Measurement data and analysis code supporting \"A large-scale integrated optical phased array with digital beamforming\" (Gurses, Sarkar, Khachaturian, Fatemi and Hajimiri, Scientific Reports, 2026).
\nThe chip is a 128-element silicon photonic optical phased array in which every element has its own coherent receiver, so the received optical field is digitized per channel and the beam is formed numerically after acquisition rather than by tuning phase shifters on the chip. This record contains the measurements underlying the published figures together with the code that produces them.
\nLICENSING
\nThis record contains material under two licenses. The measurement data (the three .zip archives) are released under a Creative Commons Attribution 4.0 International license (CC BY 4.0). The analysis and figure-generation code (dopa-digital-beamforming-v1.0.0.zip) is released under the MIT license; the full MIT text is included as LICENSE inside that archive.
\nCONTENTS
\nData, 327 MB compressed and about 1.2 GB unpacked:
\nfig4_beamforming_20221209_meas2.zip (285 MB, 1031 MB unpacked). Nine digitizer captures taken as the illuminating fiber is swept across the aperture, one per angle. These are the data from which the beam is formed digitally.
\nnoise_spectra_20221119.zip (41 MB, 172 MB unpacked). Three captures for the noise-floor analysis: received signal, EDFA-on reference, and detector dark noise.
\nchannel_characterization_128ch.zip (1.5 MB, 4.3 MB unpacked). 128 spectrum-analyzer traces, one per receiver channel, plus five later retakes under retakes/ that are not used by any published figure.
\nCode, 2.3 MB:
\ndopa-digital-beamforming-v1.0.0.zip. The analysis and figure-generation code, containing the signal-processing pipeline (bandpass filtering, Hilbert-transform amplitude and phase extraction, amplitude gating, phase calibration and array-factor evaluation), the scripts and notebooks that generate each published figure, a README describing the directory layout and the order in which scripts are run, and the MIT license. Developed at https://github.com/vgurses/dopa-digital-beamforming
\nUnzip each data archive into a directory of the same name to match the layout the code expects.
\nFILE FORMATS
\nDigitizer captures are headerless CSV, 1,000,000 rows by 8 columns, acquired with a Keysight M3100A PXIe digitizer at 100 MS/s, so each file is a 10 ms record. Values are digitizer voltages at the transimpedance-amplifier outputs (LTC6269-10, 5 kilohm transimpedance; photodiode responsivity 0.67 A/W). Column order follows the acquisition hardware rather than the array geometry. Column 1 is the timebase, column 7 is unused, and the remaining six carry receiver channels. The authoritative channel-to-column mapping is defined explicitly in the phaseCalc function inside python/figure4_beamforming.ipynb; do not infer it by reading the columns in order.
\nILLUMINATION ANGLE CONVENTION. File names give the raw motorized rotation-stage reading in degrees, not the angle reported in the paper. Normal incidence is 270 degrees, so the illumination angle is (stage reading minus 270). The 266 to 274 degree sweep therefore corresponds to -4 to +4 degrees.
\nChannel characterization traces are Keysight N9952A FieldFox exports, File_N.csv with N the channel index 1 to 128. An instrument header block precedes a BEGIN line; numeric data follows as frequency_Hz, power_dBm pairs, terminated by END. Traces are 1001 points centered at 297,799 Hz with a 10 kHz span and 1 Hz RBW.
\nThese traces were recorded with the local oscillator on and no signal coupled to the antennas, so the peak in each trace is residual LO leakage at the balanced-detector output, not a received signal. The noise floor is taken as the median of each trace, which excludes that tone; at 1 Hz RBW it is a power spectral density in dBm/Hz directly. Across the array the median is -118.9 dBm/Hz and 108 of the 128 channels lie within 1 dB of it.
\nMEASUREMENT CONDITIONS
\nWavelength 1550 nm nominal; the array-factor reconstruction evaluates at 1555.23 nm. Antenna pitch 10.85 micrometers, with waveguide grating antennas of 160 by 3.5 micrometers. The signal fiber is mounted on a motorized rotation stage at roughly 2 cm stand-off.
\nFor the beamforming captures the local oscillator is phase-modulated with a 1 MHz voltage ramp and amplified by an EDFA before being coupled onto the chip, placing the beat at 1 MHz. For the channel characterization the LO is modulated at 297.74 kHz instead, which is why the analyzer is centered at 297.799 kHz. Median LO power is 3.5 microwatts per channel.
\nSix of the 128 channels are digitized simultaneously in the beamforming measurement; the 128-channel characterization was taken channel by channel through the on-board multiplexer.
\nSCOPE
\nThis record contains the measurements underlying the published figures, the minimum dataset needed to interpret and verify the results. Further acquisition campaigns recorded during development (2022-11-20, 2022-11-21, 2022-12-09 meas1, 2022-12-14) do not appear in the paper and are available from the corresponding author on reasonable request.
", "doi": "10.22002/rp3ah-z6t97", "publisher": "CaltechDATA", "publication_date": "2026-07-28" }, { "id": "data:205zv-y9414", "collection": "data", "collection_id": "205zv-y9414", "cite_using_url": "https://data.caltech.edu/records/205zv-y9414", "type": "dataset", "title": "Measurement data and analysis code for \"A large-scale integrated optical phased array with digital beamforming\"", "author": [ { "family_name": "Gurses", "given_name": "Volkan", "orcid": "0000-0001-8184-208X" }, { "family_name": "Sarkar", "given_name": "Debjit", "orcid": "0000-0002-3066-9819" }, { "family_name": "Khachaturian", "given_name": "Aroutin", "orcid": "0000-0001-8304-3302" }, { "family_name": "Fatemi", "given_name": "Reza", "orcid": "0000-0001-9081-2608" }, { "family_name": "Hajimiri", "given_name": "Ali", "orcid": "0000-0001-6736-8019" } ], "abstract": "Measurement data and analysis code supporting \"A large-scale integrated optical phased array with digital beamforming\" (Gurses, Sarkar, Khachaturian, Fatemi and Hajimiri, Scientific Reports, 2026).
\nThe chip is a 128-element silicon photonic optical phased array in which every element has its own coherent receiver, so the received optical field is digitized per channel and the beam is formed numerically after acquisition rather than by tuning phase shifters on the chip. This record contains the measurements underlying the published figures together with the code that produces them.
\nLICENSING
\nThis record contains material under two licenses. The measurement data (the three .zip archives) are released under a Creative Commons Attribution 4.0 International license (CC BY 4.0). The analysis and figure-generation code (dopa-digital-beamforming-v1.0.1.zip) is released under the MIT license; the full MIT text is included as LICENSE inside that archive.
\nCONTENTS
\nData, 327 MB compressed and about 1.2 GB unpacked:
\nfig4_beamforming_20221209_meas2.zip (285 MB, 1031 MB unpacked). Nine digitizer captures taken as the illuminating fiber is swept across the aperture, one per angle. These are the data from which the beam is formed digitally.
\nnoise_spectra_20221119.zip (41 MB, 172 MB unpacked). Three captures for the noise-floor analysis: received signal, EDFA-on reference, and detector dark noise.
\nchannel_characterization_128ch.zip (1.5 MB, 4.3 MB unpacked). 128 spectrum-analyzer traces, one per receiver channel, plus five later retakes under retakes/ that are not used by any published figure.
\nCode, 0.9 MB:
\ndopa-digital-beamforming-v1.0.1.zip. The analysis and figure-generation code, containing the signal-processing pipeline (bandpass filtering, Hilbert-transform amplitude and phase extraction, amplitude gating, phase calibration and array-factor evaluation), the scripts and notebooks that generate each published figure, a README describing the directory layout and the order in which scripts are run, and the MIT license. Developed at https://github.com/vgurses/dopa-digital-beamforming
\nUnzip each data archive into a directory of the same name to match the layout the code expects.
\nFILE FORMATS
\nDigitizer captures are headerless CSV, 1,000,000 rows by 8 columns, acquired with a Keysight M3100A PXIe digitizer at 100 MS/s, so each file is a 10 ms record. Values are digitizer voltages at the transimpedance-amplifier outputs (LTC6269-10, 5 kilohm transimpedance; photodiode responsivity 0.67 A/W). Column order follows the acquisition hardware rather than the array geometry. Column 1 is the timebase, column 7 is unused, and the remaining six carry receiver channels. The authoritative channel-to-column mapping is defined explicitly in the phaseCalc function inside python/figure4_beamforming.ipynb; do not infer it by reading the columns in order.
\nILLUMINATION ANGLE CONVENTION. File names give the raw motorized rotation-stage reading in degrees, not the angle reported in the paper. Normal incidence is 270 degrees, so the illumination angle is (stage reading minus 270). The 266 to 274 degree sweep therefore corresponds to -4 to +4 degrees.
\nChannel characterization traces are Keysight N9952A FieldFox exports, File_N.csv with N the channel index 1 to 128. An instrument header block precedes a BEGIN line; numeric data follows as frequency_Hz, power_dBm pairs, terminated by END. Traces are 1001 points centered at 297,799 Hz with a 10 kHz span and 1 Hz RBW.
\nThese traces were recorded with the local oscillator on and no signal coupled to the antennas, so the peak in each trace is residual LO leakage at the balanced-detector output, not a received signal. The noise floor is taken as the median of each trace, which excludes that tone; at 1 Hz RBW it is a power spectral density in dBm/Hz directly. Across the array the median is -118.9 dBm/Hz and 108 of the 128 channels lie within 1 dB of it.
\nMEASUREMENT CONDITIONS
\nWavelength 1550 nm nominal; the array-factor reconstruction evaluates at 1555.23 nm. Antenna pitch 10.85 micrometers, with waveguide grating antennas of 160 by 3.5 micrometers. The signal fiber is mounted on a motorized rotation stage at roughly 2 cm stand-off.
\nFor the beamforming captures the local oscillator is phase-modulated with a 1 MHz voltage ramp and amplified by an EDFA before being coupled onto the chip, placing the beat at 1 MHz. For the channel characterization the LO is modulated at 297.74 kHz instead, which is why the analyzer is centered at 297.799 kHz. Median LO power is 3.5 microwatts per channel.
\nSix of the 128 channels are digitized simultaneously in the beamforming measurement; the 128-channel characterization was taken channel by channel through the on-board multiplexer.
\nSCOPE
\nThis record contains the measurements underlying the published figures, the minimum dataset needed to interpret and verify the results. Further acquisition campaigns recorded during development (2022-11-20, 2022-11-21, 2022-12-09 meas1, 2022-12-14) do not appear in the paper and are available from the corresponding author on reasonable request.
\nVERSION HISTORY
\nv1 (28 July 2026). Initial release; code archive v1.0.0.
\nv2 (28 July 2026). Code archive updated to v1.0.1. It repairs two metadata defects in v1.0.0, whose README and citation metadata pointed readers to a nonexistent Zenodo DOI for the data and carried an unfilled article-DOI placeholder, and updates the figure-generation scripts (python/figure4_beamforming.ipynb and matlab/phaseMaker.m) to plot Fig. 4d as 20 log10|AF|, i.e., in power decibels, matching the published figure, in which the half-power beamwidth of 1.23 degrees is read directly at -3 dB. Internal references now use the all-versions DOI, and the record-level description is included as README.md. Measurement data unchanged.
", "doi": "10.22002/205zv-y9414", "publisher": "CaltechDATA", "publication_date": "2026-07-28" }, { "id": "data:40wk6-9fa71", "collection": "data", "collection_id": "40wk6-9fa71", "cite_using_url": "https://data.caltech.edu/records/40wk6-9fa71", "type": "dataset", "title": "Measurement data and analysis code for \"A large-scale integrated optical phased array with digital beamforming\"", "author": [ { "family_name": "Gurses", "given_name": "Volkan", "orcid": "0000-0001-8184-208X" }, { "family_name": "Sarkar", "given_name": "Debjit", "orcid": "0000-0002-3066-9819" }, { "family_name": "Khachaturian", "given_name": "Aroutin", "orcid": "0000-0001-8304-3302" }, { "family_name": "Fatemi", "given_name": "Reza", "orcid": "0000-0001-9081-2608" }, { "family_name": "Hajimiri", "given_name": "Ali", "orcid": "0000-0001-6736-8019" } ], "abstract": "Measurement data and analysis code supporting \"A large-scale integrated optical phased array with digital beamforming\" (Gurses, Sarkar, Khachaturian, Fatemi and Hajimiri, Scientific Reports, 2026).
\nThe chip is a 128-element silicon photonic optical phased array in which every element has its own coherent receiver, so the received optical field is digitized per channel and the beam is formed numerically after acquisition rather than by tuning phase shifters on the chip. This record contains the measurements underlying the published figures together with the code that produces them.
\nLICENSING
\nThis record contains material under two licenses. The measurement data (the three .zip archives) are released under a Creative Commons Attribution 4.0 International license (CC BY 4.0). The analysis and figure-generation code (dopa-digital-beamforming-v1.0.0.zip) is released under the MIT license; the full MIT text is included as LICENSE inside that archive.
\nCONTENTS
\nData, 327 MB compressed and about 1.2 GB unpacked:
\nfig4_beamforming_20221209_meas2.zip (285 MB, 1031 MB unpacked). Nine digitizer captures taken as the illuminating fiber is swept across the aperture, one per angle. These are the data from which the beam is formed digitally.
\nnoise_spectra_20221119.zip (41 MB, 172 MB unpacked). Three captures for the noise-floor analysis: received signal, EDFA-on reference, and detector dark noise.
\nchannel_characterization_128ch.zip (1.5 MB, 4.3 MB unpacked). 128 spectrum-analyzer traces, one per receiver channel, plus five later retakes under retakes/ that are not used by any published figure.
\nCode, 2.3 MB:
\ndopa-digital-beamforming-v1.0.0.zip. The analysis and figure-generation code, containing the signal-processing pipeline (bandpass filtering, Hilbert-transform amplitude and phase extraction, amplitude gating, phase calibration and array-factor evaluation), the scripts and notebooks that generate each published figure, a README describing the directory layout and the order in which scripts are run, and the MIT license. Developed at https://github.com/vgurses/dopa-digital-beamforming
\nUnzip each data archive into a directory of the same name to match the layout the code expects.
\nFILE FORMATS
\nDigitizer captures are headerless CSV, 1,000,000 rows by 8 columns, acquired with a Keysight M3100A PXIe digitizer at 100 MS/s, so each file is a 10 ms record. Values are digitizer voltages at the transimpedance-amplifier outputs (LTC6269-10, 5 kilohm transimpedance; photodiode responsivity 0.67 A/W). Column order follows the acquisition hardware rather than the array geometry. Column 1 is the timebase, column 7 is unused, and the remaining six carry receiver channels. The authoritative channel-to-column mapping is defined explicitly in the phaseCalc function inside python/figure4_beamforming.ipynb; do not infer it by reading the columns in order.
\nILLUMINATION ANGLE CONVENTION. File names give the raw motorized rotation-stage reading in degrees, not the angle reported in the paper. Normal incidence is 270 degrees, so the illumination angle is (stage reading minus 270). The 266 to 274 degree sweep therefore corresponds to -4 to +4 degrees.
\nChannel characterization traces are Keysight N9952A FieldFox exports, File_N.csv with N the channel index 1 to 128. An instrument header block precedes a BEGIN line; numeric data follows as frequency_Hz, power_dBm pairs, terminated by END. Traces are 1001 points centered at 297,799 Hz with a 10 kHz span and 1 Hz RBW.
\nThese traces were recorded with the local oscillator on and no signal coupled to the antennas, so the peak in each trace is residual LO leakage at the balanced-detector output, not a received signal. The noise floor is taken as the median of each trace, which excludes that tone; at 1 Hz RBW it is a power spectral density in dBm/Hz directly. Across the array the median is -118.9 dBm/Hz and 108 of the 128 channels lie within 1 dB of it.
\nMEASUREMENT CONDITIONS
\nWavelength 1550 nm nominal; the array-factor reconstruction evaluates at 1555.23 nm. Antenna pitch 10.85 micrometers, with waveguide grating antennas of 160 by 3.5 micrometers. The signal fiber is mounted on a motorized rotation stage at roughly 2 cm stand-off.
\nFor the beamforming captures the local oscillator is phase-modulated with a 1 MHz voltage ramp and amplified by an EDFA before being coupled onto the chip, placing the beat at 1 MHz. For the channel characterization the LO is modulated at 297.74 kHz instead, which is why the analyzer is centered at 297.799 kHz. Median LO power is 3.5 microwatts per channel.
\nSix of the 128 channels are digitized simultaneously in the beamforming measurement; the 128-channel characterization was taken channel by channel through the on-board multiplexer.
\nSCOPE
\nThis record contains the measurements underlying the published figures, the minimum dataset needed to interpret and verify the results. Further acquisition campaigns recorded during development (2022-11-20, 2022-11-21, 2022-12-09 meas1, 2022-12-14) do not appear in the paper and are available from the corresponding author on reasonable request.
", "doi": "10.22002/40wk6-9fa71", "publisher": "CaltechDATA", "publication_date": "2026-07-28" } ]