dslreports logo
site
 
    All Forums Hot Topics Gallery
spc

spacer




how-to block ads


Search Topic:
uniqs
4435
share rss forum feed

spectrumhead

join:2009-05-03

SFP module for 100km

We need to transmit 1gbit constantly across a fiber which is 90km in distance.

We need 2xSFP module for each end for this and/or affordable devices to be able to do the job.

Is there any recommendations ? Quality is important.

What would be the best SFP module for this project ?


LLigetfa

join:2006-05-15
Fort Frances, ON
kudos:1

»www.transition.com/TransitionNet···ers.aspx



warwick

join:2009-06-05
Hollywood, FL

1 recommendation

reply to spectrumhead

What is the "quality" of the fiber in place?

To point you in the right direction you'd be looking for any sfp optics branded with the suffix ZX or ZR

Most sfps are oem branded and can be had "on the cheap" which makes acquiring a few at purchase for backup ideal. Although i've never had one fail. (yet).

Something similar to the GLC-ZX-SM-RGD would be ok.


spectrumhead

join:2009-05-03

1 edit

There is a -32db loss on the path.

We have tried now HP SFP J4860C but it didint link up.

Prices for transition SFPs varies too much on the internet. IS there any site where we can have reasonable prices ?

GLC-ZX-SM-RGD is also a Cisco SFP. We dont have Ciscos there.

Anyone ?



Rhaas
Premium
join:2005-12-19
Bernie, MO

1 recommendation

That seems high to me, I assume that there may be some connectors (not splices) in there?

For us, we always estimate .3db loss per KM, +.1 per splice. We are always under the calculated loss.

Dont look at the advertised distance, that is just a rule of thumb. Look at the specs of the SFP.

The one you said you tried would obviously not work:

MAX Output power: +5 dbm
Minimum Recieve Sensitivity: -23

So that gives you ~28dbm of link loss, subtract 3 dbm for fade and you have 25dbm.

The JD103A May work (couldn't find HP's specs only a compatible)
MAX Output power: +5 dbm
Minimum Recieve sesnitivity: -32 dbm

So you should have ~35dbm to work with. *If the TX power is outputting at +5*

I gotta run, but that should be enough info to get you started in the right direction.
--
I survived Hale-Bopp!


spectrumhead

join:2009-05-03

Rhaas ;

According to the above info,

I found TN-SFP-LXB162 to have 37dbm link budget.

This is the most powerful that I could find.

Do you think it may work ?



Rhaas
Premium
join:2005-12-19
Bernie, MO

Based upon the numbers you posted above, then yes. That should give you ~5db of margin and should function without error.

I would add the caution of contacting the SFP manufacturer and getting their recomendation.

When testing the fiber you will want to shoot it with the same wavelength that you plan on using to light the path.
Just like wireless at different frequencies, different wavelengths will have different losses. Also, the fiber can be manufactured so that certain freqencies and will perform better at 1550nm rather than 1310nm or 1590nm. So if you didn't shoot the fiber, you need to know what wavelength they shot the fiber at.
--
I survived Hale-Bopp!



54067323

join:2012-09-25
Tuscaloosa, AL

1 recommendation

said by Rhaas:

When testing the fiber you will want to shoot it with the same wavelength that you plan on using to light the path.

I would add to that, with dark fiber, never trust the as-built, enough crap happens with lit fiber (which gets noticed) in the day to day maintenance and changes of the plant.

lutful
... of ideas
Premium
join:2005-06-16
Ottawa, ON
kudos:1
reply to Rhaas

said by Rhaas:

Just like wireless at different frequencies, different wavelengths will have different losses.

I just wanted to comment that typical single mode optical fibers are good for both 1300nm and 1500nm range.

This spec for some really old SM fiber. I drew the loss curve just to show loss gets lower from 1300nm towards 1500nm but there is a water absorption peak in the middle.

spectrumhead

join:2009-05-03

You guys are very helpful. thanks for all the great comments.

So while checking I found out that I might be wrong considering the wrong transition network SFP.

Instead of TN-SFP-LXB162 WDM , we should consider TN-SFP-LX16.
I think LX161 and 162 are used in pairs as they have different wavelengths per transmit/receive. Instead we have to consider LX16.

Datasheet :
For LX16
»www.transition.com/TransitionNet···plex.pdf

Am I right ? Could you please help me a bit more to finalize this problem.


lutful
... of ideas
Premium
join:2005-06-16
Ottawa, ON
kudos:1

Yes, you can use a single LX16 module on each side of the fiber since it has duplex LC connectors.



54067323

join:2012-09-25
Tuscaloosa, AL

1 edit
reply to Rhaas

Delete duplicate



54067323

join:2012-09-25
Tuscaloosa, AL
reply to Rhaas

said by Rhaas:

That seems high to me, I assume that there may be some connectors (not splices) in there?

That link is long for sure, is it a bs splice or a connector(s) who knows, heck there could be a couple of biconic's in that span.

Personally I would (if possible) kick it back to the who ever placed the fiber to begin with and an OTDR to locate the loss(s) and clear them, lacking that I would push for a larger light budget, that is if the SFP’s can handle those levels.

spectrumhead

join:2009-05-03

@MatrixHDV
LX16 has a linkbudget of 37db and manufacturer states that these values are with %10 over performance to secure the specifications.

Do you think that we should consider a higher link budget than LX16 ?


lutful
... of ideas
Premium
join:2005-06-16
Ottawa, ON
kudos:1

1 edit
reply to spectrumhead

said by spectrumhead:

There is a -32db loss on the path.

We have tried now HP SFP J4860C but it didint link up.

I found basic specs for the HP and it says max 70km, so it was not suitable to begin with for your 90km link.
»h30094.www3.hp.com/product/sku/3673147

*** I am curious what tool/method you used to determine that the path has -32dB total loss. Anyway, assume that the contribution to this number from the actual 90km fiber run stays exactly the same and is out of your control.

But you do a very good job splicing the two endpoints and you also have newer design SFP modules with lower losses ... you can easily deduce that you WILL end up with a bit more system margin.

So in my opinion there is no need to buy the most expensive 200km/41dB LX20 module. The 160km/37dB LX16 you chose should be OK. Just to be safe, work out a 30-day return/exchange privilege before you buy.


spectrumhead

join:2009-05-03

@lutful
32 dbi loss was measured by the owner of the fiber pair recently. Actually the enginner told me that it is around -30 but it will be better for me to consider it like it is -32 for some reason i dont know.

The price I have got from a reseller is 1250$ per SFP. Is it too expensive or are there any cheaper options out there that I may consider ?


lutful
... of ideas
Premium
join:2005-06-16
Ottawa, ON
kudos:1
reply to spectrumhead

I am helping some high school classmates who are installing fiber all over the city where I was born. They are buying similar SFP modules from China for $500 max. So $1250 seems a bit high. But USA/Canada distributors have more overhead cost and you are also getting some assurance that they won't substitute a sub-par clone.

BTW my friends borrow Fluke test set from the local engineering university to measure the loss of the fiber length before and after install and to check the SFPs against spec. They learned the test process using Fluke's excellent instruction videos. Maybe you can do the same and determine exact losses for yourself.


spectrumhead

join:2009-05-03

Is there any link for those Chinese SFPs that I can check ? There is a big difference in price between $500 and $1250.



54067323

join:2012-09-25
Tuscaloosa, AL
reply to spectrumhead

said by spectrumhead:

We have tried now HP SFP J4860C but it didint link up. Anyone ?

By any chance do you have a light meter, the reason I ask is you might have an open fiber or two and not an SFP problem and since you have working SFP's you could shoot light down the fiber and measure it at the other end and compare what you are transmitting versus what you are receiving and if it is about 32 down you know the fibers are good, but if you are receiving nothing then your chasing the wrong dog for now going the SFP route.


54067323

join:2012-09-25
Tuscaloosa, AL

1 edit
reply to spectrumhead

said by spectrumhead:

Do you think that we should consider a higher link budget than LX16 ?

Guess that depends on the cost as you can always pad the receive down, but if the fibers come up long you cannot gain up...


54067323

join:2012-09-25
Tuscaloosa, AL
reply to spectrumhead

said by spectrumhead:

Is there any link for those Chinese SFPs that I can check ? There is a big difference in price between $500 and $1250.

Beware of seconds, they may match the specs on paper, but once in service and aging they may not turn out to be such a good deal long term and if you do not have a light meter, you will not a have a benchmark to work from if and when the link fails.


lutful
... of ideas
Premium
join:2005-06-16
Ottawa, ON
kudos:1
reply to spectrumhead

said by spectrumhead:

Is there any link for those Chinese SFPs that I can check ? There is a big difference in price between $500 and $1250.

They are Huawei "compatible" SFPs made in many different factories.

This is first link in my Google search but beware of such online stores.
»www.pc-wholesale.com/transceiver···sfp.html


TomS_
Git-r-done
Premium,MVM
join:2002-07-19
London, UK
kudos:5
reply to spectrumhead

FWIW I regularly install 1610nm SFPs capable of 120km. They shoot at approx. +2db, so watch that remaining good eye.

The manufacturer is JDSU, and they should also produce optics at other wavelengths that are equally as capable. This one in particular is a CWDM optic that we use for a management side channel on top of DWDM systems.


spectrumhead

join:2009-05-03

@TomS
I am checking their website but I couldnt seem to find any 1G SFP for long distance operation. Do you have a product list on hand ?



TomS_
Git-r-done
Premium,MVM
join:2002-07-19
London, UK
kudos:5

1 edit

1 recommendation

Try this:
»www.jdsu.com/ProductLiterature/t···c-ae.pdf

The part number for the specific optic I mentioned above is SFP-GL2LCC61AGG (edit: seems that might be an older part number?.)

Heres an example part that might work well in your situation:

SFP-GL2LKTD5DCA

1550nm rated for 120km. 1550nm usually suffers less attenuation than 1310 as indicated by lutfuls attachment earlier on, but both wavelengths are fine on single mode fibre.

We buy optics in what must be the thousands from JDSU, if thats any measure of confidence.


lutful
... of ideas
Premium
join:2005-06-16
Ottawa, ON
kudos:1

said by TomS_:

We buy optics in what must be the thousands from JDSU, if thats any measure of confidence.

If you can afford it, JDSU is still among the very best. Some of my former colleagues from Nortel do those transceiver designs in Ottawa.

spectrumhead

join:2009-05-03

My guess is that 120km SFP will not work given our loss on this Fiber. (Although RX Sensitivity and TX values are not stated on their datasheet)



TomS_
Git-r-done
Premium,MVM
join:2002-07-19
London, UK
kudos:5

I guess you'll need to bite the bullet and buy some pre-amps to bring up the rx power before piping it in to the optic at each side.

The 1610nm optic I mentioned earlier has a receive sensitivity of -32dB. With 32dB attenuation on the span and +2dB tx power, youre looking at -30 at the receiver. Probably cutting it thin, not leaving yourself any margin any how.

Do you have OTDR traces for each fibre core so you can see how much actual loss there is?



54067323

join:2012-09-25
Tuscaloosa, AL

2 edits

said by TomS_:

Do you have OTDR traces for each fibre core so you can see how much actual loss there is?

Great point, even if the OP has to rent or subcontract the work.

All that seems to be known at this point is the SFP's tried do not link, why they do not link has not been solved, is it loop loss, that could be as the fiber is already long (as built) per the OP’s specs, but could it be just open on one path or the other?

Dark or lit and unmonitored (disconnected) fibers have a tendency to get screwed up with no one knowing about it during the commissioning of new drops, or the rearrangement of existing paths.

Personally, I believe the OP needs to determine if there is a path before trying to increase power to make the path.

Unfortunately, due to the length of the loop the OP is not going to be able to stick a laser pen at one end and look for a blinking light at the other end.

switchman

join:1999-11-06

1 edit
reply to spectrumhead

said by spectrumhead:

@lutful
32 dbi loss was measured by the owner of the fiber pair recently. Actually the engineer told me that it is around -30 but it will be better for me to consider it like it is -32 for some reason i don't know.

The price I have got from a reseller is 1250$ per SFP. Is it too expensive or are there any cheaper options out there that I may consider ?

37 db is plenty of budget. I would be a lot more comfortable if they gave the specs for the SFP. If you look at the finisar site, you can see an example of what I am talking about for specs.

You could also get an amp use it. Something like this, »www.finisar.com/products/optical···cro-EDFA, in their 1RU package. Then pair it up with a DWDM sfp in you switches. This way if the SFP fails, you have a wide source to select from.

What wavelength did they shoot the fiber with, 1310 or 1550. If it was 1550, then that is not very good fiber or it has a lot of splices or connectors in the path.

When you look at optics you look at the Min TX launch power. That in combination with the min rx sensitivity gives you you link budget. The actual TX power of your SFP will be between the Min and Max value there is no guarentee as tothe actual launch power. It will probably be around+2 or +3, but you can't count on it.

JDSU also makes good stuff.