Page 1 / 3   >   >>
slider 12/4/2012 | 10:20:15 PM
re: Invites Go Out for Metro DWDM Test Has opthos been resurrected or it's just a typo
like Faul Management.
2cents 12/4/2012 | 10:20:14 PM
re: Invites Go Out for Metro DWDM Test What about Tropic?
whoisjohngault 12/4/2012 | 10:20:13 PM
re: Invites Go Out for Metro DWDM Test If you go to the Williams party, be sure to collect some coasters or cocktail napkins as memorabilia. Their faulty ATM based business model and head-in-the-sand mgmt will guarantee their extinction at their own hands in a manner of months. They will not survive much longer as WCG, but remnants of their network will.
jamesbond 12/4/2012 | 10:20:13 PM
re: Invites Go Out for Metro DWDM Test how do all optical systems measure BER?
gina 12/4/2012 | 10:20:12 PM
re: Invites Go Out for Metro DWDM Test I am amazed...Read this...
Petabit 12/4/2012 | 10:20:10 PM
re: Invites Go Out for Metro DWDM Test "how do all optical systems measure BER?"

Errm, not sure I undertsand you question. Do you mean how do all-optical systems measure BER along the link?

Bit Error Ratios on really have any meaning when you electrically terminate the signal. Then you can look at the bit stream and decide if each of the bits is corretc, based on your error correction scheme. You don't have to do this at the end of the system, you can tap off a small fraction of the data along the link, and measure the BER there.

This is not done in practice for a couple of reasons: the simplest reason is cost - a demux, detector and full line rate electronics are not cheap. Secondly, in real systems the BER is so low that it might take you several days to see an error. And then you need to monitor the next channel... Thirdly, the chromatic dispersion need to be adjusted before the detector (more cost).

So in practice you either do nothing (the most popular solution), or you find a way to measure optical signal to noise ratio. OSNR does not give you BER, and can be distorted through most of the optical effects that we face on a daily basis. Measureing system Q gives you more of an idea as to what is going on, but is not the whole answer either (not least of which no two vendors define it in the same way).

To step back for a second, you are wanting to measure the analogue performance of a link, in order to predict the digital perfomance. Not an easy job.

gea 12/4/2012 | 10:20:10 PM
re: Invites Go Out for Metro DWDM Test Petabit basically has it correct. Indeed, what he says is similar to what I wrote in Telcordia's "Notes on DWDM and Opical Networks", Chapter 5.

Fortunately, however, there's almost no such thing as a truly all-optical signal. All (or at least the vast majority of) DWDM vendors equip their systems with wavelength converting transponders. Thus, when we test a DWDM system, we inject a test signal into the "input" transponder and take it out of the "output" transponder, and try to make sure that the optical system in-between is stressed with a worst case of loss, chromatic dispersion, and OADM channel imbalance and so on (the EDFAs are particularly tricky in this case).

The validity of the BER in this case is "perfect" when you have 3R regeneration on the input and output of your DWDM system (this is increasingly common these days). When the regeneration is 2R, or when there is no output regeneration (rare now, though early systems didn't have it to cut costs), then your output eye pattern is a function of the input eye pattern, and now you have some problems, similar to what Petabit was talking about.

ESNR, OSNR and Q-factor are somewhat problematic signal quality measures in that they really only represent the highest extractable BER from the signal. Without actually hitting a detector from the same vendor as the DWDM optics, there's no knowing for sure that this is what will be seen when that system's up and running. Using an optical tap and demux and then sending the signal to a test set really only measures the test set's receiver (which I have found can be very, very different from an actual system's receiver).

This need to accurately detrmine the signal quality has, I believe, been one of the drivers away from all-optical networking, except where's there's overwhelming economic advantage. Partial proof of this can be found in the fact that just about all DWDM systems are deployed with output regenerators which monitor A1&A2, B1 and J0, even though this is pretty much unnecessary (client-layer receivers are insensitive to wavelength over a very wide band.)

My only question to Lightreading is whether they're sure that those performing the evaluations understand these issues. I've seen countless meaningless BER measurements which qualify almost nothing but the test set. And not all of this are the fault of testers: ITU has yet to setle some of the basic issues with respect to DWDM system measurement (eg, the applicability of SONET/SDH jitter specs to DWDM).

So I'm hoping they will document the results very, very carefully, so we can see a true apples-to-apples comparison.
gea 12/4/2012 | 10:20:09 PM
re: Invites Go Out for Metro DWDM Test Petabit:

You're pretty much right on. I see this all the time: they got a SONET guy (who may be very sharp)to write this test plan, and its fairly evident he/she has never seen a DWDM system before. (And having worked in the group that wrote/writes GR-253, and having previously been the guy to do DWDM system testing at Bellcore/Telcordia, I know of what I speak.)

That's not to say there's no value in some of what's being done in this test plan, but there's a lot that's totally irrelevant, and then the big issues that plague DWDM they pretty much have ignored. Some of the big issues are gain balancing and channel/node upgrade, auto-shutdown of EDFAs (when the cut is downstream of the EDFAs), transponder capabilities, and the biggest one (like you say) is fault location and isolation.

The problem is that because (as you say), from input transponder interface to output transponder interface, there's only glass, if a component or fiber span fails it can be very hard to locate the failure (or, more commonly, identify an optical misconnection). The systems that have features like pilot tones or built-in spectral monitors to address this issue have faced one of the bigger stumbling blocks in DWDM, and that test plan completely ignores it.

Oh yeah, and the OAM&P issues are make-or-break for big carriers like the RBOCs: How does their design prevent the craftsperson from plugging the output of and EDFA/OFA, for example, into the receiver of a transponder (which will almost certainly blow out that receiver)? Also, can the OFAs be serviced/replaced without taking down traffic? And what about OFA gain balancing and/or ring balancing? Does it require a platoon of craft driving around the city in trucks in order to drop additional wavelengths at a node (or worse yet to add a node)? How much does this process interrupt traffic?

Well, we're not the only ones who will point this out. When the proper SMEs from the vendors see that testplan, you'll see it change drastically or they'll start pulling their gear from the bakeoff.

Petabit 12/4/2012 | 10:20:09 PM
re: Invites Go Out for Metro DWDM Test Hi gea,

Good post. Glad to see that there are still some quality posters around.

I had a quick look at the test plan (the link is in the article), and there doesn't seem to be an optical perfomance monitoring section. They are limiting the tests to the ridiculous array of PM tests that you can perform on the SONET/SDH frame. Essentially they are testing SONET/SDH standard compliance.

I didn't see any tests that relate to fault location - that essential piece of functionality that is missing from most vendors' systems. Fault location is the tool that allows you to know where you are supposed to be sending your repair crew to. Without it you are relying on the skills of the Network Controllers to be able to correlate the alarms to determine where the fault is - and even now it is a bit of a black art.

Please correct me if I read the test plan too quickly ;)

RJC 12/4/2012 | 10:20:07 PM
re: Invites Go Out for Metro DWDM Test Gina, I'm confused. First you post an obviously well-researched timeline of the delta between management PR and reality, then you toss out something like this.

Under investigation is not the same as convicted of a crime.

Not that I'm a big Bross booster, but let's remember 8th grade civics -- a party is innocent until proven guilty, right?

Page 1 / 3   >   >>
Sign In