY-NC-ND · PubMed Central

Verification

Claims the paper makes

0 Valid · 0 Not valid · 0 Not checked · 19 Own results and interpretations

8 more claims

Create a free account to see all 19 claims and every source verdict.

See all claims — free account
  • It may state two separate claims as one. 5
  • Probably not a substantial claim. 1

Create a free account to see these claims.

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

Each contradiction between two passages listed here was confirmed by an independent check.

1 of 1

Citations

Cites

Show all 20 references

Cited by

See all 3

Metadata

Authors
Donciu, C., Temneanu, M. C., & Serea, E.
Affiliations
Gheorghe Asachi Technical University of Iași (RO)
Corresponding author
Serea, Elena
Venue
Scientific Reports
Published
2025-10-24
Volume
15
Issue
1
Pages
37269-37269
ISSN
2045-2322 (ISSN-L)
OpenAlex
W4415502789
Cited by
3
References
20
Citations by year
2026: 3
FWCI
1.39
Type
journal article
Language
en
Open access
Licence
Locations
  • Scientific Reports (publishedVersion, cc-by-nc-nd)
  • Europe PMC (PubMed Central) (publishedVersion, cc-by-nc-nd)
  • PubMed (publishedVersion)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
Topics
  • Physical Sciences › Computer Science › Signal Processing › Digital Filter Design and Implementation
  • Physical Sciences › Engineering › Electrical and Electronic Engineering › Advanced Electrical Measurement Techniques
  • Physical Sciences › Mathematics › Applied Mathematics › Mathematical Analysis and Transform Methods
Keywords
fast Fourier transform, zero-padding, computational complexity reduction, spectral resolution, frequency-domain signal analysis

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

A language model drew the claims from the paper on 2026-10-10, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-10-10 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

How to cite

Cite CoChat for the checks on this page, and the paper for its own findings.

This page
CoChat. (2026). Citation check: Direct FFT oversampling without zero-padding. https://app.cochat.ai/papers/10.1038/s41598-025-21270-5
The paper
Donciu, C., Temneanu, M. C., & Serea, E. (2025). Direct FFT oversampling without zero-padding. Scientific Reports. https://doi.org/10.1038/s41598-025-21270-5

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.