What a long haul link has to close
Long haul transport carries traffic between cities: a chain of spans, each with its own loss, joined by amplifiers, and a line system that keeps the channels inside their limit. A span of eighty kilometres and up is normal, and the distance that a channel reaches is decided by the optical signal to noise ratio at the far receiver rather than by the loss of a single span.
Coherent optics changed the economics of that layer, because one module now carries 400G or 800G on a wavelength where a 10G channel used to sit. The selection question is which rate closes on a given chain of spans, and that question is answered with a link budget rather than a data sheet. The column behind this page is coherent optics, and the rate families are collected on optics selection guides.
What makes a long haul span different
Signal to noise accumulates with every amplifier a channel passes, so a channel that crosses ten spans arrives with a worse ratio than the same channel over two. Chromatic dispersion and polarisation effects accumulate with distance in the same way, and the coherent receiver compensates for both within limits that narrow as the rate rises.
The second factor is the modulation. A higher order modulation carries more bits in the same symbol and needs a cleaner signal to do it, so an 800G channel over a long chain is usually carried with a lower order modulation and a stronger forward error correction, or it is split across two wavelengths. The third is the grid: channels are packed at fifty gigahertz, and neighbouring channels have to coexist with the launch power chosen for the whole line.
Recommended coherent optics for long haul 400G and 800G
| Module | Rate | Media / wavelength | Reach class | Where it usually lands |
|---|
| 400G QSFP-DD ZR+ pluggable | 400G | SMF, C-band tunable | Regional and long haul class | Router ports carrying wavelengths directly |
| 400G QSFP-DD ZR pluggable | 400G | SMF, C-band tunable | Metro and DCI class | Short chains and data centre interconnect |
| 400G coherent line card module | 400G | SMF, C-band tunable | Long haul span chains | Shelf based transport where clients stay grey |
| 800G QSFP-DD or OSFP ZR class | 800G | SMF, C-band tunable | Metro to regional class | New high rate links onto an existing line system |
| 800G coherent transponder card | 800G | SMF, C-band tunable | Long haul class | Shelf positions feeding the line system directly |
| 100G and 200G coherent modules | 100G / 200G | SMF, C-band tunable | Ultra long haul class | Spans where 400G will not close, and legacy line cards |
| OpenZR+ and interop modules | 100G to 400G | SMF, C-band tunable | Per mode class | Multi vendor line systems that must interoperate |
Reach classes depend on the mode chosen as much as on the module, so the working limit is the computed link budget for the chain. The pluggable route is compared on the 400G ZR DCI and coherent pluggable guides, the 800G options on 800G DR8, FR4 and ZR+, and the shelf side on 400G transponder and muxponder cards and 800G coherent cards.
Rate, modulation and the distance that closes
The same module usually offers several modes, and each mode pairs a line rate with a modulation order and a forward error correction setting. The lower order mode reaches further at the same rate, and the same hardware carries a higher rate over a shorter chain. Asking for the highest rate on the longest chain is how a project ends up with a link that runs and a commissioning report that says otherwise.
The practical approach is to compute the budget for the chain first, choose the mode that closes with margin, and only then pick the hardware that supports that mode. Where one wavelength will not carry the rate, the two wavelength arrangement is the standard answer.
Grid, channel plan and line system interoperability
A long haul line system has a fixed channel plan, and a new wavelength has to fit inside it. Tunable parts can be set to the free channel, which is one of their main advantages on a network that has already grown, but the launch power and the per channel power target still apply to the new channel.
Interoperability is the other constraint. Where the link crosses more than one platform, the interop modes are what make a mixed chain work, and the supported mode list belongs in the request rather than in the commissioning week.
Coding, standards and mixed estates
The coherent part is coded for the platform named on the request and tested on that model, and it is not manufactured by that brand. Coding in a transport network usually means a platform family and a mode list rather than a single interface, so stating both is what gets the right part the first time.
Mixed estates are normal on long haul routes: the same chain carries a new 400G channel and channels installed years ago at a lower rate, so keep the mode list and the channel per link in one record. A tunable spare covers every channel, which is why tunable parts are often chosen for the sparing argument as much as for the capacity.
How to choose, in order
- Get the span list with measured loss, amplifier positions and the number of spans.
- Compute the budget and pick the mode that closes with margin, before the hardware.
- Fix the channel plan and the grid, and check the launch power target per channel.
- Decide pluggable or shelf card per site from the client ports and the space available.
- Confirm the interop mode list where the chain crosses more than one platform.
- Set the spare rule from the channel count and the restoration window.
Frequently asked questions
Can 800G close on a long haul chain?
On shorter chains and with the right mode, yes. On the longest routes the working answer is a lower order mode or two wavelengths.
Is a coherent pluggable enough for a long haul route?
For regional and many long haul chains it is, if the port has the cage and the line system accepts the channel. Where the client ports are grey, a shelf card remains the route.
Do both ends of a link have to use the same module?
Not always, but any mixed arrangement has to be on the interop mode list of both ends, and that check belongs in the request rather than at commissioning.
How many spare modules does a route need?
One tunable spare per link family covers every channel, which is usually cheaper than a spare per channel.
What do you need to quote a long haul upgrade?
The span list with losses and amplifier positions, the current channel plan and grid, the target rate and mode, the platform at each end, the spare rule and the quantity per part.
Send the span list with losses, the channel plan and the target rate per link, and we come back with a mode that closes on the chain, the part per position and the spare lines set out separately. Related reading: coherent optics, collapsed IP over DWDM, WDM transceivers and all optics selection guides.
