GLC-LH-SM sfp to 1000baseLX GBIC won't link over MMF

[William Coffey]

Hi all,

Current setup:

We currently have 7 network closets with 2950's linked back to our 4506-E distribution switches via MMF 62.5/125 micron fiber. Both ends use 1000baseLX GBICS (WS-G5486). This configuration has been in place for many years and has been working flawlessly.

Problem:

We are getting ready to replace all the swithes in the 7 closets with 4506-E switches and 1 3560 switch and planned on using the GLC-LH-SM sfp gibcs in the new switches and keep the existing 1000baseLX gbics in the distribution switches using our existing MMF. I tried to do some testing ahead of time using a 3550 switch with a 1000baseLX gbic and connect it to one of the new 4506-E with the GLC-LH-SM gbic using a 2meter MM fiber patch cord but could not get them to link. I tried multiple gbics and cables and even tried different switches but the result was the same. The odd thing is if you pushed the fiber cable about half way to 3/4's of the way into the gbic it would link but if you pushed it all the way in you would lose the link. It did work fine with a single mode cable and also would link up if I connected sfp to sfp using MMF but would never work going 1000baseLX gbic to GLC-LH-SM. From everthing I've read this should work, has anybody seen a problem like this or know what could be the cause.

Thanks in advance for any help.

Bill

[Tom Randstrom]

When connecting the GBIC to SFP, are you using Mode Conditioning Patch Cords at each end? I would have thought the set-up would have worked given the short fiber length...

Also, just for grins, when you connect the GBIC to SFP with the 2m MMF patchcord, you might try wrapping the patch cord a couple of turns around a pencil (or insert a 5dB optical attenuator) and see if you get link.  Although the specs indicate you should be able to direct connect the ports (without additional attenuation), but you indicated link changes when the connectors were partially inserted; maybe you are getting too much light into the receivers.

Tom

[William Coffey]

Thanks for getting back to me. No I didn't use the mode conditioning cable, as you said I didn't think I would need it using a 2 meter patch cable. It is my understanding they are only needed when connecting at distances greater than 984ft or somewhere around there not sure of the exact distance.

I thought the same thing about over saturating the gbic at a short distance so I tried a 20 meter MM cable but had the same results. I also took 2 switches over to the plant where we are actually going to be installing the new switches and connected one in the MDF closet and the other one in one of the network closets about 600 feet away and again had the same results.

[Tom Randstrom · Accepted solution]

Here might be the answer...

"1. Some customers might be tempted to connect  1000BASE-LX/LH devices over short reaches of MMF jumper cables without  MCP cables. There is a risk associated to this type of nonstandard  deployment, especially when the jumper cable is an FDDI-grade or OM1  type. In such case the power coupled directly into a 62.5-micron fiber  could be as high as a few dBm (typically 5 to 6 dBm), and if the jumper  cable is short, the adjacent receiver will be saturated, causing high  bit error rate, and eventually irreversibly damaged. We recommend the  use of MCP, but if the customer remains reluctant to the deployment of  MCP, a 5-dB attenuator for 1300nm should be used and plugged at the  transmitter of the optical module on each side of the link."

Tom

[William Coffey]

Ok, I have ordered the mode conditioning patch cables for both ends. I will post back and let you know if this fixes the problem.

Thanks,

Bill

[viyuan700]

" such case the power coupled directly into a 62.5-micron fiber  could be as high as a few dBm (typically 5 to 6 dBm),"

I dont know what the person who wrote that article mean. The power spec of SFp & GBIC are

WS-G5486 and GLC-LH-SM

-9.5 dBm3 -11.5dBm4

-3 dBm5

-20 dBm

-3 dBm

8.5 dBm6 and 10.5 dBm7

Chances of getting the Transmitter of -3dBm is very less (it is the highest power of that transmitter) and the receive is also -3dBm. It is hard for me to understand that a Transmitter of -3dBm can generate 5 to 6 dBm of power even thorugh it is going through a 2M MMF patch cable whereas it is transmitting -3dBm of power.

In my opinion he need to go to a specailist if he want to keep that article on Cisco website ? If it is happening then the idea isa big  revolution of generating Optical power.

Your link probably not working becasue of Dispersion as the other end is getting too many sources with different delay.

Till the time you get Mode Conditioning patch cable you can try the link with Single mode Patch cable at both ends if your distances are short.

[Alexei Andoni]

In this setup the receivers cannot be saturated (according to 1000baseLX and GLC-LH-SM specs). The problem is due to differential mode delay, i.e. due to pulse dispersion, which happens when connecting Single-Mode interfaces via Multimode patch-cords. You can go from SM to MM. But when going from MM to SM interface/fibers, the signal degradation is huge.

The best solution is to use Single-Mode cables with 1000baseLX and GLC-LH-SM. The mode-conditioning-patch-cables would work as well, but could be more expensive.

Alex

PS: When pushing the fiber cable half way, less modes (light pathways) are inserted in SM interface/fiber from MM cable, which decreases pulse dispersion. That is why it worked.

[William Coffey]

Thanks to all who responded with suggestions.

Looks like the mode conditioning cables took care of the problem. I guess the reason I was skeptical was because I am using the same MMF over the same lengths and connections which was working fine without MCP cords. The only thing that changed was the fact we were replacing the 1000LX Gbic's with the GLC-LH-SM sfp's on one end. Oh well...live and learn.

Thanks again,

Bill

[viyuan700]

Looks like GLC-LH-SM sfp are inteliigent that they ONLYwork without MCP .

[Tom Randstrom]

Glad to see it is working but this sounds nuts to me too!

It is probably the SFPs are manufactured by a different vendor than the GBICs and the components must have different dispersion tolerances.  Maybe something was different in the fiber jumpers you were using to adapt to the LC connectors on the SFPs.

Just trying to rationalize the situation in my mind.

Thank you for letting us know the outcome!

[Tom Randstrom]

Appreciate your input... very valuable.

"The best solution is to use Single-Mode cables with 1000baseLX and  GLC-LH-SM. The mode-conditioning-patch-cables would work as well, but  could be more expensive."

But, using the existing installed MMF fiber plus the MFCs can be cheaper than having to run new singlemode fiber optic cables.  This is the difference between theoretical and reality. 

Tom

[Alexei Andoni]

"But, using the existing installed MMF fiber plus the MFCs can be cheaper than having to run new singlemode fiber optic cables.  This is the difference between theoretical and reality."

That is true. The best one is not always the cheapest :-) However it is good to have options, especially when comparing short SM patch-cords vs. MCP.

Cheers,
Alex