4.1 Channel-by-Channel Power Evolution

To understand SRS impact at the system level, it is useful to examine the power evolution of individual channels. In a fully loaded C+L system with N channels uniformly distributed across both bands, the power of the i-th channel at the end of a fiber span of length L can be expressed relative to its input power using the Raman coupling equations. For practical system design, a simplified closed-form model captures the essential behavior.

The power tilt across the entire C+L spectrum can be modeled as a linear function of frequency (in dB), where the channel at the spectral center experiences approximately zero net Raman exchange, channels at the short-wavelength edge (C-band) lose power, and channels at the long-wavelength edge (L-band) gain power. The tilt per channel pair is proportional to their frequency separation, the total system power, and the fiber Raman efficiency.

Pi,out(dBm) = Pi,in(dBm) - α.L + ΔPSRS,i

Where the SRS contribution for channel i:
ΔPSRS,i = Tilttotal x (fi - fcenter) / BWtotal

-- Channels above center frequency: negative (power loss)
-- Channels below center frequency: positive (power gain)

4.2 Multi-Span Tilt Accumulation

In a multi-span link with inline amplifiers, the SRS tilt accumulates differently depending on the amplifier tilt compensation strategy. If amplifiers fully compensate the tilt at each span boundary (the ideal case), then the tilt experienced by each channel within any given span remains bounded. However, if tilt compensation is imperfect, residual tilt accumulates span by span, leading to progressively larger power excursions across the spectrum.

For a submarine system with 130 amplifiers spanning approximately 6,500 km (50 km spans), even a small residual tilt error of 0.1 dB per amplifier can accumulate to 13 dB, completely overwhelming the system margin. This highlights why tilt control precision is a critical design parameter for C+L band systems, especially in long-haul and submarine applications.

4.3 Impact on OSNR and System Margin

The SRS tilt has asymmetric effects on the C-band and L-band channels. C-band channels lose power to the SRS process, which directly reduces their received power and degrades their OSNR. For the shortest-wavelength C-band channel, the OSNR penalty can be 3-4 dB relative to a C-band-only system at the same launch power. L-band channels gain power from SRS, which improves their received power but can also push them closer to the nonlinear threshold if the amplifier design does not properly manage the excess power.

The net effect on system capacity depends on the balance between these competing factors. Effective tilt compensation through the amplifier chain can largely restore the flat power profile, but the ASE noise contribution from higher-gain amplifier stages (needed for tilt compensation) introduces its own OSNR penalty. In practice, well-designed C+L band systems incur approximately 1-2 dB of OSNR penalty for L-band channels relative to C-band channels operating alone, primarily due to the higher EDFA noise figure in the L-band and the insertion loss of the C/L band splitter.

5.1 Split-Band Architecture

In C+L band systems, the C-band and L-band cannot be amplified simultaneously in the same erbium-doped fiber. This is because the optimal average inversion level for C-band amplification (approximately 0.7) differs significantly from the L-band optimum (approximately 0.4). Operating at a compromise inversion level would result in poor gain flatness and high noise figures for both bands.

The standard C+L amplifier architecture therefore uses a split-band approach: the incoming C+L signal is separated by a C/L band splitter (a wavelength-selective device, typically contributing approximately 0.5 dB of insertion loss), amplified in separate C-band and L-band EDFAs, and then recombined by a C/L band combiner. Each EDFA is independently optimized for its respective band in terms of erbium-doped fiber length, pump wavelength, and gain flattening filter design.

5.2 EDFA Design Differences: C-Band vs. L-Band

The fundamental reason for the split-band architecture is visible in the EDFA gain spectrum itself. As shown in Figure 4, erbium-doped fiber produces two distinct gain regions separated by a deep gain dip around 1565-1570 nm. The conventional C-band (1530-1565 nm) and the long-wavelength L-band (1570-1610 nm) each provide approximately 4.5 THz of usable optical bandwidth, but the gain dip between them makes it impossible to achieve flat amplification across both bands simultaneously in a single EDF stage. This is the physical basis for the split-band EDFA architecture.

The C-band and L-band EDFAs differ in several important design aspects. C-band EDFAs operate at a higher average inversion level (0.6-0.7), which produces a gain spectrum spanning 1530-1565 nm with a characteristic peak near 1532 nm that must be flattened by a Gain Flattening Filter (GFF). They typically achieve noise figures of 4.5-5.5 dB with standard 980 nm forward-pumping configurations.

L-band EDFAs operate at a lower average inversion level (0.3-0.4) and require significantly longer erbium-doped fiber lengths (often 2-3 times longer than C-band EDFAs) to achieve the required gain. This is because the emission cross-section in the L-band is smaller than in the C-band, necessitating more doped fiber to accumulate the same total gain. The longer fiber length, combined with the lower inversion level, results in noise figures that are typically 0.5-1.0 dB higher than C-band EDFAs (approximately 5.0-6.5 dB). Additionally, L-band EDFAs exhibit more temperature sensitivity in their gain profile, which requires more sophisticated gain equalization and temperature compensation mechanisms.

5.3 Raman Amplification for C+L Systems

Raman amplifiers play a critical complementary role in C+L band systems. Unlike EDFAs, which provide lumped amplification at discrete sites, Raman amplifiers provide distributed gain throughout the transmission fiber itself. This distributed gain characteristic offers two key advantages for C+L systems.

First, distributed Raman amplification improves the OSNR by amplifying the signal before it is significantly attenuated, effectively reducing the noise figure of the amplification chain. A backward-pumped Raman pre-amplifier can achieve effective noise figures of 0 dB or even negative values (in the distributed sense), providing a 3-5 dB improvement in OSNR compared to EDFA-only amplification.

Second, by selecting appropriate Raman pump wavelengths and powers, the Raman gain profile can be shaped to partially compensate for the SRS tilt. This is because the Raman gain itself is wavelength-dependent and can be designed to provide more gain to the shorter-wavelength (C-band) channels that suffer from SRS depletion. However, this tilt-compensating Raman design requires careful optimization, as the Raman pump wavelengths that provide gain in the C-band may also interact with L-band signals through their own SRS process.

6.1 Fiber Types for C+L Band Systems

The choice of fiber has a profound impact on C+L band system performance. The key fiber parameters affecting system design are attenuation, effective area, chromatic dispersion, and the dispersion slope.

G.652.D (Standard SMF): The most widely deployed fiber type. With a typical effective area of 80 um², attenuation of 0.19-0.20 dB/km, and chromatic dispersion of 17 ps/nm/km at 1550 nm, standard SMF is suitable for C+L band systems but produces relatively high SRS tilt due to its moderate effective area. The low water peak specification ensures consistent performance across the E-band region between C and L bands.

G.654.E (Ultra-Low-Loss, Large Effective Area): This fiber type represents the state-of-the-art for new C+L band deployments. With effective areas exceeding 110 um² and attenuation as low as 0.16 dB/km, G.654.E fiber significantly reduces both SRS tilt (by approximately 30-40% relative to G.652.D) and total span loss. The reduced loss allows for longer spans or higher OSNR at the amplifier input, while the larger effective area reduces all nonlinear interactions including SRS, SPM, XPM, and FWM.

6.2 Managing Kerr Nonlinearities in C+L Band

While SRS is the dominant new impairment in C+L systems, the Kerr nonlinear effects (SPM, XPM, FWM) remain important and their management changes in a combined-band context. The optimal per-channel launch power in a C+L system must account for the increased total power (which drives SRS) while maintaining adequate OSNR.

The optimal launch power per channel for a C+L system is typically 1-2 dB lower than for a C-band-only system, to reduce SRS tilt while keeping the OSNR above the modulation format's threshold. For QPSK modulation requiring approximately 13-15 dB OSNR, and 16-QAM requiring 21-23 dB, this launch power reduction translates to a modest reach penalty of 10-20% for C+L systems relative to C-band-only operation at the same modulation format.

7.1 The Transient Challenge

The load-dependent nature of SRS tilt creates a significant operational challenge: any change in the number of active channels on the fiber changes the SRS tilt, which in turn affects the power and OSNR of all remaining channels. In a static, fully loaded system, the tilt is predictable and can be compensated by fixed amplifier settings. In a dynamic network with ongoing channel additions and removals, the tilt changes continuously.

Consider a C+L band system with 160 channels (80 per band), with a total launch power of +20 dBm. If 48 L-band channels are simultaneously added (for example, activating a new traffic group), the SRS tilt increases by approximately 2 dB, causing existing C-band channels to lose approximately 1 dB of power within milliseconds. If the amplifier control loops take 100 ms to respond, this transient can cause temporary BER degradation on affected channels.

7.2 Orchestrated Channel Turn-Up/Turn-Down

To manage SRS transients, C+L band systems implement orchestrated traffic management. Rather than allowing channels to be added or removed at arbitrary times and in arbitrary order, the system coordinates channel changes through an automation layer that controls the sequence and timing of turn-up and turn-down operations. Key strategies include controlled ramp-up of new channels, gradual power increases rather than instant activation, coordinated add/drop across both bands to maintain power balance, pre-compensation of amplifier tilt settings before channel activation, and real-time monitoring with fast feedback loops (sub-millisecond response times for critical transients).

7.3 Amplifier Control Requirements

The amplifier control system for C+L band operation must provide fast gain and tilt adjustment, per-channel power monitoring, automatic tilt tracking, and coordinated inter-band control. Modern C+L line systems achieve tilt correction response times of less than 10 ms for small perturbations and less than 100 ms for large traffic changes (e.g., protection switching events). The tilt correction range must cover the full SRS tilt variation from empty to fully loaded conditions, typically requiring ±4 to ±6 dB of adjustable tilt per amplifier site.

8.1 C+L Band Capacity Examples

The capacity achievable on a C+L band system depends on the channel count, modulation format, baud rate, and FEC overhead. The following table provides representative capacity calculations for common deployment configurations.

8.2 Link Budget Methodology

A C+L band link budget must account for several additional factors compared to a C-band-only budget. The fundamental OSNR equation for a cascaded amplifier chain remains the starting point, but the following C+L-specific adjustments must be included.

8.2.1 Derivation in Linear Domain

The OSNR derivation starts from the physical noise generated by a single optical amplifier. An EDFA generates Amplified Spontaneous Emission (ASE) noise with a spectral power density given by:

PASE = NF x G x h x v x Bref     (watts, linear)

The OSNR at the output of a single amplifier is the ratio of the signal power to the ASE noise power within the reference bandwidth. In a system where the amplifier gain exactly compensates the span loss (G = L_span), the signal power at the amplifier output equals the channel launch power P_ch. Therefore:

OSNR1 span = Pch / PASE

OSNR1 span = Pch / (NF x G x h x v x Bref)     (linear ratio)

For a cascade of N identical amplifiers, each contributing independent ASE noise, the total noise power adds linearly. The signal power remains P_ch (each amplifier restores it), but the noise accumulates:

Pnoise,total = N x NF x Lspan x h x v x Bref

OSNRN spans = Pch / (N x NF x Lspan x h x v x Bref)     (linear ratio)

8.2.2 Conversion to dB Domain

Now we convert each term to dB by taking 10.log₁₀ of the linear OSNR expression:

OSNRdB = 10.log10(Pch) - 10.log10(N) - 10.log10(NF) - 10.log10(Lspan) - 10.log10(h.v.Bref)

-- Now identify each term:
-- 10.log10(P_ch) in watts  =  P_ch,dBm - 30   (converting W to mW: dBm = dBW + 30)
-- BUT if P_ch is already in mW, then 10.log10(P_ch_mW) = P_ch,dBm
-- 10.log10(NF)              =  NF_dB
-- 10.log10(L_span)          =  L_span,dB
-- 10.log10(N)               =  10.log10(N_spans)
-- 10.log10(h.v.B_ref)       =  a constant we must evaluate

The key constant term requires evaluating h.v.B_ref:

h   = 6.626 x 10-34 J.s
v   = 193.4 x 1012 Hz   (at 1550 nm)
Bref = 12.5 x 109 Hz    (0.1 nm at 1550 nm)

h.v.Bref = 6.626 x 10-34 x 193.4 x 1012 x 12.5 x 109
        = 6.626 x 193.4 x 12.5 x 10(-34+12+9)
        = 16,005.95 x 10-13
        = 1.6006 x 10-9 mW    (converting W to mW: multiply by 103)
        = 1.6006 x 10-6 mW    ... wait, let us redo carefully:

-- In watts:
h.v.Bref = 6.626e-34 x 1.934e14 x 1.25e10
        = 6.626 x 1.934 x 1.25 x 10(-34+14+10)
        = 16.006 x 10-10 W
        = 1.6006 x 10-9 W

-- Convert to dBm (dBm = 10.log10(P_mW) where P_mW = P_W x 10^3):
h.v.Bref = 1.6006 x 10-6 mW

10.log10(1.6006 x 10-6) = 10 x [log10(1.6006) + log10(10-6)]
                            = 10 x [0.2042 + (-6)]
                            = 10 x (-5.7958)
                            = -57.96 dBm  ≈  -58 dBm

Substituting all terms back into the dB equation:

OSNRdB, 0.1nm = 58 + Pch,out(dBm) - NF(dB) - Lspan(dB) - 10.log10(N)

-- This is the classical "58+ formula" for OSNR in cascaded amplifier chains.
-- It assumes: identical spans, G = L_span, same NF at each amplifier.

8.2.3 C+L Band Extensions

For C+L band systems, two additional penalty terms must be added to account for impairments that do not exist in single-band systems:

━━━ LINEAR DOMAIN ━━━

From Step 3, the baseline OSNR for N identical spans is:

OSNRlinear = Pch / (N x NF x Lspan x h x v x Bref)

For C+L band, two additional loss mechanisms reduce the signal
power at each amplifier input:

(a) C/L band splitter insertion loss (L_splitter ≈ 1.122, i.e. 0.5 dB)
    — sits before the EDFA at each node, attenuates signal,
    ASE noise from EDFA is unchanged → OSNR degrades.

(b) SRS depletion factor (L_SRS,i for channel i)
    — C-band channels lose power to L-band via Raman scattering,
    so signal arriving at the amplifier input is weaker than
    L_span alone predicts → OSNR degrades.
    — L-band channels gain SRS power, but tilt correction (VOA/GFF)
    attenuates the excess, and inter-band noise transfer adds penalty.

Including both factors, the effective span loss seen by the signal becomes:

Leff,i = Lspan x Lsplitter x LSRS,i

The C+L OSNR in linear domain therefore becomes:

┌─────────────────────────────────────────────────────────────────┐
│                                                                 │
│  OSNRch,i =              Pch                                  │
│          ─────────────────────────────────────────────────────  │
│          N x NFband x Lspan x Lsplitter x LSRS,i x h.v.Bref  │
│                                                                 │
│  All quantities in linear (watts, ratios)                       │
│                                                                 │
└─────────────────────────────────────────────────────────────────┘


━━━ dB DOMAIN ━━━

Taking 10.log₁₀ of the linear expression above:

OSNRdB = 10.log10(Pch) - 10.log10(N) - 10.log10(NFband) - 10.log10(Lspan)
         - 10.log10(Lsplitter) - 10.log10(LSRS,i) - 10.log10(h.v.Bref)

Substituting:
  10.log₁₀(P_ch)       = P_ch,out (dBm)
  -10.log₁₀(h.v.B_ref) = +58 (as derived in Step 5)
  10.log₁₀(NF_band)    = NF_band (dB)
  10.log₁₀(L_span)     = L_span (dB)
  10.log₁₀(L_splitter)  = ΔOSNR_splitter (dB)
  10.log₁₀(L_SRS,i)     = ΔOSNR_SRS (dB)

┌─────────────────────────────────────────────────────────────────┐
│                                                                 │
│  OSNRch = 58 + Pch,out - NFband - Lspan - 10.log10(N)      │
│         - ΔOSNRSRS - ΔOSNRsplitter                              │
│                                                                 │
│  All values in dB/dBm, referenced to 0.1 nm bandwidth          │
│                                                                 │
└─────────────────────────────────────────────────────────────────┘

8.2.4 Worked Example: 6,400 km Submarine C+L Link

Applying the formula to a trans-oceanic C+L link:

-- Given:
Cable length        = 6,400 km
Span length         = 80 km
Nspans              = 6,400 / 80 = 80 amplifiers
Fiber loss          = 0.15 dB/km
Lspan              = 0.15 x 80 = 12 dB
TOP (total output) = +18 dBm
Channels            = 100
Pch,out            = 18 - 10.log10(100) = 18 - 20 = -2 dBm
NF                  = 4.5 dB
10.log10(80)      = 19.0 dB

-- C-band only OSNR (no SRS or splitter penalty):
OSNR = 58 + (-2) - 4.5 - 12 - 19
     = 58 - 2 - 4.5 - 12 - 19
     = 20.5 dB (per 0.1 nm)

-- With C+L band penalties (L-band worst case):
NFL-band        = 6.0 dB  (instead of 4.5)
ΔOSNRSRS       = 1.5 dB  (L-band channels gain power but C-band loses)
ΔOSNRsplitter  = 0.5 dB  (C/L splitter insertion loss)

OSNRL-band = 58 + (-2) - 6.0 - 12 - 19 - 1.5 - 0.5
              = 17.0 dB  (3.5 dB worse than C-band only)

-- This 17.0 dB is sufficient for DP-QPSK (requires ~13-15 dB)
-- but marginal for DP-16QAM (requires ~21-23 dB)

9.1 Terrestrial Long-Haul: National Backbone Upgrade

Scenario: A national carrier operates a 2,400 km backbone route with existing C-band DWDM equipment at 80 km amplifier spacing on G.652.D fiber. The C-band system is fully loaded with 80 channels at 50 GHz spacing, running 200G DP-16QAM (33 GBaud), providing 10.56 Tbps total C-band capacity. Traffic projections indicate a need for 20+ Tbps within 18 months.

Solution: The operator upgrades to a C+L band line system by adding L-band EDFA cards at each amplifier site, installing C/L band splitters/combiners, and activating 80 additional L-band channels. Due to the 2,400 km distance and OSNR limitations of DP-16QAM in the L-band over this reach, the L-band channels are configured with DP-QPSK at 64 GBaud, providing 200G per wavelength.

Result: C-band: 80 channels x 200G = 16 Tbps; L-band: 80 channels x 200G = 16 Tbps. Total: 32 Tbps — effectively tripling the original C-band capacity. The C-band channels retain their original modulation while the L-band channels use a more robust modulation format to account for the higher noise figure and SRS penalties. Raman pre-amplifiers are deployed at every third amplifier site to boost L-band OSNR on the longest spans.

9.2 Metro/DCI: Data Center Interconnect

Scenario: A cloud service provider needs 100+ Tbps of capacity between two data center campuses separated by 120 km. The route has 4 dark fiber pairs available.

Solution: Using C+L band with 128 channels at 75 GHz spacing and 800G DP-16QAM with probabilistic constellation shaping (PCS) at 130 GBaud per wavelength, each fiber achieves 102.4 Tbps. Over the short 120 km distance (single span or two-span with one inline amplifier), the OSNR budget is very comfortable even with L-band penalties.

Result: A single fiber pair delivers over 100 Tbps, leaving three fiber pairs as growth capacity or protection. The total potential capacity across all four fiber pairs exceeds 400 Tbps.

9.3 Industry Landscape and Market Drivers (2025-2026)

The optical transport industry is experiencing a technology-driven transformation as AI infrastructure demand accelerates the adoption of C+L band systems. Disaggregated WDM architectures, which decouple transponders from optical line systems, are reshaping how operators build and scale their networks. Open line systems now enable independent sourcing of coherent modules, amplifiers, and ROADMs, giving operators flexibility to deploy best-of-breed technology at each layer. The transition from 400G to 800G and 1.6T coherent pluggable modules is accelerating, with ZR and ZR+ form factors bringing DWDM capabilities directly into router line cards.

The vendor landscape features several major technology providers. Ciena's WaveLogic 6 Extreme (WL6e) platform operates at 200 GBaud on 3nm process technology, delivering 1.6 Tbps single-carrier performance with DP-64QAM and probabilistic constellation shaping. Nokia, which integrated Infinera's indium phosphide photonic integrated circuit (PIC) technology into its optical portfolio, offers both silicon photonics and InP coherent approaches through its PSE-6s and successor platforms. Nokia Bell Labs has demonstrated 800 Gbps over 7,865 km and 41 Tbps unrepeated C-band transmission over 291 km. The combined Nokia-Infinera portfolio provides design flexibility across both PIC platforms.

Ribbon Communications showcased its Apollo 9408 and 9458 platforms at OFC 2025, delivering an industry-highest 25.6T line capacity in a 2RU form factor. In November 2025, Ribbon demonstrated 20 Tbps optical transmission over 10,000 km on the JUNO trans-Pacific submarine cable using its Apollo 9408 transponder (5nm, 140 GBaud) and Apollo 9608 submarine-grade line system. JUNO is planned as a 20-fiber-pair system with over 350 Tbps total capacity between Japan and the United States. In March 2025, NTT West and Cisco completed the world's first field verification of C+L band coherent pluggable optics over 60 km in Osaka, demonstrating C+L viability for metro and national backbone routes using IP-optical converged (Routed Optical Networking) architecture. Cisco's Acacia-based coherent pluggable modules have also seen strong adoption, with all major hyperscalers now among the customer base.

Disaggregated WDM architectures have become a dominant deployment model, particularly among hyperscale cloud providers. Microsoft, Amazon, and Google have driven widespread adoption of 400ZR coherent pluggable modules, with the transition to 800G ZR/ZR+ now accelerating. Nokia, Ciena, and Cisco all ship 800G pluggable coherent products, and the industry roadmap points toward standardized 1.6T ZR+ modules. Cisco's Acacia-based coherent pluggable portfolio, built on silicon photonics, supports IP-optical convergence through Routed Optical Networking (RON) architecture, bringing DWDM directly into router platforms.

10.1 Unique Challenges for Submarine Applications

Submarine cable systems present the most extreme test for C+L band technology. With cable lengths reaching 6,000-16,000 km and hundreds of inline amplifiers (repeaters), the cumulative effects of SRS tilt, gain flattening errors, and noise accumulation are dramatically amplified compared to terrestrial systems.

In a submarine C+L system with 130 repeaters (typical for a 6,500 km trans-oceanic link), the SRS tilt is compensated at each repeater. However, the C/L band splitter at each repeater introduces approximately 0.5 dB of additional insertion loss, and the L-band EDFA contributes approximately 0.5 dB higher noise figure compared to a C-band-only system. The cumulative OSNR impact is significant: L-band channels in a C+L submarine system typically have 1.0-1.5 dB worse OSNR than C-band channels in a C-band-only system of equivalent length.

Furthermore, the interaction between bands through SRS creates tight requirements for gain equalization. Laboratory measurements on a 6,600 km C+L submarine link with 130 amplifiers showed that a mere 1 dB change in the output power of one band caused measurable power variations in the other band after propagation through the full amplifier chain. This inter-band coupling requires coordinated control of both C-band and L-band amplifiers across the entire cable length, a significantly more complex problem than managing a single-band system.

10.2 C+L vs. SDM for Submarine Capacity

For submarine cable operators, the decision between C+L band expansion and Space Division Multiplexing (SDM, using additional fiber pairs or multi-core fiber) is a critical strategic choice. C+L band doubles the capacity per fiber pair but introduces the complexity and OSNR penalties described above. SDM scales capacity by adding parallel fiber paths, each operating in the C-band with proven technology, but requires more physical space in the cable and more amplifiers per repeater.

The submarine cable industry predominantly uses SDM (increasing from 4 to 8, 12, 16, or even 20+ fiber pairs per cable) for new builds, while C+L band remains essential for capacity upgrades on existing cables. The JUNO trans-Pacific cable, currently under construction between Japan and the United States, exemplifies the state of the art: planned as a 20-fiber-pair system with over 350 Tbps total capacity. In November 2025, Ribbon Communications demonstrated 20 Tbps optical transmission over 10,000 km on the JUNO cable using its Apollo 9408 transponder (5nm process, 140 GBaud) and Apollo 9608 submarine-grade line system, achieving per-wavelength rates from 400 Gbps to 1.2 Tbps with flexible modulation. Lumen Technologies was selected to provide the terrestrial backhaul connectivity for JUNO's U.S. landing.

Nokia's PSE-6s coherent engine delivers up to 1.2 Tbps per wavelength (130 GBaud, single DSP) or 2.4 Tbps per line card (dual DSP) for submarine landing terminal equipment (SLTE), with its combined Nokia-Infinera optical portfolio now offering both silicon photonics and indium phosphide approaches. Nokia Bell Labs has demonstrated 800 Gbps over a single wavelength at 7,865 km (approximately Tokyo-to-Seattle distance) and 41 Tbps over 291 km on unrepeated C-band transmission. The Pacific Light Cable Network (PLCN) was among the first submarine cables to use C+L band commercially, supporting 240 channels of 100 Gbps, and Sparkle has deployed C+L technology on its BlueMed submarine system. For upgrade scenarios on existing fiber pairs with fixed infrastructure, C+L band expansion remains the most practical and cost-effective option, providing near-2x capacity without the expense of new cable installation.

11.1 S+C+L Band Extension and Beyond

Research efforts are extending the usable optical spectrum well beyond C+L. The S-band (1460-1530 nm) would add approximately 4.8 THz of bandwidth, bringing the S+C+L total to approximately 14.4 THz (3x over C-band-only). S-band amplification is more challenging than L-band, as standard erbium-doped fibers cannot provide gain in this region. Thulium-Doped Fiber Amplifiers (TDFAs), semiconductor optical amplifiers (SOAs), and multi-pump Raman amplification are being explored as S-band amplification candidates.

At OFC 2025, NICT and Sumitomo Electric set a new world record: 1.02 petabits per second transmission over 1,808 km using a 19-core randomly coupled multi-core fiber with standard 0.125 mm cladding diameter. This achieved a capacity-distance product of 1.86 exabits per second-km, the highest ever recorded, using C+L band amplification across all 19 cores. In August 2025, NTT demonstrated 160 Tbps long-haul transmission over more than 1,000 km by opening a new ultra-long-wavelength region called the "X band" (around 1,700 nm). Using a novel ultra-broadband optical repeater based on PPLN wavelength band conversion technology, NTT expanded the total signal bandwidth to 27 THz (6.7x greater than conventional C-band systems), employing stimulated Raman scattering itself as a beneficial mechanism to extend the low-loss transmission window.

At ECOC 2025, NICT and 11 international partners demonstrated 430 Tbps on a single standard-compliant G.654 fiber (ITU-T) over 10 km, surpassing their own 2024 record of 402 Tbps with nearly 20% less overall bandwidth. The breakthrough exploited few-mode transmission in the O-band alongside single-mode E/S/C/L-band transmission, tripling the O-band capacity. Similarly, a field deployment of anti-resonant hollow-core fibers (AR-HCF) achieved 114.9 Tbps over 137 km across S+C+L bands, demonstrating terabit-scale real-time coherent transmission on novel fiber types.

11.2 Sixth-Generation Coherent DSP: 1.6 Tbps and Beyond

Sixth-generation coherent DSPs have fundamentally changed C+L band system capabilities. Ciena's WaveLogic 6 Extreme (WL6e), built on 3nm process technology and operating at 200 GBaud, is the industry's first 1.6 Tbps single-carrier coherent solution. WL6e delivers 1.6 Tbps single-carrier wavelengths for metro and short-haul ROADM deployments and 800 Gbps over the longest terrestrial links, with 50% reduction in space and power per bit versus the previous generation.

Key deployment milestones include AT&T setting a 1.6 Tbps long-distance record between Newark and Philadelphia (296 km), Telstra achieving 1.6 Tbps over 736 km in Australia, euNetworks deploying the first 1.6 Tbps wavelength in Europe, and Colt completing 1.2 Tbps transatlantic transmission on the Grace Hopper cable.

Nokia's combined PSE-7 and ICE-8 development programs (from the Nokia and Infinera legacy portfolios respectively) target 1.6 Tbps-class single-carrier performance, challenging Ciena's current lead. Both companies bring different PIC technologies: Nokia contributes silicon photonics while Infinera adds indium phosphide expertise, giving the combined entity design flexibility similar to Ciena's "technology agnostic" approach. Cisco's Acacia-based pluggable coherent portfolio has seen rapid adoption, with CEO Chuck Robbins noting that all major hyperscalers are now customers. Ribbon Communications' Apollo platform supports integrated and modular C+L band approaches with its patented Fast SRS Suppression technology. Looking ahead, the industry roadmap points toward 3.2 Tbps single-carrier capacity within the next generation, using 256QAM and symbol rates exceeding 200 GBaud.

11.3 Machine Learning-Assisted SRS Management

Next-generation coherent DSP engines are being designed with explicit C+L band awareness. Machine learning algorithms can predict SRS tilt changes based on traffic patterns and pre-adjust amplifier settings before transients occur. Ciena's WL6e DSP processes 1,600 tera operations per second, capable of supporting sampling rates as fast as 1 microsecond for in-service streaming telemetry, enabling real-time detection of external events such as lightning, earthquakes, or fiber intrusion attempts. Additionally, advanced nonlinear compensation techniques (such as digital back-propagation extended across both bands) can partially mitigate the inter-band Raman interaction at the receiver, recovering some of the OSNR lost to SRS.

11.4 Multi-Core Fiber with C+L Band

The combination of multi-core fiber (MCF) with C+L band operation represents the ultimate capacity scaling path, already demonstrating petabit-per-second fiber capacities in laboratory settings. At OFC 2025, NICT and Sumitomo Electric achieved 1.02 Pbps over 1,808 km using a 19-core randomly coupled MCF with standard 0.125 mm cladding diameter, operating across C+L bands. Earlier, the same team demonstrated 1.7 Pbps over 63.5 km using the same fiber type, transmitting 381 wavelength channels across 19 cores with 64QAM modulation. These results represent more than an order of magnitude higher capacity than current operational single-mode fiber systems, which are limited to approximately 250 Tbps per fiber.

The practical challenges of inter-core crosstalk, per-core amplification, and SRS management across all cores and bands remain active research areas. However, the fact that standard cladding diameter (0.125 mm) MCF can be spliced and handled with existing cable infrastructure makes this technology a realistic path to deployment. Coupled MCF designs deliberately allow controlled signal mixing between cores, then use MIMO digital signal processing at the receiver to separate the signals. This approach reduces the per-core space requirements, enabling higher core counts within the standard cladding. As C+L band technology and multi-core fiber converge with advanced DSP, petabit-per-second commercial systems within the next decade appear increasingly achievable.

C+L band DWDM technology represents the most capital-efficient path to doubling the capacity of existing fiber infrastructure. By extending DWDM operation from the C-band into the L-band, operators can access approximately 9.6 THz of usable optical bandwidth, supporting 160 channels at 50 GHz spacing (80 per band). C+L band systems achieve total capacities from 21.1 Tbps (DP-QPSK at 33 GBaud, 132 Gbps per channel) to 48 Tbps (16-QAM) and up to 72 Tbps (64-QAM) based on the modulation format and reach requirements. With sixth-generation coherent DSPs now delivering 800G to 1.2 Tbps per wavelength, these aggregate capacities continue to scale as AI infrastructure demand drives rapid technology advancement.

The engineering challenges of C+L band design center on managing Stimulated Raman Scattering, which transfers energy from C-band to L-band channels and creates a power tilt of up to 8 dB per fiber span. Successful C+L deployment requires split-band EDFA architectures with precise tilt compensation, orchestrated traffic management to handle dynamic SRS transients, careful fiber selection (with G.654.E offering significant advantages for new builds), and hybrid Raman+EDFA amplification for long-haul applications where OSNR margins are tight.

For network operators evaluating capacity expansion options, the following recommendations apply. For terrestrial networks with existing fiber, C+L band upgrade provides the most efficient capacity doubling path, leveraging existing fiber infrastructure without new route construction. For new greenfield builds, specifying G.654.E fiber maximizes C+L band performance through reduced SRS tilt and lower attenuation. For submarine systems, SDM (increased fiber pair counts, up to 20+ pairs on new cables like JUNO) is favored for new builds, with C+L band remaining essential for capacity upgrades on existing infrastructure. The PLCN (Pacific Light Cable Network) remains the only transoceanic cable deployed with C+L band EDFA technology to date. For data center interconnects, C+L band with sixth-generation coherent transceivers at 800G per wavelength on flex-grid spacing can deliver very high aggregate capacities over typical DCI distances of 50-200 km, where OSNR budgets are comfortable even with L-band penalties.

The optical networking industry is advancing rapidly, driven by the convergence of AI-driven bandwidth demand, sixth-generation coherent optics (Ciena WL6e, Nokia PSE-7/ICE-8, Cisco/Acacia pluggables), modular C+L solutions (Ribbon Apollo), and advanced fiber technology. C+L band deployment is now both technically mature and widely available. Disaggregated WDM architectures with open line systems enable operators to independently source transponders, amplifiers, and ROADMs. Research breakthroughs at 430 Tbps per single fiber (NICT, ECOC 2025), 1.02 Pbps over 1,808 km on multi-core fiber (OFC 2025), and NTT's 160 Tbps over 1,000+ km using the new X-band demonstrate that the technology trajectory extends far beyond C+L into multi-band, multi-mode, and multi-core dimensions. Understanding the design principles, impairment management techniques, and capacity planning methodologies presented in this article is essential for optical network engineers building the next generation of high-capacity transport infrastructure.