DWDM Systems Explained: C-Band, L-Band, and the 400G/800G Era Interview Prep

DWDM Grid Options — Fixed vs Flex

Sample interview questions

1. A 100 GHz fixed-grid DWDM line system carries 32 GBaud DP-16QAM (200G) channels. Why does moving to a 50 GHz fixed-grid double the channel count without changing the line system fundamentally?
   • A. 50 GHz spacing halves the per-channel data rate, so the total capacity stays the same.
   • B. 50 GHz spacing fits twice as many ITU-T grid channels in the C-band. A 32 GBaud channel occupies roughly 32–37 GHz of spectrum (signal plus a few GHz of guard band), so it fits inside a 50 GHz slot. The line system reuses the same EDFAs, filters, and ROADMs; only the channel plan changes. This is the classic upgrade path from 100 GHz spacing (~44 channels in C-band) to 50 GHz spacing (~88 channels in C-band). Higher baud rates (64+ GBaud) no longer fit 50 GHz and drive the move to 75/100 GHz fixed-grid or flex-grid. ✓
   • C. 50 GHz spacing requires entirely new optical filters and amplifiers.
   • D. 50 GHz spacing only works with on-off-keyed modulation, not coherent.

   Option B is correct. The ITU-T frequency grid (G.694.1) defines channel anchors at 50 GHz, 100 GHz, and other spacings. A 100 GHz grid in the C-band (1530–1565 nm, roughly 4400 GHz of usable bandwidth) accommodates ~44 channels; a 50 GHz grid doubles that to ~88. Moving to 50 GHz spacing roughly doubles capacity per fiber without changing the underlying line system — same EDFAs, same ROADMs, same fiber plant. The fit constraint at 50 GHz is that the channel's spectral footprint must fit in 50 GHz minus a guard band. A 64 GBaud signal with raised-cosine pulse shaping (roll-off 0.1–0.2) has a 3-dB optical bandwidth around 64–67 GHz — too wide for 50 GHz spacing without pre-DSP spectral shaping. But Nyquist pulse shaping (sharp rectangular spectrum) at 32 GBaud fits 50 GHz comfortably; 32 GBaud DP-16QAM is 200G; 32 GBaud DP-QPSK is 100G. Higher baud rates (64+ GBaud) need 75 GHz or 100 GHz fixed-grid slots, which is why flex-grid (variable-slot allocation) was added — see /topics/optical-fiber-impairments and the flex-grid sections of this page. Option A treats line rate and channel rate as one number — they are not. Option C is wrong — the same EDFAs and ROADMs work, only the channel-filter shaping needs to support 50 GHz selectivity. Option D is wrong — 50 GHz fixed-grid is the standard for coherent operation. Production reality: 50 GHz fixed-grid is the most common deployed DWDM configuration; flex-grid is the upgrade path for systems carrying 600G+ per channel where 50 GHz no longer fits.

2. What is the operational reason for using L-band amplification in addition to C-band, and why is C+L not just "twice the C-band"?
   • A. L-band is identical to C-band; the naming is just a regulatory distinction.
   • B. C-band (~1530–1565 nm) and L-band (~1565–1625 nm) together roughly double the per-fiber capacity available from C-band alone. But the bands cannot share the same EDFA — erbium-doped fiber amplifiers have peak gain around 1530–1560 nm, and L-band EDFAs use longer or differently doped fiber to provide gain at the longer wavelengths. Operators using C+L deploy separate amplifier chains, often combined by interleavers, and must manage two gain spectra plus inter-band power balance. The capacity roughly doubles but the system complexity and cost grow more than 2×. ✓
   • C. L-band is only used in submarine deployments; terrestrial DWDM uses C-band exclusively.
   • D. C+L operation cancels nonlinear effects entirely.

   Option B is correct. The C-band spans roughly 1530–1565 nm (about 4.4 THz of optical bandwidth) and is the band where erbium-doped fiber amplification has its strongest, flattest gain. L-band spans roughly 1565–1625 nm (another ~4.8 THz of optical bandwidth) and requires a differently-engineered EDFA (longer erbium-doped fiber and/or modified composition) because erbium's natural gain peak is centered in the C-band. C+L deployment roughly doubles per-fiber capacity from the ~17–24 Tb/s typical of C-only systems to ~35–45 Tb/s C+L. But the system is not "twice the C-band" — operators must: - Deploy separate EDFA chains for C and L bands, often interleaved - Manage two gain spectra and the gain tilt of each band independently - Balance inter-band power to control nonlinear Raman crosstalk (the higher-power C-band can transfer energy to L-band via stimulated Raman scattering) - Coordinate channel-power equalization across both bands The cost and complexity grow more than 2×, but capacity per fiber roughly doubles, so the per-bit cost still improves enough to justify the upgrade in fiber-constrained routes (urban backbones, submarine systems, dense metro). New deployments increasingly default to C+L from day one rather than upgrading later. Option A misses the gain-spectrum distinction. Option C is wrong — both terrestrial and submarine deployments use C+L. Option D is wrong — nonlinear effects accumulate within each band and inter-band Raman crosstalk is an added complication. Production reality: 800G-era deployments increasingly assume C+L as a baseline; the new fiber types (G.654, hollow-core) provide larger effective area and lower nonlinearity to better support C+L power budgets. See /topics/optical-fiber-standards-itu for the fiber-side context.

3. A flex-grid DWDM system replaces fixed-grid spacing with variable-slot allocation. What does this enable, and what new operational complexity does it introduce?
   • A. Flex-grid is identical to fixed-grid; the naming is a marketing distinction.
   • B. Flex-grid divides the band into 12.5 GHz "slots" (the granularity defined by ITU-T G.694.1 supplement) and allocates each channel a variable number of contiguous slots based on its spectral width. A 32 GBaud DP-16QAM channel uses ~3 slots (~37.5 GHz); a 64 GBaud DP-64QAM channel uses ~6 slots (~75 GHz); a superchannel grouping multiple carriers might use 8+ slots. This enables right-sizing each channel's spectrum to its baud rate and capacity, packing the band more efficiently. The new complexity: spectrum management becomes a non-trivial allocation problem ("spectrum defragmentation"), ROADM filters must support variable passbands, and channel adds/drops must avoid creating un-allocatable gaps. ✓
   • C. Flex-grid only works at FR2 frequencies.
   • D. Flex-grid eliminates the need for ROADMs.

   Option B is correct. Flex-grid (ITU-T G.694.1 supplement) replaces the fixed-50-GHz or fixed-100-GHz grid with a finer-granularity slot architecture: the band is divided into 12.5 GHz slots, and each channel is allocated a contiguous block of slots sized to its spectral width. A 32 GBaud DP-16QAM channel needs roughly 3 slots (~37.5 GHz). A 64 GBaud DP-64QAM channel needs roughly 6 slots (~75 GHz). A superchannel — multiple sub-carrier signals jointly modulated — might span 8 or more slots. Why flex-grid is a real win at 400G+: at 64 GBaud and higher, the channel's spectral width exceeds 50 GHz, so it doesn't fit a 50 GHz fixed-grid slot anyway. Fixed-grid 100 GHz slots waste capacity for narrower channels; flex-grid packs both wide and narrow channels into the band without wasted spectrum. The new operational complexity: - Spectrum becomes a fragmentation problem. Channels added at different times and sizes can leave un-allocatable gaps. Spectrum defragmentation procedures (similar to disk defrag) shift channels to consolidate free spectrum. - ROADM filters must support variable passbands. Wavelength-Selective Switches (WSS) with multi-channel-spectrum capability replace fixed-grid WSS. - The management plane must track per-channel slot allocation; SDN-based optical control is increasingly the way operators run flex-grid systems. Option A misses real architectural differences. Option C is wrong — flex-grid is an optical-domain concept, not an FR concept. Option D is wrong — flex-grid actually requires more sophisticated ROADMs, not none. See /topics/roadm-wavelength-switching (queued) for the ROADM-side context. Production reality: most newly-deployed 400G-and-above DWDM is flex-grid; mixed-grid migrations (fixed + flex on the same fiber) are common as operators upgrade segment-by-segment.

Frequently asked questions

What is DWDM?

DWDM (Dense Wavelength Division Multiplexing) is the optical-transport technology that carries many independent channels on the same fiber by assigning each channel a distinct wavelength (equivalently, a distinct frequency slot in the ITU-T frequency grid). A typical DWDM line system in 2026 carries 80–96 channels in the C-band at 50 GHz spacing, plus another 80+ channels in the L-band, with each channel carrying 100–800 Gb/s of payload depending on the coherent transceiver generation. The aggregate per-fiber capacity reaches 20–45 Tb/s for C-only systems and up to 70 Tb/s for C+L systems. DWDM is the workhorse of long-haul, regional, metro, and submarine optical transport, and increasingly of high-throughput DCI (data-center interconnect) links between hyperscalers.

What is the difference between DWDM and CWDM?

CWDM (Coarse Wavelength Division Multiplexing) uses wider channel spacing (typically 20 nm, ~2500 GHz) and looser laser-wavelength tolerances; this lowers the optical filter and laser cost but limits the channel count to about 18 channels across the S+C+L bands. DWDM uses narrow channel spacing (50 GHz or 100 GHz, occasionally 25 GHz) and tightly-controlled DFB or external-cavity lasers; this enables 80+ channels per band but raises the per-channel cost. CWDM dominates short-reach metro-access deployments (~80 km) where high channel count is not needed; DWDM dominates everywhere capacity matters.

What is the ITU-T frequency grid?

The ITU-T G.694.1 recommendation defines the channel anchor frequencies for DWDM. The grid is defined in frequency (not wavelength) with anchors spaced at fixed intervals: 100 GHz, 50 GHz, 25 GHz, 12.5 GHz (for flex-grid), and finer. The reference frequency is 193.1 THz (about 1552.52 nm in the C-band). 50 GHz grid is the most common fixed-grid deployment because it balances per-channel spectrum against C-band capacity (about 96 channels in the C-band). 100 GHz grid is used in older systems and in deployments where the channel rate is low enough to leave usable margin. Flex-grid (12.5 GHz granularity with variable slot widths) is the upgrade path for 400G+ systems where channels exceed 50 GHz spectral width.

What is the difference between C-band and L-band?

The C-band spans roughly 1530–1565 nm (~4.4 THz of optical bandwidth) and is the band where erbium-doped fiber amplifiers have their natural gain peak. The L-band spans roughly 1565–1625 nm (~4.8 THz) and requires a differently-engineered EDFA (longer or differently-doped erbium fiber) because erbium's gain falls off above 1565 nm. C-band alone supports about 20–24 Tb/s per fiber at 400G channels on a 50 GHz grid; C+L roughly doubles that to 35–45 Tb/s by using both bands. C+L operation requires separate EDFA chains, separate gain-tilt management per band, and careful inter-band Raman crosstalk control (the C-band can pump energy into L-band channels via stimulated Raman scattering).

What is a flex-grid DWDM system?

Flex-grid replaces fixed channel spacing with variable-slot allocation. The band is divided into 12.5 GHz slots (the granularity defined in the ITU-T G.694.1 supplement), and each channel is allocated a contiguous block of slots sized to its spectral width. A 32 GBaud channel uses ~3 slots; a 64 GBaud channel uses ~6 slots; a superchannel (multiple sub-carriers jointly modulated) might span 8 or more slots. Flex-grid is the standard for 400G+ DWDM because channels above 64 GBaud do not fit a 50 GHz fixed-grid slot. The operational complexity is non-trivial: spectrum defragmentation becomes a real management procedure, ROADMs need variable-passband WSS, and the management plane must track per-channel slot allocation across the line system.

What is a superchannel?

A superchannel groups multiple closely-spaced optical carriers into a single jointly-managed channel that is added and dropped as one unit. A typical superchannel might combine 4 carriers of 32 GBaud DP-16QAM (each 200 Gb/s) into a 800 Gb/s superchannel, or a similar arrangement with PCS-QAM at 1.6 Tb/s. The carriers are spaced at Nyquist sub-channel spacing (sometimes called Nyquist WDM, or N-WDM) so they pack tightly with minimal guard band. Superchannels exist because (a) coherent transceiver line rates increasingly exceed what a single carrier can carry per Hz under deployed OSNR, and (b) managing the carriers as one unit simplifies ROADM and management-plane operation. The trade-off is reduced flexibility (the operator commits the full superchannel slot rather than per-carrier allocation), which is fine for high-utilization backbone routes.

How are 400ZR and 800ZR pluggable transceivers related to DWDM?

400ZR is the OIF-defined coherent pluggable transceiver standard for 400G over DWDM, packaged in QSFP-DD or OSFP form factor with reach typically up to 120 km amplified or unamplified depending on the variant. 800ZR/800ZR+ is the follow-on standard for 800G coherent in similar form factors, with reach extensions for amplified spans. The significance is that these pluggables put coherent DWDM transceivers into router/switch front panels — the same boxes that historically used grey (non-coherent) short-reach optics now plug in coherent DWDM optics directly without a separate transponder shelf. This is the architectural shift sometimes called "router-integrated coherent" or "IPoDWDM" that is collapsing the traditional IP/optical layering. Note this is a different debate from intra-datacenter switch fabrics, where the packaging question is whether to keep the same QSFP-DD/OSFP pluggables at all or move the optical engine onto the switch ASIC — see /topics/cpo-vs-pluggable-optics.

How does channel-power equalization work across cascaded amplifiers?

A DWDM line system passes through many EDFAs in series. Each EDFA has a gain that varies slightly with wavelength (gain tilt) and with input power (gain saturation effects). After 5–10 cascaded EDFAs, the channel power across the band can be tilted by several dB — some channels arrive much stronger than others, which causes nonlinear-effect imbalance and Q-factor degradation. Channel-power equalization uses either (a) variable optical attenuators (VOAs) per channel inside each ROADM degree, or (b) wavelength-selective switches (WSS) that can apply per-channel attenuation. The control plane typically runs a power-equalization loop that monitors per-channel power at each ROADM, calculates the required per-channel attenuation, and applies it to keep channel powers within a target range across the whole line system. See /topics/edfa-vs-raman-vs-soa for the amplifier-side context.

Related topics