[NANOG] DWDM More Details


Do you want

      CWDM - Course Wave Division Multiplexing - > 100 nm optical spacing 1 - 10 x 2.5 - 10 Gbps lambdas
      DWDM - Dense Wave Division Multiplexing - = 50 nm optical spacing 20 - 40 x 2.5 - 10 Gbps
    UDWDM - Ultra Dense Wave Division Multiplexing - < 50 nm optical spacing 40 - 80 x 2.5 - 10 Gbps

What size area do you want

    CAN - Campus Area Network 1 - 5 Miles Short Optics, No amplifiers
   MAN - Metro Area Network 10 - 40 Miles Medium Optics, No amplifiers
   WAN - Wide Area Network 150 Miles between In Line Amplifiers (ILA) Long haul optics, EDFA or Ramiun Amps
ULHWAN - Ultra Long Haul 600 - 800 Miles between ILA, ULH optics, Ramiun Amps

What type of fiber do you want to use?

SMF, Zero dispersion, phase shifted, etc.. This one you usually cannot control fiber since it depends on who you lease or buy the fiber from.

Running the fiber cost the money.

There are now a large number of regional fiber providers with Level (3) having the "most available" dark or lite fiber nationally in the US.

Alcaltel/Lucent, ADVA, Ciena, Cisco and the Chinese are normal list of optical equipment providers and Siemens, Ericson and others.

Subscribe to Lightwave (at no charge) and look at the back issues for networks. Show up at Supercom or OFC or what is replacing them and get the latest on ROADM, full channel tunable lasers and maintenance costs.

What size of network do you want to grow to before replacing the optical link equipment including ILAs?

Most any org can cost justify a CWDM / CAN since you can add one fiber pair at a time and one lambda per fiber pair.

DWDM gear is much more expensive and is aimed at 20 to 40 lambdas per fiber for service providers while UDWDM and ULHWAN are aimed at trans oceanic links and are very very expensive.

John (ISDN) Lee

DWDM gear is not expensive. Passive muxes cost little. Active
transceivers cost money but not very expensive at all.

Check out these two presentations (by yours truly et al):



In your talk, I agree that the CAN with your CWDM is not that expensive but you also mention that the tighter DWDM with long haul optics is expensive ie "Everybody knows how to do (active) xWDM by giving a lot of money to (insert vendor of choice]:"

When you talk about the tighter itu spacing for "real" DWDM and the lasers with fiber that can handle the power, jitter, chromatic dispersion et al. the optics you mention will not handle that.

We have all duct taped optical systems on campus for the lab "and across the state of Georgia" see the Peach Net map.

What is the largest number of lambdas you have actually run on a single fiber with your duct tape system and how bad was the optical cross talk?


I'd be curious to ask reverse question, did anyone *have* real problems
deploying duct tape systems, or power jitter chromatic dispersion is
vendor mumbo jumbo designed to make you buy their gear?

(within the distance limits spec'd, 80km dwdm etc)


What I think you are referring to is known as the
Gordon-Haus effect. This is related to the length
of a strand to the 3rd power, so I would guess for
modern optics / filters, it is a non-issue for 80k.


Yes, but in very specific cases such as older ZD, NZD fiber and SMF with 20 or more lambdas (at 2.5 Gbps or less) and with Ultra long haul and ultra DWDM 600 - 800 miles, 25 nm spacing and fiber with too high a water vapor content and/or higher impurities in the fiber. If you have less than optimum fiber you need the additional optics and electronics to make it perform.

So my question is the same as the underliying one in your presentation which for Scott is what are you trying to do and there are several ways to reduce the cost or if you need the super neat, highest speed, longest run stuff you will pay for it.

Side note, I liked your two presentations.