Ch 10 — Submarine and Subsea Optical Systems

Submarine and Subsea Optical Systems

Subsea optical systems carry the majority of intercontinental Internet traffic over cables that span 6 000–13 000 km between landing stations, with amplifiers sitting on the ocean floor in pressure-rated housings powered from shore through a kilovolt DC current down the cable’s central conductor. This chapter is scope-capped to the optical layer plus the Power-Feed Equipment (PFE) — cable mechanical design, armouring, burial, and landing-station civil works are explicitly out of scope. The engineering distinction from terrestrial DWDM is not the technology stack but the margin discipline: every 0.1 dB of OSNR matters, distributed Raman is mandatory rather than optional, and Space-Division Multiplexing (SDM) has overtaken spectral expansion as the dominant capacity-scaling lever.

Concept	What it says
Repeatered systems	A subsea cable with periodic in-line optical amplifier pods (typically every 60–90 km) on the seabed, series-powered from shore-end PFE. Reaches up to 13 000+ km on transpacific routes with modern coherent transponders and SD-FEC.
Unrepeatered systems	A subsea cable with no in-line amplifiers — only end-station equipment with high-power boosters, low-NF preamps, and Remote Optical Pumping (ROP) of erbium-doped fibre installed near each end. Reaches typically 300–550 km, used for festoon and short-haul links.
Power-Feed Equipment (PFE)	Shore-end DC supply that injects high-voltage (kV-range) current into the cable’s central conductor; both shore ends inject from opposite polarity with the seawater providing the return path. Manages soft-start, earth-return current, and fault localisation.
Space-Division Multiplexing (SDM)	The current-generation capacity strategy: instead of widening the optical band, deploy more parallel paths — multi-fibre-pair cables (up to 24+ pairs) and, in research, multi-core fibre. Reduces per-pair capacity but increases total cable capacity dramatically and improves electrical-power efficiency per bit.

Repeatered vs Unrepeatered Systems

The two architectures are not competing options for the same route — they target different reach envelopes and the choice is driven by geography. Once the route exceeds approximately 500 km, in-line amplification becomes mandatory.

Parameter	Unrepeatered	Repeatered
Typical reach	300–550 km	1 000–13 000+ km
Wet-plant amplifiers	None	Erbium-doped repeaters every 60–90 km
Shore-end optics	High-power booster (+20 dBm), low-NF preamp	Standard transponder + booster
Remote pumping	ROPA (Remote Optically Pumped Amplifier) common	Local pumping in each repeater
Cable conductor	DC required only if ROPA is used	High-voltage DC mandatory (PFE)
Typical applications	Festoons, short hops (e.g. island chains, intra-Mediterranean)	Transpacific, transatlantic, equatorial backbones
Cost driver	Shore terminal optics	Wet-plant repeater count and PFE voltage rating

Key Insight

Unrepeatered does not mean “no amplification anywhere along the cable” — many unrepeatered systems include a Remote Optically Pumped Amplifier (ROPA) located 80–150 km from the shore station, with the pump light fed down a separate fibre from the shore. The cable still has no electronics underwater; only a passive erbium fibre coil that lights up when illuminated by the shore-fed pump.

Submarine Fibre Types

The fibre choice for new subsea builds is overwhelmingly G.654 cut-off-shifted single-mode fibre, specifically the G.654.B/C/D/E sub-categories optimised for low loss and large effective area. G.652 is rarely used subsea — its attenuation is too high to support the per-span OSNR margins required over thousands of kilometres.

Fibre type	Attenuation @ 1550 nm	Effective area Aeff	Subsea use
G.652.D (terrestrial workhorse)	0.20 dB/km typical	~80 µm²	Almost never — too lossy
G.654.B	0.17–0.18 dB/km	80–110 µm²	Legacy subsea
G.654.C	0.16–0.17 dB/km	110–130 µm²	Modern long-haul subsea
G.654.D / .E	0.15–0.17 dB/km (≈0.16 dB/km production median) [1]	~110–130 µm² typical (≤150 µm² in best samples) [1]	New transpacific / transatlantic builds

Rule of Thumb

A 0.04 dB/km reduction in fibre attenuation (e.g. G.652.D → G.654.E) saves roughly 3.2 dB over an 80 km span. Across an 80-span (6 400 km) cable that is 256 dB of cumulative loss budget recovered — usually enough to step the modulation up one level (PDM-QPSK → DP-8QAM) or cut amplifier count by 15–20 %.

The reason G.654 dominates subsea is its combination of low attenuation (achieved via pure-silica core rather than germanium-doped) and large effective area (which raises the nonlinear power threshold — see 05-fibre-nonlinearities-and-power-optimisation). The cut-off wavelength is shifted longer than G.652, which gives up some bend tolerance at short wavelengths — irrelevant subsea where the fibre is laid straight along the seabed and never operates outside the C/L bands.

Erbium-Doped Repeaters with Remote Optical Pumping

A subsea repeater is a pressure-rated titanium housing containing erbium-doped fibre coils, pump-laser diodes, gain-flattening filters, isolators, and tap monitors — functionally an EDFA, but mechanically engineered to survive at 8 km depth for a 25-year service life with no possibility of physical maintenance.

flowchart LR
    subgraph SHORE_A ["Shore station A"]
        TX["Transmitter<br/>+ booster"]
        PFE_A["PFE<br/>kV DC supply"]
    end
    subgraph WET ["Wet plant — seabed"]
        REP1["Repeater 1<br/>EDFA + Raman pump<br/>~80 km mark"]
        REP2["Repeater 2<br/>EDFA + Raman pump"]
        REPN["Repeater N<br/>EDFA + Raman pump"]
    end
    subgraph SHORE_B ["Shore station B"]
        RX["Preamp<br/>+ receiver"]
        PFE_B["PFE<br/>kV DC supply"]
    end
    TX -->|"G.654 fibre<br/>~80 km"| REP1
    REP1 -->|"~80 km"| REP2
    REP2 -.->|"..."| REPN
    REPN -->|"~80 km"| RX
    PFE_A -->|"+ kV DC<br/>via cable conductor"| REP1
    PFE_B -->|"- kV DC<br/>seawater return"| REPN
    style TX fill:#378ADD,stroke:#185FA5,color:#fff
    style RX fill:#378ADD,stroke:#185FA5,color:#fff
    style PFE_A fill:#D85A30,stroke:#993C1D,color:#fff
    style PFE_B fill:#D85A30,stroke:#993C1D,color:#fff
    style REP1 fill:#1D9E75,stroke:#0F6E56,color:#fff
    style REP2 fill:#1D9E75,stroke:#0F6E56,color:#fff
    style REPN fill:#1D9E75,stroke:#0F6E56,color:#fff

The cable’s central copper conductor delivers shore-fed DC to every repeater. The seawater itself completes the return path. A single fault in the central conductor takes every repeater offline simultaneously.

Each repeater pod typically contains amplifiers for all fibre pairs in the cable — modern 16–24 pair cables route fibres into a single shared mechanical housing to minimise pressure-vessel count and manufacturing cost. Pump-laser redundancy is the dominant reliability investment: pumps are dual-redundant per stage, and all redundant electronics operate continuously to avoid a “cold start” decade after deployment.

Warning

A subsea repeater cannot be repaired — only replaced via cable-ship retrieval (multi-week, multi-million-dollar operation). Reliability targets are 25 years of continuous operation with single-digit failure probability across the entire cable’s repeater chain. This drives extreme component derating and dual-redundant pump architectures, both of which inflate per-repeater power consumption — a critical constraint when total wet-plant power is shore-supply-limited.

Power-Feed Equipment

The wet plant is series-powered from shore by high-voltage DC injected into the cable’s central conductor — kV-range — with shore-end PFE rectifying mains, managing earth-return, and providing soft-start. Both shore ends drive opposite polarity with the seawater completing the circuit, so a mid-cable fault still allows partial powering from each end up to the break point. PFE current is the gate that limits how many repeaters and how many fibre pairs a single cable can carry — every additional repeater adds resistive voltage drop and constant-current consumption, eventually saturating the PFE supply rating.

That single paragraph is the entirety of this chapter’s PFE coverage; cable mechanical design, electrode systems, and shore-station civil engineering are out of scope.

SDM — Space-Division Multiplexing

Through the 2010s, subsea capacity scaling came from spectral expansion (more channels) and modulation-format upgrades (more bits per symbol). Both have run out of headroom: the C-band is already filled with 96 channels at 50 GHz spacing, L-band adds ~50 % more but suffers worse OSNR, and modulation-format upgrades cost reach exponentially because higher-order QAM demands higher OSNR.

SDM is the response: instead of pushing harder on a single fibre pair, deploy more pairs. Modern transpacific cables ship with 16, 20, or 24 fibre pairs in a single cable structure; research cables target 32+ pairs and multi-core fibre is in field trials.

Capacity strategy	Per-fibre throughput	Total cable capacity	Power efficiency (W/Tbps)	Practical limit
Single-pair, C-band only, PDM-QPSK	~10 Tbps	10 Tbps	high (electrical)	OSNR-bound
Single-pair, C+L, DP-16QAM	~30–40 Tbps	30–40 Tbps	high	OSNR + nonlinear
8-pair SDM, C-band, PDM-QPSK	~10 Tbps	80 Tbps	improving	PFE current
16-pair SDM, C-band, PDM-QPSK + PCS	~12 Tbps	~200 Tbps	best	PFE current
24-pair SDM (frontier)	~10–12 Tbps	250–300 Tbps	excellent	PFE / cable mechanics

Key Insight

SDM beats further C-band widening because amplifier power scales linearly with the number of fibre pairs but non-linearly with band width — wideband amplifiers cost disproportionately more electrical power per bit. With shore-fed PFE budget the limiting resource, “more pairs at lower per-pair power” wins decisively over “fewer pairs at higher per-pair power”.

The trade-off is per-pair capacity. SDM cables typically run lower per-channel launch power (–3 to 0 dBm per channel rather than +1 dBm in C-only systems) and simpler modulation (PDM-QPSK with Probabilistic Constellation Shaping, rather than DP-16QAM). The reach × spectral-efficiency product per pair drops, but total cable capacity and W/Tbps improve dramatically.

Line-System Design — Subsea vs Terrestrial

The subsea engineering envelope is so different from terrestrial that whole categories of margin assumptions are inverted. The table below summarises the deltas; each row warrants a paragraph of justification in a subsea design review.

Parameter	Terrestrial long-haul	Subsea repeatered
Span loss budget	18–24 dB (80–120 km)	11–15 dB (60–80 km) — shorter spans, lower losses
Number of cascaded amplifiers	10–30	80–150+
Amplifier NF target	5.0–5.5 dB acceptable	4.0–4.5 dB mandatory
Launch power per channel	0 to +1 dBm	–3 to 0 dBm — nonlinearity-driven
Distributed Raman	Optional (long spans only)	Mandatory on every span — shore-pumped near landings, in-line via co-pumping at repeaters
Per-stage OSNR budget	~25 dB acceptable	~30 dB needed — every 0.1 dB matters
Per-channel data rate	Typically 200–400G	Typically 100–200G with PCS
Modulation	DP-16QAM common, DP-64QAM short reach	PDM-QPSK or DP-8QAM with PCS, very rarely 16QAM
OSNR margin at end-of-life	2–3 dB	1.5–2 dB (cannot be repaired)
Component aging assumption	15–20 years with replaceable cards	25 years, no replacement possible

The shorter span lengths in subsea systems are not because longer spans cannot be amplified — they reflect the cumulative-OSNR arithmetic. After 100+ amplifier stages, per-stage OSNR penalty multiplies. The cascade formula OSNR_total = OSNR_stage − 10·log10(N) shows that 100 cascaded 30 dB-OSNR stages produce a 10 dB end-to-end penalty: 30 − 20 = 10 dB OSNR at the receiver, barely enough for PDM-QPSK with SD-FEC. Lengthening a span by 20 km might save one repeater, but the per-stage OSNR drop typically more than offsets the saving.

Warning

Subsea margin discipline is fundamentally different. A terrestrial system can lose 1 dB to a degraded splice and survive for years until truck-rolled to fix it. A subsea system loses 1 dB and the operator has roughly 18 months before the modulation must be downgraded — there is no truck roll. End-of-life margin is whatever was over-engineered at install, minus the steady decline from component aging. This is why subsea capacity is sometimes upgraded mid-life by replacing only the shore-end transponders with newer-generation coherent ASICs; the wet plant is fixed for its lifetime.

Reach Envelopes by Modulation

Vendor-specific, but the order of magnitude is robust across the industry. Required OSNR is at 0.1 nm reference bandwidth.

Modulation + FEC	Required OSNR @ 0.1 nm	Bits per symbol (dual-pol)	Typical subsea reach
PDM-QPSK + SD-FEC	~12.5 dB	4	12 000–13 000+ km
PDM-QPSK + SD-FEC + PCS	~12 dB (PCS shaping gain on QPSK is small, <0.5 dB — QPSK is already near-Gaussian-optimal at low order) [10]	4 (effective ~3.6)	13 000+ km, transpacific
DP-8QAM + SD-FEC + PCS	~14–15 dB	6	8 000–10 000 km
DP-16QAM + SD-FEC	~16–17 dB	8	5 000–6 500 km — transatlantic
DP-16QAM + SD-FEC + PCS	~15 dB	8 (effective ~6.5)	6 500–8 000 km
DP-32QAM	~21 dB	10	2 500–3 500 km — short subsea
DP-64QAM	~25 dB	12	<1 500 km — DCI-class subsea

Rule of Thumb

Probabilistic Constellation Shaping (PCS) buys roughly 1–1.5 dB of OSNR margin compared to uniform constellations of the same nominal order, at the cost of 5–15 % spectral-efficiency reduction. For long subsea routes, that 1.5 dB equates to ~2 000 km of additional reach — PCS is now the default on every modern subsea transponder.

See Also

• 01-optical-physics-and-link-engineering
• 03-optical-amplifiers-edfa-raman-cl-band
• 05-fibre-nonlinearities-and-power-optimisation
• 06-coherent-dsp-internals

References

Standards (ITU-T)

1. 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
2. ITU-T G.652 — Characteristics of a single-mode optical fibre and cable (11/2016). https://www.itu.int/rec/T-REC-G.652
3. 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
4. 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
5. ITU-T G.977 — Characteristics of optically amplified optical fibre submarine cable systems (07/2015). https://www.itu.int/rec/T-REC-G.977

Standards (IEC / ISO)

6. IEC 61280-2-9 — Fibre optic communication subsystem test procedures — Digital systems — Optical signal-to-noise ratio measurement for dense wavelength-division multiplexed systems. https://webstore.iec.ch/

Books

7. J. Chesnoy (Ed.), Undersea Fiber Communication Systems, 2nd ed., Academic Press, 2016.
8. G. P. Agrawal, Fiber-Optic Communication Systems, 5th ed., Wiley, 2021.
9. C. Headley & G. P. Agrawal (Eds.), Raman Amplification in Fiber Optical Communication Systems, Academic Press, 2005.

Papers

10. J. Cho et al., “Probabilistic Constellation Shaping for Optical Fiber Communications,” J. Lightwave Technol. 37, 1590 (2019).
11. P. J. Winzer, D. T. Neilson, A. R. Chraplyvy, “Fiber-optic transmission and networking: the previous 20 and the next 20 years,” Opt. Express 26, 24190 (2018).
12. O. V. Sinkin et al., “SDM for Power-Efficient Undersea Transmission,” J. Lightwave Technol. 36, 361 (2018).