Strange CWDM Observation

[Fraud]

I was hoping someone may be able to shed some light on a problem me and some colleagues came up against recently.

We were installing a CWDM service for a customer that involved going via multiple sites, and a total of 6 CWDM muxs were used. The total path length would have been ~30km. There was one mux at each end point, and 2 at each of the 2 sites inbetween (some wavelengths had to be dropped along the way)... to make this easier to explain, the path went site a to site b, b to c, and c to d.

when we were commissioning the service and doing testing, a to b was fine, b to c was fine, and c to d was fine (all tested using power meters, and CWDM gbics installed in an ethernet reflector). Then when we try to test A to D, it was no good (2 wavelengths had gone completely dead, and one was at about -40dB).

so then we tested from A to C, everything was fine, the loss was well within spec (about -13dB iirc)... test from C to D again, no problems. we tried everything.. we swapped out patch leads, we swapped out the mux's, and still no result. In the end we gave the customer dark fibre for the last hop (between C and D), because we were done scratching our heads.

for the record the overall path loss between the two sites (when using the dark fibre for the last hop) was about -19dB. Also for the record, when each of the mux's were tested independently, they worked fine on all wavelengths.

So there was some kind of problem we came up against when using those two extra mux's... Everything on its own functioned correctly, but when we tried to go end to end, it just wouldn’t work. Can anyone think of what it may have been? I can't find any documentation to indicate that there are limitations on this sort of thing.

We've been meaning to do a bench test to replicate the scenario ever since, but we tend not to keep that much gear just sitting around in stock, because its not cheap! The main theory that we've come up with so far is polarisation.. We’re not sure where, or how…. but its a theory...

Also, if this description is too confusing, let me know and I’ll draw a diagram! Any help is greatly appreciated.

Thanks,

Nick

 

[2darkpark]

a few stupid questions, is your optical power meter calibrated for CWDM wavelengths or just the two standard FP wavelengths?

Is the loss the same at each site (a,b,c,d) for all wavelengths? The mux/demux's typically have different levels of attenuation (depending on the filter order) so it's important that they're matched.

How are you measuring optical power at the D site with all the different wavelengths turned on? Are you using an optical spectrum analyzer or the power meters?

 

[Fraud]

hi, sorry I only jsut saw thsi reply, thanks.

all measurements were taken with power meters, with the meters set to 1550 (it has 4 modes, 650, 1300, 1310 and 1550)

we were trying to turn up 1570, 1590 and 1610.

when testing along the mux path, at every point (until the very end) loss was within ~1db on all wavelegths. at teh very end 1610 was at ~-40, and the others were lower than -55 (our power meters aren't sensitive any lower than that)

no OSA was used at all.

I actually do have an update on this though. We got some extra equipment in for a few jobs, and i had some spare time one afternoon, so i bench tested the exact same scenario just with 6 muxes and some LC patch leads, and there was no problem with it. I even put a packet tester over it (with an ethernet reflector on one end, and usign 10db attenuators so i didnt saturate the GBICs), and a 10 minute test showed no errors.

so its pointing to the probelm being something silly like a busted patch lead or something.. but we definatley swapped them all out... any other ideas?