Code and data for: Depth-adaptive off-grid equivalent-source NAH using SBL and variable projection

Code and experimental data accompanying the manuscript:

Lippincott, W., Malahowski, R. J., Nowicki, K., and Mann, C. J. (2026). Depth-adaptive off-grid equivalent-source near-field acoustic holography using sparse Bayesian learning and variable projection. Journal of Sound and Vibration.

This archive contains everything needed to reproduce the simulation and experimental results reported in the paper.

CODE

simulation_sbl_varpro_v5.m — Self-contained MATLAB Monte Carlo simulation reproducing Section 3 of the paper and Sections S1 to S7 of the supplement: the eight test cases (A1 to A5, B1 to B3) at SNR = 30 dB across N = 100 trials, the four-method NMSE / lateral / depth metrics in Tables 2a and 2b, and the BIC kappa sensitivity sweep on case B3. All randomness is deterministically seeded for bit-exact reproducibility. Approximately 2 to 3 hours runtime on a desktop CPU; no toolbox dependencies beyond base MATLAB.

process_experimental_nah.m — MATLAB experimental processing pipeline reproducing Section 4 of the paper. Loads Sorama CAM64 raw recordings, extracts complex pressure at 4395 Hz, and runs Tikhonov, FISTA-L1, IRLS-L1, on-grid SBL, SBL-VarPro(xy), and SBL-VarPro(xyz) on the same hologram.

DATA (15 zip files, approx 1.5 GB total)

Single-source recordings: A3, B1.

Case A5 (two sources on the matched ESM plane): A5_S1_separate, A5_S2_separate, A5_take2_simultaneous.

Case B2 (three sources at 15 mm depth offset): B2_S1_separate, B2_S2_separate, B2_S3_separate.

Case B3 (two sources at mixed depth, paper Figs. 8 and 9): B3_S1_separate, B3_S2_separate, B3_take2_simultaneous.

Controlled depth-scaling check (paper Section 4.5, Table 4): B3_take3_metadata50mm_sep15mm, B3_take3_metadata65mm_sep15mm, B3_take4_metadata50mm_sep25mm, B3_take4_metadata75mm_sep25mm.

Each zip preserves Sorama's native Containers / RecordingGroup / UUID and Containers / Recording / UUID directory structure. The Recording / UUID / sound blob in each zip contains the 64-channel Int32 hologram (approximately 120 MB per recording, 10 seconds at 46,875 Hz).

DOCUMENTATION

README — top-level archive guide and quick-start instructions.

HOW_TO_LOAD_SORAMA_DATA — Sorama directory format, full UUID inventory mapping each zip to its case, take, and configuration, and step-by-step instructions for the experimental pipeline.

Sorama Recording VarPro spreadsheet — colour-coded source placement table (canonical inventory).

LICENSE: MIT.

CITATION: If you use this code or data, please cite the JSV manuscript above and this Zenodo record.