Skip to content

Oscillator linewidth

The frequency spread of a selected physical oscillator output line, reported with a specified spectral estimator and width convention.

Version
v1 · 2026-10-25 · History
Domain-specific #
14427
Domain group
Natural Sciences
Origin domain
Physics
Subdomains
Oscillators, Spectral Metrology → Physics

Core Idea

Oscillator linewidth is the frequency spread assigned to a selected output line of a physical oscillator. A useful linewidth report says which line was studied, how its spectrum was measured or inferred, and what width statistic was used. A fitted Lorentzian component, a full observed spectral peak, and a width inferred from a frequency-noise model are related measurements, but they are not automatically the same number. In an integrated laser, Li and colleagues reported a 15.0 kHz fitted Lorentzian full width and a separate 11.3 kHz intrinsic Lorentzian linewidth inferred from a measured noise floor.[1] In a spin-torque oscillator, Keller and colleagues found that a width inferred from phase noise was about 70% of the spectrum-analyzer width in two measured modes.[2]

The identity is a property of the oscillator's selected output line under a reporting convention, not a name for one noise mechanism. Under a white-frequency-noise phase-diffusion model, a Lorentzian full width can be calculated from a frequency-noise density; Keller and colleagues explicitly state the one-sided density convention for their conversion.[2] For their other studied spin-torque devices, nonwhite low-offset frequency noise made spectral width depend weakly on measurement time.[3] These results explain why line shape, estimator, and material observation conditions belong with the value.

Structural Signature

Sig role-phrases: physical oscillator; selected output line; spectral representation; width statistic and model; material measurement conditions.

  • Physical oscillator. A laser, electrical oscillator, or magnetic auto-oscillator supplies a recurring physical signal. Li's integrated laser and Keller's spin-torque nanocontact are unlike carriers of this role.[1][2]
  • Selected output line. The claim must identify which emission or voltage mode is being measured. Keller reports separate 142.5 and 128 MHz modes; their width values are not interchangeable.[2] A broad spectrum without an identified oscillator output line does not fill this role.
  • Spectral representation. A measured peak or a justified phase/frequency-noise-to-spectrum calculation makes a width available. Li uses delayed self-heterodyne fitting and, separately, correlated phase-noise analysis; Keller compares zero-crossing phase analysis with a spectrum analyzer.[1][2] An unconverted phase-noise curve is evidence about noise, not yet a reported line width.
  • Width statistic and model. The report identifies FWHM or HWHM, the line shape or fitted component, and whether the value describes the observed peak or a modeled contribution. Li's figure labels its fitted Lorentzian full width as \(2\gamma=15.0\,\mathrm{kHz}\), with \(\gamma\) the Lorentzian HWHM.[1] Keller calls its phase-derived and analyzer results FWHM.[2]
  • Material measurement conditions. Apparatus and time or resolution conditions are qualified where they change interpretation. Li uses different delay lines for its two procedures; those lengths are not themselves a common linewidth observation interval.[1] Keller's later nonwhite-noise study shows a case in which measurement time changes the reported width.[3] This is a conditional qualifier, not a universal fifth prerequisite.

What It Is Not

  • Not phase-noise density itself. Phase and frequency fluctuations can support a line-shape inference under a declared model. Keller's phase-only inferred FWHM remains below its analyzer FWHM in both measured modes.[2]
  • Not necessarily one Lorentzian width. A measurement may be described with more than one line-shape component; Li fits the recorded data with a combination of Lorentzian and Gaussian distributions without specifying how the components are combined.[1] Nonwhite noise can make a measured width depend on observation time in the studied spin-torque devices.[3]
  • Not HWHM and FWHM at once. For a Lorentzian proportional to \(\gamma/[\gamma^2+(f-f_0)^2]\), half maximum occurs at \(f_0\pm\gamma\), so its FWHM is \(2\gamma\). The convention must be named. Li states both \(\gamma\) and \(2\gamma\) explicitly.[1]
  • Not a universal diagnosis of phase noise alone. Keller attributes roughly 70% of the measured width to its phase-noise inference and presumes an amplitude-noise contribution for the remainder; the paper also says amplifier noise masks intrinsic amplitude variations in its waveform record.[2]
  • Not every spectral-line width. An atomic transition line without an oscillator output may have a width, but it lacks the carrier that defines this entry. Nor does a mere list of harmonic frequencies specify the widths of those lines.

Scope of Application

The pattern applies literally to physical oscillators whose selected output line can be examined in frequency. Its two fully mapped settings here are an integrated III-V/lithium-niobate optical laser and a voltage-output spin-torque nanocontact in vortex mode.[1][2] Optical emission and electrical voltage require different instruments and physical accounts, yet each supplies a recurring signal, a selected line, and a qualified width.

The entry does not prescribe one line shape for all oscillators. For white-frequency-noise diffusion in Keller's stated one-sided PSD convention, the authors give Δν = πSν for the phase-noise contribution to FWHM after neglecting amplitude noise.[2] That equation belongs to those assumptions. Li's separate 15.0 and 11.3 kHz procedures should be reported as separate estimates under their own methods.[1] A bilateral normalized Lorentzian with units of inverse hertz is not already a single-sideband phase-noise value in dBc/Hz. The side convention, carrier normalization, and decibel conversion must be supplied before the quantities can be compared.

Clarity

The phrase “the oscillator linewidth” can hide at least three choices: which output line, which estimate, and which width rule. Li's 15.0 kHz number comes from a fitted Lorentzian component of a delayed self-heterodyne result; the 11.3 kHz number is inferred from a correlated phase/frequency-noise measurement and its white floor.[1] Calling both simply “the measured linewidth” erases a method difference. In Keller's case, 0.72 versus 1.05 MHz and 0.54 versus 0.78 MHz distinguish phase-only inference from analyzer FWHM for the two modes.[2]

A clear report keeps frequency offsets and line width separate from carrier frequency and from phase-noise units. A half-width parameter such as \(\gamma\) must be named as HWHM if that is the displayed Lorentzian form; FWHM is twice that parameter. A normalized bilateral line profile integrating to one is not automatically a single-sideband phase-noise ratio, nor is a linear per-hertz density already expressed in dBc/Hz.

Manages Complexity

Oscillators can differ in gain medium, nonlinear dynamics, noise source, detection chain, and environmental drift. The linewidth abstraction compresses that detail into a small reportable structure: selected line, spectral or noise-based estimator, line-shape model, width statistic in frequency units, and any material time or resolution condition. The structure permits useful comparison while requiring the qualifications that prevent false equality between different measurements.

It also separates property from cause. The spin-torque example has a spectrum-analyzer width greater than the phase-noise-only width; the difference prompts a further noise investigation, but the width value alone cannot identify its cause.[2] The 2010 spin-torque measurements add a distinct low-offset, nonwhite-noise condition that a single stationary white-noise number would miss.[3]

Abstract Reasoning

To use the abstraction, first select one oscillator output and one line. Next ask whether the proposed number comes from the line itself, a fitted component, or an inference from phase or frequency noise. If it is inferred, inspect the noise model and the one-sided or two-sided PSD convention. Then name HWHM or FWHM, the units, and any material measurement-time or resolution condition. Only after those checks compare widths or infer spectral coherence.

This sequence gives a diagnostic implication without overclaiming cause. If two valid procedures give different widths for the same device, the first question is whether they estimate the same component under the same convention. Li's two reported kHz figures do not by themselves prove an experimental contradiction; they arise from different procedures.[1] If a phase-only width falls below an analyzer width, as in Keller's modes, the gap calls for investigation of additional contributions and apparatus effects rather than a silent substitution of one value for the other.[2]

Knowledge Transfer

The reporting structure transfers literally from optical lasers to spin-torque voltage oscillators: in both, a selected recurring output has a spectral line whose width can be fitted or inferred, and the result needs a width convention.[1][2] The instruments and noise physics do not transfer unchanged. Keller explicitly notes that a classical circuit phase-noise model does not apply directly to the nonlinear spin-torque oscillator.[2]

“Width of a distribution” in a non-oscillator field is a broad analogy. It does not by itself carry an oscillator output line, phase/frequency-noise model, or the spectral reporting problem that constitutes this domain-specific identity.

Examples

Canonical: integrated Pockels laser

Li and colleagues study light from a hybrid integrated III-V/lithium-niobate laser.[1] They split and delay its output for self-heterodyne characterization and fit the recorded result with a combination of Lorentzian and Gaussian distributions.[1] Figure 2e gives a Lorentzian component with \(2\gamma=15.0\,\mathrm{kHz}\), where \(\gamma\) is HWHM.[1] A second, correlated phase-noise procedure yields a frequency-noise white floor near \(1.8\times10^3\,\mathrm{Hz^2/Hz}\) around 3 MHz, from which the authors infer an 11.3 kHz intrinsic Lorentzian linewidth.[1] The two numbers remain method-qualified estimates on one laser, not two independent instances.

Mapped back: physical oscillator → integrated laser; selected output line → its lasing emission; spectral representation → fitted delayed self-heterodyne result and separately noise-derived inference; width statistic/model → 15.0 kHz fitted Lorentzian FWHM versus 11.3 kHz inferred intrinsic linewidth; material conditions → the distinct delay-line procedures, without treating delay length as a shared observation interval.[1]

Applied / In Practice: spin-torque nanocontact

Keller and colleagues record voltage from a vortex-mode spin-torque oscillator.[2] Zero crossings of a four-microsecond waveform yield phase fluctuations, and the authors calculate a phase-noise-limited FWHM using their one-sided white-frequency-noise density.[2] For the 142.5 MHz mode they report \(0.72\pm0.10\,\mathrm{MHz}\) from phase noise and \(1.05\pm0.14\,\mathrm{MHz}\) on the analyzer.[2] For the 128 MHz mode the corresponding widths are \(0.54\pm0.08\) and \(0.78\pm0.11\,\mathrm{MHz}\).[2] They summarize the phase-noise contribution as about 70% of measured width.[2]

Mapped back: physical oscillator → spin-torque nanocontact; selected output line → either reported voltage mode; spectral representation → zero-crossing phase inference compared with spectrum-analyzer line; width statistic/model → phase-only FWHM versus full analyzer FWHM for each mode; material conditions → different waveform and analyzer acquisition times, without claiming they are one common linewidth interval.[2]

Structural Tensions

A phase-derived estimate and a directly measured spectrum answer related but different questions. They can be reported together, so choosing one does not force the other to be sacrificed. The practical issue is to state enough model, width, and measurement context for a reader to compare them without treating them as interchangeable.[1][2]

Structural–Framed Character

Spectrum placement: mostly structural within physical spectral metrology: the same roles can be recognized in a laser and a spin-torque oscillator. Evaluative weight: a narrower line may be desirable for a particular application, but narrowness is not a membership condition. Human-practice dependence: an investigator selects the line, procedure, line-shape fit, and width convention; the output spectrum is a physical observation, while its reported value is method-qualified. Institutional origin: conventions and instruments shape the reported number, but no single laboratory or vendor owns the identity. Vocabulary travel: “width” and “noise” travel broadly, whereas oscillator output, FWHM, PSD side convention, and spectral line have precise roles here. Import versus recognition: a case qualifies by its selected oscillator line and width estimator, not by borrowing the term. Portable skeleton: a target property measured with a method and unit resembles the broader Measurement Prime, but it loses the physical line and frequency-noise context if lifted wholesale. Its character: a measurable domain-specific property with a stable reporting structure and context-dependent value.

Structural Core vs. Domain Accent

A portable skeleton would select an attribute, apply an estimator, and report a statistic under a convention. That skeleton is a possible cross-domain abstraction for separate study, not a claim established here as a Prime. The Measurement Prime is related but has a fuller instrument–procedure–scale–unit–uncertainty chain; the three-part skeleton alone does not prove a strict parent relation. Oscillator linewidth retains a narrower carrier and mechanism: recurring physical output, a line in a frequency spectrum, possible phase/frequency diffusion, and width conventions such as FWHM and HWHM. The optical and spin-torque cases stay in oscillator physics even though their carriers differ.[1][2]

This is why the present entry remains domain-specific. A generic spread in survey responses or a broad resonance curve could reuse the word “width,” but it would not instantiate this oscillator output-line identity. A cross-domain Prime would require a different, independently evidenced abstraction rather than a metaphor based on similar vocabulary.

The Measurement Prime is a broad interpretive neighbor: linewidth reporting maps an oscillator output attribute to a frequency value through a procedure and width convention. Its fuller signature includes scale, instrument, unit, frame, and uncertainty. The resemblance is useful without equating the two identities. Resonance is a nearby Prime because resonators can shape oscillator dynamics, but response to a driving frequency is not identical to the emitted output-line width. Coherence Breakdown under External Interaction describes one broad route by which coherence can degrade; linewidth does not require that every cause be external. The Harmonic Spectrum entry concerns the placement and weighting of harmonic components, not the width of a chosen component.

Neighborhood in Abstraction Space

Oscillator linewidth sits in a sparse region of the domain-specific corpus (72nd percentile for distinctiveness): few abstractions share its structure, so a faithful description tends to retrieve it precisely.

Family — Unclustered & Miscellaneous (2769 abstractions)

Nearest neighbors

Computed from structural-signature embeddings · 2026-10-25

Not to Be Confused With

  • A bare phase-noise plot: it can support a linewidth inference only when the conversion model and PSD convention are justified.[2]
  • A line's carrier frequency: 128 MHz names one Keller oscillator mode; 0.78 MHz is its reported analyzer FWHM.[2]
  • A Lorentzian half-width parameter: the parameter \(\gamma\) in Li's fitted Lorentzian is half of the plotted \(2\gamma=15.0\,\mathrm{kHz}\) full width.[1]
  • A single-sideband phase-noise density: a bilateral normalized spectral shape cannot be turned into dBc/Hz by relabeling it. Side and logarithmic conversion conventions require their own derivation.
  • An atomic transition line or an arbitrary spectrum: both can have widths, but without a selected oscillator output they are outside this entry's bounded carrier.

References

[1] Mingxiao Li et al. “Integrated Pockels laser”. Nature Communications 13, 5344 (2022). Fig. 2d–f and caption, pp. 3–4, and “Laser linewidth measurement,” Methods, pp. 7–8, support the two qualified laser widths and methods; full author-hosted article available at the published PDF. registry ↩a ↩b ↩c ↩d ↩e ↩f ↩g ↩h ↩i ↩j ↩k ↩l ↩m ↩n ↩o ↩p ↩q ↩r ↩s ↩t

[2] Mark W. Keller, A. B. Kos, T. J. Silva, W. H. Rippard, and M. R. Pufall. “Time domain measurement of phase noise in a spin torque oscillator”. Applied Physics Letters 94, 193105 (2009). Original full author manuscript, pp. 1–3, especially Figs. 1, 3–4 and the FWHM comparison on p. 3, supports the one-sided model and two mode comparisons; the NIST published copy is also available. registry ↩a ↩b ↩c ↩d ↩e ↩f ↩g ↩h ↩i ↩j ↩k ↩l ↩m ↩n ↩o ↩p ↩q ↩r ↩s ↩t ↩u ↩v ↩w ↩x ↩y

[3] Mark W. Keller, Matthew R. Pufall, William H. Rippard, and Thomas J. Silva. “Non-white frequency noise in spin torque oscillators and its effect on spectral linewidth”. Physical Review B 82 (2010). The NIST author abstract supports the bounded measurement-time result; the full paper was not used for detailed equations or figures. registry ↩a ↩b ↩c ↩d