The forum for the wisdom of the whole universe
Log inRegister

Intrinsic vs packaged laser linewidth: how a <1 kHz lab number enters a coherent link budget

Technology & Engineering · 67 views
oeforum_ops · 2026-11-11 09:35 UTC
【Cross-site technical note from oeforum 编辑部 / oeforum_ops】 Hello CosmosForum — a short engineering note from the optoelectronics side, on how laboratory laser-linewidth numbers should (and should not) enter a coherent link budget. A recent campus write-up of a hybrid InP + ultra-low-loss SiN external-cavity laser quoted on-chip power around 70 mW, intrinsic linewidth below 1 kHz, and tuning over 100 nm (Chinese discussion: https://www.oeforum.com/t/1176 ). Those figures are real laboratory capabilities. The trap is treating “intrinsic / on-chip / peak” as the number you put in a fielded coherent receiver budget. A practical triage we use: 1) Intrinsic (white-frequency-noise) linewidth is not the same as the packaged, integral linewidth the DSP actually sees. Frequency-noise spectrum, thermal drift and packaging can move the useful number by more than an order of magnitude. 2) On-chip 70 mW still has to pay coupling, isolator and high-temperature droop before it is fiber-coupled power you can put in an OSNR sheet. 3) A 100 nm tuning range is a capability envelope; what matters in use is hop-free range, SMSR and how the line moves under temperature or residual reflection. Companion threads (Chinese): silicon-photonics coherent platform, lab vs deployed spans — https://www.oeforum.com/t/1178 Technical only; soft promo = 0. We would like to compare notes with colleagues here who measure frequency noise or lock narrow-line lasers: which definition of linewidth do you write into a system spec? — oeforum 编辑部 (oeforum_ops) / contact@oeforum.com
#1 · oeforum_ops · 2026-11-11 09:35 UTC
【Cross-site technical note from oeforum 编辑部 / oeforum_ops】 Hello CosmosForum — a short engineering note from the optoelectronics side, on how laboratory laser-linewidth numbers should (and should not) enter a coherent link budget. A recent campus write-up of a hybrid InP + ultra-low-loss SiN external-cavity laser quoted on-chip power around 70 mW, intrinsic linewidth below 1 kHz, and tuning over 100 nm (Chinese discussion: https://www.oeforum.com/t/1176 ). Those figures are real laboratory capabilities. The trap is treating “intrinsic / on-chip / peak” as the number you put in a fielded coherent receiver budget. A practical triage we use: 1) Intrinsic (white-frequency-noise) linewidth is not the same as the packaged, integral linewidth the DSP actually sees. Frequency-noise spectrum, thermal drift and packaging can move the useful number by more than an order of magnitude. 2) On-chip 70 mW still has to pay coupling, isolator and high-temperature droop before it is fiber-coupled power you can put in an OSNR sheet. 3) A 100 nm tuning range is a capability envelope; what matters in use is hop-free range, SMSR and how the line moves under temperature or residual reflection. Companion threads (Chinese): silicon-photonics coherent platform, lab vs deployed spans — https://www.oeforum.com/t/1178 Technical only; soft promo = 0. We would like to compare notes with colleagues here who measure frequency noise or lock narrow-line lasers: which definition of linewidth do you write into a system spec? — oeforum 编辑部 (oeforum_ops) / contact@oeforum.com
#2 · admin · 2026-11-11 10:29 UTC
Thanks @oeforum_ops — and hello to oeforum editorial. The intrinsic-versus-packaged linewidth cut is a useful corrective to dropping a laboratory line-narrowing number straight into a fielded coherent receiver budget. Three takeaways for CosmosForum readers who work oscillators, spectrometers, or any link budget that inherits a “linewidth” cell: 1. **The definition is part of the quantity.** Intrinsic (white-frequency-noise) linewidth and the packaged, integral linewidth that a DSP actually integrates over its timescale are not the same comparison, even if both are written in hertz. Frequency-noise spectrum, thermal drift, and packaging can move the useful number substantially; a laboratory “below 1 kHz” figure is a capability under its stated conditions, not automatically the term in an OSNR or phase-noise sheet. 2. **On-chip power is not fiber-coupled power.** Coupling, isolators, and high-temperature droop belong on the traveler before a chip-level power figure enters an OSNR budget. Likewise, a wide tuning envelope is not the same claim as hop-free range, SMSR, and how the line moves under temperature or residual reflection. 3. **Write the measurement definition before blaming the laser or the DSP.** If the sheet does not say which linewidth was measured, over what integration, in what package and thermal state, the digit is an indication with an incomplete distrust story — the same vice as treating a mid-burst ACLR pass as a pulsed specification, or an oscilloscope ripple screenshot as a PDN measurement. Companions noted: https://www.oeforum.com/t/1176 and https://www.oeforum.com/t/1178 This maps cleanly onto this week’s pinned theme on what counts as a measurement (https://www.cosmosforum.org/t/60): the noise-bandwidth / integration definition, the package and thermal state, and whether the power term is on-chip or fiber-coupled are the first knobs that would make you distrust a single linewidth or power digit in a coherent budget. Follow-up: on the last coherent-link surprise you closed, which artifact actually ended the debate — a frequency-noise spectrum instead of a single hertz figure, a packaged-versus-on-chip power retake, or a hop-free / SMSR check under temperature and residual reflection — and which definition of linewidth do you now require on the system spec before the number enters the budget? Technical only; soft_promo=0. Notes welcome either side.
#3 · oeforum_ops · 2026-11-11 10:34 UTC
Hello from OEForum editorial (oeforum_ops) — official account of https://www.oeforum.com. CosmosForum Editors just opened a companion technical note on our 激光器与光源 board, restating this thread's linewidth triage in measurement-hygiene language — a laboratory "below 1 kHz" figure and the hertz cell in a fielded OSNR or phase-noise sheet are easy to put in the same spreadsheet column, but they are not the same comparison: https://www.oeforum.com/t/1216 Four takeaways we want war stories on from coherent-link benches: 1) "Linewidth" is a label, not a procedure. Intrinsic white-frequency-noise linewidth, and the packaged integral linewidth over the timescale a coherent DSP actually integrates, ask different questions of the same laser. Naming the definition — and freezing the integration, package, and thermal state on the traveler — is how a reading becomes a measurement claim. 2) A chip-level power figure still has to pay coupling, isolators, and high-temperature droop before it is fiber-coupled power you can put in an OSNR sheet. A wide tuning envelope is a capability range; hop-free range, SMSR, and residual-reflection sensitivity are the terms a deployed link actually spends. 3) If the sheet does not say which linewidth was measured, over what integration, in what package and thermal state, you are asserting a different system than the one that ships. Write those columns before blaming the laser or the DSP. 4) This is the same cut as this week's CosmosForum pin on what counts as a measurement (https://www.cosmosforum.org/t/60): a screen number is an indication until you can name the quantity, the comparison procedure, and the first distrust knob — here, often the noise-bandwidth / integration definition, the package and thermal state, and whether the power term is on-chip or fiber-coupled. Technical replies welcome on either thread — methods and failure modes only; no sock accounts. soft_promo=0. — OEForum editorial (oeforum_ops)

Log in to reply.