Optical Transport Networks — DWDM, OTN, and SDH

Every packet traversing a service-provider backbone ultimately rides a beam of infrared light through glass. This chapter covers the photonic underlay that sits beneath the MPLS layer: the physics that govern fibre transmission, the decibel mathematics essential for link engineering, the Dense Wavelength Division Multiplexing architecture that packs dozens of terabits onto a single fibre pair, the amplifier technologies that keep signals alive across continental distances, the OTN digital wrapper that provides carrier-grade monitoring and grooming, and the legacy SDH layer that OTN replaces. Two worked network-design examples — a regional long-haul ring and a metro data-centre interconnect — tie the theory together into deployable practice.

Key Concepts at a Glance

Term	One-Line Definition
DWDM	Dense Wavelength Division Multiplexing — packs 96+ independent data channels onto a single fibre by assigning each a distinct infrared wavelength in the C/L-band, spaced as little as 50 GHz apart on the ITU grid [1].
OTN	Optical Transport Network (ITU-T G.709) — the digital framing layer that wraps client signals in ODU containers, adds FEC, and provides per-segment monitoring through up to six TCM levels [3].
EDFA	Erbium-Doped Fibre Amplifier — the long-haul workhorse; amplifies all C-band wavelengths simultaneously in the optical domain via stimulated emission in a short coil of Er³⁺-doped fibre pumped at 980 or 1480 nm [12, 24].
Coherent Optics	Modern transponders that recover amplitude and phase on both polarisations, enabling DP-QPSK through DP-16QAM, DSP-based dispersion compensation, and adaptive capacity-vs-reach optimisation [21, 28].
ROADM	Reconfigurable Optical Add-Drop Multiplexer — uses a Wavelength Selective Switch to remotely add, drop, or pass through individual wavelengths at a node.
OSNR	Optical Signal-to-Noise Ratio — the single most critical metric in optical link design; determines whether a given modulation format can be sustained over a given distance.

Fibre Physics and Transmission Bands

Key Insight

Optical networking operates in the infrared spectrum (1260-1625 nm) where silica glass exhibits its lowest attenuation [7, 21]. The C-band (1530-1565 nm) dominates DWDM because it sits at the absolute minimum of the fibre loss curve and aligns with the EDFA gain window.

Silica glass fibre is not uniformly transparent across all wavelengths. The useful transmission window is divided into five bands, each with distinct loss characteristics and roles in modern networks.

Band	Wavelength Range	Attenuation	Role in DWDM Systems
O-band	1260-1360 nm	~0.35 dB/km	Short-reach and legacy systems
E-band	1360-1460 nm	Variable	Historically blocked by the water peak; G.652D low-water-peak fibre opens it for future use
S-band	1460-1530 nm	~0.25 dB/km	Raman pump wavelengths; expansion band for next-generation systems
C-band	1530-1565 nm	~0.20 dB/km	Primary DWDM band — lowest loss, EDFA gain window
L-band	1565-1625 nm	~0.22 dB/km	DWDM extension band adding ~96 channels beyond C-band

C-band and L-band together provide approximately 192 channels — the backbone of modern long-haul transport [1]. The “water peak” near 1383 nm in the E-band was an artefact of hydroxyl (OH-) ions trapped during older fibre manufacturing processes [7]. Modern G.652D low-water-peak fibre virtually eliminates this absorption, opening the E-band and S-band for future DWDM expansion as traffic demands grow beyond C+L capacity.

graph LR
    subgraph Fiber Optical Bands
        O["O-band<br/>1260-1360 nm<br/>0.35 dB/km"]
        E["E-band<br/>1360-1460 nm<br/>Water peak"]
        S["S-band<br/>1460-1530 nm<br/>~0.25 dB/km"]
        C["C-band<br/>1530-1565 nm<br/>0.20 dB/km"]
        L["L-band<br/>1565-1625 nm<br/>0.22 dB/km"]
    end
    O --> E --> S --> C --> L
    style C fill:#7F77DD,stroke:#534AB7,color:#fff
    style L fill:#D85A30,stroke:#993C1D,color:#fff
    style O fill:#378ADD,stroke:#185FA5,color:#fff
    style E fill:#888780,stroke:#5F5E5A,color:#fff
    style S fill:#BA7517,stroke:#854F0B,color:#fff

C-band and L-band (highlighted) are where DWDM operates — the lowest-attenuation window of silica fibre.

The continuous attenuation curve below makes the band table concrete and shows why the water peak shaped legacy band assignments:

xychart-beta
    title "SMF Fibre Attenuation vs Wavelength (G.652D)"
    x-axis "Wavelength (nm)" [1260, 1300, 1340, 1380, 1420, 1460, 1500, 1530, 1550, 1565, 1590, 1625]
    y-axis "Attenuation (dB/km)" 0.15 --> 0.45
    line "G.652D SMF" [0.35, 0.33, 0.31, 0.27, 0.26, 0.23, 0.21, 0.20, 0.19, 0.20, 0.21, 0.23]
    line "Legacy (water peak)" [0.35, 0.33, 0.34, 0.45, 0.30, 0.23, 0.21, 0.20, 0.19, 0.20, 0.21, 0.23]

The absolute minimum sits near 1550 nm at ~0.19 dB/km — the reason C-band dominates DWDM. Attenuation rises at both extremes: shorter wavelengths suffer Rayleigh scattering (∝ 1/λ⁴), longer wavelengths hit the silica lattice’s infrared absorption tail [21, 23]. The L-band’s ~0.02 dB/km penalty over C-band is negligible for backbone designs — when evaluating fibre plant for C+L deployment, the real constraint is amplifier availability rather than fibre loss.

Tip

Every additional 0.05 dB/km of attenuation costs 5 dB of budget on a 100 km span — the difference between needing 12 amplifier sites versus 10 on a 1000 km route. Translate fibre loss directly into amplifier site count when comparing fibre vendors or routes.

Core Optical Terminology

Term	Definition
Lambda	A single wavelength of light. One lambda equals one data channel in DWDM.
Frequency (v)	Inversely related to wavelength: v = c / lambda, where c is approximately 3 x 10^8 m/s.
ITU Grid	The international channel plan (ITU-T G.694.1) [1]. Channels are spaced by frequency — fixed-grid spacings are 12.5 / 25 / 50 / 100 GHz; 75 GHz is supported only on the flex-grid (G.694.1 §7) [1]. ITU anchor: 193.1 THz (1552.524 nm); 1550.12 nm = 193.4 THz is one valid grid channel near band centre [1].
OSNR	Optical Signal-to-Noise Ratio. The single most critical metric in optical link design — it determines whether a given modulation format can be sustained over a given distance.
BER	Bit Error Ratio. The fraction of bits received incorrectly. Pre-FEC BER thresholds gate whether a link is viable.

Decibel Mathematics — dB vs dBm

Key Insight

dB is a dimensionless ratio between two power levels. dBm is an absolute power measurement referenced to 1 milliwatt. The distinction matters because every optical link budget mixes both: component losses and gains (dB) applied to actual signal power (dBm).

The Distinction

dB (decibel) expresses a ratio. It answers “how much larger or smaller is one power relative to another?” Fibre loss, splice loss, amplifier gain, noise figure, and OSNR are all ratios, so they are all expressed in dB.

dB = 10 x log10(P2 / P1)

dBm (decibel-milliwatt) expresses absolute power referenced to 1 mW. It answers “how much optical power is actually present?” Launch power, received power, and receiver sensitivity are real measurements, so they use dBm.

dBm = 10 x log10(P / 1 mW)

dBm-to-Milliwatt Conversion Reference

dBm	Milliwatts	Practical Context
+20	100 mW	High-power EDFA output
+10	10 mW	Strong signal
+3	2 mW	Double the 1 mW reference
0	1 mW	Reference point
-3	0.5 mW	Half the reference
-10	0.1 mW	1/10 of reference
-20	10 uW	Weak signal
-30	1 uW	Very weak — approaching receiver sensitivity

Conversion formulae: P(mW) = 10^(dBm/10) and dBm = 10 x log10(P in mW).

Why dB and dBm Arithmetic Works

In the linear domain, output power is the product of input power and gain: P_out = P_in x G. Taking 10 x log10 of both sides converts multiplication into addition:

dBm_out = dBm_in + dB_gain    (amplification)
dBm_out = dBm_in - dB_loss    (attenuation)

The result stays in dBm because the starting point was an absolute reference. This is the entire reason the optical industry uses logarithmic units — a multi-span link budget with dozens of loss and gain elements reduces to simple addition and subtraction.

flowchart LR
    A["Launch power<br/>+1 dBm<br/>(1.26 mW)"] -->|"Fiber loss<br/>-24 dB<br/>(div 251)"| B["After fiber<br/>-23 dBm<br/>(0.005 mW)"]
    B -->|"EDFA gain<br/>+25 dB<br/>(x316)"| C["After amp<br/>+2 dBm<br/>(1.58 mW)"]
    C -->|"More fiber<br/>-12 dB<br/>(div 15.8)"| D["At receiver<br/>-10 dBm<br/>(0.1 mW)"]
    style A fill:#1D9E75,stroke:#0F6E56,color:#fff
    style B fill:#E24B4A,stroke:#A32D2D,color:#fff
    style C fill:#1D9E75,stroke:#0F6E56,color:#fff
    style D fill:#BA7517,stroke:#854F0B,color:#fff

In dB math: +1 - 24 + 25 - 12 = -10 dBm. In linear math: 1.26 mW / 251 x 316 / 15.8 = 0.1 mW. Same answer — dB is simpler.

Per-Component Loss and Gain Values

Component	dB Value	Linear Ratio	Physical Meaning
SMF loss/km (C-band)	0.20 dB	x0.955/km	95.5% of power passes each kilometre [7]
SMF loss/km (L-band)	0.22 dB	x0.951/km	Slightly higher loss than C-band
SMF loss/km (O-band)	0.35 dB	x0.923/km	Highest loss among the useful bands
Fusion splice	0.05 dB	x0.989	98.9% of power passes through
Connector pair	0.30 dB	x0.933	93.3% of power passes through
EDFA gain (typical)	20-30 dB	x100-1000	Amplifies signal 100- to 1000-fold [12, 13, 24]
EDFA noise figure	4.5-6 dB	x2.8-4.0	Degrades SNR by this factor
MUX/DEMUX loss	4-7 dB	x0.20-0.40	20-40% of power passes through
ROADM node loss	8-12 dB	x0.06-0.16	6-16% of power passes through
System margin	3-5 dB	x0.32-0.50	Reserve for ageing and degradation

Essential Optical Engineering Formulas

Every optical link design revolves around a small set of formulas. Each is presented below with its derivation context and a worked example.

Wavelength-Frequency Conversion

v (THz) = 299,792.458 / lambda (nm)
lambda (nm) = 299,792.458 / v (THz)

Parameter	Value	Notes
lambda	1548.52 nm	Transponder card reading
v	299,792.458 / 1548.52 = 193.600 THz	ITU channel C21 on the 100 GHz grid

Span Loss Calculation

Span_loss (dB) = (alpha x L) + (N_splices x splice_loss) + (N_connectors x conn_loss)

Variable	Example Value	Description
alpha	0.20 dB/km	G.652D fibre attenuation in C-band [7]
L	65 km	Span length
N_splices x splice_loss	3 x 0.05 = 0.15 dB	Fusion splice losses
N_connectors x conn_loss	2 x 0.30 = 0.60 dB	Connector pair losses
Total span loss	13.75 dB	Only 4.2% of original power arrives

Link Budget (Power Budget)

P_received (dBm) = P_launch (dBm) - total_losses (dB) - margin (dB)

If the received power falls below the receiver sensitivity threshold, the link fails and amplification is required.

Unamplified example: 65 km span, launch power +2 dBm, receiver sensitivity -22 dBm, margin 4 dB.

Step	Value
Span loss	13.75 dB
P_received	+2 - 13.75 - 4 = -15.75 dBm (26.6 uW)
Sensitivity check	-15.75 > -22 — link works without amplification
Remaining headroom	6.25 dB beyond minimum sensitivity

Amplified example: 120 km span, launch +1 dBm, sensitivity -18 dBm, margin 3 dB. Span loss = 24.75 dB. Without amplification: received = -26.75 dBm — below threshold. With a mid-span EDFA at km 60 providing 25 dB gain: received = +1 - 12.3 + 25 - 12.45 - 3 = -1.75 dBm — passes comfortably.

The diagram below traces optical power node-by-node through a 3-span amplified link. Each EDFA restores power with a deliberate 0.2 dB over-compensation per span to account for connector ageing and splice degradation across the 15–20 year life of the plant:

flowchart LR
    A["+1 dBm<br/>1.26 mW<br/>Launch"] -->|"80 km fibre<br/>-16.8 dB"| B["-15.8 dBm<br/>26 µW"]
    B --> C["EDFA 1<br/>+17 dB"]
    C -->|"Output"| D["+1.2 dBm<br/>1.32 mW"]
    D -->|"80 km fibre<br/>-16.8 dB"| E["-15.6 dBm<br/>28 µW"]
    E --> F["EDFA 2<br/>+17 dB"]
    F -->|"Output"| G["+1.4 dBm<br/>1.38 mW"]
    G -->|"80 km fibre<br/>-16.8 dB"| H["-15.4 dBm<br/>29 µW"]
    H --> I["Pre-amp<br/>+17 dB"]
    I --> J["+1.6 dBm<br/>At receiver"]
    style A fill:#1D9E75,stroke:#0F6E56,color:#fff
    style B fill:#E24B4A,stroke:#A32D2D,color:#fff
    style C fill:#BA7517,stroke:#854F0B,color:#fff
    style D fill:#1D9E75,stroke:#0F6E56,color:#fff
    style E fill:#E24B4A,stroke:#A32D2D,color:#fff
    style F fill:#BA7517,stroke:#854F0B,color:#fff
    style G fill:#1D9E75,stroke:#0F6E56,color:#fff
    style H fill:#E24B4A,stroke:#A32D2D,color:#fff
    style I fill:#BA7517,stroke:#854F0B,color:#fff
    style J fill:#7F77DD,stroke:#534AB7,color:#fff

Point	dBm	mW	Event
Launch	+1.0	1.26	Transponder output
After span 1	-15.8	0.026	16.8 dB loss in 80 km fibre
After EDFA 1	+1.2	1.32	17.0 dB gain
After span 2	-15.6	0.028	16.8 dB loss
After EDFA 2	+1.4	1.38	17.0 dB gain
After span 3	-15.4	0.029	16.8 dB loss
After pre-amp	+1.6	1.45	17.0 dB gain, receiver-ready

Note

The 0.2 dB/span over-compensation is standard practice. It builds in margin for the inevitable increase in splice and connector losses as the fibre plant ages — typically 0.1-0.3 dB over a 15-20 year life cycle.

OSNR — Definition and Engineering Approximation

The fundamental definition is a pure ratio of signal power to noise power, both measured within a standard 0.1 nm (12.5 GHz) reference bandwidth:

OSNR (dB) = 10 x log10(P_signal / P_noise)

For planning cascaded EDFA chains with identical spans, a shortcut approximation applies:

OSNR ~ 58 + P_ch - NF - 10 x log10(N_spans) - span_loss

Note

The constant “58” is not arbitrary — it derives from the physics of spontaneous emission noise. Each EDFA adds ASE noise proportional to P_ASE = 2 × n_sp × h × v × Δv × (G − 1), where h is Planck’s constant (6.626 × 10⁻³⁴ J·s) and v ≈ 193.4 THz [21, 24]. Evaluated in dBm, the term 10 × log₁₀(2 × h × v × Δv) equals approximately −58 dBm.

Worked example: 8-span amplified link, 80 km spans, channel power 0 dBm, noise figure 5.5 dB.

Parameter	Value
Span loss	(0.2 x 80) + 0.2 + 0.6 = 16.8 dB
OSNR	58 + 0 - 5.5 - 10 x log10(8) - 16.8 = 26.67 dB
Linear SNR	10^(26.67/10) = 464:1
DP-16QAM threshold	~22 dB
Margin	26.67 - 22 = 4.67 dB — Pass

Note

OSNR thresholds quoted in this chapter (e.g., DP-16QAM ~22 dB) are conservative SD-FEC planning numbers [21, 28]. Modern oFEC / OpenZR+ implementations require approximately 17–20 dB OSNR for DP-16QAM at 200 G — vendor data sheets remain the authoritative source for any specific transponder generation [18, 19].

Every doubling of span count costs exactly 3 dB of OSNR — a logarithmic penalty that explains why subsea systems with 60+ amplifier hops must use DP-QPSK while metro links of 2–4 spans can sustain DP-16QAM. The cascade and design lookup below operationalise the formula for quick feasibility checks:

flowchart LR
    TX["TX<br/>P = 0 dBm<br/>OSNR = ∞"] -->|"Span 1<br/>-17 dB"| A1["EDFA 1<br/>G = 17 dB<br/>adds ASE"]
    A1 -->|"Span 2<br/>-17 dB"| A2["EDFA 2<br/>G = 17 dB<br/>adds ASE"]
    A2 -->|"Span 3<br/>-17 dB"| A3["EDFA 3<br/>G = 17 dB<br/>adds ASE"]
    A3 -->|"..."| AN["EDFA N"]
    AN --> RX["RX<br/>P = 0 dBm<br/>OSNR = ?"]
    style TX fill:#1D9E75,stroke:#0F6E56,color:#fff
    style A1 fill:#BA7517,stroke:#854F0B,color:#fff
    style A2 fill:#BA7517,stroke:#854F0B,color:#fff
    style A3 fill:#BA7517,stroke:#854F0B,color:#fff
    style AN fill:#BA7517,stroke:#854F0B,color:#fff
    style RX fill:#D85A30,stroke:#993C1D,color:#fff

OSNR by span count, assuming NF = 5.5 dB, span loss = 17 dB, channel power = 0 dBm:

Spans	10·log₁₀(N)	OSNR (dB)	Linear Ratio	Supports
1	0.0	35.5	3548:1	Any format
4	6.0	29.5	891:1	DP-16QAM (needs ~22 dB)
8	9.0	26.5	447:1	DP-16QAM (with 4.5 dB margin)
12	10.8	24.7	295:1	DP-8QAM (needs ~18 dB)
16	12.0	23.5	224:1	DP-8QAM (with 5.5 dB margin)
24	13.8	21.7	148:1	DP-QPSK (needs ~12 dB)
40	16.0	19.5	89:1	DP-QPSK (with 7.5 dB margin)

Key Insight

Every doubling of span count costs exactly 3 dB of OSNR. This logarithmic scaling is why subsea systems with 60+ amplifier hops must use DP-QPSK while metro links of 2-4 spans can sustain DP-16QAM — the modulation ceiling falls predictably with hop count.

Tip

Use this lookup as the working design aid: count amplifier hops, read the OSNR, and verify the target modulation has at least 3 dB of margin before committing to a line-system design.

Shannon Capacity Limit

C = B x log2(1 + SNR_linear)
SNR_linear = 10^(OSNR_dB / 10)

A 400G channel occupying 75 GHz bandwidth at 25 dB OSNR reaches a theoretical Shannon limit of 623 Gbps [28, 30]. The actual 400G payload uses 64% of this limit — typical for production coherent systems, reflecting the gap between ideal and achievable coding.

System Capacity and Fibre Latency

Formula	Example	Result
Total capacity = N_channels x per_channel_rate	88 ch x 200G DP-QPSK	17.6 Tbps per fibre pair
Same fibre upgraded	88 ch x 400G DP-16QAM	35.2 Tbps — doubled, same infrastructure
One-way latency = distance x 5 us/km	Riyadh-Jeddah, 950 km	4.75 ms one-way, ~9.52 ms RTT

Note

Light in glass travels at roughly 200,000 km/s — two-thirds of vacuum speed due to the refractive index (~1.47). The 5 µs/km rule accounts for this and is accurate enough for all planning purposes.

Modulation Formats and Coherent Detection

Key Insight

Every modulation format trades capacity for reach [21, 28]. More bits per symbol means higher throughput per channel but demands a cleaner signal (higher OSNR), which limits transmission distance. Modern transponders resolve this tradeoff dynamically through probabilistic constellation shaping [29].

The fundamental relationship is: bits per symbol = log2(constellation size). Dual polarisation (DP) doubles the effective bits per symbol by transmitting two independent signals on orthogonal polarisations (X and Y) through the same fibre simultaneously.

Modulation	Constellation Points	Bits/Symbol	With DP (x2)	Typical Rate	Reach	Required OSNR
OOK/NRZ	2	1	N/A	10G	Long	~10 dB
QPSK	4	2	DP-QPSK = 4	100-200G	2000-6000 km	~12 dB
8QAM	8	3	DP-8QAM = 6	300-400G	800-1500 km	~18 dB
16QAM	16	4	DP-16QAM = 8	400-800G	80-500 km	~22 dB

Coherent detection is the enabling technology behind all modern modulation formats [21, 28]. Unlike legacy direct-detect systems that measured only light intensity (on/off), a coherent receiver mixes the incoming signal with a local oscillator laser and recovers amplitude, phase, and polarisation state simultaneously — accessing all four dimensions of the optical field. This allows the receiver’s DSP to compensate chromatic dispersion, polarisation mode dispersion, and nonlinear impairments digitally, eliminating the need for external dispersion-compensating fibre.

Current-generation transponders (Ciena WaveLogic 6, Nokia PSE-6s, Infinera ICE7) auto-select the optimal modulation based on real-time channel conditions. The technique — probabilistic constellation shaping — places constellation points non-uniformly to approach the Shannon limit more closely, squeezing maximum capacity from each link without manual configuration [29].

graph TD
    subgraph "Modulation — reach vs capacity"
        QPSK["DP-QPSK<br/>4 bits/sym<br/>100-200G<br/>Reach: 6000 km"]
        QAM8["DP-8QAM<br/>6 bits/sym<br/>300-400G<br/>Reach: 1500 km"]
        QAM16["DP-16QAM<br/>8 bits/sym<br/>400-800G<br/>Reach: 500 km"]
    end
    QPSK -->|"Higher capacity<br/>shorter reach"| QAM8
    QAM8 -->|"Higher capacity<br/>shorter reach"| QAM16
    style QPSK fill:#1D9E75,stroke:#0F6E56,color:#fff
    style QAM8 fill:#BA7517,stroke:#854F0B,color:#fff
    style QAM16 fill:#D85A30,stroke:#993C1D,color:#fff

The same tradeoff plotted as bar charts makes the non-linear scaling obvious — required OSNR roughly doubles between adjacent formats while reach drops by 3–4×:

xychart-beta
    title "Maximum Reach by Modulation Format (400G, standard fibre)"
    x-axis "Modulation" ["DP-QPSK", "DP-8QAM", "DP-16QAM", "DP-64QAM"]
    y-axis "Approximate Reach (km)" 0 --> 7000
    bar [6000, 1500, 500, 80]

xychart-beta
    title "Required OSNR by Modulation Format"
    x-axis "Modulation" ["DP-QPSK", "DP-8QAM", "DP-16QAM", "DP-64QAM"]
    y-axis "Required OSNR (dB)" 0 --> 30
    bar [12, 18, 22, 28]

Key Insight

As required OSNR roughly doubles between adjacent formats, reach drops by 3-4×. Always select the highest modulation the link’s OSNR can sustain with at least 3 dB of margin — that maximises capacity per wavelength.

Constellation Layouts

Each format places symbols on the complex I/Q plane. More symbols means more bits per symbol, but tighter symbol spacing demands cleaner OSNR to avoid decision errors:

OOK / NRZ                         QPSK
                                       Q
  I axis →                             │
  ●─────────────────●             01 ● │ ● 00
  0                 1                  │
                                  ─────┼─────── I
  2 points = 1 bit/symbol              │
                                  11 ● │ ● 10
                                       │
                                  4 points = 2 bits/symbol
                                  DP-QPSK = 4 bits/symbol

8QAM                              16QAM
       Q                                 Q
       │                                 │
    ●  │  ●                       ●  ● │ ●  ●
  ●    │    ●                     ●  ● │ ●  ●
  ─────┼─────── I                 ─────┼─────── I
  ●    │    ●                     ●  ● │ ●  ●
    ●  │  ●                       ●  ● │ ●  ●
       │                                 │
  8 points = 3 bits/symbol         16 points = 4 bits/symbol
  DP-8QAM = 6 bits/symbol          DP-16QAM = 8 bits/symbol

Dual-Polarisation Multiplexing

The “DP” prefix doubles spectral efficiency by transmitting two independent symbol streams on orthogonal polarisation states of the same wavelength. The coherent receiver separates them using a polarisation beam splitter and a local oscillator:

flowchart LR
    LASER["Laser<br/>source"] --> SPLIT["Polarisation<br/>beam splitter"]
    SPLIT -->|"X polarisation"| MODX["Modulator X<br/>e.g. QPSK = 2 bits"]
    SPLIT -->|"Y polarisation"| MODY["Modulator Y<br/>e.g. QPSK = 2 bits"]
    MODX --> COMBINE["Polarisation<br/>beam combiner"]
    MODY --> COMBINE
    COMBINE --> FIBER["Single λ on fibre<br/>carries 2 + 2 = 4 bits/symbol"]
    style LASER fill:#378ADD,stroke:#185FA5,color:#fff
    style MODX fill:#1D9E75,stroke:#0F6E56,color:#fff
    style MODY fill:#1D9E75,stroke:#0F6E56,color:#fff
    style COMBINE fill:#7F77DD,stroke:#534AB7,color:#fff
    style FIBER fill:#BA7517,stroke:#854F0B,color:#fff

DWDM System Architecture

A DWDM line system consists of a small number of component types chained together in a repeating pattern [16, 26, 27]. Understanding each component’s role and its contribution to the link budget is the foundation of optical network design.

flowchart LR
    TX["Transponders<br/>lambda1...lambda96"] --> MUX["MUX"]
    MUX --> BA["EDFA<br/>Booster"]
    BA --> F1["Fiber<br/>Span 1<br/>~80 km"]
    F1 --> ILA["EDFA<br/>ILA"]
    ILA --> F2["Fiber<br/>Span 2<br/>~80 km"]
    F2 --> ROADM["ROADM<br/>Add/Drop"]
    ROADM --> F3["Fiber<br/>Span 3"]
    F3 --> PA["EDFA<br/>Pre-amp"]
    PA --> DMX["DEMUX"]
    DMX --> RX["Receivers<br/>lambda1...lambda96"]
    ROADM -.->|"Drop lambda5"| LOCAL["Local<br/>client"]
    style TX fill:#378ADD,stroke:#185FA5,color:#fff
    style RX fill:#378ADD,stroke:#185FA5,color:#fff
    style MUX fill:#7F77DD,stroke:#534AB7,color:#fff
    style DMX fill:#7F77DD,stroke:#534AB7,color:#fff
    style BA fill:#1D9E75,stroke:#0F6E56,color:#fff
    style ILA fill:#1D9E75,stroke:#0F6E56,color:#fff
    style PA fill:#1D9E75,stroke:#0F6E56,color:#fff
    style ROADM fill:#BA7517,stroke:#854F0B,color:#fff
    style LOCAL fill:#D85A30,stroke:#993C1D,color:#fff

End-to-end DWDM system: transponders feed the MUX, amplified fibre spans carry the composite signal, ROADM nodes add/drop individual wavelengths, and the far-end DEMUX delivers each channel to its receiver.

Component	Function	Key Detail
Transponder	Converts client signals (100GbE, 400GbE) into DWDM wavelengths [17]	Contains coherent DSP, DAC/ADC, modulator, and tuneable laser. Modern “universal” transponders handle all modulation formats in software.
MUX / DEMUX	Combines (or separates) multiple wavelengths onto a single fibre [27]	Technologies: AWG (Arrayed Waveguide Grating) and thin-film filters. Typical insertion loss 4-7 dB.
EDFA	Amplifies all wavelengths simultaneously in the optical domain [12, 13, 24]	Three roles: booster (post-transmitter), inline/ILA (mid-span every 80-120 km), pre-amplifier (pre-receiver). Adds ASE noise.
ROADM	Remotely selects wavelengths to add, drop, or pass through [27]	Uses WSS (Wavelength Selective Switch) technology [23]. Modern configurations: 1x9 or 1x20 degree for mesh topologies. Enables software-defined optical networking.
OSC	Carries management data on a dedicated out-of-band channel	Operates at ~1510 nm (outside C-band). Transports amplifier status, alarms, firmware updates, remote configuration.
DCM	Compensates chromatic dispersion	Legacy systems used DCF (dispersion-compensating fibre). Modern coherent DSP handles dispersion digitally — no external DCM required.

The end-to-end client-to-client path makes the layering concrete. Both client routers see nothing but a 100GbE Ethernet link; everything between them — OTN framing, DWDM multiplexing, coherent modulation, optical amplification — is completely transparent:

flowchart LR
    CLIENT1["Client router<br/>100GbE port"] -->|"100GBASE-LR4"| TXP1["Transponder<br/>Maps to ODU4<br/>Wraps in OTU4<br/>Modulates λ42"]
    TXP1 --> MUX["MUX<br/>Combines<br/>96 λ"]
    MUX --> BOOST["EDFA<br/>Booster"]
    BOOST --> SPAN1["80 km<br/>fibre"]
    SPAN1 --> ILA1["EDFA<br/>ILA"]
    ILA1 --> SPAN2["80 km"]
    SPAN2 --> ILA2["EDFA<br/>ILA"]
    ILA2 --> SPAN3["80 km"]
    SPAN3 --> PREAMP["EDFA<br/>Pre-amp"]
    PREAMP --> DEMUX["DEMUX<br/>Splits<br/>96 λ"]
    DEMUX --> TXP2["Transponder<br/>Demodulates λ42<br/>Unwraps OTU4<br/>Extracts ODU4"]
    TXP2 -->|"100GBASE-LR4"| CLIENT2["Client router<br/>100GbE port"]
    style CLIENT1 fill:#378ADD,stroke:#185FA5,color:#fff
    style CLIENT2 fill:#378ADD,stroke:#185FA5,color:#fff
    style TXP1 fill:#7F77DD,stroke:#534AB7,color:#fff
    style TXP2 fill:#7F77DD,stroke:#534AB7,color:#fff
    style MUX fill:#534AB7,stroke:#3D2F99,color:#fff
    style DEMUX fill:#534AB7,stroke:#3D2F99,color:#fff
    style BOOST fill:#1D9E75,stroke:#0F6E56,color:#fff
    style ILA1 fill:#1D9E75,stroke:#0F6E56,color:#fff
    style ILA2 fill:#1D9E75,stroke:#0F6E56,color:#fff
    style PREAMP fill:#1D9E75,stroke:#0F6E56,color:#fff

Key Insight

The client routers at both ends see nothing but a 100GbE Ethernet link. Everything between them — OTN framing, DWDM multiplexing, coherent modulation, optical amplification — is completely transparent.

Walking the transmit side: the client router sends 100GbE frames over a short-reach optic [17]; the transponder maps the payload into an ODU4 (adding PM and TCM overhead) [3], wraps it in OTU4 (adding FEC, typically oFEC or SD-FEC) [14, 19], and modulates onto a specific ITU lambda (e.g. λ42 = 193.200 THz) using the format appropriate for the link distance; the MUX combines this lambda with up to 95 others; a booster EDFA amplifies the combined signal for the first span; ILAs restore power after each subsequent span; the pre-amp, DEMUX, and remote transponder reverse the entire process at the far end.

Optical Amplifiers — EDFA and Raman

EDFA — Erbium-Doped Fibre Amplifier

An EDFA contains a short coil (10-30 metres) of fibre doped with erbium (Er3+) ions [12, 24]. A pump laser at 980 nm or 1480 nm excites these ions to a higher energy state, creating population inversion. When C-band signal photons (1530-1565 nm) enter the excited fibre, they trigger stimulated emission: each signal photon causes an erbium ion to release an identical photon — same wavelength, same phase, same direction. Every wavelength within the erbium gain window is amplified simultaneously and transparently.

The unavoidable cost is ASE (Amplified Spontaneous Emission). Some erbium ions decay spontaneously rather than being stimulated by a signal, generating broadband noise across the entire C-band. This ASE noise accumulates with every amplifier in the chain and ultimately limits the achievable OSNR [21].

flowchart LR
    PUMP["Pump laser<br/>980 or 1480 nm"] -->|"Excites Er3+ ions"| EDF["Erbium-doped<br/>fibre coil<br/>10-30 m"]
    SIG_IN["Signal in<br/>C-band lambdas"] --> EDF
    EDF -->|"Stimulated<br/>emission"| SIG_OUT["Amplified<br/>signal out<br/>+ ASE noise"]
    style PUMP fill:#D85A30,stroke:#993C1D,color:#fff
    style EDF fill:#1D9E75,stroke:#0F6E56,color:#fff
    style SIG_IN fill:#378ADD,stroke:#185FA5,color:#fff
    style SIG_OUT fill:#7F77DD,stroke:#534AB7,color:#fff

EDFA: pump laser excites erbium ions, signal photons trigger stimulated emission, all wavelengths amplified simultaneously.

Placement roles:

Booster — immediately after the transmitter, maximises launch power into the fibre.
Inline amplifier (ILA) — every 80-120 km to compensate span loss. The most common deployment.
Pre-amplifier — before the receiver, lifts weak signals above the detection threshold.

Warning

EDFA gain must match span loss [12, 13]. A 30 dB gain EDFA injects roughly 10× more ASE noise power than a 20 dB gain EDFA (because P_ASE is proportional to G − 1). Over a multi-span chain, excess gain at each site compounds into several dB of unnecessary OSNR degradation — potentially forcing a modulation downgrade that halves per-channel capacity. Modern systems use automatic gain control (AGC) to lock each amplifier’s gain to its span loss within 0.5 dB.

Design parameters:

Noise figure: typically 4.5-6 dB; 5.5 dB is the standard planning value [12, 24].
Gain flattening filter (GFF) inside the EDFA compensates for non-uniform gain across the C-band.
Transient control (AGC) prevents surviving-channel power overshoot when wavelengths are added or removed.

The two plots below quantify the gain-vs-noise tradeoff that motivates AGC. ASE noise rises linearly with gain on the dB scale (each extra 10 dB of gain multiplies noise power by 10×), and end-of-chain OSNR drops in lockstep:

xychart-beta
    title "ASE Noise Power vs EDFA Gain (NF = 5.5 dB)"
    x-axis "EDFA Gain (dB)" [10, 12, 14, 16, 18, 20, 22, 24, 26, 28, 30]
    y-axis "ASE Noise per amp (dBm)" -32 --> -18
    line "ASE noise" [-31, -29.5, -28.0, -26.5, -25.2, -24.0, -22.8, -21.8, -20.8, -19.9, -19.0]

xychart-beta
    title "End-of-chain OSNR vs EDFA Gain (10 spans, P_ch = 0 dBm)"
    x-axis "EDFA Gain (dB)" [10, 12, 14, 16, 18, 20, 22, 24, 26, 28, 30]
    y-axis "OSNR after 10 spans (dB)" 20 --> 32
    line "OSNR" [31.0, 29.5, 28.0, 26.5, 25.2, 24.0, 22.8, 21.8, 20.8, 19.9, 19.0]

A worked comparison makes the cost of misconfiguration concrete. With a span loss of 17 dB, every dB of gain beyond 17 is excess — it amplifies noise without adding signal:

Parameter	Gain = 17 dB (matched)	Gain = 25 dB (excess)	Difference
ASE per amplifier	-25.5 dBm (2.8 µW)	-19.0 dBm (12.6 µW)	4.5× more noise
Total ASE (10 amps)	-15.5 dBm (28 µW)	-9.0 dBm (126 µW)	4.5× more noise
OSNR at receiver	25.5 dB	21.0 dB	-4.5 dB penalty
Supported modulation	DP-16QAM (needs ~22 dB)	Marginal — may force DP-8QAM	Half the capacity

Warning

Excess gain is the most common misconfiguration in amplified DWDM lines. A 4.5 dB OSNR penalty can force a modulation downgrade from DP-16QAM to DP-8QAM, halving per-channel capacity. AGC must lock each amplifier’s gain to its measured span loss within 0.5 dB — verify on every commissioning turn-up.

Gain Spectrum and Gain Flattening Filter

An EDFA’s gain spectrum is governed by the energy-level structure of erbium ions in silica glass [24]. The raw (unfiltered) gain curve exhibits two dominant features: a sharp peak near 1530-1535 nm — the strongest erbium emission line, where gain can exceed the band edges by 6-8 dB — and a gradual roll-off toward 1565 nm as the erbium emission cross-section decreases. Without correction, a 96-channel DWDM system would see massive inter-channel power divergence after even a single amplifier.

xychart-beta
    title "EDFA Gain vs Wavelength (C-band, no GFF)"
    x-axis "Wavelength (nm)" [1525, 1530, 1535, 1538, 1542, 1545, 1548, 1550, 1553, 1556, 1558, 1560, 1563, 1565, 1568]
    y-axis "Relative Gain (dB)" 15 --> 30
    line "Raw EDFA gain" [18, 22, 28, 26, 23, 22, 21.5, 21, 20.8, 20.5, 20.2, 20, 19.5, 19, 18]
    line "After GFF" [21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21]

The Gain Flattening Filter (GFF) is a passive optical filter inside the EDFA that attenuates peak wavelengths more aggressively than trough wavelengths, producing flat gain (typically ±0.5 dB) across the entire C-band [12, 24]. Every modern commercial EDFA includes a GFF as a standard internal component.

Warning

Gain-tilt accumulation is multiplicative across spans. Even ±1 dB of residual ripple per amplifier compounds to ±10 dB over a 10-span chain — enough to push edge channels below the receiver’s OSNR threshold while centre channels saturate. The GFF flatness specification (typically ±0.5 dB) is one of the most critical parameters on an EDFA datasheet.

Gain Shape Dependence on Operating Point

The EDFA gain profile is not static — it changes with the erbium population inversion level, which is controlled by pump power. At high inversion (high gain settings), the 1530 nm peak dominates. At low inversion (low gain), the peak flattens and the long-wavelength roll-off becomes less severe. This behaviour is called gain tilt versus operating point, and it means variable-gain EDFAs must re-optimise their GFF or employ a Variable Optical Attenuator (VOA) to maintain flatness across the operating range.

Channel loading also affects per-channel gain. When additional wavelengths are lit on a fibre, the EDFA’s total output power is shared among more channels. Modern EDFAs operate in constant-gain (AGC) mode — pump power is adjusted automatically to maintain the same per-channel gain regardless of how many channels are present, preventing surviving-channel power excursions during wavelength provisioning.

EDFA Internal Architecture

The functional blocks inside a modern EDFA follow a fixed signal path:

flowchart LR
    subgraph "Inside an EDFA"
        INPUT["Signal in<br/>96 lambdas"] --> ISO1["Isolator"]
        ISO1 --> EDF["Erbium-doped<br/>fibre coil"]
        PUMP["Pump laser<br/>980 nm"] -->|"WDM coupler"| EDF
        EDF --> GFF["Gain flattening<br/>filter (GFF)"]
        GFF --> VOA["Variable optical<br/>attenuator (VOA)"]
        VOA --> ISO2["Isolator"]
        ISO2 --> MON["Tap + monitor<br/>photodiode"]
        MON --> OUTPUT["Signal out<br/>flat gain"]
    end
    style INPUT fill:#378ADD,stroke:#185FA5,color:#fff
    style PUMP fill:#D85A30,stroke:#993C1D,color:#fff
    style EDF fill:#1D9E75,stroke:#0F6E56,color:#fff
    style GFF fill:#BA7517,stroke:#854F0B,color:#fff
    style VOA fill:#7F77DD,stroke:#534AB7,color:#fff
    style OUTPUT fill:#378ADD,stroke:#185FA5,color:#fff

The input and output isolators prevent backward reflections that would cause parasitic lasing. The tap-monitor photodiode feeds the AGC control loop, and the VOA provides fine output-power adjustment to hit the exact target gain.

Raman Amplifier

Raman amplification exploits the transmission fibre itself as the gain medium [22, 25]. A high-power pump laser (~1450 nm for C-band gain) is injected counter-directionally from the far end of the span. Through Stimulated Raman Scattering (SRS), pump photons transfer energy to signal photons — the energy difference corresponds to a vibrational mode of the silica molecules.

The critical advantage is that Raman gain is distributed across the last 20-30 km of fibre rather than concentrated at a single point. The signal never drops as low as it would without Raman, yielding a 2-4 dB OSNR improvement over EDFA alone — a significant margin in long-haul design.

flowchart RL
    SIG["Signal<br/>propagation"] --> FIBER["Transmission fibre<br/>80+ km<br/>Gain happens here"]
    RPUMP["Raman pump<br/>~1450 nm<br/>counter-propagating"] --> FIBER
    FIBER --> OUT["Amplified<br/>signal"]
    style SIG fill:#378ADD,stroke:#185FA5,color:#fff
    style FIBER fill:#1D9E75,stroke:#0F6E56,color:#fff
    style RPUMP fill:#D85A30,stroke:#993C1D,color:#fff
    style OUT fill:#7F77DD,stroke:#534AB7,color:#fff

Raman: pump laser injects energy into the fibre itself. Distributed amplification across the last 20-30 km keeps the signal above the noise floor.

Eye-Safety

Raman pump lasers operate at high power (300-500 mW). Automatic Laser Shutdown (ALS) must be verified during commissioning — an open fibre connector under full Raman pump power is an eye-safety hazard.

Design considerations:

Raman gain is wavelength-dependent (shorter wavelengths gain more), requiring tilt management via the NMS.
Different fibre types (G.652D, G.655, G.654E) produce different Raman gain profiles [7, 8, 9].
Double Rayleigh Backscatter (DRB) limits practical Raman gain to approximately 15-20 dB.
Hybrid Raman + EDFA is the most common production deployment for long spans — see Hybrid Raman + EDFA — Simulation and Engineering below for the full quantitative analysis.

Raman Gain Profile and Multi-Pump Flattening

Raman gain arises from the vibrational modes of the silica lattice rather than from doped ions, producing a fundamentally different spectral shape from EDFAs [22, 25]. Peak gain occurs at approximately 13 THz (~100 nm) below the pump wavelength — for a pump at 1450 nm, the gain peaks near 1550 nm, precisely targeting the C-band. The useful gain bandwidth from a single pump spans roughly 30-40 nm, and the profile is broad but asymmetric: a gradual rise from the pump wavelength to the peak, followed by a sharper drop-off beyond it.

xychart-beta
    title "Raman Gain vs Wavelength (single pump at 1450 nm)"
    x-axis "Wavelength (nm)" [1470, 1480, 1490, 1500, 1510, 1520, 1530, 1540, 1545, 1550, 1555, 1560, 1570, 1580, 1590]
    y-axis "Raman Gain (dB)" 0 --> 12
    line "Single pump gain" [1, 2, 3.5, 5, 7, 9, 10.5, 11.5, 11.8, 12, 11, 9, 6, 3, 1]

The single-pump gain curve is triangular — shorter C-band wavelengths receive more gain than longer ones — producing a pronounced tilt that must be compensated in multi-span systems.

Production Raman amplifiers therefore deploy 2-4 pump wavelengths simultaneously. Each pump generates its own gain curve peaked ~100 nm above, and the individual profiles sum to a composite that can be made flat to within ±1 dB across the C-band. Pumps at 1425 nm and 1450 nm, for example, produce overlapping gain profiles whose sum is substantially flatter than either alone:

xychart-beta
    title "Multi-pump Raman — flattened C-band gain"
    x-axis "Wavelength (nm)" [1525, 1530, 1535, 1540, 1545, 1550, 1555, 1560, 1565]
    y-axis "Gain (dB)" 0 --> 14
    line "Pump 1 (1425 nm)" [3, 5, 7, 8, 9, 8.5, 7, 5, 3]
    line "Pump 2 (1450 nm)" [1, 2, 3, 4, 5, 6, 7, 7, 6]
    line "Combined" [8, 9, 10, 10, 10, 10, 10, 9.5, 9]

EDFA vs Raman Flatness Control

EDFA gain flatness depends on a fixed passive GFF — a physical filter designed for a specific gain shape. Raman gain flatness is tuned by adjusting pump powers and wavelengths in software or firmware, making it inherently more flexible but also more complex to optimise. This distinction matters during field commissioning: EDFA flatness is factory-set, while Raman flatness requires per-span characterisation.

Rule of Thumb

Raman either extends usable span length by ~20 km or buys ~3 dB of OSNR at the same distance — often enough to step up one modulation level (DP-16QAM instead of DP-8QAM) on a marginal link.

Hybrid Raman + EDFA — Simulation and Engineering

Hybrid Raman + EDFA amplification is the standard architecture for long-haul and subsea DWDM spans [21, 25]. This subsection quantifies the OSNR advantage through a 170 km span simulation comparing three amplification strategies, explains why standalone Raman is not viable beyond ~75 km spans, details the physical co-location of Raman pump and EDFA at a single site, and provides an engineering decision guide based on span length and terrain.

The OSNR Argument — Why Distributed 12 dB Outperforms Lumped 28 dB

Key Insight

OSNR degradation is determined by the point of lowest signal power along the span, not by the total gain applied at the end. Amplifying a noise-degraded signal produces a louder noise-degraded signal — the damage is irreversible.

Raman gain of 10-12 dB is far less than a typical EDFA’s 20-30 dB, yet Raman consistently delivers superior OSNR on long spans. The reason is spatial: Raman gain is distributed across the last 30-35 km of the fibre itself through Stimulated Raman Scattering, preventing the signal from reaching the deep power minimum that occurs with end-of-span EDFA-only amplification.

An EDFA at the far end restores the signal to any desired power level, but it cannot undo the OSNR penalty accumulated while the signal sat near the noise floor. Raman eliminates that penalty at its source.

Metric	EDFA Only (16 dB gain at end)	Raman 12 dB + EDFA 4 dB at end
Signal at weakest point	-15 dBm (32 µW)	-7 dBm (200 µW)
Effective system NF	5.5 dB	~-1 dB (effective)
OSNR contribution per span	26.5 dB	29.5 dB
OSNR improvement	Baseline	+3 dB better

That 3 dB margin is the difference between DP-8QAM and DP-16QAM on a marginal link — literally doubling per-wavelength capacity, or extending reach by 25-30 % at the same modulation format.

Water-Bucket Analogy

Consider carrying water in a leaky bucket across a desert. The EDFA approach lets the bucket drain completely, then provides a large refill at the destination — severe dehydration occurred over the last 30 km and caused permanent damage. The Raman approach has someone walking backwards from the destination, dripping water into the bucket for the final 30 km — less total water added, but the critical low point was avoided entirely. The smaller refill does not matter because the damage never happened.

Why Raman Is Never Used Alone — DRB Revisited

Raman gain is limited to approximately 10-15 dB before Double Rayleigh Backscatter (DRB) creates a delayed copy of the signal that interferes destructively with itself [22, 25]. DRB arises because a small fraction of the amplified signal is Rayleigh-scattered backward, then Rayleigh-scattered forward again — arriving at the receiver as a delayed, incoherent replica that degrades the signal.

For any span exceeding ~75 km (i.e., more than ~15 dB of loss), Raman alone cannot fully compensate the attenuation. The production deployment is therefore always hybrid Raman + EDFA:

Component	Role	Typical Contribution
Raman pump	Distributed gain in the fibre	10-12 dB
EDFA	Discrete gain at the receiving site	Remainder to match total span loss
Combined	Full span-loss compensation	Matches fibre loss exactly

Physical Architecture — Single Equipment Shelf

Both the Raman pump and the EDFA are co-located in the same equipment shelf at the receiving end of the span (Site B). There is one rack, one power feed, one management connection — no additional infrastructure beyond what a standard EDFA-only site requires.

flowchart LR
    SITEA["Site A<br/>Transmitter<br/>+2 dBm"] -->|"170 km fibre<br/>34 dB total loss"| WDM["WDM<br/>coupler"]
    WDM --> EDFA["EDFA<br/>+22 dB gain"]
    RAMAN["Raman pump<br/>~1450 nm<br/>300-500 mW"] -->|"Pump light<br/>counter-propagating"| WDM
    EDFA --> OUT["Output<br/>+2 dBm<br/>next span"]
    subgraph SITEB ["Site B — single equipment shelf"]
        WDM
        RAMAN
        EDFA
    end
    style SITEA fill:#378ADD,stroke:#185FA5,color:#fff
    style RAMAN fill:#D85A30,stroke:#993C1D,color:#fff
    style EDFA fill:#1D9E75,stroke:#0F6E56,color:#fff
    style OUT fill:#7F77DD,stroke:#534AB7,color:#fff

Signal arrives from the fibre into a WDM coupler. The Raman pump (~1450 nm) is injected backward through the same coupler. After distributed Raman gain over the last ~35 km, the partially amplified signal enters the EDFA for discrete amplification to the target output power.

170 km Span Simulation — Three Amplification Strategies

Key Insight

A 170 km span at 0.2 dB/km incurs 34 dB of total loss. Three amplification strategies produce dramatically different minimum signal powers, OSNR outcomes, and infrastructure requirements.

Approach 1 — EDFA Only at the Far End

Launch power is +2 dBm. The signal decays continuously for 170 km, arriving at +2 - 34 = -32 dBm (0.6 µW). A single EDFA provides 34 dB of gain to restore the signal. The problem: the signal spent the last 50 km below -20 dBm, deep in the noise floor. An EDFA producing 34 dB of gain also injects massive ASE noise (P_ASE is proportional to G - 1).

xychart-beta
    title "Approach 1: EDFA only at end — signal power vs distance"
    x-axis "Distance (km)" [0, 20, 40, 60, 80, 100, 120, 135, 150, 170]
    y-axis "Power (dBm)" -35 --> 5
    line "Signal" [2, -2, -6, -10, -14, -18, -22, -25, -28, -32]

Approach 2 — Mid-Span EDFA at km 85

The first 85 km drops the signal to +2 - 17 = -15 dBm. A mid-span EDFA restores to +2 dBm. The second 85 km drops the signal to -15 dBm again. An end EDFA restores to +2 dBm. The minimum power is only -15 dBm — 17 dB better than Approach 1.

The cost: a powered shelter at km 85 requiring land lease, power supply (often solar + batteries in remote areas), environmental protection, security, and ongoing maintenance access. On desert routes this can cost hundreds of thousands of dollars.

xychart-beta
    title "Approach 2: Mid-span EDFA at km 85 — signal power vs distance"
    x-axis "Distance (km)" [0, 20, 40, 60, 85, 85, 100, 120, 140, 170]
    y-axis "Power (dBm)" -20 --> 5
    line "Signal" [2, -2, -6, -10, -15, 2, -1, -5, -9, -15]

Approach 3 — Raman + EDFA at the Far End

The signal decays identically to Approach 1 for the first 135 km, reaching -25 dBm (3 µW). At km 135, the Raman pump’s gain zone begins. Over the last 35 km, Raman provides ~12 dB of distributed gain while fibre loss continues at 0.2 dB/km, yielding a net +5 dB improvement. The signal arrives at -20 dBm (10 µW).

The EDFA then provides 22 dB of gain (34 - 12 = 22), restoring the signal to +2 dBm. The minimum power of -25 dBm is 7 dB better than Approach 1, with no mid-span infrastructure required.

xychart-beta
    title "Approach 3: Raman + EDFA at end — signal power vs distance"
    x-axis "Distance (km)" [0, 20, 40, 60, 80, 100, 120, 135, 145, 155, 165, 170]
    y-axis "Power (dBm)" -30 --> 5
    line "Signal" [2, -2, -6, -10, -14, -18, -22, -25, -23, -22, -21, -20]

All Three Approaches Overlaid

xychart-beta
    title "170 km span — signal power comparison"
    x-axis "Distance (km)" [0, 20, 40, 60, 85, 100, 120, 135, 150, 170]
    y-axis "Power (dBm)" -35 --> 5
    line "EDFA only" [2, -2, -6, -10, -15, -18, -22, -25, -28, -32]
    line "Mid-span EDFA" [2, -2, -6, -10, 2, -1, -5, -9, -12, -15]
    line "Raman+EDFA" [2, -2, -6, -10, -15, -18, -22, -25, -23, -20]

The three curves diverge after km 85. EDFA-only continues its monotonic decay. Mid-span EDFA resets at km 85, creating a sawtooth. Raman + EDFA bends upward at km 135 as distributed gain fights fibre loss. The vertical gap between curves at km 170 represents the OSNR difference.

Quantitative Comparison

Approach	Equipment Sites	Minimum Signal	OSNR per Span	Practical Notes
EDFA only at end	1	-32 dBm (0.6 µW)	~19 dB — marginal	May force DP-QPSK (reduced capacity)
Mid-span EDFA at km 85	2	-15 dBm (32 µW)	~26 dB — excellent	Requires mid-span powered shelter
Raman + EDFA at end	1	-25 dBm (3 µW)	~23 dB — good	No mid-span site, moderate cost

When Each Approach Wins

Decision Guide

The choice between amplification strategies is driven by span length and mid-span accessibility, not by which technology produces more gain.

EDFA-only at the far end is viable for spans under ~100 km where total loss stays below 20 dB. Signal minimum is manageable, EDFA gain is moderate, and ASE accumulation is acceptable. This is the standard for most terrestrial backbone spans.

Mid-span EDFA delivers the best OSNR but at the highest cost. It is appropriate when the route passes through accessible areas where building a shelter is practical (urban or semi-urban corridors), or on very long spans (over 150 km) where even Raman cannot compensate enough loss.

Raman + EDFA at the far end is the sweet spot for long spans (100-170 km) through remote areas where mid-span construction is prohibitively expensive:

Deployment Context	Why Hybrid Raman + EDFA
Desert routes (e.g., Saudi Arabia, Australia, Sahara crossings)	No power or physical access for hundreds of km
Subsea cables	Every repeater pod on the ocean floor is hybrid Raman + EDFA; mid-span access is physically impossible; spans are 60-80 km but over 6000+ km of cable, every dB of OSNR matters
Mountain crossings	Remote terrain where shelters are extremely expensive to build and maintain

EDFA Gain-Reduction Benefit in Hybrid Deployments

Key Insight

In a hybrid setup, the EDFA operates at significantly lower gain because Raman has already contributed 10-12 dB. Lower gain means dramatically less ASE noise injection.

A secondary but substantial advantage of the hybrid architecture: the EDFA runs at lower gain. For the 170 km simulation, the EDFA produces 22 dB instead of 34 dB. Since ASE noise power is proportional to (G - 1):

Parameter	EDFA-Only (34 dB gain)	Hybrid (22 dB EDFA gain)	Ratio
Linear gain	2512	158	—
P_ASE proportional to (G - 1)	2511	157	16× less noise

The EDFA in the hybrid setup injects 16 times less ASE noise than the EDFA-only approach. Combined with the higher minimum signal power from distributed Raman gain, the total OSNR benefit is substantial — the two effects reinforce each other.

Warning

This gain-reduction benefit applies to the individual amplifier’s ASE contribution. Over a multi-span chain, the cumulative effect is even more pronounced: lower ASE per stage means the noise floor rises more slowly, enabling longer total reach or higher modulation formats at the same distance.

C+L Wideband Amplification

Why a Single EDFA Cannot Cover Both Bands

As C-band capacity fills, operators turn to L-band (1565-1625 nm) to approximately double channel count [21, 24]. Erbium’s emission spectrum exhibits a natural dip between the C and L windows, and achieving gain across L-band requires a much longer erbium fibre at lower population inversion than C-band. No single EDFA design can efficiently span both bands, so production C+L systems employ a parallel architecture with separate C-band and L-band amplifiers.

Parallel C+L EDFA Architecture

flowchart LR
    INPUT["C+L signal<br/>in"] --> SPLIT["C/L band<br/>splitter"]
    SPLIT -->|"C-band<br/>1530-1565"| CEDFA["C-band<br/>EDFA<br/>short EDF<br/>high inversion"]
    SPLIT -->|"L-band<br/>1565-1625"| LEDFA["L-band<br/>EDFA<br/>long EDF<br/>low inversion"]
    CEDFA --> COMBINE["C/L band<br/>combiner"]
    LEDFA --> COMBINE
    COMBINE --> OUTPUT["C+L signal<br/>out"]
    style INPUT fill:#378ADD,stroke:#185FA5,color:#fff
    style SPLIT fill:#BA7517,stroke:#854F0B,color:#fff
    style CEDFA fill:#7F77DD,stroke:#534AB7,color:#fff
    style LEDFA fill:#D85A30,stroke:#993C1D,color:#fff
    style COMBINE fill:#BA7517,stroke:#854F0B,color:#fff
    style OUTPUT fill:#378ADD,stroke:#185FA5,color:#fff

The C/L band splitter and combiner are passive dichroic filters that separate or merge the two bands at ~1567-1568 nm with approximately 0.5 dB insertion loss. Each EDFA is independently optimised for its band:

Parameter	C-band EDFA	L-band EDFA
EDF length	10-15 m	50-100 m
Pump wavelength	980 nm (primary)	1480 nm (primary)
Population inversion	High (~70%)	Low (~40%)
Gain per metre	High	Low (compensated by fibre length)
Noise figure	4.5-5.5 dB	5.0-6.5 dB
Gain flatness (with GFF)	±0.5 dB	±1.0 dB
Power consumption	Lower	Higher (longer fibre, more pump power)
Typical gain (C+L with SRS compensation)	20-22 dB	16-18 dB

Note

L-band EDFAs have both worse noise figure and lower gain setting (to avoid over-amplifying SRS-boosted channels). Combined, L-band channels carry approximately 2-3 dB less OSNR than C-band at the same span count. This may force a modulation downgrade on marginal links — a span supporting DP-16QAM on C-band might only sustain DP-8QAM on L-band. Link budgets must be calculated separately for each band.

C+L Gain Profile — A Step, Not a Slope

Each band’s EDFA is independently flattened by its own GFF, so the combined C+L gain profile exhibits a discrete step at the band boundary rather than a continuous slope. The C-band EDFA typically operates at 20-22 dB gain; the L-band EDFA at 16-18 dB. The 4-5 dB offset compensates for bulk SRS power transfer from C-band to L-band during fibre propagation [22] — it is not a deficiency of the L-band amplifier but a deliberate design choice.

xychart-beta
    title "C+L EDFA gain — SRS-compensated operating point"
    x-axis "Wavelength (nm)" [1530, 1535, 1540, 1545, 1550, 1555, 1560, 1565, 1570, 1575, 1580, 1585, 1590, 1595, 1600, 1610, 1620]
    y-axis "Gain (dB)" 14 --> 24
    line "C-band EDFA (higher)" [22, 22, 22, 22, 22, 22, 22, 22, 0, 0, 0, 0, 0, 0, 0, 0, 0]
    line "L-band EDFA (lower)" [0, 0, 0, 0, 0, 0, 0, 0, 17, 17, 17, 17, 17, 17, 17, 16.5, 16]

Common expectation	Actual behaviour	Reason
One continuous gain curve sloping from C to L	Two separate flat curves with a step between them	C and L are physically separate EDFAs with independent GFFs
Smooth slope compensates SRS	Step offset between band gains compensates bulk SRS	Each EDFA is independently flattened
EDFA handles all SRS compensation	Three layers work together: EDFA offset + pre-tilt + DGE	SRS tilt is nonlinear and varies with channel loading

SRS Compensation — Three Layers

Stimulated Raman Scattering transfers approximately 3-4 dB of power from C-band channels to L-band channels per 80 km span [22]. Compensation is distributed across three mechanisms:

Layer 1 — EDFA Gain Offset Between Bands

The C-band EDFA is set to a higher gain than the L-band EDFA (e.g. 22 dB vs 17 dB). This 5 dB offset accounts for C-band channels arriving weaker and L-band channels arriving stronger after SRS. Within each band, gain remains flat — the offset is a step, not a tilt.

Layer 2 — Pre-Tilt at Launch

Transponders or per-channel VOAs at the transmit side launch C-band channels 2-3 dB higher than L-band channels. After SRS redistributes power during propagation, channels arrive roughly equalised at the far-end amplifier.

Layer 3 — Dynamic Gain Equalisation (DGE)

At ROADM nodes, a Wavelength Selective Switch measures every channel’s power individually and adjusts each within ±0.5 dB. This per-channel fine-tuning catches residual tilt that the first two layers cannot perfectly correct — particularly important because SRS tilt is not perfectly linear across the band and varies with channel loading.

flowchart LR
    TX["Transmitter<br/>Pre-tilt:<br/>C launched higher<br/>L launched lower"] -->|"80 km fibre<br/>SRS: C loses ~3 dB<br/>L gains ~3 dB"| AMP["C/L EDFAs<br/>C-EDFA: 22 dB<br/>L-EDFA: 17 dB"]
    AMP --> DGE["DGE / WSS<br/>Per-channel<br/>fine adjustment<br/>±0.5 dB"]
    DGE --> NEXT["To next span<br/>channels equalised"]
    style TX fill:#378ADD,stroke:#185FA5,color:#fff
    style AMP fill:#1D9E75,stroke:#0F6E56,color:#fff
    style DGE fill:#7F77DD,stroke:#534AB7,color:#fff
    style NEXT fill:#BA7517,stroke:#854F0B,color:#fff

Wideband Raman for C+L

Raman amplification handles C+L more naturally than EDFA because gain bandwidth is determined by pump count rather than doped-fibre chemistry [25]. A 4-pump Raman architecture covers the full 1525-1620 nm range without requiring a band splitter or combiner — the gain occurs in the transmission fibre itself, which naturally carries both bands:

Pump Wavelength	Gain Peak	Band Coverage
~1425 nm	~1525 nm	Short C-band
~1450 nm	~1550 nm	Mid C-band
~1470 nm	~1570 nm	Long C / short L
~1495 nm	~1595 nm	Mid L-band

Independent adjustment of the four pump powers shapes the composite gain to be flat across the entire C+L range. This flexibility is a key reason hybrid Raman + EDFA is preferred for C+L systems: Raman provides broadband pre-amplification across both bands in a single fibre, then split C/L EDFAs supply the remaining discrete gain.

SRS Power Tilt — The C+L Engineering Problem

SRS in the transmission fibre causes shorter wavelengths to transfer energy to longer wavelengths — the same physical effect exploited by Raman amplifiers, but here it is uncontrolled and unwanted [22]. In a loaded C+L system, the effect is dramatic:

xychart-beta
    title "SRS Power Tilt — shorter wavelengths lose power to longer wavelengths"
    x-axis "Wavelength (nm)" [1530, 1540, 1550, 1560, 1570, 1580, 1590, 1600, 1610, 1620]
    y-axis "Relative Power Change (dB)" -4 --> 4
    line "SRS tilt after 80 km" [-3.5, -2.8, -2.0, -1.2, -0.3, 0.5, 1.3, 2.0, 2.8, 3.5]

The shortest C-band channels (1530 nm) lose 3-4 dB of power per span, while the longest L-band channels (1620 nm) gain 3-4 dB. The total tilt across C+L can reach 6-8 dB per 80 km span. Left uncorrected over 10 spans, this compounds into a massive power imbalance that renders edge channels unusable.

In a C-band-only system, SRS tilt is modest — approximately 1-2 dB across the 35 nm C-band, easily handled by the EDFA’s GFF and per-channel power adjustments. Doubling the spectral width to ~90 nm for C+L roughly triples the SRS interaction bandwidth and doubles the channel count, pushing per-span tilt to 6-8 dB — far beyond static GFF compensation.

Parameter	C-band Only	C+L Band	Impact
Spectral width	~35 nm	~90 nm	2.5× more SRS interaction
SRS tilt per span	~1-2 dB	~6-8 dB	Requires active per-channel management
Number of channels	~96	~192	More channels = more total power = more SRS
Pre-tilt requirement	Minimal	2-4 dB across C+L	Must be calculated per link
Amplifier complexity	Single C-band EDFA	Split C + L EDFAs	Double the amplifier count and cost
OSNR penalty	Negligible	0.5-1.5 dB for L-band	L-band channels have slightly worse OSNR
Nonlinear threshold	Higher per channel	Must reduce per-channel power	Total fibre power constrained

SRS is not the only nonlinear concern. Self-Phase Modulation (SPM), Cross-Phase Modulation (XPM), and Four-Wave Mixing (FWM) all worsen as total fibre power increases [22, 30]. With 192 channels in C+L, total launch power can reach +22 to +24 dBm — approaching the fibre’s nonlinear regime (~+18 to +21 dBm composite for typical SMF, channel-count and spacing dependent; +23 dBm is on the high end for standard G.652D).

The design must balance per-channel power high enough for adequate receiver OSNR against total power low enough to avoid nonlinear distortion. Typical per-channel launch power in C+L systems is -3 to +1 dBm, chosen so the sum of all channels remains below the nonlinear threshold. Modern planning tools (Ciena MCP / OnePlanner, Nokia WaveSuite) solve this optimisation automatically.

Practical C+L Upgrade Example

A Tier-1 SP upgrading an existing 88-channel C-band backbone (88 × 400G = 35.2 Tbps) to C+L faces the following engineering sequence:

L-band EDFAs are added at every amplifier site in parallel with existing C-band EDFAs, connected through C/L band splitters and combiners.
SRS pre-tilt is calculated by the NMS for each span based on measured fibre parameters.
Per-channel launch power is reduced from 0 dBm (C-only) to approximately -2 dBm (C+L) to keep total fibre power below the nonlinear threshold.
The 2 dB per-channel power reduction plus L-band EDFA’s ~1 dB worse noise figure costs L-band channels approximately 3 dB of OSNR. Links running DP-16QAM on C-band may require DP-8QAM on L-band.
Existing C-band transponders continue operating unchanged. New transponders are added for L-band channels.
ROADM WSS modules are upgraded to C+L variants if the existing hardware does not cover L-band.

The result: capacity approximately doubles from 35 Tbps to ~65 Tbps on the same fibre pair, at the cost of more complex amplification and reduced per-channel performance on L-band.

Noise Figure — Deep Dive

NF Definition

Every optical amplifier degrades the signal-to-noise ratio [12, 24]. Noise Figure quantifies this degradation:

NF = SNR_input / SNR_output    (linear ratio)
NF (dB) = 10 × log10(SNR_in / SNR_out)

An EDFA with NF = 5.5 dB degrades the SNR by a factor of 3.55. The quantum-mechanical minimum NF for any phase-insensitive optical amplifier is 3 dB (factor of 2) — this limit is fundamental because stimulated emission (useful gain) is always accompanied by spontaneous emission (unavoidable noise) [21].

NF and ASE — The Exact Formula

NF determines the ASE noise power injected by each amplifier:

P_ASE = 2 × NF_linear × h × v × dv × (G - 1)

Variable	Value	Description
NF_linear	10^(NF_dB / 10)	e.g. 10^(5.5/10) = 3.55
h	6.626 × 10⁻³⁴ J·s	Planck’s constant
v	193.4 × 10¹² Hz	C-band centre frequency
dv	12.5 × 10⁹ Hz	0.1 nm reference bandwidth (ITU)
G	10^(Gain_dB / 10)	e.g. 10^(20/10) = 100

The factor of 2 accounts for ASE in both polarisation modes. This formula connects directly to the OSNR approximation in Essential Optical Engineering Formulas above, where the constant “58” derives from 10 × log10(2 × h × v × dv) evaluated in dBm.

Worked Example — NF = 5.5 dB, Gain = 20 dB

Step	Calculation	Result
Convert NF to linear	10^(5.5/10)	3.55
Convert gain to linear	10^(20/10)	100
Calculate P_ASE	2 × 3.55 × 6.626e-34 × 193.4e12 × 12.5e9 × 99	1.125 µW = -29.5 dBm
OSNR after 1 amplifier	10 × log10(1000 µW / 1.125 µW)	29.5 dB
OSNR after 10 amplifiers	10 × log10(1000 / 11.25)	19.5 dB

Key Insight

Each amplifier in a cascade adds equal ASE noise power. After N identical amplifiers, OSNR degrades by exactly 10 × log10(N) — hence the 10 dB penalty going from 1 to 10 amplifiers (29.5 dB to 19.5 dB in the example above).

NF by Technology

NF (dB)	Linear Factor	SNR Degradation	Typical Technology
3.0	2.0×	Quantum limit	Theoretical minimum — unachievable in practice
4.0	2.5×	Near-quantum-limited	Research laboratory prototypes
4.5	2.8×	Excellent	Premium low-noise EDFA, 980 nm pump
5.0	3.2×	Very good	High-quality commercial EDFA
5.5	3.5×	Standard planning value	Typical commercial EDFA
6.0	4.0×	Acceptable	Budget EDFA or L-band EDFA
6.5	4.5×	Mediocre	Older or budget L-band EDFA
7.0	5.0×	Poor	End-of-life equipment

The 1 dB NF Rule

Every 1 dB improvement in noise figure translates directly to 1 dB better OSNR at the receiver — always, regardless of span count, gain setting, or channel count. Upgrading from NF 6.0 to NF 4.5 across all amplifier sites yields 1.5 dB of OSNR margin. That 1.5 dB might enable DP-16QAM where only DP-8QAM was previously feasible (doubling per-channel capacity) or extend reach by approximately 15-20 km (equivalent to eliminating one amplifier site).

Friis Formula — Why the First Amplifier Dominates

In a cascade of amplifiers, the overall noise figure is dominated by the first stage [21]:

NF_total = NF1 + (NF2 - 1)/G1 + (NF3 - 1)/(G1 × G2) + ...

Because G1 is large (e.g. 100× for 20 dB gain), the second and subsequent terms become negligible. The first amplifier’s NF sets the noise floor for the entire chain. This is why service providers deploy a premium low-noise pre-amplifier (NF ~4.5 dB) at the receiver end where the signal is weakest — even when inline amplifiers are cheaper models with NF 5.5-6 dB.

Warning

A common cost-saving mistake is deploying budget amplifiers everywhere, including the pre-amplifier position. The Friis formula shows that saving 1 dB of NF at the pre-amplifier is worth far more than saving 1 dB at any inline site. Premium pre-amplifiers pay for themselves in reach and capacity.

Raman Effective NF Below 3 dB

A standalone EDFA cannot achieve NF below 3 dB — this is a quantum-mechanical limit. However, hybrid Raman + EDFA configurations routinely achieve an effective combined NF of 0 to -2 dB [25]. This does not violate physics.

The mechanism: Raman provides distributed gain in the fibre before the signal reaches the EDFA. The signal arrives at the EDFA 10-15 dB stronger than it would without Raman pre-amplification. The EDFA injects the same absolute ASE noise power, but relative to the much stronger incoming signal, the effective SNR degradation is smaller. When the combined system is characterised as a single “black box” amplifier, its effective NF falls below the 3 dB quantum limit — because the measurement convention assumes the signal enters at its un-amplified level, not at the Raman-boosted level.

This is the fundamental physical reason hybrid Raman + EDFA improves OSNR by 2-4 dB over EDFA-only designs, and it explains why every long-haul and subsea system employs distributed Raman amplification.

Reading NF on Amplifier Datasheets

When evaluating EDFA specifications, four curves and trade-offs require attention:

Datasheet Characteristic	What to Check	Typical Variation
NF vs gain	NF at the actual operating gain, not just the headline value	0.5-1 dB across the gain range
NF vs wavelength	NF at band edges vs centre (lowest near 1540-1550 nm)	±0.5 dB across C-band
NF at low input power	NF may increase when per-channel input approaches the noise floor	Check at actual per-channel input power
Single-stage vs dual-stage	Dual-stage (with mid-stage GFF/DCM/VOA access) has slightly higher NF	0.5-1 dB penalty, offset by better gain-flatness control

Amplifier Selection Guide

Scenario	Recommendation
Standard terrestrial spans under 100 km	EDFA only — sufficient and simplest to operate
Long spans (100-150 km) or OSNR-sensitive routes	Hybrid Raman + EDFA
Subsea cable systems	Always Raman + EDFA — every fraction of a dB matters
Metro DCI under 80 km	Often unamplified — 400ZR+ pluggable optics handle the loss budget
L-band or S-band expansion	Raman required — standard EDFAs cover C-band only; dedicated L-band EDFAs exist but Raman offers more flexibility

The decision-tree below provides the engineering rationale behind those rows, walking from “what bands do you need” through the appropriate amplifier architecture and SRS-management requirements:

flowchart TD
    START["What band(s)<br/>do you need?"] -->|"C-band only"| CONLY{"Span<br/>length?"}
    CONLY -->|"Under 100 km"| EDFA_ONLY["EDFA only<br/>simplest, cheapest"]
    CONLY -->|"100-150 km"| HYBRID["Hybrid Raman + EDFA<br/>better OSNR"]
    CONLY -->|"Over 150 km<br/>(subsea)"| HYBRID_MAX["Raman + EDFA<br/>+ extended fibre types<br/>(G.654E)"]
    START -->|"C+L band"| CL{"Amplifier<br/>architecture"}
    CL --> SPLIT_EDFA["Split C/L EDFAs<br/>+ band splitter/combiner<br/>at every amp site"]
    CL --> WIDEBAND_RAMAN["Wideband Raman (4 pumps)<br/>+ split C/L EDFAs<br/>for maximum reach"]
    SPLIT_EDFA --> SRS_MGMT["SRS management required:<br/>pre-tilt + DGE + power optimisation"]
    WIDEBAND_RAMAN --> SRS_MGMT
    style EDFA_ONLY fill:#1D9E75,stroke:#0F6E56,color:#fff
    style HYBRID fill:#BA7517,stroke:#854F0B,color:#fff
    style HYBRID_MAX fill:#D85A30,stroke:#993C1D,color:#fff
    style SPLIT_EDFA fill:#7F77DD,stroke:#534AB7,color:#fff
    style WIDEBAND_RAMAN fill:#7F77DD,stroke:#534AB7,color:#fff
    style SRS_MGMT fill:#E24B4A,stroke:#A32D2D,color:#fff

Rule of Thumb

Below 100 km with C-band only: EDFA is sufficient and simplest to operate. Above 100 km or when deploying C+L: hybrid Raman + EDFA is the standard choice. Subsea systems always use Raman — every fraction of a dB matters when amplifier sites are separated by hundreds of kilometres.

OTN — Optical Transport Network (G.709)

Key Insight

OTN is the digital wrapper that turns raw DWDM wavelengths into carrier-grade transport circuits [3, 5]. It adds FEC to extend reach, ODU containers for sub-wavelength grooming, and tandem connection monitoring for multi-carrier fault isolation — capabilities that DWDM alone cannot provide [4].

Layer Model

OTN nests three container layers around client payloads:

OCh (Optical Channel) — one lambda on the fibre. The outermost optical container.
OTU (Optical Transport Unit) — adds FEC and section-layer framing. Modern soft-decision FEC variants — SD-FEC and oFEC (e.g., OpenZR+/oFEC) — correct raw BER from ~10^-2 down to 10^-15, enabling coherent links to operate well beyond their uncorrected error threshold [15, 19]. Classic G.975 hard-decision RS(255,239) is far weaker, correcting only from a pre-FEC BER of ~1.5×10^-4 [14].
ODU (Optical Data Unit) — the path layer. Carries client payload plus performance monitoring, tandem connection monitoring (up to 6 levels), and overhead. ODUs can be switched, multiplexed, and protected independently of the optical layer.

graph TB
    subgraph OCh ["OCh — Optical Channel (one lambda)"]
        subgraph OTU ["OTU — adds FEC + framing"]
            subgraph ODU ["ODU — digital path layer"]
                PAYLOAD["Client payload<br/>+ PM + TCM overhead"]
            end
        end
    end
    style OCh fill:#FAECE7,stroke:#D85A30
    style OTU fill:#EEEDFE,stroke:#7F77DD
    style ODU fill:#E6F1FB,stroke:#378ADD
    style PAYLOAD fill:#E1F5EE,stroke:#1D9E75

OTN nesting: client payload wrapped in ODU (path), then OTU (FEC + frame), then mapped to an OCh (lambda).

OTN Rate Hierarchy

Container	Line Rate	Client Signal	OTU Pair	Multiplexing
ODU0	1.244 Gbps	GbE (1000BASE-X)	None (tributary only)	Mapped into ODU1 or higher [3]
ODU1	2.499 Gbps	STM-16 (or 2× ODU0 / GbE via multiplexing)	OTU1 (2.667 Gbps)	2 x ODU0
ODU2	10.037 Gbps	10GbE, STM-64	OTU2 (10.709 Gbps)	4 x ODU1 or 8 x ODU0
ODU2e	10.399 Gbps	10GbE (exact rate)	OTU2e	Variant for exact 10GbE mapping
ODU3	40.319 Gbps	40GbE, STM-256	OTU3 (43.018 Gbps)	4 x ODU2 or 16 x ODU1
ODUflex	Variable	FC, sub-rate Ethernet	Mapped into ODU2/3/4	Arbitrary rate in ~1.25G steps
ODU4	104.794 Gbps	100GbE	OTU4 (111.810 Gbps)	10 x ODU2 or 80 x ODU0
ODUCn	n x ~100G	400GbE, FlexE	OTUCn	FlexO bonding for beyond-100G [17]

Note

There is no OTU0. ODU0 exists only as a tributary — it must be multiplexed into a higher-order ODU that provides the OTU wrapper with FEC before transmission on a wavelength. ODUflex extends the hierarchy to arbitrary client rates by allocating a variable number of tributary slots, enabling efficient transport of Fibre Channel (8G/16G/32G), FlexEthernet, and other non-standard payloads.

A single 100G lambda’s container nesting visualised as a tree — ODU4 carries 10 × ODU2, each ODU2 carries 4 × ODU1, each ODU1 carries 2 × ODU0, for up to 80 individually-monitored GbE services per wavelength:

graph TD
    ODU4["ODU4<br/>~105 Gbps<br/>100GbE client"] --> ODU2a["ODU2 slot 1<br/>~10G"]
    ODU4 --> ODU2b["ODU2 slot 2"]
    ODU4 --> ODU2c["..."]
    ODU4 --> ODU2d["ODU2 slot 10"]
    ODU2a --> ODU1a["ODU1 slot 1<br/>~2.5G"]
    ODU2a --> ODU1b["ODU1 slot 2"]
    ODU2a --> ODU1c["ODU1 slot 3"]
    ODU2a --> ODU1d["ODU1 slot 4"]
    ODU1a --> ODU0a["ODU0 slot 1<br/>~1.25G<br/>= 1 GbE"]
    ODU1a --> ODU0b["ODU0 slot 2<br/>= 1 GbE"]
    style ODU4 fill:#7F77DD,stroke:#534AB7,color:#fff
    style ODU2a fill:#378ADD,stroke:#185FA5,color:#fff
    style ODU2b fill:#378ADD,stroke:#185FA5,color:#fff
    style ODU2c fill:#378ADD,stroke:#185FA5,color:#fff
    style ODU2d fill:#378ADD,stroke:#185FA5,color:#fff
    style ODU1a fill:#1D9E75,stroke:#0F6E56,color:#fff
    style ODU1b fill:#1D9E75,stroke:#0F6E56,color:#fff
    style ODU1c fill:#1D9E75,stroke:#0F6E56,color:#fff
    style ODU1d fill:#1D9E75,stroke:#0F6E56,color:#fff
    style ODU0a fill:#BA7517,stroke:#854F0B,color:#fff
    style ODU0b fill:#BA7517,stroke:#854F0B,color:#fff

This sub-lambda grooming is one of OTN’s most powerful capabilities — expensive wavelengths get filled efficiently rather than wasting an entire lambda on a small customer circuit.

Why OTN Matters

Per-wavelength monitoring — BER, delay, and quality measured at every node without costly optical-to-electrical-to-optical conversion.
FEC — extends usable reach far beyond what raw coherent detection can sustain, correcting errors that would otherwise make a link unusable [14, 15].
Sub-lambda grooming — packs multiple smaller services (e.g., 4 x 10G customers) into a single 100G lambda using ODU multiplexing, analogous to SDH time-slot grooming but at modern rates [11].

Tandem Connection Monitoring (TCM)

OTN provides six independent TCM levels (TCM1-TCM6) in the ODU overhead [3]. Each level carries its own monitoring fields:

Field	Width	Purpose
BIP-8	8 bits	Parity check calculated at TCM ingress, verified at egress — detects errors within the segment
BEI	4 bits	Backward Error Indication — reports the error count back to the ingress node
BDI	1 bit	Backward Defect Indication — signals a fault condition upstream
STAT	3 bits	Signal state: normal, maintenance, or fault
DMp/DMti	Variable	Delay measurement timestamps for one-way propagation delay across the segment

Multi-carrier example — London to Tokyo:

A financial institution purchases a 100G wavelength traversing three carriers. Each carrier is assigned a distinct TCM level: BT monitors TCM1 across the UK domestic segment (London to Cornwall), the subsea carrier monitors TCM2 across the ocean floor, and NTT monitors TCM3 across the Japan domestic segment. The customer uses TCM6 for end-to-end SLA visibility spanning the entire path.

When errors spike, the TCM overhead immediately identifies the degraded segment — if TCM2 shows errors while TCM1 and TCM3 are clean, the submarine segment is the problem. No ambiguity, no inter-carrier blame — the protocol-level evidence is carried in the OTN frame itself.

flowchart LR
    A["Carrier A<br/>London<br/>TCM1"] -->|"Handoff"| B["Carrier B<br/>Subsea<br/>TCM2"]
    B -->|"Handoff"| C["Carrier C<br/>Tokyo<br/>TCM3"]
    E2E["Customer end-to-end: TCM6"]
    A ~~~ E2E
    E2E ~~~ C
    style A fill:#378ADD,stroke:#185FA5,color:#fff
    style B fill:#7F77DD,stroke:#534AB7,color:#fff
    style C fill:#D85A30,stroke:#993C1D,color:#fff
    style E2E fill:#1D9E75,stroke:#0F6E56,color:#fff

Each carrier monitors its own TCM level independently. The customer uses TCM6 for end-to-end SLA verification.

Note

TCM overhead is transparent to intermediate carriers. Each carrier reads and writes only its assigned level — the other levels pass through unmodified. TCM level assignment is agreed during service ordering and forms the contractual basis for SLA enforcement. The BIP-8 error counts provide objective, protocol-level evidence of whether each carrier met its segment quality commitment.

The fault-isolation procedure when customer-level errors appear is a simple decision tree — walk the carrier-level monitors in order until the offending segment is identified:

flowchart TD
    START["Errors detected<br/>on customer TCM6"] --> Q1{"TCM1<br/>errors?"}
    Q1 -->|Yes| FAULT_A["Fault in<br/>Carrier A segment"]
    Q1 -->|No| Q2{"TCM2<br/>errors?"}
    Q2 -->|Yes| FAULT_B["Fault in<br/>Carrier B segment"]
    Q2 -->|No| Q3{"TCM3<br/>errors?"}
    Q3 -->|Yes| FAULT_C["Fault in<br/>Carrier C segment"]
    Q3 -->|No| FAULT_IX["Fault at interconnect<br/>or terminal equipment"]
    style START fill:#E24B4A,stroke:#A32D2D,color:#fff
    style FAULT_A fill:#378ADD,stroke:#185FA5,color:#fff
    style FAULT_B fill:#7F77DD,stroke:#534AB7,color:#fff
    style FAULT_C fill:#D85A30,stroke:#993C1D,color:#fff
    style FAULT_IX fill:#BA7517,stroke:#854F0B,color:#fff

Warning

When multiple TCM levels show errors simultaneously, the fault is likely at or near a carrier handoff — a cable cut at a landing station, for instance, will affect both adjacent segments at once. Treat concurrent multi-level alarms as a handoff problem until proven otherwise.

SDH — Legacy Transport and Sunset Path

SDH (Synchronous Digital Hierarchy, ITU-T G.707/G.783) dominated transport networks from the 1990s through the 2010s [11]. North America deployed the equivalent SONET standard. While SDH is being phased out, understanding it remains relevant because migration planning is an active operational concern for many carriers.

SDH Rate Hierarchy

Level	Rate	SONET Equivalent
STM-1	155.52 Mbps	OC-3 [11]
STM-4	622.08 Mbps	OC-12
STM-16	2.488 Gbps	OC-48
STM-64	9.953 Gbps	OC-192
STM-256	39.813 Gbps	OC-768

Each level multiplexes four times the rate of the level below it.

Strengths and Limitations

SDH’s enduring contribution was its 50 ms protection switching (MSP ring, SNCP) — the benchmark that every successor technology has been measured against. It also provided robust OAM, deterministic timing for TDM voice circuits, and mature management ecosystems.

However, SDH was designed for an era when 2.5 Gbps was cutting-edge and TDM voice was the dominant payload. Modern networks are Ethernet-centric, demanding 100G+ per interface [17]. SDH’s rigid multiplexing structure cannot efficiently pack arbitrary Ethernet rates into STM containers, its per-payload overhead (~5%) is high by current standards, and all major vendors have announced product discontinuation.

Migration Path

flowchart LR
    FREEZE["Phase 1<br/>Freeze SDH"] --> GROOM["Phase 2<br/>Groom circuits"]
    GROOM --> MIGRATE["Phase 3<br/>Migrate to<br/>OTN/Ethernet"]
    MIGRATE --> DECOM["Phase 4<br/>Decommission"]
    style FREEZE fill:#E24B4A,stroke:#A32D2D,color:#fff
    style GROOM fill:#BA7517,stroke:#854F0B,color:#fff
    style MIGRATE fill:#378ADD,stroke:#185FA5,color:#fff
    style DECOM fill:#1D9E75,stroke:#0F6E56,color:#fff

Freeze — no new SDH circuits; all new services provisioned on OTN or Ethernet.
Groom — audit existing SDH circuits to identify live traffic versus zombie circuits consuming resources.
Migrate — move remaining services to OTN (ODU0/ODU1 for legacy E1/STM-n tributaries) or directly to Ethernet. Circuit Emulation Service (CES) over Ethernet/MPLS bridges timing-sensitive applications (SCADA, railway signalling, legacy banking). SAToP provides structure-agnostic E1 emulation.
Decommission — remove SDH equipment, reclaim rack space, power, and fibre pairs for DWDM expansion.

Warning

Some legacy services depend on SDH-derived synchronisation. IEEE 1588 PTP (Precision Time Protocol) provides equivalent timing distribution over packet networks, but regulatory environments may require proving timing equivalence before decommission approval.

Network Design Examples

Two designs illustrate how the optical concepts above come together in practice — one long-haul regional backbone and one short-reach metro data-centre interconnect.

Regional SP Backbone — 5-City Ring

Requirements: five cities forming a ring (Riyadh - Jeddah - Madinah - Tabuk - Ha’il - Riyadh), current peak demand 4 Tbps, 30% annual growth over five years, 99.999% availability.

Capacity planning: 4 x (1.3)^5 = 14.8 Tbps. Design target: 16 Tbps. At 400G per channel, this requires 40 wavelengths. C-band at 75 GHz spacing provides ~64 channels — ample headroom for growth without touching the line system.

graph TD
    RUH["Riyadh<br/>ROADM"] -->|"950 km<br/>DP-8QAM"| JED["Jeddah<br/>ROADM"]
    JED -->|"420 km"| MED["Madinah<br/>ROADM"]
    MED -->|"680 km"| TUK["Tabuk<br/>ROADM"]
    TUK -->|"640 km"| HAI["Ha'il<br/>ROADM"]
    HAI -->|"650 km"| RUH
    style RUH fill:#7F77DD,stroke:#534AB7,color:#fff
    style JED fill:#7F77DD,stroke:#534AB7,color:#fff
    style MED fill:#378ADD,stroke:#185FA5,color:#fff
    style TUK fill:#378ADD,stroke:#185FA5,color:#fff
    style HAI fill:#378ADD,stroke:#185FA5,color:#fff

5-city DWDM ring with ~64 C-band channels at 75 GHz spacing. Ring protection ensures 99.999% availability.

Link budget (Riyadh-Jeddah, 950 km — the longest and most challenging segment):

Parameter	Value
Span length	80 km, 12 spans total
Per-span loss	(0.2 x 80) + 0.2 + 0.6 = 16.8 dB
EDFA gain / NF	17.6 dB / 5.5 dB
Channel power	-1 dBm
OSNR	58 + (-1) - 5.5 - 10.8 - 16.8 = 23.9 dB
DP-8QAM threshold	~18 dB
Available after 4 dB margin	19.9 dB — Pass

Protection: ring with DWDM 1+1 path protection or OTN SNCP [4, 6]. A single fibre cut reroutes traffic in the opposite ring direction within 50 ms. With only 40 of 96 channels in use, the working capacity fits comfortably within the protection path.

Equipment summary: 5 ROADM sites, 11 ILA sites, 200 transponders initially. Growth to full 96-channel capacity requires only adding transponders — no line-system changes.

Metro DCI — 3-Site Data Centre Mesh

Requirements: three data centres, 15-25 km apart, 12.8 Tbps per link, sub-0.5 ms latency, 400GbE native interfaces, cost-optimised.

Amplification analysis: at 25 km, total path loss = fibre (5 dB) + connectors (0.6 dB) + ROADM (20 dB) = 25.6 dB. 400ZR/ZR+ pluggable optics handle up to ~30 dB — no inline amplifiers needed [18, 19].

graph LR
    DC1["DC 1<br/>400ZR+"] ---|"25 km<br/>No amps"| DC2["DC 2<br/>400ZR+"]
    DC2 ---|"20 km<br/>No amps"| DC3["DC 3<br/>400ZR+"]
    DC3 ---|"15 km<br/>No amps"| DC1
    style DC1 fill:#1D9E75,stroke:#0F6E56,color:#fff
    style DC2 fill:#1D9E75,stroke:#0F6E56,color:#fff
    style DC3 fill:#1D9E75,stroke:#0F6E56,color:#fff

Metro DCI: amplifier-free mesh using 400ZR+ pluggable optics. 32 channels x 400G = 12.8 Tbps per link.

Metric	Value
Channel plan	32 channels at 75 GHz (C-band fits 64 — room to double)
OSNR	~33 dB effective (no inline ASE) vs 22 dB threshold = 11 dB margin
Latency	25 km x 5 us/km + ~5 us DSP = 130 us (0.13 ms) — well under target
Cost advantage	400ZR+ QSFP-DD optics cost ~1/3 of traditional transponders; eliminating the amplified line system saves 50-60% on total optical cost [20]

Regional vs Metro Design Comparison

Aspect	Regional Backbone	Metro DCI
Distance	80-950 km	15-25 km
Amplification	EDFA every 80 km	Not required
Modulation	DP-8QAM / DP-16QAM (adaptive)	DP-16QAM fixed
Protection	Ring OTN SNCP, 50 ms	Link-level LAG/ECMP
Transponder type	Embedded line-system	400ZR+ pluggable (QSFP-DD)
Management layer	Full OTN with TCM	Minimal, router-managed
Cost driver	Amplifiers + fibre plant	Transponder density
Growth model	Add wavelengths to existing line	Add pluggable optics to switches

References

Standards (ITU-T)

ITU-T G.694.1 — Spectral grids for WDM applications: DWDM frequency grid (10/2020). https://www.itu.int/rec/T-REC-G.694.1
ITU-T G.694.2 — Spectral grids for WDM applications: CWDM wavelength grid (12/2003). https://www.itu.int/rec/T-REC-G.694.2
ITU-T G.709/Y.1331 — Interfaces for the Optical Transport Network (06/2020). https://www.itu.int/rec/T-REC-G.709
ITU-T G.798 — Characteristics of optical transport network hierarchy equipment functional blocks (12/2017). https://www.itu.int/rec/T-REC-G.798
ITU-T G.872 — Architecture of optical transport networks (10/2017). https://www.itu.int/rec/T-REC-G.872
ITU-T G.873.1 — Optical Transport Network — Linear protection (03/2022). https://www.itu.int/rec/T-REC-G.873.1
ITU-T G.652 — Characteristics of a single-mode optical fibre and cable (11/2016). https://www.itu.int/rec/T-REC-G.652
ITU-T G.654 — Characteristics of cut-off shifted single-mode optical fibre and cable (03/2020). https://www.itu.int/rec/T-REC-G.654
ITU-T G.655 — Characteristics of non-zero dispersion-shifted single-mode optical fibre and cable (11/2009). https://www.itu.int/rec/T-REC-G.655
ITU-T G.657 — Characteristics of a bending-loss insensitive single-mode optical fibre and cable (11/2016). https://www.itu.int/rec/T-REC-G.657
ITU-T G.707/Y.1322 — Network node interface for the synchronous digital hierarchy (SDH) (01/2007). https://www.itu.int/rec/T-REC-G.707
ITU-T G.661 — Definitions and test methods for the relevant generic parameters of optical amplifier devices and subsystems (07/2018). https://www.itu.int/rec/T-REC-G.661
ITU-T G.662 — Generic characteristics of optical amplifier devices and subsystems. https://www.itu.int/rec/T-REC-G.662
ITU-T G.975 — Forward error correction for submarine systems (10/2000). https://www.itu.int/rec/T-REC-G.975
ITU-T G.975.1 — Forward error correction for high bit-rate DWDM submarine systems (02/2004). https://www.itu.int/rec/T-REC-G.975.1
ITU-T G.698.2 — Amplified multichannel dense wavelength division multiplexing applications with single channel optical interfaces. https://www.itu.int/rec/T-REC-G.698.2

Standards (IEEE / OIF / MSA)

IEEE 802.3 — Ethernet; specifically 802.3bs (400 GbE), 802.3ck, 802.3df (800 GbE). https://standards.ieee.org/standard/802_3-2022.html
OIF 400ZR Implementation Agreement (OIF-400ZR-01.0). https://www.oiforum.com/documents/
OpenZR+ MSA specification. https://www.openzrplus.org/
OIF Common Management Interface Specification (CMIS). https://www.oiforum.com/documents/

Books

G. P. Agrawal, Fiber-Optic Communication Systems, 5th ed., Wiley, 2021.
G. P. Agrawal, Nonlinear Fiber Optics, 6th ed., Academic Press, 2019.
I. P. Kaminow, T. Li, A. E. Willner (Eds.), Optical Fiber Telecommunications VI-A & VI-B, Academic Press, 2013.
P. C. Becker, N. A. Olsson, J. R. Simpson, Erbium-Doped Fiber Amplifiers — Fundamentals and Technology, Academic Press, 1999.
C. Headley & G. P. Agrawal (Eds.), Raman Amplification in Fiber Optical Communication Systems, Academic Press, 2005.
V. Alwayn, Optical Network Design and Implementation, Cisco Press, 2004.
R. Ramaswami, K. N. Sivarajan, G. H. Sasaki, Optical Networks: A Practical Perspective, 3rd ed., Morgan Kaufmann, 2009.

Papers

K. Roberts et al., “Beyond 100 Gb/s: Capacity, Flexibility, and Network Optimization,” J. Lightwave Technol. 35 (2017).
J. Cho et al., “Probabilistic Constellation Shaping for Optical Fiber Communications,” J. Lightwave Technol. 37, 1590 (2019).
A. D. Ellis et al., “Performance limits in optical communications due to fiber nonlinearity,” Adv. Opt. Photon. 9, 429 (2017).

Transport Network Guide

Explorer

Optical Transport Networks — DWDM, OTN, and SDH

Optical Transport Networks — DWDM, OTN, and SDH

Key Concepts at a Glance

Fibre Physics and Transmission Bands

Core Optical Terminology

Decibel Mathematics — dB vs dBm

The Distinction

dBm-to-Milliwatt Conversion Reference

Why dB and dBm Arithmetic Works

Per-Component Loss and Gain Values

Essential Optical Engineering Formulas

Wavelength-Frequency Conversion

Span Loss Calculation

Link Budget (Power Budget)

OSNR — Definition and Engineering Approximation

Shannon Capacity Limit

System Capacity and Fibre Latency

Modulation Formats and Coherent Detection

Constellation Layouts

Dual-Polarisation Multiplexing

DWDM System Architecture

Optical Amplifiers — EDFA and Raman

EDFA — Erbium-Doped Fibre Amplifier

Gain Spectrum and Gain Flattening Filter

Gain Shape Dependence on Operating Point

EDFA Internal Architecture

Raman Amplifier

Raman Gain Profile and Multi-Pump Flattening

Hybrid Raman + EDFA — Simulation and Engineering

The OSNR Argument — Why Distributed 12 dB Outperforms Lumped 28 dB

Why Raman Is Never Used Alone — DRB Revisited

Physical Architecture — Single Equipment Shelf

170 km Span Simulation — Three Amplification Strategies

Approach 1 — EDFA Only at the Far End

Approach 2 — Mid-Span EDFA at km 85

Approach 3 — Raman + EDFA at the Far End

All Three Approaches Overlaid

Quantitative Comparison

When Each Approach Wins

EDFA Gain-Reduction Benefit in Hybrid Deployments

C+L Wideband Amplification

Why a Single EDFA Cannot Cover Both Bands

Parallel C+L EDFA Architecture

C+L Gain Profile — A Step, Not a Slope

SRS Compensation — Three Layers

Layer 1 — EDFA Gain Offset Between Bands

Layer 2 — Pre-Tilt at Launch

Layer 3 — Dynamic Gain Equalisation (DGE)

Wideband Raman for C+L

SRS Power Tilt — The C+L Engineering Problem

Practical C+L Upgrade Example

Noise Figure — Deep Dive

NF Definition

NF and ASE — The Exact Formula

Worked Example — NF = 5.5 dB, Gain = 20 dB

NF by Technology

The 1 dB NF Rule

Friis Formula — Why the First Amplifier Dominates

Raman Effective NF Below 3 dB

Reading NF on Amplifier Datasheets

Amplifier Selection Guide

OTN — Optical Transport Network (G.709)

Layer Model

OTN Rate Hierarchy

Why OTN Matters

Tandem Connection Monitoring (TCM)

SDH — Legacy Transport and Sunset Path

SDH Rate Hierarchy

Strengths and Limitations

Migration Path

Network Design Examples

Regional SP Backbone — 5-City Ring

Metro DCI — 3-Site Data Centre Mesh

Regional vs Metro Design Comparison

See Also

References

Standards (ITU-T)

Standards (IEEE / OIF / MSA)