Diffused-Beam Laser-Diode LiFi Under Realizable Receiver, Noise, and Safety Constraints: Design-Space Analysis and an Open Cross-Verified Simulation Framework

JOURNAL OF LIGHTWAVE TECHNOLOGY

1

Diffused-Beam Laser-Diode LiFi Under Realizable Receiver, Noise, and Safety Constraints: Design-Space Analysis and an Open Cross-Verified Simulation Framework

arXiv:2608.15236v1 [eess.SP] 15 Aug 2026

Hussain Ahmad, Saleem Aslam, Ammara Nasim, Syed Muhammad Talha Gillani, and Toheed Omer,

Abstract—Link-budget studies of indoor optical wireless systems frequently assume receiver parameter sets—large photodetector area, large transimpedance, and wide bandwidth simultaneously—that violate basic circuit constraints, and noise budgets that omit the amplifier and laser noise terms that dominate at exactly those design points. This paper develops a realizability-constrained design-space analysis of a diffused-beam laser-diode (LD) LiFi link anchored to a hardware prototype built by the authors. The analysis couples the generalizedLambertian channel of a holographic-diffuser source to a receiver model that enforces the transimpedance-amplifier gainbandwidth/capacitance constraint and carries a complete noise budget: shot, feedback-resistor thermal, input current noise, capacitance-driven voltage-noise gain, and laser relative intensity noise (RIN). Against this budget we evaluate unipolar M -PAM under two specified FEC tiers (7%-overhead hard-decision at 3.8 × 10−3 , 20%-overhead soft-decision at 2 × 10−2 ), firstbounce diffuse multipath computed by patch integration, and a quantitative extended-source eye-safety assessment. The full model predicts 140 Mb/s net at the prototype’s demonstrated 14m range with 6.7 dB of link margin (OOK, HD tier), 240 Mb/s at the zero-margin 4-PAM/SD reach boundary of 14.0 m, and 480–558 Mb/s at 5 m—a factor 3.9–6.6 below what the same link yields under a naive textbook budget with an unconstrained receiver, quantifying how strongly idealized assumptions inflate LiFi rate projections. First-bounce analysis shows the downfacing-source/up-facing-receiver geometry confines multipath to a worst-case LOS-to-diffuse ratio of 4.2 dB and delay spreads below 0.13 ns, and the diffused 500-mW source remains a factor ≥7.8 under the Class-1 extended-source emission limit at all viewing distances. All models are released as an ns-3 module and an independent Python engine whose mutual agreement is enforced by automated tests; these predictions are explicitly model-based, and the paper delineates what the megabit-class prototype does and does not validate. Index Terms—Visible light communication, LiFi, laser diode, holographic diffuser, intensity modulation, transimpedance amplifier, relative intensity noise, multipath, laser safety, ns-3.

I. I NTRODUCTION ASER diodes offer indoor optical wireless systems an order-of-magnitude modulation-bandwidth advantage over phosphor white LEDs, demonstrated by 9-Gb/s singlechannel GaN links [1], illumination-grade white-laser transmitters [2], [3], and real-time Gb/s transceivers [4]; LED

L

H. Ahmad, S. M. T. Gillani, T. Omer, and S. Aslam are with the Department of Electrical Engineering, Bahria University, Islamabad, Pakistan (e-mail: hussainbuic@gmail.com). Simulation code (ns-3 module and Python link-level engine) is available with this article.

systems have answered with heavily engineered spectral efficiency, reaching 15.73 Gb/s with off-the-shelf devices [5]. Between these laboratory flagships and deployable LiFi systems [6]–[8] lies a less glamorous question: given a buildable transmitter and a buildable receiver, what does a diffused-beam LD link actually support, at what range, under which safety constraints, and with which honest error bars? This paper addresses that question quantitatively, and its starting point is self-critical. An earlier version of this work [9] specified the receiver as a large-area photodiode (1 cm2 ), a large transimpedance (10 kΩ), and a wide bandwidth (250 MHz) simultaneously, with a noise budget containing only shot noise and the feedback resistor’s 4kT /Rf . Such a receiver cannot be built: a 1-cm2 fully-depleted silicon PIN presents ∼35 pF, placing the transimpedance pole near 0.5 MHz, and no realistic amplifier gain-bandwidth closes the gap [10], [11]. The omitted amplifier terms—input current noise and the capacitance-driven voltage-noise gain rising as f 2 , which integrates to B 3 [11]—then dominate the budget at exactly the wide-bandwidth design points assumed, and laser RIN caps the SNR exactly where high-order constellations are claimed. We are deliberately careful about how far this generalizes. The canonical VLC noise model of [12] does carry a capacitance- and transconductance-dependent term, and studies that adopt it in full inherit that term [13]. The pattern we point to is narrower and is visible in the parameter tables of widely cited work: a 1-cm2 detector tabulated together with 40–200 MHz of bandwidth and a noise figure given only as a power spectral density, with no amplifier gain-bandwidth, feedback resistance, or voltage-noise density stated [14]–[16]. In network-layer studies such as these the abstraction is entirely legitimate—the receiver is not the object of analysis. The failure mode arises when the same convention is carried into link-level rate prediction, as our own earlier version [9] did, because the unstated constraint is then the binding one. A. Contributions 1) A realizability-constrained receiver model (Section V): the gain-bandwidth/capacitance constraint is enforced explicitly, the étendue trade between collection area and field of view is carried through the optics, and the noise budget includes shot, 4kT /Rf , input current

JOURNAL OF LIGHTWAVE TECHNOLOGY

noise i2n B, the (2πCT vn )2 B 3 /3 term, and RIN [17], [18]. A design-frontier chart makes infeasible parameter sets visible at a glance. 2) Quantified idealization penalty (Section IX): at the anchor operating point (14 m, 20◦ diffuser) the complete budget yields 5.1 dB less SNR than the naive budget at the same parameters—and 3.9–6.6× lower net rate than the earlier unconstrained receiver—providing a concrete, reproducible measure of how idealized assumptions inflate projections. 3) First-bounce multipath by patch integration (Section IV), following [19], [20]: for the ceiling-down/deskup geometry the worst-case LOS-to-diffuse power ratio in a 5 × 5 × 3-m room is 4.2 dB (wide diffuser, wide FoV, room corner) with first-bounce rms delay spread ≤ 0.13 ns—i.e., τrms Rs < 0.02 at the 150-MBd symbol rate, quantifying rather than assuming the LOS approximation, within stated limits (first bounce); tiltedreceiver cases are computed and shown to fail on power (outage) before ISI. 4) A quantitative extended-source eye-safety assessment (Section VII) in the manner of IEC 60825-1 [21]: the 25-mm, 20◦ holographic diffuser keeps the 7-mmpupil collected power a factor ≥ 7.8 below the Class1 extended-source limit at every viewing distance from 10 cm, closing the loop between safety and the link budget that uses the same 450-mW radiated power. 5) Specified FEC tiers and modulation (Section VI): unipolar M -PAM at symbol rate = B (no subcarrier, hence no double-sideband bandwidth penalty and no unaccounted bias power split), evaluated against a G.975.1class 7%-overhead HD-FEC (3.8 × 10−3 ) and a 20%overhead SD-FEC (2 × 10−2 ) [22]. 6) An open, cross-verified simulation framework (Section VIII): an ns-3 module (channel, PHY, ARQ MAC, NetDevice) and an independent Python engine implementing the same physics, with automated tests enforcing agreement. We are explicit about epistemic status: this machinery verifies the implementations against each other and against Monte-Carlo statistics; it does not validate the physical assumptions, which only wideband measurement can do (Section X). The study is anchored to the authors’ hardware prototype [23], [24]: a 500-mW LD switched through a CH340 USB–TTL bridge, diffused holographically, and received by a threshold photodetector over a 14-m indoor LOS path. (Priorpublication statement: the prototype hardware, its demonstrated transfers, and the interface-ceiling data of Fig. 13 originate in the thesis [23]; all modeling, the receiver design, the simulation framework, and every figure and result in this article are new.) The prototype’s role is strictly delimited: it establishes end-to-end feasibility (data, real-time voice, and image transfer) at interface-limited rates ≤ 1.59 Mb/s, and it anchors geometry and component classes. Every rate above that figure in this paper is a model prediction, labeled as such, whose experimental confirmation requires the wideband front end specified in Section X.

2

II. R ELATED W ORK AND P OSITIONING Every result cited below is labelled measured (a physical demonstration), modelled (analysis or simulation), or review. The convention is deliberate: the gap this paper addresses is the one between what has actually been measured and what indoor link-budget models routinely assume. Broad context for optical wireless in the 6G programme is set by recent surveys and roadmaps [25]–[28], which we do not duplicate; we concentrate on the six threads our contributions touch. A. Laser-Diode Optical Wireless: The Measured State of the Art At device level the blue GaN laser diode is no longer the limiting element. A c-plane narrow-ridge short-cavity mini-LD has been measured with a −3 dB modulation bandwidth of 5.9 GHz and a 1.02 W/A slope efficiency, supporting 20.06 Gb/s with bit- and power-loaded DMT [29] (measured); InGaN quantum barriers push the intrinsic bandwidth beyond 8 GHz at 1.85 W/A slope efficiency and 29% wall-plug efficiency, with 36.5 Gb/s reported [30] (measured); and an InGaN mini-LD with 8.4 GHz measured bandwidth reaches 33 Gb/s using PAM-4 rather than multicarrier modulation [31] (measured). The underlying epitaxial and cavity-geometry trends are consolidated in two recent reviews [32], [33]. System-level records, however, are set in geometries very unlike a room. A ten-wavelength packaged module reached 113.175 Gb/s with a sparse-structured reservoir-computing post-equaliser [34]; a 50-channel WDM intra-connect reached 534.51 Gb/s [35]; and 601.46 Gb/s has been reported over a 1 m MMF–1 m free-space–1 m MMF path with constellations up to 1024-QAM [36] (all measured). Underwater, 170.1 Gb/s was obtained with a five-wavelength module and a reciprocal differential receiver whose learned equaliser is itself worth 16.1 Gb/s [37] (measured). The closest prior art to an illuminating laser luminaire is the laser-based SMD white source of [38], which delivers 450 lm at over 1000 cd/mm2 brightness while a ten-channel WDM link exceeds 100 Gb/s aggregate with Volterra equalisation, plus 4.8 Gb/s over a 500 m outdoor point-to-point path (measured). These are short, aligned or guided paths sustained by heavy digital signal processing; none is a room-scale broadcast link operated under an eyesafety cap. That regime, not the record table, is the subject of this paper. B. Eye Safety, Speckle, and the Diffused-Source Premise The diffused-beam transmitter has an explicit experimental precedent: a phosphorous diffuser bonded to a blue GaN laser diode both broadens the luminescent spectrum and diverges the beam spot for room illumination while carrying 5.2 Gb/s 16-QAM OFDM over 60 cm at BER 3.1 × 10−3 [39] (measured). Laser-pumped phosphor lighting is itself a reviewed, commercially deployed source class [40]. The two standing objections to laser illumination have been quantified and answered: additively blended RGB laser white achieves speckle contrast ≤ 5% at 150 lm/W luminous efficacy while carrying multi-Gb/s data [41] (measured), cascaded diffusing

JOURNAL OF LIGHTWAVE TECHNOLOGY

stages have been measured to suppress speckle contrast from 0.32 (stationary diffuser) through 0.14 and 0.09 to 0.05 for a complete module with phosphor plate [42] (measured), and the design variables that control residual speckle in a phosphorconverted laser source have been isolated systematically [43] (measured). For safety, the tutorial treatment of IEC 60825-1 for optical wireless links [44] (review) establishes the extendedversus point-source distinction, apparent source size, and the exposure limits that bound transmit power. The specific classification machinery our analysis uses — characterising the apparent-source angular subtense, scaling the emission limit by C6 once the source exceeds the minimum angle, and classifying at the “most restrictive position” rather than the closest one — is set out in [45] (modelled), which also warns that a product need not have a single unique apparent source. A committee-side account of the standard’s development [46] (review) documents the difficulty of keeping classification aligned with fast-moving source technology, which is why we present our margin as conditional on a diffuser-integrity interlock rather than as a settled classification. C. LED Systems and the Matched-Condition Comparison Any laser-versus-LED claim must be stated at matched distance, matched emitter count, and matched receiver-side processing, or it is not a comparison at all. The canonical deployable-LED benchmark remains 15.73 Gb/s over 1.6 m from four off-the-shelf LEDs with OFDM and adaptive bit loading [5] (measured). The headline aggregate figure, 25.20 Gb/s, uses eight emitters spanning 276–655 nm over 25 cm; the visible-band-only portion is 20.11 Gb/s [47] (measured). An all-visible RGBP module reaches 23.43 Gb/s over 1 m, but only after transmitter-side pre-equalisation raises perchannel bandwidths to 756.6–929 MHz [48] (measured). At 2 m, 16.6 Gb/s is obtained from tricolour mini-LEDs with percolour bandwidths of 668–859 MHz and a five-layer neuralnetwork receiver replacing conventional estimation, equalisation and demodulation [49] (measured). Room-scale LED evidence is thinner and more instructive. A 3 × 3 violet micro-LED array with 13.4 mW transmitted power and only 350 MHz of device bandwidth delivers 10.23, 10.10 and 9.51 Gb/s at 0.2, 1 and 10 m below the 3.8 × 10−3 threshold, using bit loading with distanceadaptive pre-equalisation [50] (measured); a single 60 µm c-plane micro-LED reaches 10.547 Gb/s at the optimum of an emitter-size sweep and still delivers 8.649 Gb/s at 10 m [51] (measured). Elsewhere the rate is bought with processing or with bespoke devices: 8.75 Gb/s from a 75 µm blue micro-LED of 1025 MHz bandwidth using an artificial neural-network equaliser [52], 19.3 Gb/s over 1.2 m from a spatially multiplexed array at over 9.5 Gb/s per emitter [53], and 15.64 Gb/s at 0.3 m using a red GaN micro-LED as the photodetector [54] (all measured). Adaptive loading with an explicit SNR gap [55] (measured) and physics-informed reservoir-computing equalisation, worth roughly a 10% rate gain with an SNR advantage of 1.2 dB over a Volterra equaliser and 1.5 dB over conventional reservoir computing [56] (both),

3

show that LED baselines are not static. At the opposite extreme of emitter area, RGB-single-chip OLEDs top out at 3.2 Gb/s with −6 dB bandwidths of 171–255 MHz [57] (measured). We therefore report no rate record and make no laser-beats-LED claim; we compare architectures at stated distance and stated receiver complexity. D. Receiver-Side Limits: Where Our Frontier Sits That receiver aperture cannot be increased for free is established prior art, not our contribution. The rate–FOV tradeoff arising from the compounding of photodetector area– bandwidth and optical gain–FOV constraints is derived in [58] (modelled), where an “array of arrays” imaging receiver of 2 × 2 cm is optimised to roughly 24 Gbps at 15◦ FOV; the non-imaging counterpart optimises multi-tier angle-diversity receivers under both constraints jointly and reaches 12 Gb/s at 30◦ half-angle over a 3 m laser link with height ≤ 0.5 cm and effective area ≤ 0.5 cm2 [59] (modelled). The informationtheoretic statement that an interior optimum detector area exists — too large costs bandwidth, too small costs captured power — is given in [60] (modelled), and the practice of treating the front end as a frequency-dependent gain-to-noise profile modelled by a pole–zero transfer function, rather than a single 3 dB number, is developed in [61] (both). The known escape routes each substitute one ceiling for another. Luminescent solar concentrators evade the etendue limit but are shown analytically to be equivalent to an RC lowpass network, with fluorescence lifetime and self-reabsorption as the dominant bottlenecks [62] (both); 50 × 50 mm luminescent antennas nonetheless outperform a bare photodiode under white light [63] (both), three such antennas benchmarked under both laser and LED sources reach above 70 Mb/s with NRZ-OOK [64] (measured), and a fluorescent slab of 245 MHz bandwidth coupled to a SiPM yields 1.4 Gb/s over 30 cm under 500 lux with 20-fold ambient rejection [65] (measured). Segmentation is an active route: an APD array laid out to match a fluorescent antenna’s side-emission profile achieves higher power gain at comparable bandwidth to a single detector of nearly the same total area [66] (both). Photon-counting front ends trade bandwidth for sensitivity — 30 pW, or 41.5 photons per bit, at 1 Mbit/s [67] (measured) — with nonlinearity and pulse width as the binding limits [68] (modelled) and dead-time blocking and ISI as the PAM-specific failure mode [69] (modelled). Several of these works do carry front-end effects into the achievable rate, notably the jointly constrained optimisation of [59] and the pole–zero gain-to-noise treatment of [61]. The amplifier side of the problem is, separately, a mature body of theory that the optical-wireless modelling literature does not always meet. The bound tying transimpedance to bandwidth through the amplifier gain–bandwidth product and the total input capacitance is the transimpedance limit [10], treated at book length in [70] and derived in equivalent 2 form as RT ≤ AωA /(CT ω3dB ) in [11] (measured), whose CMOS front end also states explicitly that the dominant inputreferred noise densities rise as f 2 and therefore dominate once integrated — the B 3 behaviour our budget carries. The

JOURNAL OF LIGHTWAVE TECHNOLOGY

lineage runs back to the classical optical-receiver analyses of [71]–[73], and VLC-specific front-end work confronts the same trade directly, whether by regulated-cascode topologies sized for high-capacitance photodiodes [74] (measured) or by bandwidth-extension techniques for visible-light receivers [75] (measured). We therefore claim no circuit-theoretic novelty whatever. What we add is narrower and lies at the seam between the two literatures: tying the assumed detector area of a link-budget study to a specific gain–bandwidth product and feedback resistance through the implied junction capacitance, and propagating the resulting voltage-noise term into reach and rate, so that a parameter set which cannot be built is visible as such before any rate is quoted. E. Modulation and Coding for IM/DD The choice of unipolar M -PAM over optical OFDM is supported rather than asserted. For indoor multipath IM/DD links under an average-optical-power constraint, unipolar M PAM with MMSE decision-feedback equalisation is the most power-efficient scheme over the practical range, with OFDM variants requiring several dB more average optical power at equal spectral efficiency [76] (modelled). The normalisedpower-budget framing of [77] (modelled) makes the boundary quantitative: optical OFDM becomes advantageous only above roughly a 60 dB power budget, DCO-OFDM beats PAM in the 30–60 dB range only with adaptive bit loading, without bit loading PAM is about 2.5 dB better at 60 dB, and below 30 dB PAM is always preferred. Capacity bounds for the non-negativity-constrained Gaussian IM/DD channel under average- and peak-power constraints [78] (review) supply the correct ceiling against which unipolar rates should be reported. Emitter-side clipping and static nonlinearity, which penalise high-PAPR waveforms in a peak-constrained source, are quantified in [79] (both), and multilevel PAM with equalisation is what practice actually uses to reach multi-Gb/s: 8.8 Gb/s PAM4 with an external modulator and a neural-network equaliser [80] (measured). Laser RIN is the term most often omitted. For LD-based indoor VLC, RIN emerges as the dominant noise source once transmitted optical power is raised [81] (modelled), a conclusion extended to random channels in [82] (modelled). Because RIN is signal-dependent, the signal-independent Gaussian error formula is not exact, and a closed-form SER for the optimal RIN-aware detector making no assumption on constellation geometry is available [83] (modelled). The magnitude matters for visible GaN devices specifically: mode-hopping and modecompetition dynamics drive InGaN edge-emitter RIN roughly 9 dB above comparable AlGaAs lasers [84] (measured), which is why we sweep RIN rather than adopt a telecom value, and why our 16-PAM tier is flagged as RIN-fragile. Shaping is the natural next step but is out of scope here; we note that probabilistically shaped unipolar 4-PAM under an explicit peakpower constraint is compatible with a PAS/FEC chain [85] (both), that per-subcarrier shaping has measured 10.81 Gb/s in a WDM VLC link, 25% above adaptive bit-and-power loading under identical channel conditions [86] (measured), and that probabilistic shaping has been benchmarked against geometric shaping for bandwidth-limited VLC [87] (modelled).

4

We report rates at two specified FEC tiers rather than at an unlabelled “FEC limit”. The 7% hard-decision and ∼20% soft-decision overhead classes, with their net coding gains and input BER thresholds, come from the 100G transport literature [88] (review). We also state the caveat: a channel-independent soft-decision FEC threshold is invalid for soft-decision bitwise decoding and can underestimate spectral efficiency by up to 20% at low code rates, with generalised mutual information the consistent predictor [89] (both). The conditions under which quoting uncoded BER against a threshold is legitimate, and the mismatched-receiver alternative, are given as implementable recipes in [90] (review); our claims ledger follows that discipline. F. Indoor Channel Modelling: Multipath, Mobility, Orientation The taxonomy of indoor VLC channel-modelling methods, from recursive and Monte Carlo approaches to non-sequential ray tracing, is surveyed in [91] (review), and the ray-traced reference impulse responses adopted by IEEE 802.11bb for industrial, medical, enterprise and residential usage models are documented in [92] (modelled). Published delay-spread magnitudes bound what any first-bounce treatment must be consistent with: an industrial ray-tracing study reports RMS delay spreads of 6–11 ns under a single transmitter, rising to means of 13.79 ns and 14.78 ns at 1.5 m and 2.5 m receiver height under multiple transmitters [93] (modelled), while a 3D non-stationary geometry-based stochastic model validates its simulated 3 dB channel bandwidth against existing measurement data [94] (both). Our truncation after the first bounce is a stated approximation, and the literature indicates its direction of error rather than excusing it: the measurementvalidated hybrid of [95] (both) treats early reflections explicitly in the frequency domain and adds an integrating-sphere term for higher orders, precisely because truncation alone is not sufficient in the general case; its sub-5% relative mean square error is reported for the LOS component. Higher-order bounces therefore add a weaker, longer-delayed tail that our first-bounce figures do not capture, and we treat our delayspread numbers as a lower bound whose margin against the symbol period (below 2% at 150 MBd) is large enough to absorb a several-fold increase. The regime in which the diffuse component stops being a correction at all — blocked LOS and large access-point separation — is characterised in [96] (modelled). Our fixed-normal geometry is a stated scope limit, not an oversight. Measured device orientation from 40 participants is well fitted by a Laplace distribution for static users and a Gaussian for mobile users [97] (measured), with closed-form channel-gain statistics (modified truncated Laplace, modified Beta, and their sums) derived in [98] (both); these supply the distributions for a stochastic-orientation extension. Measured rotation and movement traces put the LiFi channel coherence time on the order of tens of milliseconds and show that mobile operation requires at least two differently oriented photodiodes [99] (both), which bounds the interval over which our static link budget is valid.

JOURNAL OF LIGHTWAVE TECHNOLOGY

5

TABLE I P OSITIONING AGAINST THE STRONGEST RECENT measured DEMONSTRATIONS . T HE FINAL COLUMN IS THE WEDGE : WHAT EACH SOURCE REPORTS ABOUT THE ANALOG RECEIVER FRONT END . Work

Emitter / scheme

Rate / reach

Evidence

Receiver detail reported

Zhou et al. [36]

50-ch WDM LD, bit/power-loaded DMT, up to 1024-QAM Laser-based white SMD source, 10-ch WDM

601.46 Gb/s over 1 m MMF + 1 m free space + 1 m MMF >100 Gb/s aggregate indoor; 4.8 Gb/s over 500 m outdoor

Measured

Hu et al. [31]

InGaN mini-LD, PAM-4

Measured

Chi et al. [39]

Blue LD + phosphorous diffuser (diverged beam) Blended R/G/B laser white light

33 Gb/s; 8.4 GHz measured device bandwidth 5.2 Gb/s 16-QAM OFDM over 0.6 m, BER 3.1 × 10−3 Multi-Gb/s; speckle contrast ≤ 5%; 150 lm/W 15.73 Gb/s over 1.6 m

Measured

Differential pilot coding; no front-end bandwidth or noise budget stated Volterra nonlinear equalisation; illumination reported (450 lm, >1000 cd/mm2 ); front end not stated Emitter bandwidth reported; receiver front end not stated Not stated

Measured

Illumination metrics reported; front end not stated

Measured

Not stated

Measured

Distance-adaptive pre-equalisation (Tx side); receiver noise not stated Detector device characterised; bespoke, noncommodity front end Fluorophore bandwidth 245 MHz; 20× ambient rejection (200× filtered) p GBW/(2πRf CT ) TIA closed-loop B ≈ frontier vs. PD area/capacitance; fiveterm noise budget (shot, 4kT /Rf , i2n B, (2πCT vn )2 B 3 /3, RIN)

Chen et al. [38]

Lazzaro et al. [41] Bian et al. [5] Jin et al. [50] Ai et al. [54] Ali et al. [65] This work

4 off-the-shelf LEDs, WDM OFDM, adaptive bit loading 3 × 3 violet micro-LED array, 13.4 mW, 350 MHz Red GaN micro-LED used as the photodetector 405 nm laser Tx; fluorescent antenna + SiPM Diffused-beam LD luminaire, unipolar M -PAM / OOK

Measured

10.23 / 10.10 / 9.51 Gb/s at 0.2 / 1 / 10 m 15.64 Gb/s over 0.3 m

Measured

1.4 Gb/s over 0.3 m under 500 lux

Measured

558 Mb/s @ 5 m; 279 Mb/s @ 12 m; 140 Mb/s @ 14 m; OOK reach 20.7 m (HD) / 23.6 m (SD)

Computed, crossverified; anchored to a 14 m UARTlimited (1.59 Mb/s) prototype

G. Networking, Standards, and Simulation Tooling Two IEEE light-communication standards are directly relevant. IEEE 802.11bb reuses the 802.11 MAC and OFDMbased PHYs over optical links in the 800–1000 nm band with MAC-SAP throughput specified from 10 Mb/s to 9.6 Gb/s, and its stated open problems include interoperability across solid-state sources with differing modulation bandwidths [100] (review), with its reference channels in [92]. That open problem concerns the emitter side; the constraint we add is its receiver-side counterpart, and the two together bound what a heterogeneous light-communication deployment can assume of an arbitrary endpoint. IEEE Std 802.15.13 specifies a deterministic dynamic-TDMA MAC with two PHYs and has been implemented and experimentally tested for industrial use [101] (measured), and ITU-T G.9991 specifies a third track for in-premises optical networking [102]. Our rates fall inside the 802.11bb MAC-SAP envelope, and we report them against that envelope rather than against interconnect records. The attocell abstraction and its challenge list — inter-cell interference, uplink asymmetry, mobility and handover — are defined in [103] (review), with the hybrid LiFi/WiFi context in [104] (review), the laser-based “LiFi 2.0” roadmap that names receiver front-end design as a limiting subsystem in [105] (review), and the IoT deployment pull in [106] (review). Learning-based load balancing and mobility management define the current network-layer baseline [107], [108] (modelled), handover thresholds keyed to optical gain and incidence angle are given in [109] (modelled), and a COTS LiFi/WiFi testbed with conveyor-belt-emulated motion reduces handover outage to under one second at 20 Mb/s [110] (measured) — a useful reminder of the distance between standardised capability and deployed behaviour. On tooling, the precedent for open-source ns-3 optical modules is [111] (both), the first such VLC module, vali-

dated against a software-defined-radio testbed with phosphorconverted white LEDs. The closest existing artefact is the ns-3 LiFi framework of [112] (modelled), which already implements LOS plus first-order NLOS reflections, SNR/BER for OOK, PAM and QAM, orientation-aware mobility, a TDMA MAC and vertical handover, with source released. We state the delta plainly: that framework models neither a laser-diode source with RIN, nor a TIA-limited receiver whose bandwidth is tied to photodiode capacitance, nor the amplifier voltageand current-noise terms that dominate our budget. Our module is an extension of that modelling layer, not a replacement for it.

H. Positioning Table I places this work against the strongest recent measured demonstrations. We claim no rate record and no new device. What we add is narrower and, we hope, more durable: (i) a receiver realizability constraint that ties assumed bandwidth to a specific TIA gain–bandwidth product, feedback resistance and photodiode capacitance. That the constraint bites is not hypothetical: an earlier version of this work [9] assumed a 1 cm2 detector, a 10 kΩ feedback resistance and 250 MHz of bandwidth simultaneously, a combination the frontier of Section V shows to be unbuildable by a factor above five, and on which its headline rates depended. We correct that here rather than defend it, and we note that the receiver column of Table I shows how commonly the analog front end is simply not reported, which is what allows such combinations to pass unexamined. (ii) A noise budget in which the amplifier terms, including the capacitance-driven (2πCT vn )2 B 3 /3 contribution, account for roughly 69% of the variance at our anchor point, so that a shot-plus-thermal budget flatters SNR by 5.1 dB. (iii) Rates reported at two

JOURNAL OF LIGHTWAVE TECHNOLOGY

6

named FEC tiers under unipolar M -PAM, following the reporting discipline of [89], [90]. (iv) A first-bounce diffuse analysis giving a worst-case LOS/diffuse ratio of 4.2 dB and τrms ≤ 0.13 ns, from which tilted receivers are shown to fail on power before ISI; higher-order bounces are not modelled, and [95] indicates the form the residual takes. (v) An eye-safety margin computed with the extended-source machinery of [44], [45], stated as conditional on a diffuser-integrity interlock. (vi) Two independent implementations cross-verified against each other, with every number labelled measured, computed or predicted. Cross-verification is not validation, and we do not describe it as such. III. S YSTEM A RCHITECTURE Fig. 1 shows the modeled chain. Two receiver profiles span the terminal design space: Desk profile (default): a fixed terminal aimed at the ceiling luminaire; field-of-view semi-angle Ψc = 15◦ , which lets the étendue-limited optics concentrate g = n2 / sin2 Ψc = 33.6 onto the small photodiode, and a 10-nm interference filter with narrow FoV suppressing ambient photocurrent to Ibg = 20 µA. Pointing budget: inside the FoV cone the cos ψ penalty is benign (≤0.2 dB at ±12◦ , 0.3 dB at the 15◦ edge), so the requirement is capture within a ±15◦ cone—coarse aiming for a fixed terminal—while the FoV boundary itself is a hard cutoff; the aligned-boresight numbers quoted later carry this stated tolerance. Mobile profile: Ψc = 60◦ (g = 3.0), Ibg = 100 µA; used for room-coverage analysis under a lamp. The prototype [23] anchors the transmit side (500-mWclass LD, 25-mm holographic diffuser, 14-m indoor path) and the demonstrated payloads; its CH340 interface bounds its measurable goodput at 1.59 Mb/s (Section X), and its threshold receiver anchors no parameter of the wideband receiver model, which is specified from component classes in Section V. IV. O PTICAL C HANNEL : LOS AND F IRST-B OUNCE M ULTIPATH A. LOS Gain The diffused emission is modeled as a generalized Lambertian lobe of order m = − ln 2/ ln cos Φ1/2 ; the diffuser transmission Td = 0.9 is carried explicitly, so the radiated power is Pt = Td PLD = 450 mW. For detector area A, filter transmission Ts , concentrator gain g(ψ), emission angle ϕ and incidence ψ ≤ Ψc [12], [113]: HLOS =

(m + 1)A cosm ϕ Ts g(ψ) cos ψ. 2πd2

(1)

B. First-Bounce Diffuse Component Reviewer-facing honesty requires the LOS assumption to be quantified, not asserted. We compute the first-bounce response by patch integration over the walls and ceiling (ρ = 0.7, 25-cm patches) following [19]: each patch receives (m+1) cosm ϕ1 cos θ1 /2πd21 from the source and re-radiates as a first-order Lambertian toward the receiver, subject to the receiver FoV. Fig. 2 maps the resulting LOS-to-diffuse power

ratio K and rms delay spread τrms (power-weighted, LOS included) across the desk plane for the widest configuration (source Φ1/2 = 60◦ , mobile receiver Ψc = 60◦ )—the case most favorable to multipath. Three conclusions. (i) Even in the widest case, first-bounce power remains 4.2 dB below LOS at the worst corner and >15 dB below over the central two-thirds of the room; for the desk profile (Ψc = 15◦ ) no first-bounce path falls inside the FoV at all (floor bounces arrive from below an upfacing detector). The physical reason is geometric: a downpointing source illuminates predominantly the floor, whose bounce cannot couple into an up-facing receiver. (ii) Delay spread stays below 0.13 ns, so intersymbol interference at the 150-MBd symbol rate is negligible (τrms Rs < 0.02; <0.2 even if higher-order bounces multiplied τrms several-fold). (iii) The diffuse term slightly aids corner coverage power. (iv) Tilted receivers, evaluated with the same patch model: a corner terminal tilted 45◦ toward the luminaire improves K to 7.1 dB; tilted 30–45◦ away, the luminaire leaves the FoV entirely and the link becomes diffuse-only—received power then falls below the OOK threshold long before ISI matters (diffuse-only τrms ≤ 0.34 ns, τrms Rs ≤ 0.05). Tilt is therefore a coverage/outage phenomenon to be handled by handover at network level, not an ISI regime. Scope: first bounce only; higher-order bounces add a weaker, longer tail and are the remaining extension of the released channel-model class. V. R ECEIVER R EALIZABILITY AND THE C OMPLETE N OISE B UDGET A. The Gain-Bandwidth/Capacitance Constraint A transimpedance stage of feedback resistance Rf over total input capacitance CT (photodiode + amplifier input) built on an amplifier of gain-bandwidth product GBW achieves closedloop bandwidth [10], [11], [70] s GBW . (2) B≈ 2πRf CT We claim no novelty for (2): it is the transimpedance limit of the broadband-circuits literature [10], the same bound that 2 appears as RT ≤ AωA /(CT ω3dB ) in [11] and descends from the classical optical-receiver analyses of [71]–[73]. What is novel here is only its use as a binding admissibility test on a published link budget, applied before any rate is quoted. VLC-specific front ends that confront the same constraint— regulated-cascode topologies for high-capacitance photodiodes [74] and bandwidth-extension techniques for visiblelight receivers [75]—show that the trade is live in this field and not merely textbook. Fig. 3(a) plots this frontier against photodiode area (fully depleted 150-µm PIN, Cd = εSi A/w). The parameter set assumed in the earlier version of this work—A = 1 cm2 , Rf = 10 kΩ, B = 250 MHz—sits a factor >5 above the frontier even at GBW = 8 GHz and is unrealizable; this error class is common in linkbudget studies and is exactly what the frontier chart exposes. The revised design point, A = 7 mm2 (Cd = 4.9 pF, CT = 6.5 pF), Rf = 5 kΩ, B = 150 MHz, requires GBW = 2πRf CT B 2 = 4.6 GHz—within reach of modern

JOURNAL OF LIGHTWAVE TECHNOLOGY

Data source

Unipolar M-PAM + FEC encoder

Linear LD driver

7

500-mW LD (650 nm)

Holographic diffuser Φ1/2 , Td

LOS + first-bounce channel

7-mm2 PIN + lens/concentrator

TIA Rf , CT in , vn

PAM decision + FEC decoder

Data sink

optical, free space

Fig. 1. Modeled transceiver chain. The hardware prototype realized this chain with a CH340 UART driver and threshold receiver (megabit class); the model assumes the wideband linear chain whose realizability Section V enforces. The return direction for acknowledgements is a symmetric link on a separated wavelength (Section VIII).

DSRZHUUDWLR 1/2 = 60  c = 60

EUPVGHOD\VSUHDGILUVWERXQFH   K /26GLIIXVHG%

\P









 



rms QV



 











 



 



[P









[P





Fig. 2. First-bounce multipath over the desk plane of a 5 × 5 × 3-m room (source: ceiling center, Φ1/2 = 60◦ ; receiver: up-facing, Ψc = 60◦ ; ρ = 0.7): (a) LOS-to-diffuse power ratio K; (b) rms delay spread. Worst case K = 4.2 dB at the room corner with τrms = 0.13 ns, i.e. τrms Rs < 0.02 at 150 MBd.

wideband amplifiers—and recovers collection area optically: the étendue-limited concentrator of the 15◦ desk profile yields an effective aperture A g = 235 mm2 .

B. Noise Budget With photocurrent Iph = RPr and background current Ibg , the input-referred noise over bandwidth B is [11], [18], [113] [70, Chs. 4–5] σ 2 = 2q(Iph +Ibg )B + +

4kB T B + i2n B Rf

(2πCT vn )2 3 2 B + RIN · Iph B, 3

(3)

√ with amplifier input noise densities in = 3 pA/ Hz, vn = √ 1 nV/ Hz (representative of wideband TIAs), and laser RIN = −120 dB/Hz as the default for a multimode visible LD, swept over −110 to −130 dB/Hz in Section IX because vendor data for low-cost visible LDs is scarce—an uncertainty we propagate rather than hide. The RIN term imposes a distanceindependent SNR ceiling 1/(RIN·B): 38.2 dB at −120 dB/Hz but only 28.2 dB at −110 dB/Hz, which brackets the 16-PAM thresholds and is why the highest-order results carry an explicit RIN sensitivity figure. Fig. 3(b) shows the budget at the 14-m signal level: at B = 150 MHz the amplifier terms contribute 69% of the total variance and the naive budget overstates SNR by 5.1 dB. Table II lists all parameters.

TABLE II S YSTEM PARAMETERS (R EVISED , R EALIZABILITY-C HECKED ) Symbol

Meaning

Value

PLD Td Φ1/2 Ds RIN A R Ts n, Ψc B Rf , CT GBW in , vn Ibg FEC

LD optical power diffuser transmission diffuser semi-angle diffuser (source) diameter laser intensity noise photodiode area (150-µm PIN) responsivity (650 nm) filter transmission concentrator index, FoV receiver bandwidth = symbol rate TIA feedback R, input C required amplifier GBW (2) TIA input noise densities background photocurrent HD tier / SD tier thresholds

500 mW 0.9 20◦ (desk) / 60◦ 25 mm −120 dB/Hz (−110.. − 130) 7 mm2 (Cd =4.9 pF) 0.45 A/W 0.85 1.5; 15◦ /60◦ 150 MHz 5 kΩ, 6.5 pF 4.6 GHz √ 3 pA, 1 nV/ Hz 20 (desk) / 100 µA 3.8×10−3 / 2×10−2

VI. M ODULATION AND FEC A. Unipolar M-PAM The PHY uses baseband unipolar M -PAM at symbol rate Rs = B. This choice is deliberate: intensity levels are directly the signal (no electrical subcarrier), so there is no double-sideband bandwidth penalty and no separatelyaccounted DC bias—the unipolarity cost is inherent in the average-optical-power constraint of the constellation. With levels 0, . . . , (M −1) δ, mean fixed at RPr , and electrical SNR

JOURNAL OF LIGHTWAVE TECHNOLOGY

8

EQRLVHEXGJHWDWWKHPVLJQDO VKRW 4kT/Rf

 

1RLVHYDULDQFH$2

$FKLHYDEOH7,$EDQGZLGWK0+]

DEDQGZLGWKYVGHWHFWRUDUHD

 *%: *+]Rf N *%: *+]Rf N *%: *+]Rf N YDVVXPSWLRQLQIHDVLEOH WKLVZRUN

 



in2 B

(2 CT vn )2 B 3 /3 5,1 WRWDO

     







5HFHLYHUEDQGZLGWKB0+]

3KRWRGLRGHDUHDPP2

Fig. 3. (a) TIA-achievable bandwidth versus photodiode area ((2), 150-µm PIN): the v1 assumption sits far above every frontier; the revised design point lies on the 8-GHz/5-kΩ frontier with margin. (b) Complete noise budget versus bandwidth at the 14-m received signal: above ∼120 MHz the capacitance-driven (2πCT vn )2 B 3 /3 term dominates; the naive shot+4kT /Rf budget (bottom two curves) captures less than a third of the total at 150 MHz.

B. Specified FEC Tiers Two concrete tiers, with the code family, overhead, and threshold stated: (i) HD tier: G.975.1-class concatenated harddecision FEC, 7% overhead, pre-FEC threshold 3.8×10−3 for < 10−15 output [22]; (ii) SD tier: LDPC-class soft-decision FEC, 20% overhead, threshold 2×10−2 —the pairing standard in optical transport. Net rate of modulation M under tier t: Rnet = B log2 M (1 − ρt ) when Pb ≤ thrt . Every analytic curve is checked by Monte-Carlo simulation with error-countbased 95% confidence intervals (Fig. 4); points are plotted only where at least ten error events were counted. VII. E YE S AFETY OF THE D IFFUSED S OURCE The diffuser converts the LD into an extended source of diameter Ds = 25 mm. Following the extended-source provisions of IEC 60825-1 [21], we compare the power collectable by a 7-mm pupil at viewing distance r, Ppup = (m+1)Pt Apup /2πr2 , against the Class-1 accessible-emission limit 0.39 mW × C6 with C6 = α/1.5 mrad, α = clamp(Ds /r, 1.5, 100) mrad. Fig. 5 shows the result for both diffuser angles: the pupil power tracks a factor ≥7.8 below the limit at all distances ≥10 cm—the 20◦ diffuser’s worst margin occurs nearest the source, where the α cap binds. This is an engineering assessment, not certification (aided viewing/Class1M conditions remain for a product process), but it establishes that the 450-mW radiated power used throughout the link budget is compatible with Class-1-style naked-eye exposure,

  6')(&



%LWHUURUUDWH

γ = (RPr )2 /σ 2 , Gray mapping gives  √  γ 2(M − 1) Pb ≈ Q , (4) M log2 M M −1 √ which reduces to Q( γ) for OOK (M =2). The (M − 1) penalty is steep—each doubling of M costs ∼6–7 dB—which is the honest price of IM/DD unipolarity; comparisons with DCO-OFDM bit-loading, which can trade this differently at the cost of bias and clipping analysis [114], [115], are future work on top of the released framework.



+')(&

    

22. 3$0 3$0 3$0



 











(OHFWULFDO615G%





Fig. 4. Unipolar M -PAM BER versus electrical SNR: analysis (lines) and Monte-Carlo counts with 95% confidence intervals (markers; ≥10 error events per point, up to 2 × 107 bits). Dashed/ dotted lines: HD and SD FEC thresholds.

closing the safety-performance loop quantitatively. One singlefault condition must be stated explicitly: a cracked or detached diffuser leaves a collimated 500-mW visible emitter—Class 4. Any physical realization therefore requires a diffuser-integrity interlock that removes drive current on fault, and every “eyesafer” statement in this paper is conditioned on that interlock. Two further scope notes follow the classification literature. First, a laser-illuminated diffuse source may fall to be assessed as a lamp rather than as a laser product, under IEC 62471 [116] and its visible-source part [117], and the boundary between the two regimes for laser-illuminated luminaires is itself the subject of dedicated analysis [118], [119]; a product realization must determine which applies. Second, eye-safe operation of laser white lighting at multi-Gb/s is not hypothetical—it has been demonstrated with tricolour R/G/B diodes [120].

JOURNAL OF LIGHTWAVE TECHNOLOGY

9

)XOOPRGHOGHVN5; c = 15  1DLYHEXGJHWVKRW4kT/Rf RQO\ )XOOPRGHOPRELOH5; c = 60





(OHFWULFDO615G%

3RZHUP:

  SXSLOSRZHU 1/2 = 20 SXSLOSRZHU 1/2 = 60 &ODVV$(/H[WHQGHGVRXUFH

   

3$0+'

  

22.+'



  







9LHZLQJGLVWDQFHP





Fig. 5. Extended-source eye-safety assessment: 7-mm-pupil collected power versus the Class-1 extended-source AEL as a function of viewing distance, for both diffuser angles at Pt = 450 mW. Minimum margin 7.8× (at 10 cm, 20◦ ). Applications, IPv4/UDP/TCP, FlowMonitor (stock ns-3) LifiNetDevice LifiMac stop-and-wait ARQ (lower bound) lifi module LifiPhy full noise (3), PAM BER LifiChannel + LifiOpticalChannelModel

Mobility Models

Fig. 6. ns-3 lifi module (v2). The MAC is an analyzable lower bound, not a standards model; Section VIII relates it to IEEE 802.11bb and ITU-T G.9991.

VIII. C ROSS -V ERIFIED S IMULATION F RAMEWORK The framework comprises the ns-3 module of Fig. 6 and an independent Python engine implementing (1)–(4), the patch-integration multipath, and the safety assessment. Automated tests pin the two implementations together at 0.1% on reference values (channel gain, SNR, thresholds). We state its epistemic status precisely, in the established sense of the verification-and-validation literature [121]–[125]: agreement between the implementations, and between analysis and Monte-Carlo counting, is verification—evidence that the equations are solved correctly, and protection against transcription and coding error. It is not validation, which concerns whether the equations describe the physical system (Lambertian diffusion, linear LD response above threshold, the RIN figure, speckle averaging), which only measurements can provide (Section X). Duplexing and MAC. For the fixed desk terminals that carry all network results, the return link is specified as a symmetric diffused-LD transceiver on a separated wavelength (e.g., 850 nm): same radiated power class, same up-facing-luminaire geometry, hence the same link budget and the same extendedsource safety assessment as the downlink—the prototype itself

 







/LQNGLVWDQFHdP





Fig. 7. Electrical SNR versus distance (Φ1/2 = 20◦ ): full budget (3) versus the naive shot+4kT /Rf budget at identical parameters (5.1-dB gap), and the wide-FoV mobile profile. Dotted: OOK and 16-PAM HD thresholds.

was symmetric in exactly this sense. Acknowledgements therefore travel at the downlink line rate, and the saturation formula below uses that specification. (Wide-FoV mobile terminals would need a reduced-rate infrared uplink whose budget and terminal-side safety are future work; the network results are stated for fixed-terminal cells only.) The stop-and-wait MAC with bounded retransmission is chosen as a fully analyzable lower bound whose saturation goodput tpkt /(tpkt +tack +2τp ) any practical aggregating protocol (802.11bb block-ack [126], G.9991 data-link layer [102]) will exceed. The module is a physical-layer research vehicle beneath standard stacks, not a standards implementation; mapping the PHY abstractions onto an 802.11bb MAC inside ns-3’s WiFi model is planned and the module’s interfaces were designed for it. IX. R ESULTS A. SNR: the Cost of Realism Fig. 7 quantifies the central methodological point: at 14 m the full budget yields 15.3 dB where the naive budget claims 20.3 dB. Five decibels is the difference between 4-PAM and OOK at range—rate factors, not rounding. Compared with the earlier unconstrained receiver (which additionally assumed an unrealizable 250-MHz/1-cm2 /10-kΩ front end), the honest accounting lowers projected net rate at 14 m from 930 to 140–240 Mb/s: a factor 3.9–6.6 inflation that the realizability discipline of Section V exists to prevent. B. Error Rate, Reach, and Rate Figs. 8 and 9 give the revised staircases. At the prototype’s demonstrated 14 m: OOK/HD 140 Mb/s with 6.7 dB of margin, or 4-PAM/SD 240 Mb/s—the latter at zero margin, since the 4-PAM/SD reach is exactly 14.0 m (a 1-dB implementation allowance retracts it to 13.2 m); the robust 14-m operating point is therefore the OOK/HD figure. At 5 m: 16-PAM 480 (SD) to 558 (HD) Mb/s. OOK persists to 20.7 m (HD) / 23.6 m (SD). The Shannon gap of 2.5–4 b/s/Hz locates the headroom

JOURNAL OF LIGHTWAVE TECHNOLOGY

10



1HWUDWH+')(&0ELWV

3UH)(&ELWHUURUUDWH

       

22. 3$0 3$0 3$0

   









/LQNGLVWDQFHdP





Fig. 8. Pre-FEC BER versus distance per modulation (desk profile). HD-tier reaches: 5.3 m (16-PAM), 8.0 m (8-PAM), 12.1 m (4-PAM), 20.7 m (OOK); SD tier extends these by 14–22% (23.6, 14.0, 9.4, and 6.5 m).



1HWGDWDUDWH0ELWV

5,1QR5,1 5,1G%+] 5,1G%+] 5,1G%+]

 6' +'

     







/LQNGLVWDQFHdP





Fig. 10. HD-tier net rate versus distance as laser RIN varies from −130 to −110 dB/Hz. At −110 dB/Hz the RIN ceiling (28.2 dB) eliminates 16-PAM entirely; measuring the actual device RIN is therefore a gating step for any high-order deployment.

source supports OOK essentially wall to wall. Multi-luminaire tiling and handover atop these cells are exactly the studies the ns-3 module exists to host.

 E. Fidelity and Network Behavior





6KDQQRQERXQG0+] $GDSWLYH3$0+')(&2+ $GDSWLYH3$06')(&2+ 3URWRW\SH8$57OLPLWHG









/LQNGLVWDQFHdP





Fig. 9. Model-predicted net rate versus distance under both FEC tiers, with the Shannon bound and the prototype’s measured UART-limited point. All staircase values are predictions per Section X.

accessible to DCO-OFDM bit-loading and stronger codes; unlike the earlier version we do not treat that observation as a contribution, merely as the map’s edge. C. RIN Sensitivity Fig. 10 propagates the least-known parameter. The conclusions that survive the full RIN range: OOK/4-PAM reaches move by <0.5 m; the 16-PAM tier is fragile (absent at −110 dB/Hz). We flag the highest-order predictions accordingly rather than assert them. D. Coverage Fig. 11 maps the room with the mobile profile. The wideFoV terminal cannot close a 14-m boresight link (Fig. 7); its regime is the under-luminaire cell, where the 20◦ source supports high-order PAM over a ∼1-m-radius spot and the 60◦

The image experiment of the prototype is reproduced bitexactly through the PAM chain (Figs. 12–13). Network-level curves (Figs. 14–15) carry the stated MAC lower bound over the revised PHY rates: saturation at 93–96% of line rate and sub-millisecond 99th-percentile latency at 70% load. These inherit the PHY model’s epistemic status; they demonstrate the framework and bound the protocol overhead, and they will sharpen (upward in throughput) under 802.11bb-style aggregation. F. A Fair Emitter Comparison At matched radiated power (450 mW, eye-safe for both emitter types), matched receiver, and honest bandwidths— 15 MHz for a phosphor-white LED lamp versus the LD’s (receiver-limited) 150 MHz— the model gives 14-m HD rates of ∼28 Mb/s (LED, 4-PAM at 15 MHz) versus 140 Mb/s (LD): a factor ∼5 attributable to bandwidth, not a contrived power gap. We note explicitly that engineered LED systems with WDM and adaptive loading exceed both figures at short range [5]; the LD case rests on retaining its bandwidth advantage at range and modest complexity, not on beating LED flagships. X. W HAT I S M EASURED , W HAT I S P REDICTED , AND W HAT I S N EXT Table III is the paper’s contract with the reader: every headline number labeled by evidence class. The prototype validates feasibility and geometry; it cannot validate the wideband model, whose observation of zero errors at UART rates is consistent with but not discriminating between

JOURNAL OF LIGHTWAVE TECHNOLOGY

11

D 1/2 = 20 PRELOH5;



E 1/2 = 60 PRELOH5;













22

    



.+



'



(OHFWULFDO615G%

\P







 









[P















[P







Fig. 11. Desk-plane SNR maps for the mobile receiver profile (Ψc = 60◦ ): (a) Φ1/2 = 20◦ concentrates a small high-rate cell under the luminaire; (b) Φ1/2 = 60◦ trades peak SNR for wall-to-wall OOK coverage. The dashed contour is the OOK/HD boundary.

5HFRQVWUXFWLRQ3615G%



TABLE III C LAIMS L EDGER : E VIDENCE C LASS OF E VERY H EADLINE N UMBER

   

3$0 3$0 3$0

 







(OHFWULFDO615G%



Claim

Class

Basis

End-to-end data/voice/image over 14 m Goodput ceiling 1.59 Mb/s (CH340 path) Zero observed errors at 14 m, UART rates

measured measured/analytic

SNR 15.3 dB at 14 m, B=150 MHz 140–240 Mb/s net at 14 m 480–558 Mb/s net at 5 m

predicted predicted predicted (RINfragile) computed

prototype [23] UART framing (Fig. 16) model predicts BER < 10−12 there full budget (3) PAM + FEC tiers Fig. 10

Multipath K ≥ 4.2 dB, τrms ≤ 0.13 ns Class-1-style margin ≥7.8×

measured; consistent

computed

first-bounce integration Section VII



Fig. 12. Bit-exact reconstruction PSNR of the prototype’s 69 × 71 test image versus SNR (uncoded PAM chain).

of static holographic diffusers under coherent illumination; higher-order reflection bounces; the mobile-terminal infrared uplink; and an 802.11bb MAC mapping. XI. C ONCLUSION

parameterizations—a limitation we state rather than blur. The experimental path this framework specifies for its own confirmation, in order of information gained per effort: (i) measured end-to-end frequency response of an LD-driver/ PIN-TIA chain built to Table II; (ii) measured RIN of the deployed LD class; (iii) BER-versus-distance at 50–150 MBd OOK/ 4-PAM, off the error floor; (iv) received-plane scans versus the Lambertian fit of the actual diffuser, including speckle statistics. Items (i)–(iii) require only commodity instruments (VNA-lite, BER tester at 150 MBd); none require the multiGHz apparatus of record-chasing demonstrations. Model scope that remains open: LD nonlinearity and clipping for high-order PAM; speckle and modal noise statistics

This paper asked what a diffused-beam LD LiFi link supports when every optimistic assumption is replaced by an enforceable constraint: a buildable TIA, a complete noise budget with RIN, unipolar PAM without spectral sleight of hand, specified FEC, computed multipath, and a quantitative safety margin at the actual operating power. The answer — 140–240 Mb/s at 14 m, half a gigabit at 5 m, OOK to 21 m, with the top tier explicitly RIN-fragile—is smaller than idealized budgets suggest by factors of 3.9–6.6, and that correction, made auditable and reproducible through an open cross-verified ns-3/Python framework, is the contribution. The framework’s claims ledger separates its one measured anchor from its predictions, and its specified confirmation experi-

JOURNAL OF LIGHTWAVE TECHNOLOGY

12

E3$0G% %(5 H3615 G%

DVRXUFHLPDJH

F3$0G% %(5 H3615 G%

G3$0G% %(5 H3615 G%

Fig. 13. Uncoded 16-PAM image transfer at three SNRs (b–d); with either FEC tier above threshold, delivery is bit-exact, matching the prototype’s observed error-free transfers at its (deeply threshold-clearing) UART rates.



d P3$00EV d P3$00EV d P22.0EV

0ELWVDV\PSWRWH



*RRGSXW0ELWV

*RRGSXW0ELWV





  

 &+#0EDXG &+#NEDXG

   





 











2IIHUHGORDGIUDFWLRQRI3+<UDWH



Fig. 14. Goodput versus offered load, stop-and-wait lower bound, full-duplex ACK wavelength, 1500-B packets, post-FEC residual BER 10−9 .





3D\ORDGVL]HE\WHV



Fig. 16. The prototype’s measured regime: CH340 UART goodput ceiling (1.59 Mb/s at 1500-B payloads). Timing-based estimates exceeding this (the thesis’s 127 Mb/s [23]) are processing-side artifacts, not link throughput.

wireless modeling literature, where the gap this paper closes for one system remains open for many.



&')

R EFERENCES

  d P d P d P

 

 

3DFNHWODWHQF\PV



Fig. 15. Packet-latency CDFs at 70% load. 99th-percentile: 0.17 ms (5 m), 0.33 ms (10 m), 0.67 ms (14 m).

ments are deliberately modest. We believe this accounting discipline—realizability frontiers, complete budgets, evidenceclass labeling—transfers directly to the broader indoor-optical-

[1] Y.-C. Chi, D.-H. Hsieh, C.-T. Tsai, H.-Y. Chen, H.-C. Kuo, and G.R. Lin, “450-nm GaN laser diode enables high-speed visible light communication with 9-Gbps QAM-OFDM,” Optics Express, vol. 23, no. 10, pp. 13 051–13 059, 2015. [2] T.-C. Wu, Y.-C. Chi, H.-Y. Wang, C.-T. Tsai, Y.-F. Huang, and G.-R. Lin, “White-lighting communication with a Lu3 Al5 O12 :Ce3+ /CaAlSiN3 :Eu2+ glass covered 450-nm InGaN laser diode,” Journal of Lightwave Technology, vol. 36, no. 9, pp. 1634–1643, 2018. [3] Y.-C. Chi, Y.-F. Huang, T.-C. Wu, C.-T. Tsai, L.-Y. Chen, H.-C. Kuo, and G.-R. Lin, “Violet laser diode enables lighting communication,” Scientific Reports, vol. 7, no. 1, p. 10469, 2017. [4] Y.-L. Gao, Z.-Y. Wu, Z.-K. Wang, and J. Wang, “A 1.34-Gb/s realtime Li-Fi transceiver with DFT-spread-based PAPR mitigation,” IEEE Photonics Technology Letters, vol. 30, no. 16, pp. 1447–1450, 2018. [5] R. Bian, I. Tavakkolnia, and H. Haas, “15.73 Gb/s visible light communication with off-the-shelf LEDs,” Journal of Lightwave Technology, vol. 37, no. 10, pp. 2418–2424, 2019. [6] H. Haas, L. Yin, Y. Wang, and C. Chen, “What is LiFi?” Journal of Lightwave Technology, vol. 34, no. 6, pp. 1533–1544, 2016. [7] H. Haas, L. Yin, C. Chen, S. Videv, D. Parol, E. Poves, H. Alshaer, and M. S. Islim, “Introduction to indoor networking concepts and challenges in LiFi,” Journal of Optical Communications and Networking, vol. 12, no. 2, pp. A190–A203, 2020.

JOURNAL OF LIGHTWAVE TECHNOLOGY

[8] T. Koonen, “Indoor optical wireless systems: Technology, trends, and applications,” Journal of Lightwave Technology, vol. 36, no. 8, pp. 1459–1467, 2018. [9] H. Ahmad, S. M. T. Gillani, T. Omer, and S. Aslam, “Laser-diode LiFi with diffused-beam optics: System-level modeling and a cross-validated ns-3 simulation framework,” arXiv:2608.10950 [eess.SP], 2026, earlier version of the present work; superseded by this article. [10] E. Säckinger, “The transimpedance limit,” IEEE Transactions on Circuits and Systems I: Regular Papers, vol. 57, no. 8, pp. 1848–1856, 2010. [11] S. S. Mohan, M. del Mar Hershenson, S. P. Boyd, and T. H. Lee, “Bandwidth extension in CMOS with optimized on-chip inductors,” IEEE Journal of Solid-State Circuits, vol. 35, no. 3, pp. 346–355, 2000. [12] T. Komine and M. Nakagawa, “Fundamental analysis for visiblelight communication system using LED lights,” IEEE Transactions on Consumer Electronics, vol. 50, no. 1, pp. 100–107, 2004. [13] P. Mach, Z. Becvar, M. Najla, and S. Zvanovec, “Combination of visible light and radio frequency bands for device-to-device communication,” in Proc. Proc. IEEE 28th Annual International Symposium on Personal, Indoor and Mobile Radio Communications (PIMRC), 2017, pp. 1–7. [14] O. Maraqa, S. Aboagye, and T. M. N. Ngatched, “Optical STAR-RISaided VLC systems: RSMA versus NOMA,” IEEE Open Journal of the Communications Society, vol. 5, pp. 430–441, 2024. [15] A. Surampudi and R. K. Ganti, “Interference characterization in downlink Li-Fi optical attocell networks,” Journal of Lightwave Technology, vol. 36, no. 16, pp. 3211–3228, 2018. [16] H. Tabassum and E. Hossain, “Coverage and rate analysis for coexisting RF/VLC downlink cellular networks,” IEEE Transactions on Wireless Communications, vol. 17, no. 4, pp. 2588–2601, 2018. [17] E. Säckinger, Broadband Circuits for Optical Fiber Communication. Wiley, 2005. [18] K. Petermann, Laser Diode Modulation and Noise. Kluwer Academic, 1988. [19] J. R. Barry, J. M. Kahn, W. J. Krause, E. A. Lee, and D. G. Messerschmitt, “Simulation of multipath impulse response for indoor wireless optical channels,” IEEE Journal on Selected Areas in Communications, vol. 11, no. 3, pp. 367–379, 1993. [20] J. B. Carruthers and J. M. Kahn, “Modeling of nondirected wireless infrared channels,” IEEE Transactions on Communications, vol. 45, no. 10, pp. 1260–1268, 1997. [21] IEC, “Safety of laser products—part 1: Equipment classification and requirements,” IEC 60825-1:2014, 2014. [22] ITU-T, “Forward error correction for high bit-rate DWDM submarine systems,” Recommendation G.975.1, 2004. [23] H. Ahmad, S. M. T. Gillani, and T. Omer, “High speed LiFi transceiver,” BS thesis, Department of Electrical Engineering, Bahria University, Islamabad, 2019. [24] H. Ahmad, S. M. T. Gillani, T. Omer, T. Hassan, S. Aslam, and S. U. Ali, “Futuristic short range optical communication: A survey,” in Proc. International Conference on Information Science and Communication Technology (ICISCT), 2019. [25] A. Celik, I. Romdhane, G. Kaddoum, and A. M. Eltawil, “A top-down survey on optical wireless communications for the internet of things,” IEEE Communications Surveys & Tutorials, vol. 25, no. 1, pp. 1–45, 2023. [26] Y. Ata, F. M. Al-Sallami, M. C. Gokce, A. M. Vegni, S. Rajbhandari, and Y. Baykal, “Optical wireless communication in atmosphere and underwater: Statistical models, improvement techniques, and recent applications,” IEEE Communications Surveys & Tutorials, vol. 28, pp. 4248–4284, 2026. [27] C.-W. Chow, “Recent advances and future perspectives in optical wireless communication, free space optical communication and sensing for 6G,” Journal of Lightwave Technology, vol. 42, no. 11, pp. 3972– 3980, 2024. [28] A. Krishnamoorthy, H. Safi, O. Younus, H. Kazemi, I. N. O. Osahon, M. Liu, Y. Liu, S. Babadi, R. Ahmad, A. Ihsan, B. Majlesein, Y. Huang, J. Herrnsdorf, S. Rajbhandari, J. J. D. McKendry, I. Tavakkolnia, H. Caglayan, H. Helmers, G. Turnbull, I. D. W. Samuel, M. D. Dawson, R. Schober, and H. Haas, “Optical wireless communications: Enabling the next-generation network of networks,” IEEE Vehicular Technology Magazine, vol. 20, no. 2, pp. 20–39, 2025. [29] J. Wang, J. Hu, C. Guan, Y. Hou, Z. Xu, L. Sun, Y. Wang, Y. Zhou, B. S. Ooi, J. Shi, Z. Li, J. Zhang, N. Chi, S. Yu, and C. Shen, “Highspeed GaN-based laser diode with modulation bandwidth exceeding 5 GHz for 20 Gbps visible light communication,” Photonics Research, vol. 12, no. 6, pp. 1186–1193, 2024.

13

[30] H. Jia, J. Hu, Z. Xu, Z. Gu, Z. Yang, Z. Li, Y. Zhou, J. Shi, Z. Li, J. Zhang, N. Chi, and C. Shen, “High-speed blue laser diodes with InGaN quantum barrier for beyond 36 Gbps visible light communications,” Laser & Photonics Reviews, vol. 19, no. 11, p. 2401751, 2025. [31] J. Hu, Z. Gu, H. Jia, Z. Yang, Z. Li, J. Wu, L. Sun, A. Sun, O. Huang, C. Xia, B. S. Ooi, J. Shi, Z. Li, J. Zhang, S. Yu, N. Chi, and C. Shen, “InGaN-based high-speed mini laser diode surpasses PAM-4 visible light links by over 30 Gbps,” Science China Information Sciences, vol. 69, no. 3, p. 132405, 2026. [32] Z. Gu, Z. Yang, F. Du, D. Fan, B. Xiao, J. Hu, H. Jia, S. Wang, Y. Pan, J. Zhang, N. Chi, A. Kafar, P. Perlin, and C. Shen, “Advances of highspeed GaN-based laser diodes for emerging applications,” Progress in Quantum Electronics, vol. 108, p. 100633, 2026. [33] Y. Hou, Y. Wang, Z. Li, M. Liu, S. Yi, X. Wang, L. Xia, G. Liu, J. Shi, Z. Li, J. Zhang, N. Chi, T. K. Ng, B. S. Ooi, and C. Shen, “Tutorial on laser-based visible light communications [Invited],” Chinese Optics Letters, vol. 22, no. 9, p. 092502, 2024. [34] Z. Luo, X. Lin, Z. Lu, J. Shi, C. Shen, Z. Li, J. Zhang, Z. He, H. Chen, S. Yu, and N. Chi, “113Gbps rainbow visible light laser communication system based on 10λ laser WDM emitting module in fiber-free spacefiber link,” Optics Express, vol. 32, no. 2, pp. 2561–2573, 2024. [35] N. Chi, W. Niu, Y. Zhou, J. Wang, H. Chen, Z. He, J. Li, Z. Xu, X. Lin, Z. Luo, Z. Lu, J. Zhang, C. Shen, Z. Li, J. Shi, and S. Yu, “Enabling technologies to achieve beyond 500 Gbps optical intra-connects based on WDM visible light laser communication,” Journal of Lightwave Technology, vol. 43, no. 4, pp. 1843–1854, 2025. [36] Y. Zhou, X. Lin, Z. Xu, Z. Lu, Y. Huang, Z. Luo, H. Zhang, L. Sun, J. Ye, H. Chen, Z. He, J. Li, J. Zhang, C. Shen, S. Yu, and N. Chi, “Beyond 600 Gbps optical interconnect utilizing wavelength division multiplexed visible light laser communication [Invited],” Chinese Optics Letters, vol. 23, no. 5, p. 050002, 2025. [37] Z. Lu, Z. Li, X. Lin, J. Cai, F. Li, Z. Xu, L. Wang, Y. Zhou, C. Shen, J. Zhang, and N. Chi, “170 Gbps PDM underwater visible light communication utilizing a compact 5-λ laser transmitter and a reciprocal differential receiver,” Photonics Research, vol. 13, no. 6, p. 1654, 2025. [38] C. Chen, S. Das, S. Videv, A. Sparks, S. Babadi, A. Krishnamoorthy, C. Lee, D. Grieder, K. P. Hartnett, P. Rudy, J. W. Raring, M. Najafi, V. K. Papanikolaou, R. Schober, and H. Haas, “100 Gbps indoor access and 4.8 Gbps outdoor point-to-point LiFi transmission systems using laser-based light sources,” Journal of Lightwave Technology, vol. 42, no. 12, pp. 4146–4157, 2024. [39] Y.-C. Chi, D.-H. Hsieh, C.-Y. Lin, H.-Y. Chen, C.-Y. Huang, J.-H. He, B. Ooi, S. P. DenBaars, S. Nakamura, H.-C. Kuo, and G.-R. Lin, “Phosphorous diffuser diverged blue laser diode for indoor lighting and communication,” Scientific Reports, vol. 5, p. 18690, 2015. [40] F. Rahman, “Diode laser-excited phosphor-converted light sources: a review,” Optical Engineering, vol. 61, no. 6, p. 060901, 2022. [41] J. P. Lazzaro, X. Wang, C. J. Carver, N. R. Shade, H. Schwartz, Y. Liu, X. Zhou, J. Liu, and E. R. Fossum, “Multi-Gbps visible light communication with high-efficiency, low-speckle contrast white laser light,” Optical Review, vol. 33, no. 1, pp. 135–141, 2026. [42] V. Kumar, P. Sharma, and D. S. Mehta, “Design and development of speckle-free high-power laser-driven phosphor converted compact automotive headlamp module,” Journal of Physics: Photonics, vol. 6, no. 2, p. 025008, 2024. [43] O. B. Jensen, J. Xu, and M. L. Jakobsen, “Dependence of speckle contrast in laser lighting on laser and phosphor characteristics,” Journal of Luminescence, vol. 286, p. 121415, 2025. [44] M. D. Soltani, E. Sarbazi, N. Bamiedakis, P. de Souza, H. Kazemi, J. M. H. Elmirghani, I. H. White, R. V. Penty, H. Haas, and M. Safari, “Safety analysis for laser-based optical wireless communications: A tutorial,” Proceedings of the IEEE, vol. 110, no. 8, pp. 1045–1072, 2022. [45] K. Schulmeister, “Classification of extended source products according to IEC 60825-1,” in ILSC 2015: Proceedings of the International Laser Safety Conference. Laser Institute of America, 2015, pp. 271–280. [46] T. Wheatley, Y. Meng, Y. Wang, W. Bi, K. Gutoehrlein, and A. Frederiksen, “Challenges, development, and use of the laser safety standard IEC 60825-1:2014,” Journal of Laser Applications, vol. 38, no. 1, p. 012043, 2026. [47] P. Qiu, S. Zhu, Z. Jin, X. Zhou, X. Cui, and P. Tian, “Beyond 25 Gbps optical wireless communication using wavelength division multiplexed LEDs and micro-LEDs,” Optics Letters, vol. 47, no. 2, pp. 317–320, 2022.

JOURNAL OF LIGHTWAVE TECHNOLOGY

[48] L. Tang, Y. Wu, Z. Cheng, D. Teng, and L. Liu, “Over 23.43 Gbps visible light communication system based on 9 V integrated RGBP LED modules,” Optics Communications, vol. 534, p. 129317, 2023. [49] X. Liu, Z. Wei, M. Li, L. Wang, Z. Liu, C. Yu, L. Wang, Y. Luo, and H. Y. Fu, “Experimental investigation of 16.6 Gbps SDM-WDM visible light communication based on a neural network receiver and tricolor mini-LEDs,” Optics Letters, vol. 46, no. 12, pp. 2888–2891, 2021. [50] Z. Jin, L. Yan, S. Zhu, X. Cui, and P. Tian, “10-Gbps visible light communication in a 10-m free space based on violet series-biased micro-LED array and distance adaptive pre-equalization,” Optics Letters, vol. 48, no. 8, pp. 2026–2029, 2023. [51] Z. Rao, X. Shan, G. Wang, Z. Jin, R. Lin, X. Cui, R. Liu, K. Xu, and P. Tian, “10.5 Gbps visible light communication systems based on Cplane freestanding GaN micro-LED,” Journal of Lightwave Technology, vol. 42, no. 13, pp. 4360–4364, 2024. [52] Z. Wei, Z. Liu, X. Liu, L. Wang, L. Wang, C. Yu, and H. Y. Fu, “8.75 Gbps visible light communication link using an artificial neural network equalizer and a single-pixel blue micro-LED,” Optics Letters, vol. 46, no. 18, pp. 4670–4673, 2021. [53] Z. Jin, H. Ying, Z. Zhang, R. Lin, E. Gu, and P. Tian, “Toward 20-Gbps single-chip space division multiplexing visible light communication based on a low-crosstalk micro-LED array,” Journal of Lightwave Technology, vol. 43, no. 24, pp. 10 960–10 967, 2025. [54] J. Ai, Z. Zhang, T. Deng, D. Shen, Y. Xiao, Z. Jin, T. Ren, X. Cui, and P. Tian, “15.64 Gbps high-speed visible light communication and multiple-user secure communication employing red GaN microLED based semi-transparent photodetector,” Journal of Lightwave Technology, vol. 43, no. 15, pp. 7058–7065, 2025. [55] Z. Jin, R. Lin, X. Liu, D. Sun, B. Liu, T. Tao, F. Xu, Z. Fang, X. Cui, and P. Tian, “Green micro-LED with a bandwidth exceeding 2 GHz for 9-Gbps visible light communication based on SNR gap-dependent bit and power loading,” Journal of Lightwave Technology, vol. 43, no. 2, pp. 472–480, 2025. [56] Y. Dong, Y. Huang, X. Deng, S. Li, X. Zou, W. Pan, P. Bienstman, and L. Yan, “LED model-driven reservoir computing for signal equalization in OFDM-VLC system,” Journal of Lightwave Technology, vol. 44, no. 9, pp. 3500–3512, 2026. [57] K. Yoshida, C. Chen, H. Haas, G. A. Turnbull, and I. D. W. Samuel, “RGB-single-chip OLEDs for high-speed visible-light communication by wavelength-division multiplexing,” Advanced Science, vol. 11, no. 47, p. 2404576, 2024. [58] M. Dehghani Soltani, H. Kazemi, E. Sarbazi, T. E. H. El-Gorashi, J. M. H. Elmirghani, R. V. Penty, I. H. White, H. Haas, and M. Safari, “High-speed imaging receiver design for 6G optical wireless communications: A rate-FOV trade-off,” IEEE Transactions on Communications, vol. 71, no. 2, pp. 1024–1043, 2023. [59] E. Sarbazi, H. Kazemi, M. Crisp, T. El-Gorashi, J. Elmirghani, R. V. Penty, I. H. White, M. Safari, and H. Haas, “Design and optimization of high-speed receivers for 6G optical wireless networks,” IEEE Transactions on Communications, vol. 72, no. 2, pp. 971–990, 2024. [60] M. S. Bashir, Q. Z. Ahmed, and M.-S. Alouini, “Optimal photodetector size for high-speed free-space optics receivers,” IEEE Transactions on Wireless Communications, vol. 23, no. 11, pp. 16 390–16 403, 2024. [61] X. Liu, J.-P. M. G. Linnartz, and T. E. B. Cunha, “Modeling and modulation optimization for OWC limited by electronic and photonic bandwidth,” arXiv:2605.03976, 2026. [62] M. Portnoi, P. A. Haigh, T. J. Macdonald, F. Ambroz, I. P. Parkin, I. Darwazeh, and I. Papakonstantinou, “Bandwidth limits of luminescent solar concentrators as detectors in free-space optical communication systems,” Light: Science & Applications, vol. 10, no. 1, p. 3, 2021. [63] M. Meucci, S. Doria, A. M. Umair, D. Franchi, M. Fattori, M. Di Donato, A. Picchi, A. Pucci, M. Calamante, and J. Catani, “Efficient whitelight visible light communication with novel optical antennas based on luminescent solar concentrators,” Journal of Lightwave Technology, vol. 42, no. 7, pp. 2235–2244, 2024. [64] M. Meucci, M. Aresti, G. Cossu, L. Gilli, L. Oliviero, M. Bartolini, A. Pucci, E. Ciaramella, and J. Catani, “Comparative test of novel fluorescent optical antennas for LED- and laser-based optical wireless communications,” Advanced Optical Materials, vol. 13, no. 13, p. 2402367, 2025. [65] W. Ali, P. P. Manousiadis, D. C. O’Brien, G. A. Turnbull, I. D. W. Samuel, and S. Collins, “A gigabit VLC receiver that incorporates a fluorescent antenna and a SiPM,” Journal of Lightwave Technology, vol. 40, no. 16, pp. 5369–5375, 2022.

14

[66] Y. Wang, J. Xiong, L. Qi, Y. Qi, J. Li, H. Dong, and L. Zhang, “Design and implementation of an avalanche photodiode array receiver for fluorescent-antenna-based visible light communication systems,” Applied Optics, vol. 65, no. 11, pp. 3483–3490, 2026. [67] F. Liu, J. Farmer, A. Schreier, G. Faulkner, H. Chun, W. Matthews, Z. Wang, and D. O’Brien, “Ultra-sensitive UV solar-blind optical wireless communications with an SiPM,” Optics Letters, vol. 48, no. 20, pp. 5387–5390, 2023. [68] W. Matthews and S. Collins, “A roadmap for gigabit to terabit optical wireless communications receivers,” Sensors, vol. 23, no. 3, p. 1101, 2023. [69] C. Wang, Z. Xu, J. Wang, J. Li, W. Mou, and H. Zhu, “Unified analytical framework for SPAD array receivers with dead-time-induced blocking loss and inter-symbol interference in PAM-OWC systems,” arXiv:2605.28560, 2026. [70] E. Säckinger, Analysis and Design of Transimpedance Amplifiers for Optical Receivers. Wiley, 2017. [71] S. D. Personick, “Receiver design for digital fiber optic communication systems, I,” Bell System Technical Journal, vol. 52, no. 6, pp. 843–874, 1973. [72] J. L. Hullett and T. V. Muoi, “A feedback receive amplifier for optical transmission systems,” IEEE Transactions on Communications, vol. 24, no. 10, pp. 1180–1185, 1976. [73] T. V. Muoi, “Receiver design for high-speed optical-fiber systems,” Journal of Lightwave Technology, vol. 2, no. 3, pp. 243–267, 1984. [74] A. Kassem and I. Darwazeh, “A high bandwidth modified regulated cascode TIA for high capacitance photodiodes in VLC,” in Proc. IEEE International Symposium on Circuits and Systems (ISCAS), Sapporo, Japan, 2019. [75] J. L. Cura and L. N. Alves, “Bandwidth improvements in transimpedance amplifiers for visible-light receiver front-ends,” in Proc. 2013 IEEE 20th International Conference on Electronics, Circuits, and Systems (ICECS), 2013, pp. 831–834. [76] D. J. F. Barros, S. K. Wilson, and J. M. Kahn, “Comparison of orthogonal frequency-division multiplexing and pulse-amplitude modulation in indoor optical wireless links,” IEEE Transactions on Communications, vol. 60, no. 1, pp. 153–163, 2012. [77] S. Mardanikorani, X. Deng, and J.-P. M. G. Linnartz, “Optimization and comparison of M-PAM and optical OFDM modulation for optical wireless communication,” IEEE Open Journal of the Communications Society, vol. 1, pp. 1721–1737, 2020. [78] A. Chaaban, Z. Rezki, and M.-S. Alouini, “On the capacity of intensitymodulation direct-detection Gaussian optical wireless communication channels: A tutorial,” IEEE Communications Surveys & Tutorials, vol. 24, no. 1, pp. 455–491, 2022. [79] X. Deng, S. Mardanikorani, Y. Wu, K. Arulandu, B. Chen, A. M. Khalid, and J.-P. M. G. Linnartz, “Mitigating LED nonlinearity to enhance visible light communications,” IEEE Transactions on Communications, vol. 66, no. 11, pp. 5593–5607, 2018. [80] J. Shi, Y. Wei, Z. Luo, Z. Li, C. Shen, J. Zhang, and N. Chi, “8.8 Gbps PAM-4 visible light communication link using an external modulator and a neural network equalizer,” Optics Letters, vol. 48, no. 20, pp. 5193–5196, 2023. [81] M. Yaseen, M. Elamassie, S. S. Ikki, and M. Uysal, “Signal-dependent shot and relative intensity noise in channel estimation of laser diodebased indoor VLC systems,” IEEE Transactions on Communications, vol. 73, no. 1, pp. 498–509, 2025. [82] S. H. ElFar, S. Ikki, and A. R. Sebak, “Performance analysis of VLC systems over random channels with signal-dependent shot noise and relative intensity noise,” IEEE Photonics Journal, vol. 18, no. 4, p. 7301615, 2026. [83] F. Villenas, K. Wu, Y. C. Gültekin, J. Riani, and A. Alvarado, “Beyond 200 Gb/s/lane: An analytical approach to optimal detection in shaped IM-DD optical links with relative intensity noise,” arXiv:2506.19684, 2025. [84] A. Congar, K. Hussain, C. Pareige, R. Butté, N. Grandjean, P. Besnard, and S. Trebaol, “Impact of mode-hopping noise on InGaN edge emitting laser relative intensity noise properties,” IEEE Journal of Quantum Electronics, vol. 54, no. 1, p. 1100107, 2018. [85] T. Wiegart, F. Da Ros, M. P. Yankov, F. Steiner, S. Gaiarin, and R. D. Wesel, “Probabilistically shaped 4-PAM for short-reach IM/DD links with a peak power constraint,” Journal of Lightwave Technology, vol. 39, no. 2, pp. 400–405, 2021. [86] T. Z. Gutema, H. Haas, and W. O. Popoola, “WDM based 10.8 Gbps visible light communication with probabilistic shaping,” Journal of Lightwave Technology, vol. 40, no. 15, pp. 5062–5069, 2022.

JOURNAL OF LIGHTWAVE TECHNOLOGY

[87] A. Kafizov, A. Elzanaty, and M.-S. Alouini, “Probabilistic constellation shaping for enhancing spectral efficiency in NOMA VLC systems,” IEEE Transactions on Wireless Communications, vol. 23, no. 8, pp. 9958–9971, 2024. [88] F. Chang, K. Onohara, and T. Mizuochi, “Forward error correction for 100 G transport networks,” IEEE Communications Magazine, vol. 48, no. 3, pp. S48–S55, 2010. [89] A. Alvarado, E. Agrell, D. Lavery, R. Maher, and P. Bayvel, “Replacing the soft-decision FEC limit paradigm in the design of optical communication systems,” Journal of Lightwave Technology, vol. 34, no. 2, pp. 707–721, 2016. [90] E. Agrell, M. Secondini, A. Alvarado, and T. Yoshida, “Performance prediction recipes for optical links,” IEEE Photonics Technology Letters, vol. 33, no. 18, pp. 1034–1037, 2021. [91] F. Miramirkhani and M. Uysal, “Channel modelling for indoor visible light communications,” Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences, vol. 378, no. 2169, p. 20190187, 2020. [92] F. Miramirkhani, T. Baykas, M. Elamassie, and M. Uysal, “IEEE 802.11bb reference channel models for light communications,” IEEE Communications Standards Magazine, vol. 7, no. 4, pp. 84–89, 2023. [93] Y. Tong, P. Tang, J. Zhang, S. Liu, Y. Yin, B. Liu, and L. Xia, “Channel characteristics and link adaption for visible light communication in an industrial scenario,” Sensors, vol. 23, no. 7, p. 3442, 2023. [94] X. Zhu, C.-X. Wang, J. Huang, M. Chen, and H. Haas, “A novel 3D non-stationary channel model for 6G indoor visible light communication systems,” IEEE Transactions on Wireless Communications, vol. 21, no. 10, pp. 8292–8307, 2022. [95] S. Maravanchery Mana, K. G. K. Gabra, S. Mohammadi Kouhini, P. Hellwig, J. Hilt, and V. Jungnickel, “An efficient multi-link channel model for LiFi,” in 2021 IEEE 32nd Annual International Symposium on Personal, Indoor and Mobile Radio Communications (PIMRC), 2021, pp. 1–6. [96] C. Chen, D. A. Basnayaka, A. A. Purwita, X. Wu, and H. Haas, “Wireless infrared-based LiFi uplink transmission with link blockage and random device orientation,” IEEE Transactions on Communications, vol. 69, no. 2, pp. 1175–1188, 2021. [97] M. D. Soltani, A. A. Purwita, Z. Zeng, H. Haas, and M. Safari, “Modeling the random orientation of mobile devices: Measurement, analysis and LiFi use case,” IEEE Transactions on Communications, vol. 67, no. 3, pp. 2157–2172, 2019. [98] M. A. Arfaoui, M. D. Soltani, I. Tavakkolnia, A. Ghrayeb, C. M. Assi, M. Safari, and H. Haas, “Measurements-based channel models for indoor LiFi systems,” IEEE Transactions on Wireless Communications, vol. 20, no. 2, pp. 827–842, 2021. [99] S. Ma, H. Sheng, J. Sun, H. Li, X. Liu, C. Qiu, M. Safari, N. Al-Dhahir, and S. Li, “Feasibility conditions for mobile LiFi,” IEEE Transactions on Wireless Communications, vol. 23, no. 7, pp. 7911–7923, 2024. [100] E. Khorov and I. Levitsky, “Current status and challenges of Li-Fi: IEEE 802.11bb,” IEEE Communications Standards Magazine, vol. 6, no. 2, pp. 35–41, 2022. [101] K. L. Bober, A. Ebmeyer, F. Dressler, R. Freund, and V. Jungnickel, “LiFi for industry 4.0: Main features, implementation and initial testing of IEEE Std 802.15.13,” IEEE Open Journal of Vehicular Technology, vol. 5, pp. 1625–1636, 2024. [102] ITU-T, “High-speed indoor visible light communication transceiver— system architecture, physical layer and data link layer specification,” Recommendation G.9991, 2019. [103] H. Haas, L. Yin, C. Chen, S. Videv, D. Parol, E. Poves, H. Alshaer, and M. S. Islim, “Introduction to indoor networking concepts and challenges in LiFi,” IEEE/OSA Journal of Optical Communications and Networking, vol. 12, no. 2, pp. A190–A203, 2020. [104] X. Wu, M. D. Soltani, L. Zhou, M. Safari, and H. Haas, “Hybrid LiFi and WiFi networks: A survey,” IEEE Communications Surveys & Tutorials, vol. 23, no. 2, pp. 1398–1420, 2021. [105] M. D. Soltani, A. A. Qidan, S. Huang, B. Yosuf, S. Mohamed, R. Singh, Y. Liu, W. Ali, R. Chen, H. Kazemi, E. Sarbazi, B. Berde, D. Chiaroni, B. Bechadergue, F. Abdel-Dayem, H. Soni, J. Tabu, M. Perrufel, N. Serafimovski, T. E. H. El-Gorashi, J. Elmirghani, M. Crisp, R. Penty, I. H. White, H. Haas, and M. Safari, “Terabit indoor laser-based wireless communications: LiFi 2.0 for 6G,” IEEE Wireless Communications, vol. 30, no. 5, pp. 36–43, 2023. [106] J.-P. M. G. Linnartz, C. R. B. Corrêa, T. E. B. Cunha, E. Tangdiongga, T. Koonen, X. Deng, M. Wendt, A. A. Abbo, P. J. Stobbelaar, P. Polak, M. Müller, D. Behnke, M. Martı́nez, S. Vicent, T. Metin, M. Emmelmann, S. M. Kouhini, K. L. Bober, C. Kottke, and V. Jungnickel, “ELIoT: enhancing LiFi for next-generation Internet of things,”

15

EURASIP Journal on Wireless Communications and Networking, vol. 2022, no. 1, p. 89, 2022. [107] H. Ji, X. Wu, Q. Wang, S. J. Redmond, and I. Tavakkolnia, “Adaptive target-condition neural network: DNN-aided load balancing for hybrid LiFi and WiFi networks,” IEEE Transactions on Wireless Communications, vol. 23, no. 7, pp. 7307–7318, 2024. [108] H. Ji and X. Wu, “Resource and mobility management in hybrid LiFi and WiFi networks: A user-centric learning approach,” IEEE Transactions on Wireless Communications, vol. 24, no. 2, pp. 1293– 1305, 2025. [109] S. S. Murad, S. Yussof, W. Hashim, and R. Badeel, “Three-phase handover management and access point transition scheme for dynamic load balancing in hybrid LiFi/WiFi networks,” Sensors, vol. 22, no. 19, p. 7583, 2022. [110] E. Ancillotti, L. Pescosolido, and A. Passarella, “Toward hybrid COTSbased LiFi/WiFi networks with QoS requirements in mobile environments,” arXiv:2511.00210, 2025. [111] A. Aldalbahi, M. Rahaim, A. Khreishah, M. Ayyash, and T. D. C. Little, “Visible light communication module: An open source extension to the ns3 network simulator with real system validation,” IEEE Access, vol. 5, pp. 22 144–22 158, 2017. [112] S. Ullah, S. U. Rehman, and P. H. J. Chong, “A comprehensive open-source simulation framework for LiFi communication,” Sensors, vol. 21, no. 7, p. 2485, 2021. [113] J. M. Kahn and J. R. Barry, “Wireless infrared communications,” Proceedings of the IEEE, vol. 85, no. 2, pp. 265–298, 1997. [114] M. S. Islim and H. Haas, “Modulation techniques for Li-Fi,” ZTE Communications, vol. 14, no. 2, pp. 29–40, 2016. [115] M. T. Niaz, F. Imdad, S. Kim, and H. S. Kim, “Total least-squarebased receiver for asymmetrically clipped optical-OFDM visible light communication system,” IET Optoelectronics, vol. 11, no. 4, pp. 129– 133, 2017. [116] International Electrotechnical Commission, “Photobiological safety of lamps and lamp systems,” IEC 62471:2006 (CIE S 009:2002), Ed. 1.0, Geneva, Switzerland, 2006. [117] ——, “Photobiological safety of lamps and lamp systems—part 7: Light sources and luminaires primarily emitting visible radiation,” IEC 62471-7:2023, Ed. 1.0, Geneva, Switzerland, 2023. [118] K. Schulmeister and J. Daem, “Classification of laser illuminated light sources under IEC 60825-1 edition 3,” in Proc. ILSC 2017: Proceedings of the International Laser Safety Conference, Laser Institute of America, 2017, pp. 174–180. [119] ——, “Risk of retinal injury from ”risk group 2” laser illuminated projectors,” Journal of Laser Applications, vol. 28, no. 4, p. 042002, 2016. [120] T.-C. Wu, Y.-C. Chi, H.-Y. Wang, C.-T. Tsai, Y.-F. Huang, and G.R. Lin, “Tricolor R/G/B laser diode based eye-safe white lighting communication beyond 8 gbit/s,” Scientific Reports, vol. 7, no. 11, p. 11, 2017. [121] W. L. Oberkampf and C. J. Roy, Verification and Validation in Scientific Computing. Cambridge University Press, Cambridge, UK, 2010. [122] R. G. Sargent, “Verification and validation of simulation models,” Journal of Simulation, vol. 7, no. 1, pp. 12–24, 2013. [123] P. J. Roache, “Verification of codes and calculations,” AIAA Journal, vol. 36, no. 5, pp. 696–702, 1998. [124] American Institute of Aeronautics and Astronautics, “Guide for the verification and validation of computational fluid dynamics simulations,” AIAA Standard G-077-1998, Reston, VA, 1998. [125] American Society of Mechanical Engineers, “Standard for verification and validation in computational solid mechanics,” ASME V&V 102019 (R2025), New York, NY, 2019. [126] IEEE, “IEEE standard for information technology—wireless LAN MAC and PHY specifications—amendment 7: Light communications,” IEEE Std 802.11bb-2023, 2023.

Related documents

• Diffused-Beam Laser-Diode LiFi Under Realizable Receiver, Noise, and Safety Constraints: Design-Space Analysis and an Open Cross-Verified Simulation Framework#612655
• Diffused-Beam Laser-Diode LiFi Under Realizable Receiver, Noise, and Safety Constraints: Design-Space Analysis and an Open Cross-Verified Simulation Framework#611178
• Anomalous relative intensity noise transfer in ultralong random fiber lasers#374605
• Compact and electrically driven active mode-locked laser at 10 GHz repetition rate based on a graphene electro-absorption modulator#677134
• Transmissive silicon photonic dichroic filters with spectrally selective waveguides#955266