dect,

Don't believe all the half-truths given here as answers. In the 1980s and 90s I worked professionally in the area of fibre-optics for data communications and even designed and specified composite fibre bundles myself. I can therefore inform you without hesitation that optical fibre does indeed have a per-unit distance loss. Indeed, all media, not just optical media, exhibit unit distance loss in the propagation of electromagnetic energy. There's a formula for it, based on the speed-of-light in a vacuum and which has a multiplication factor in the case of optical fibre that takes into account the refractive index of the core, the impurities in the medium, and unit distance losses through the cladding. Back in the 80s/90s, so-called singlemode fibre had a typical loss of around 0.5dB max per km, though this did, and does, also depend on wavelength. I don't think much has changed in that respect in the interim. Maybe that figure has, with more advanced manufacturing over the years, dropped a little? A good-quality singlemode, which certainly in those days was being used by BT for longhaul datacomms/phonelines, was perhaps 0.2 - 0.35dB per km. Transit losses in the fibre itself have nothing to do with launch power, of course.

All fibre splices and all jointing that involves optical connectors also have associated losses, and even a good-quality, in-the-field fusion-splice will render an attenuation loss of typically 0.2dB, and mated screw-together connectors typically 0.5dB. Yes, I had experience doing fusion-splicing too.

Engineers with the requisite professional equipment can check the integrity of an optical line by launching a calibrated optical test signal into it and viewing the graphical results on an optical time-domain reflectometer (OTDR). The latter is a piece of sophisticated kit vaguely resembling a portable oscilloscope. Essentially, the OTDR plots the attenuation (loss of the light energy) of the light-signal (of a particular wavelength) against distance, so you can see precisely what's happening along the entire length. Beside the unit distance loss of the fibre itself, the TDR will also show in the plot each fusion-splice loss as a small step. Calibrated scaling on the TDR enables the engineer to read off either the individual losses or the overall end-to-end loss. Prior to the advent of fibre, ie. the days of communication on long lines that involved copper-only, there were just 'plain vanilla' TDR's for copper lines. BT would use them for finding exactly where jointing issues in copper lines existed in hard-to-access situations. In fact, later fibre-based ones were just an optical version of the same principle being employed, that of transmission-line theory.

Although fusion-splicing has become more common these days it still requires a good deal of skill to lessen the misalignment between the two butt ends of the fibre at the required joint. The fibre ends need careful preparation too, as the fibre will have a micro-thin inherent cladding or outer coating on it and this needs to be stripped off at the proposed joint and thoroughly cleaned, before setting the two butt ends into a jig on the splicer, that holds them in place for their fusing together. Any contamination of the joint, as it's being done, results in a high loss and bad joint that then has to be re-done. The operator uses a microscope to check the cleanliness of the ends and to see that they're in optimum alignment in the jig, before pressing the button to evoke an 'arc-welding' of the two butt-ends. To protect the joint from thereon in, a cover already slid on to the fibre is then brought over the finished joint and is crimped on or otherwise sealed on.

Hope this puts you in the picture.

In reply to a post by RobertoS:

You mean you are installing some part of our FTTP circuits and not doing a proper light test? Do you work for Openreach, or one of the shoddy sub-contractors who are the cause of a high proportion of faults?

Such as perhaps the 32-connection splitter cockup here?

Nope I work as an admin for a HPC (otherwise known as a supercomputer) at a UK university. I have been working with fibre optic for ethernet, fibre channel and Infiniband now for around 15 years.

What I find is that if I have an issue with link and I am not sure if I am getting light or whether the patch lead is a straight or crossover then a mobile phone is great for a quick check. Much quicker than going and digging the light meter out for an actual loss measurement. Most of the time the issue is either a duff patch cable, a patch cable is not seated properly or a dead laser.

The closest I have come to a fibre install was stringing some pre-terminated 24 core cable between a couple of rows of racks and that was about a decade ago. On the other hand I regularly deal with fibre optic links. Right now I am about to go over to the data centre to try and work out why my new 30m active optic 40Gbps QSFP leads don't work. They where supposed to replace the hack with two 15m ones and an Infiniband switch in the middle but I am not getting link I am hoping I have not damage the cables but they allegedly support a 7.5mm minimum bend radius which I am confident I have not exceeded installing them.

Hi meditator

Many thanks for your detailed response, it was a very interesting read and has given me a better understand.

I think the point is that sure the signal degrades over distance with fibre optic. However unlike xDSL technology where the degraded signal leads to a lower data rate, with a fibre optic link you either have the link and it operates at full speed, or you don't have a link at all. Further given the much lower losses of the signal with single mode fibre you can be 20km from the exchange and still get full speed. This is a fundamental difference and why people tend to describe fibre links as suffering no loss.

However note it is not quite true that you either get link or you don't. If you are right on the cusp of signal loss you may get link but then suffer a lot of RX/TX errors and only get a fraction of the advertised link speeds. Don't ask how I know...