Topic: Fiber Optic Installation Deliverables Table of Contents: The FOA Reference Guide To Fiber Optics

Fiber Optic Cable Plant - Acceptance Of The Finished Product - Deliverables

What is involved in the specification and acceptance of a cable plant at the end of a installation project and what are reasonable specifications for a cable plant. FOA has a lot of documentation on a project involving  designing and installing a cable plant in the FOA Online Guide and our Textbooks, but the acceptance process is relegated to a few paragraphs. Let's look at a project and include a few links to FOA tech documents in case you want to investigate further.

The Project "Deliverables"
Fiber optic projects start with a design that creates project paperwork - the scope of work (SOW), request for proposal or quote (RFP/Q) and a contract with the builder/installer. A "Scope of Work" document is created by the initiator of a project to describe the work to be performed or the services to be provided by a contractor. It describes tasks to be performed, directs methods to be used, and defines the period of performance. It should contain design and performance requirements. A scope of work for communications cabling or fiber optics may be part of a larger building project document that is based on a standardized format called "MasterFormat" in the US and Canada.

A well written scope of work can do more for the success of a contract than any other part of the contracting process. A good scope of work is clear, complete, and logical enough to be understood by the respondent and the personnel who will administer it. Because it describes the details of performance, it is the yardstick against which the respondent's performance is measured. That is why the user's requester, contract administrator and/or subject matter expert should be the focal point for developing the scope of work.

What we are discussing here is the final product - the "deliverables" - that define the final product that the end user expects to have when the installation is complete and ready to use, or in some cases already has the communications equipment installed and operating. Of course the deliverables include the physical cable plant described in the SOW, but also must include full documentation and test results, and maybe even a warranty.

Cable Plant Specifications
What are reasonable specifications for a cable plant. We've often heard stories of specs that are too stringent and others too lax. Since specifications for the installed cable plant are up to the person specifying the cable plant, they may be confusing because very few standards exist for the design and specification of the cable plant. OSP networks have traditionally been specified and owned by sophisticated users who have a history of what specs can be expected. Premises cabling systems tend to use component specs from TIA or ISO/IEC standards that are generally too lax, much higher than what should be expected.

Let's do the executive summary. The cable plant should be specified for loss using a loss budget. Network speed may dictate specifications for component types or bandwidth. Here is a summary of FOA's "reasonable specs" to use for cable plant loss budgets. Below we'll get into testing with loss budgets based on those specs.

Fiber: G.652 attenuation 0.4dB/km at 1310nm, 0.25dB/km @ 1550nm
Splices: Average 0.1dB, reject @ 0.2dB
Average loss 0.3dB, reject @ 0.5dB, Reflectance better than -40dB.
Long haul fibers like G.654 will have slightly better attenuation specs, ~0.2dB/km.

Fiber: Multimode OM3/OM4 attenuation 3dB/km @ 850nm, Bend insensitive fibers
Singlemode OS1/OS2 attenuation 0.5dB/km @ 1310nm, Bend insensitive fibers
Splices: Average 0.3dB, rare in premises
Average loss 0.5dB, SM reflectance better than

Cables types, of course, are specified according to the requirements of the project and it's physical locations.

The final documentation delivered to the customer must be comprehensive, with full route information including GIS (geographic information system) data on the location of the cable and every component - cables, manholes/handholes, splice locations and full descriptions, plus test data. The physical component and location information is obvious, but what is not is test data, which we elaborate on below.

What's sometimes missing, based on inquiries we get from end users, is understandable documentation. Managers who may not be familiar with fiber optics can be given reams of documentation which they are expected to use to sign off on a project. There are many stories about problems at this stage: signing off on a data center installation where all 4,000 connectors were failures, getting test data on a OSP network where every OTDR test was the same, you get the idea. Before signing off on a project, someone who knows fiber optics and was involved in the project should review the documentation and test data and verify that it is correct and valid.

Cable Plant Test Data
To prove the cable plant was installed properly requires test data, of course. During the design phase, loss budgets calculated for each cable run should provide an estimate of the expected loss of the fibers in each cable link to compare to actual test results.

Short fiber optic premises cabling networks are generally tested in three ways, connector inspection/cleaning with a microscope, insertion loss testing with a light source and power meter or optical loss test set, and polarity data, meaning that the routing of fibers is confirmed so that when connecting equipment the tech can identify fiber pairs for transmit and receive. Polarity testing generally can be done with a visual fault locator to confirm that fibers are connected per the documented cable diagrams.

Outside plant (OSP) testing is more complex. If the cable plant includes cables concatenated with splices, it's expected to add OTDR testing to the connector inspection, insertion loss and polarity testing. If the link has passive devices like FTTH splitters or WDMs, those need to be tested and documented also.

There is one thing that whoever is reviewing the data - and going back to the design phase, whoever writes the test specifications based on the loss budgets in the first place - needs to understand: none of these are absolute numbers. The loss budget which is created early in the design phase estimates the loss of the cable plant based on estimates of component loss and therefore is not an absolute number, but an estimate to be used to compare to test data.

Test data is created by instruments and related components that make measurements which have  measurement errors. There are always factors in making measurements that cause the instrument reading to be inaccurate - only an approximation of the real value - and the real value is unknowable because of measurement errors. (If you are curious, look up the Heisenberg uncertainty principle.)

Let's look at this symbolically:


The loss budget is not exact, nor is the testing, so there is a range of measurements that should be acceptable. Some judgement is needed to determine if a particular fiber's test results are acceptable.
In our experience, those two factors cause more stress between managers and installers than just about any other factor in a
cable plant project. Consider these examples of the issues with loss budgets and testing errors.


OSP Cable Plant
Here is the situation a CFOT found themselves in when they called the FOA. They were an 30+ year experienced splicer with a half-million splices of experience. A customer wanted to specify a long cable plant (~50 miles/80km) with splices that averaged 0.05dB and any splices above 0.15dB was not acceptable. Testing of the splice loss would be done with an OTDR with bidirectional measurements and averaged.

What made this call particularly interesting is this tech had some real world data, the kind you do not see often. On a past job, he had spliced a ~60 mile (100km) 288 fiber cable plant at 18 splice locations, that's 5,184 splices. His test records showed that 60% of the splices were in the range of 0.02 to 0.08dB and 40% were in the range of 0.08 to 0.15dB. Only 17 splices were over 0.15dB.

OSP cable plant

The customer noted that the manufacturer of the splicer used by the tech quoted a splice loss capability of 0.02dB, so a field spec of 2.5 times that should be easily achievable in the field. What the customer did not understand was that 0.02dB spec for the splicer was data taken in a laboratory on a new or perfectly set up machine. The splices were made by breaking a fiber and splicing it back together. Every splice used in determining the splicer capability used identical fibers - they were the same fiber.

In the field, when splicing cables together, the environment is not like a lab. Machines are used to splice thousands of fibers. The fibers in the cable can be from numerous production runs and will have variations in mode field diameter (MFD) and geometry. Assuming a long haul network like this one is using G.654 fiber, we can look at the ITU standard for G.654 fibers and we find these specifications:

Mode Field Diameter: 9.5-10.5microns
Core Concentricity Errors: 0.8micron

That variation in fiber geometry and MFD can produce a real difference in splice loss that will be directional. That difference can be 0.05 to 0.1dB. That is independent of how well the splicer can align and fuse fibers. Even if aligned
perfectly and spliced perfectly the differences in the fiber will cause directional splice differences - higher in one direction, lower in the other.

The can be a 0.20 to 0.25dB difference in directional splice loss when measured by an OTDR 
caused by MFD variation in the fibers (data from Corning ap note AN3060). This is something which many techs are familiar with, but that method of bi-directional testing merely removes the OTDR scattering error and gives the average of loss from each direction. It can't compensate for the actual directional splice loss caused by the difference in MFD. Let's repeat that: Bi-directional OTDR testing removes the OTDR error caused by differences in fiber MFD or backscattering, but cannot compensate for the actual directional difference in splice loss caused by the difference in MFD.

Back to the customer's spec. They wanted an average splice loss of 0.05dB and no splices over 0.15dB - which was unacceptable. Using some math, we can analyze the data the tech had from the prior 60 mile (100km) job. The average splice loss on that cable plant was ~0.07dB. And only 17 were larger than 0.15dB, so the reject rate would have been 0.3%.

If we compare the results of that job to the specs the customer wants on the new job, the difference would be 0.02dB/splice at 18 splice points. The total loss difference in the
60 mile (100km) cable plant would be 0.02dB X 18 or 0.36dB - and that is on a 60 mile (100km) run where the fiber loss is 0.20dB/km X 100km or 20dB. And the loss of the original 18 splices was only ~1.33dB! The difference is negligible and the measurement uncertainty of the OTDR test of end-to-end loss is much bigger than the difference.

Premises Network
In premises fiber optic networks, the TIA standards allow for connections to have a loss of 0.75dB - that is two connectors mated to create a connection.
A fiber optic connector has no loss, per se, because it is not being used. When in use, it is mated to another connector creating a joint between two fibers, and that joint is what has loss - a "connection" loss.

That number has been in the standards for at least 30 years, but even then typical connectors with ceramic ferrules were much better than that. That 0.75 dB loss was needed for early connectors like SMAs and Biconics, so it became the standard, Later, although everyone knew that the typical ST, SC, FC and then LC connector was much better, the industry saw the introduction of array connectors (MPOs) where the 0.75 dB loss was needed, so rather than have different values for single fiber and array connectors, it was left at 0.75dB.

premises cable plant

If you do a loss budget for a premises network with an intermediate patch panel like the one above, your loss budget would include 4 connection losses, the two in the patch panel and the ones in the outlets at each end where you connect the patchcords to the LAN gear. The 4 connection losses using the TIA model would allow a loss budget for connections of 4 X 0.75dB = 3.0dB. But if typical
connections are less than 0.5dB, you could have 3 connections at 0.5dB and 1 connections could be 1.5dB. If you had good connections of 0.3dB, that fourth connection could be 2.1dB!

When we look at fiber losses, TIA allows f
iber losses of 3.0 to 3.5dB/km at 850nm for multimode fiber. Actual fiber is now less than 3dB/km, but since links are typically short, ~100meters, the error due to fiber being better than the standard is only tenths of a dB. That is too small to matter. 

If our link above is 100m, the loss budget using TIA numbers would be:

Fiber 0.1km X 3.0dB/km = 0.3dB
Connectors 4x 0.75dB = 3.0dB
Link Loss Budget = 3.3dB

With more realistic numbers, say 0.5dB connections, it would be:

Fiber 0.1km X 3.0dB/km = 0.3dB
Connectors 4x 0.5dB = 2.0dB
Link Loss Budget = 2.3dB

And with really good
connections, say 0.2dB:

Fiber 0.1km X 3.0dB/km = 0.3dB
Connectors 4x 0.2dB = 0.8dB
Link Loss Budget = 1.1dB

That's a 2.2dB difference in a 100m multimode network; that's a big uncertainty! What would we choose for a GO/NO-GO loss? Our judgement would be the link should be under 2.3dB

Now what happens when we test this link?

We use a LED test source at ~850nm, a meter calibrated at 850nm reading in dB, and two reference cables to make a double-ended test. In a short link like this the cause of measurement uncertainty is the loss of the connections. Variations in modal fill from the test source and launch cable can result in 0.2dB variations, which has resulted in an international standard for mode fill, called "encircled flux" which most multimode test sources today meet, but early sources are unknowns and add to the uncertainty, The launch and receive cables also add to the uncertainty, since fiber standards allow up to +/-5% variation in core size, which can cause loss variations at connections depending on the direction of the light.

There are so many variables in making an insertion loss test of multimode fiber that they fill a giant table in the FOA page on "metrology" or the science of fiber optic measurements. The generally accepted number for uncertainty of this kind of measurement is ~0.2 to 0.5dB.

So exactly what is a acceptable test result for this fiber link? If we measure a loss of 1.5dB, no question it passes. If it measures 3.5dB, that's certainly a problem. But what if it measures 2.5dB? That's 0.2dB higher than the loss budget estimate we used, but we know the measurement is uncertain by +/-
0.2 to 0.5dB, and 2.3dB is within the uncertainty of the measurement. It's probably OK.


More Topics On Fiber Optic Installation

Table of Contents: The FOA Reference Guide To Fiber Optics

(C)1999-2020, The Fiber Optic Association, Inc.