Tse and Chan (IEEE J. Sel. Top. Quantum Electron. 18(2), 654–661, 2012) proposed a variable all-optical packet buffer in which a passive fiber loop clones each input packet into successively attenuated copies and a two-stage power-dependent filter — self-phase-modulation (SPM) spectral broadening in nonlinear fiber followed by offset bandpass filtering, twice — transmits only the copy whose peak power matches its transfer peak. The input power level alone therefore selects the number of loop circulations and hence the delay: the proof of concept showed three to four discrete 2-ns delay steps at a 10-GHz pulse rate, with a simulated selector transfer peak near 15-dBm input peak power. The paper left the concept's practical envelope uncharacterized, and this record poses the open problem those items jointly constitute: quantify the achievable performance envelope of power-encoded SPM-based optical buffering — the minimum selector power and its reduction with high-nonlinearity fibers, the impact of coherence degradation and supercontinuum onset in the anomalous-dispersion regime on packet integrity, the maximum number of delay steps under amplifier noise and cumulative distortion, the comparative standing of alternative selector implementations, and bit-rate/format generality — which together decide the concept's viability for optical packet switching. No published work characterizes this architecture's practical envelope.

Background

In optical packet switching (OPS), data packets are routed entirely in the optical domain, and when two packets contend for the same output port one of them must be delayed — buffered — without optical-electrical-optical conversion. The classical buffering approaches each carry a structural drawback. Fiber delay lines (FDLs) provide only a fixed, pre-provisioned set of delays. Wavelength conversion followed by propagation through dispersive fiber maps wavelength to delay but requires very long dispersive spans or widely tunable converters; one systematic analysis of a self-phase-modulation-based variant of that route — broadening the signal spectrum by SPM and slicing an offset wavelength from it before dispersive delay — established quantitative limitations on its delay efficiency and signal fidelity for buffer design (Pant, Stenner, and Neifeld, Appl. Opt. 47, 5051, 2008). Slow-light buffers, which reduce the group velocity of light in a resonant or structured medium, are subject to fundamental delay-bandwidth, loss, and noise tradeoffs that bound their usable buffering capacity (Tucker, Ku, and Chang-Hasnain, J. Lightwave Technol. 23, 4046, 2005). Recirculating-loop buffers pass a packet through a fiber loop NN times, but selecting NN conventionally requires synchronized active optical switching inside the loop.

Tse and Chan (IEEE J. Sel. Top. Quantum Electron. 18(2), 654–661, 2012) proposed a recirculating-loop buffer in which the number of circulations is selected passively by the packet's own optical power. A 3-dB coupler with a 2-ns fiber delay clones each input packet into a train of copies spaced by τ=2\tau = 2 ns, each successive copy attenuated by about 3 dB. The copy train then passes through a two-stage power-dependent filter. Each stage consists of self-phase modulation (SPM) — the instantaneous Kerr-nonlinearity phase shift ϕNL=γPLeff\phi_{\mathrm{NL}} = \gamma P L_{\mathrm{eff}} that a pulse of peak power PP imposes on itself while propagating a length LeffL_{\mathrm{eff}} of fiber with nonlinear coefficient γ\gamma, which chirps and spectrally broadens the pulse — followed by an offset optical bandpass filter (OBPF) whose passband sits at a wavelength the pulse spectrum reaches only after sufficient broadening. Because the broadening grows with peak power, stage transmission becomes a sharply power-dependent function: Stage 1 (1 km of dispersion-flattened highly nonlinear fiber, HNLF, with γ=10.8 W−1km−1\gamma = 10.8\ \mathrm{W}^{-1}\mathrm{km}^{-1} and zero-dispersion wavelength (ZDW) 1550 nm, plus a 0.8-nm OBPF offset to 1541 nm) shows a transmission peak near 15-dBm input peak power in simulation, and Stage 2 (a second SPM fiber plus offset filter) steepens the transfer curve, dropping more than 20 dB for a 2-dB deviation from its optimum input power. Only the copy whose power lies near the transfer peak survives, so raising the input power by one 3-dB step promotes the next delayed copy and increases the output delay by τ\tau. In the experiment (2-ps pulses at a 10-GHz repetition rate, return-to-zero format), input powers of 0.4, 3.85, 6.25, and 8.75 dBm at the Stage-1 entry produced packet delays of 0, 2, 4, and 6 ns. Because the selector uses only the signal's own SPM, it needs no auxiliary continuous-wave probe laser and, to first order, no polarization control.

The Stage-1 dispersion regime introduces a further tradeoff that the paper mapped only qualitatively. Operating with the signal wavelength in the anomalous-dispersion regime (ZDW shorter than the signal wavelength) compresses the pulses, strengthens the SPM sidelobes, and sharpens the power-filtering peak, but risks pulse-to-pulse coherence degradation and supercontinuum formation; operating deep in the normal regime preserves coherence but weakens and broadens the transfer peak. The authors recommend operating near the ZDW on the normal-dispersion side, without quantifying the penalty either way. The demonstration thus established feasibility — three to four discrete delay steps at 10 Gb/s — while leaving the concept's practical envelope, and hence its viability for practical OPS deployment, uncharacterized.

Problem Statement

Determine the achievable performance envelope of the power-encoded, SPM-based variable all-optical packet buffer of Tse and Chan (2012) — the set of jointly linked quantities that decides whether the concept is viable for practical optical packet switching. Concretely, the unresolved problem is:

  1. Power budget: the minimum selector input peak power at which the selected copy is separated from its ~3-dB-spaced neighbours by an extinction ratio of at least 20 dB (the contrast the paper's two-stage selector exceeds at a 2-dB detuning), and how far it can be reduced below the ~15-dBm simulated Stage-1 transfer peak by systematic use of higher-γ\gamma fibers (e.g., bismuth-oxide or photonic-crystal fibers with γ>1000 W−1km−1\gamma > 1000\ \mathrm{W}^{-1}\mathrm{km}^{-1}), shorter pulses, or alternative nonlinear media.

  2. Signal quality: quantify how the operating point and Stage-1 dispersion regime — in particular the pulse-coherence degradation and supercontinuum onset that accompany anomalous-dispersion pulse compression — affect packet integrity (bit-error characteristics, timing jitter, and waveform distortion) at the demonstrated 10-Gb/s rate and at higher rates.

  3. Granularity and scalability: the maximum number of discrete delay steps achievable, reported as an explicit function of the available input-power span (steps versus span in dB), and the practical limits imposed by amplifier noise — the amplified spontaneous emission (ASE) of the erbium-doped fiber amplifiers (EDFAs) used between stages — and by cumulative distortion through many loop copies (the paper's 2-dB-per-copy design optimization rests on simulation only).

  4. Comparative standing: how the two-stage SPM-plus-offset-filter selector compares, in input power, extinction ratio (the power contrast between the transmitted copy and the rejected ones), pulse fidelity, and integrability, against the alternative selector implementations the paper nominates — soliton self-frequency shift with femtosecond pulses (the Raman-driven continuous red-shift of a soliton's center wavelength, whose magnitude depends on pulse peak power), triangular/sawtooth-pulse-induced wavelength shifts, four-wave-mixing-based thresholding (thresholding via the parametric Kerr nonlinearity, in which two pump photons convert into signal and idler photons), nonlinear amplifying loop mirrors (fiber-loop interferometers whose transmission depends on pulse peak power through nonlinear phase accumulation), and saturable absorbers (devices whose absorption drops at high optical intensity).

  5. Format and rate generality: performance for non-return-to-zero formats, higher symbol rates (the 40-Gb/s-and-beyond regime), and multi-wavelength operation, and the buffer's influence on real OPS packet formats.

These five items are the mandatory constituent sub-questions of one scientific objective — establishing the power-to-delay performance envelope of power-encoded SPM-based buffering — so that the concept's practicality can be judged rather than merely asserted. Each item is resolved on its own terms — a numeric minimum meeting the stated extinction floor (item 1), a quantified signal-quality result at a stated bit rate (item 2), a maximum step count as a function of the power span (item 3), a defined comparison against the named alternatives (item 4), and a stated format and rate range (item 5) — and partial resolutions are credited per item, while the envelope as a whole is established only when all five are resolved.

References

  1. Kam-Hon Tse and Chun-Kit Chan, "A Novel Fiber-Based Variable All-Optical Packet Buffer Based on Self-Phase-Modulation-Induced Spectral Broadening", IEEE Journal of Selected Topics in Quantum Electronics 18(2), 654–661 (2012). DOI 10.1109/JSTQE.2011.2135336, https://doi.org/10.1109/jstqe.2011.2135336
  2. Xiao-wei Dong and Wen-kai Liu, "All-optical buffer and shaper based on complex-modulated long-period-grating coupler and self-closed fiber loop", Optoelectronics Letters 8(6), 418–421 (2012). DOI 10.1007/s11801-012-2286-6, https://doi.org/10.1007/s11801-012-2286-6
  3. Junhyuk Choi, Ghang-Ho Lee, Taekook Kim, Yeongjun Kim, and Jinwoo Park, "All-optical packet buffer system based on controllable-delay recirculating loop", Microwave and Optical Technology Letters 56(9), 2058–2061 (2014). DOI 10.1002/mop.28512, https://doi.org/10.1002/mop.28512
  4. S. Shinada, H. Furukawa, and N. Wada, "Investigation of Optical Buffer Capacity using Large-scale Fiber Delay Lines for Variable-length Optical Packet Switching", 39th European Conference and Exhibition on Optical Communication (ECOC 2013), 672–674 (2013). DOI 10.1049/cp.2013.1497, https://doi.org/10.1049/cp.2013.1497
  5. Vaibhav Shukla, Ramesh Kumar Sonkar, and Rajiv Srivastava, "Performance comparison between recirculating loop buffer-based optical packet switch architectures", Indian Journal of Physics 94(7), 1085–1096 (2020; published online 2019). DOI 10.1007/s12648-019-01540-5, https://doi.org/10.1007/s12648-019-01540-5
  6. A. V. Polyakov, "Simulation of Fiber-Optic Buffer Loop Memory with All-Optical 2R Regeneration", Optical Memory and Neural Networks 29(2), 100–109 (2020). DOI 10.3103/S1060992X20020101, https://doi.org/10.3103/S1060992X20020101
  7. R. S. Tucker, Pei-Cheng Ku, and C. J. Chang-Hasnain, "Slow-light optical buffers: capabilities and fundamental limitations", Journal of Lightwave Technology 23(12), 4046–4066 (2005). DOI 10.1109/JLT.2005.853125, https://doi.org/10.1109/jlt.2005.853125
  8. Ravi Pant, Michael D. Stenner, and Mark A. Neifeld, "Limitations of self-phase-modulation-based tunable delay system for all-optical buffer design", Applied Optics 47(27), 5051 (2008). DOI 10.1364/AO.47.005051, https://doi.org/10.1364/ao.47.005051
  9. Kam-Hon Tse and Chun-Kit Chan, "A Path Tracing Scheme for All-Optical Packet-Switched Networks", Journal of Lightwave Technology 30(11), 1625–1631 (2012). DOI 10.1109/JLT.2012.2188495, https://doi.org/10.1109/jlt.2012.2188495