EQUATIONS AND GUIDE — STEADY STATE
t = (A − iΔ)/(κ/2 − iΔ) with A = (κ₀ + κ₂ − κ₁)/2, and d = −√(κ₁κ₂)/(κ/2 − iΔ) at the drop port. On resonance Tmin = (2A/κ)², the drop peak is 4κ₁κ₂/κ², and what is left, 4κ₀κ₁/κ², is lost in the ring. Inside, the light builds up to (2/π)·η·finesse times the input.
Close the gap. Drag BUS GAP down and watch the dip: it widens all the way, but it deepens only until critical coupling (κ₁ = κ₀: every photon is lost in the ring, none comes out), then fills in again. Two different rings give the same dip, which is what the MEASURE tab is about. The phase screen tells them apart: under-coupled is a small wiggle with negative group delay, over-coupled winds through a full 2π and delays the light by τg(0) = 4κ₁/(κ₁² − κ₀²). In the complex plane t(Δ) is a circle through 1 with radius η = κ₁/κ, and it swallows the origin at the same moment.
ADD-DROP bolts on a second waveguide. From the bus it looks like extra loss, so critical coupling moves to κ₁ = κ₀ + κ₂. BACKSCATTER lets the clockwise wave leak into the counter-clockwise one: the line splits by γ, light comes back out of the input, and the zero of transmission moves to κ₁ = √((κ₀+κ₂)² + γ²). IDEALITY below 1 means the coupler itself wastes light, which caps how far you can over-couple. The span stays put while you turn knobs, so you see the line broaden; FIT SPAN re-centres the view.
EQUATIONS AND GUIDE — MEASURE
A transmission dip hands you two numbers: its width κ and its floor Tmin = ((κ₀ + κ₂ − κ₁)/κ)². Solving for the rates gives κ₀, κ₁ = (κ/2)(1 ± √Tmin): two answers, and nothing in the power spectrum picks one. With a drop port the twin of (κ₀, κ₁, κ₂) is (κ₀κ₁/K, K, κ₁κ₂/K) with K = κ₀ + κ₂, and even the drop spectrum agrees.
PHASE: the group delay on resonance, τg(0) = 4κ₁/(κ₁² − K²), changes sign. GAP: closing the gap raises κ₁; an under-coupled dip deepens, an over-coupled one fills in. FLASH: switch the laser off and the ring empties into the bus, 4η²·Pin at the first instant. That is more than you were putting in once η > ½. A resolved backscatter doublet needs none of this: its floor ((κ₀ + κ₂)/κ)² only goes one way.
FIT: Lorentzian dip, or doublet, on a sloped baseline by least squares, with 1σ error bars from the residuals. Both assignments are returned. Close to critical coupling the error bars on Qi and Qex grow, because √Tmin is then barely constrained.
EQUATIONS AND GUIDE — DYNAMICS
ȧ = (iΔ(t) − κ/2)·a + √κ₁·sin holds exactly for a swept laser. The speed that matters is s = 2πV/κ²: at s = 1 the laser crosses one linewidth per photon lifetime. Faster than that, the ring is still ringing at ω₀ while the laser has moved on, and the two beat at the through port, above the input level when the ring is well coupled. Fitting that chirped beat gives κ₀ and κ₁ separately (Dumeige et al., JOSA B 25, 2073). Try the silica toroid: it rings at sweep speeds a real laser reaches. The SiN ring does not.
SWITCHED INPUT, on resonance: U(t) = Uss(1 − e−κt/2)² on the way up, e−κt on the way down, and a through-port flash of 4η²·Pin at switch-off. An over-coupled ring's through port dips to exactly zero on the way up, at t = (2/κ)·ln(2η/(2η − 1)). Move DETUNING off resonance and the transient oscillates at Δ.
EQUATIONS AND GUIDE — BROADBAND
Ring picture. Follow the field once around: couplers with self-coupling r₁, r₂, round-trip amplitude a, round-trip phase φ. t = (r₁ − r₂·a·eiφ)/(1 − r₁r₂·a·eiφ). With rj = e−κj/2FSR this is exactly sinh[(x₀+x₂−x₁−iφ)/2] / sinh[(x₀+x₁+x₂−iφ)/2], and the Lorentzian of the other tabs is its small-argument limit. At the real finesse the dips are hairlines, so LOSS × lets you fatten every loss channel and watch the Lorentzian approximation give way.
Across the band. Resonances sit at ωμ = ω₀ + D₁μ + D₂μ²/2. Longer wavelengths reach further out of the waveguide, so the bus coupling grows roughly as κ₁ ∝ eα(λ−λ₀) while Qi stays put. One gap is therefore critical at one wavelength only: under-coupled to the blue of it, over-coupled to the red. COUPLING SLOPE × scales α from the evanescent estimate; real couplers are often steeper. Click a resonance to zoom.
EQUATIONS AND GUIDE — NONLINEAR
Stored light pulls its own resonance to the red: ω₀ → ω₀ − δKU − δthΔT, with the Kerr shift δK = ω₀·c·n₂/(ng²·Veff) instantaneous and the temperature following Cth·dΔT/dt = κabsU − ΔT/Rth. In units of half a linewidth the steady states obey ρ(1 + (ζ − ρ)²) = f², the same curve the comb bench draws, with f² = (1 + r)·P/Pth and r the thermal-to-Kerr ratio. Above f² = 8/(3√3) ≈ 1.54 it folds over: sweep to the red and the laser drags the resonance along (the thermal triangle), sweep back and it only catches the short side. Sweep faster than τth and the triangle shrinks to the Kerr tilt.
Threshold. Pth = κ³/(8κ₁δK) is where the comb bench reaches f² = 1. It is lowest at κ₁ = (κ₀+κ₂)/2, under-coupled: weaker coupling stores light longer than it costs in injection. At critical coupling you pay 19 % more, and a drop port raises everything by (1 + κ₂/κ₀)². Assumes anomalous dispersion with a phase-matched sideband. Rth and τth are order-of-magnitude values.
EQUATIONS AND GUIDE — QUANTUM
Everything quantum that leaves the ring is filtered by the escape efficiency η = κport/κ: the odds that a photon exits through the port you collect from, not into scattering or the other port. It caps the heralding efficiency of a pair source and the squeezing you can observe: S(Ω) = 1 − η·4x/((1 + x)² + (2Ω/κ)²) with x = P/Pth, at best 1 − η at threshold. In a Kerr ring the gain follows the pump power, hence x and not √x.
Pairs. Below threshold the ring emits R = (κ/2)·x²/(1 − x²) pairs per second into one signal–idler pair of modes. At fixed cw power that is ∝ κ₁²/κ⁵, best at κ₁ = ⅔K. Ask for one photon out of the bus and the optimum moves to critical coupling; ask for both and it moves to 4⁄3 K. Chase η → 1 and the rate collapses: there is no free lunch (Vernon, Liscidini & Sipe 2016). PULSED PUMP switches to their case, fixed pulse energy with the pulse matched to the cavity lifetime (∝ κ₁²/κ⁴), where the optima sit at 1, 1.5 and 2 K.
All of this is the ideal cavity output: pump on the Kerr-shifted resonance, one phase-matched pair, equal linewidths, no detection loss, no thermorefractive or Raman noise. SEND TO COMB BENCH opens the LLE simulator with this ring's f² = P/Pth, ζ = −2Δ/κ and d₂ = D₂/κ, clamped to its slider ranges.