A new arXiv preprint reports a numerical route for making a wideband radar transmit pattern—the shape of modeled power across angle and frequency—match a chosen target more closely. In an illustrative multiple-input, multiple-output (MIMO) radar design, the optimized waveform set had a narrower transition from its main beam to sidelobes, lower sidelobe peaks and less ripple across the main beam than the starting set.
The finding comes from one analytical and numerical example, not from a radar trial. It does not show better target detection, localization, tracking, communications or field performance.
A waveform tuned to the target
The study tunes MTSFM waveforms—signals whose modulation coefficients can be adjusted to shape their spectral content. The coefficients are chosen so the frequency-dependent cross-spectral density matrix, or CSDM, a mathematical description of how the waveforms relate across frequency, approximates the desired wideband MIMO transmit beampattern. The design keeps the waveforms constant in magnitude and spectrally compact.
The wideband calculation approximates a continuous beampattern integral by adding modeled power densities over angle and a selected grid of discrete frequencies.
The illustrative setup used a uniform linear array with 10 sensors, 16 modulation indices per waveform and a time-bandwidth product of 64. Its original narrowband configuration had a fractional bandwidth of γ = 0.01. The target was a flat response within |u| ≤ 0.3, where u is the sine of the angle, and zero elsewhere; total power across u was normalized to the array dimensionality.
The objective was to minimize an MMSE mismatch, a numerical measure of the difference between the synthesized and desired patterns. The constraints kept each waveform’s root-mean-square bandwidth close to its starting value, preserved total energy and required the CSDM at each frequency to be positive semidefinite.
Wide bandwidth exposes the mismatch
At γ = 0.01, the modeled pattern was indistinguishable from the narrowband MIMO formulation. As fractional bandwidth widened, the response became more distorted: low-frequency components broadened the mainbeam and smeared and raised sidelobes, while high-frequency components concentrated near u = 0 and increased mainbeam ripple.
The optimized set was closer to the target than the initial set. Its mainbeam-to-sidelobe transition was narrower, sidelobe heights were lower and mainbeam ripple was reduced. It also showed substantially less low-frequency content and less variation in transition width across frequency.
The authors say the resulting waveform sets evenly distribute transmit power across the desired mainbeam and operational frequency band. They describe the work as a first attempt to optimize MTSFM wideband MIMO beampatterns at γ = 0.5, a very large fractional-bandwidth setting.
The result remains a model
The numerical problem was solved with MATLAB’s fmincon using an interior-point algorithm. That approach finds a local minimum, so it does not establish that the reported coefficients are globally best, and the paper does not present the routine as computationally efficient.
The comparison was with an initial waveform set rather than an independently matched algorithmic control, and only one illustrative design was reported. The reported gains are qualitative: no numerical MMSE mismatch, beamwidth, sidelobe, ripple or uncertainty estimates are provided.
The document is an arXiv preprint, version 1, dated 20 August 2026. It reports funding from the Naval Undersea Warfare Center’s In-House Laboratory Independent Research program of the Office of Naval Research under award N0001424WX00177.
Paper data and sources
Original title: Wideband MIMO Beampattern Synthesis using Adaptive Frequency Modulated Waveforms
Authors: David A. Hague
Journal/Repository: arXiv
Status: Preprint, not yet peer-reviewed
First online: 2026-08-20
DOI: Not available
Original paper · Full text