Understanding The Math Of Insertion Loss Testing
In
order to test “insertion loss” or the direct loss of a fiber optic
cable or cable plant using a light source and power meter (LSPM in most
international standards or optical loss test set – OLTS – in many
articles), one must make an initial measurement to determine the “0 dB”
reference point with the LSPM and reference cables to be used in making
the measurement.
It
has become apparent in technical and standards meetings that what these
measurements actually measure are not well understood, even by experts,
because of the casual way the reference and test measurements are
often described. The purpose of this paper is to use simple math is to
clear up this issue.
Testing the cable plant
Standard
test methods, use a light source [LS] and reference “launch” cable on
one end of the “cable to test” and a “receive” reference cable
connected to a power meter [PM] on the far end. The test is intended to
measure the loss of the connections of the connectors on either end to
the reference test cables and the loss of the rest of the cable (which
may include splices or additional connections in addition to the fiber.)
OFSTP-14/7: [LS]------ref cable------[A]------cable to test------[B]------ref cable------[PM]
Three ways to set the reference
All
standards offer three methods of setting a “0 dB” reference using
one two or three reference cables. It’s assumed that all reference
cables are short enough (generally less than 2m or 6 feet) that the
loss of the fiber in them is ignorable and that the connectors are high
quality – that is to say low loss. It’s also assumed that the power
meter has a large enough detector that it gathers all the light from
the end of a fiber so it has a consistent connection to the fiber being
connected to its interface.
1 Cable Reference : [LS]-------ref cable-------[PM]
2 Cable Reference: [LS]-------ref cable-------[1]-------ref cable-------[PM]
3 Cable Reference: [LS]-------ref cable-------[1]-------ref cable-------[2]-------ref cable-------[PM]
Why
do we have three different methods? Mostly we believe it came about
from three issues: 1) Test equipment with fixed connectors that do not
match the connectors on the cable to test cannot be used for the 1 cable reference,
2) Techs working at either end of a length of cable and set the
reference with a local light source and power meter and 3) Belief that
since you need a launch and receive reference cable, you should make
the reference setting using both of them.
The 3 cable reference has become popular with plug/jack connectors like the MT-RJ or MTP
which are generally incompatible with test instruments, but LC
connectors may require a similar approach if the power meter or OLTS
does not have adapters that allow connecting different types of
connectors.
If you look at the schematic drawings
of the three methods, you can see that when you set the reference, one
or two connections are included. That means that the loss of those
connections are included in setting the 0 dB reference.
This
has led to the misunderstanding that the 3 cable reference , which includes two
connections when the reference is set, does not measure the loss of the
connectors on the ends of the cable because two connections were
included in setting the 0 dB reference and the 1 cable reference (the old Method B) only measures
one connector because one is included in setting the reference.
Unfortunately, it’s not that simple - that's an improper
generalization. Any version of the test measures all the connector
losses in the cable under test, but subtracts the loss of connections
included when setting the “0 dB reference.” Let’s do the math for each
method and you will see what we mean.
The 1 Cable Reference, Formerly Method B
Set reference by Method B:
[LS]-------ref cable-------0-[PM]
where “0” equals the output of the reference cable measured by [PM].
No
connections are included in the reference test (see next paragraph), so
when we do the test, we are connecting the connectors of the ref cables
directly to the connections on the cable to test and that loss will be
included in the measurement.
We consider the
connection to the meter to have "no loss" which may be true with a
power meter with a large area detector capturing ligth directly from
the end of a connected optical fiber cable. However, many OLTS,
especially those designed to make bidirectional tests, may have that
connection be fiber-to-fiber and may even have a couple internal to the
OLTS. That test set has a double uncertainty, the connection to the
fiber under test and the variability of the coupler due to mode power
distribution in MM fiber.
OFSTP-14/7: [LS]------ref cable------[A]------cable to test------[B]------ref cable------[PM]
Thus,
with the power meter, we measure loss “L”, we measure connection loss
[A], the loss of the fiber and any intermediate connections or splices
in the cable to test [CTT] and connection [B] or
L = 0 – ([A]+CTT+[B]) = loss of the cable
The 2 Cable Reference, Formerly Method A
Set the reference with method A but do not change the output of the LS.
2 cable reference : [LS]-------ref cable------0-[1]-------ref cable-------(0-[1])[PM]
where
“0” equals the output of the launch reference cable measured by [PM] as
in the 1 cable reference, decreased by the loss of the connection [1] between the
reference cables, so the [PM] now measures (0-[1]).
OFSTP-14/7: [LS]------ref cable------[A]------cable to test------[B]------ref cable------[PM]
Thus,
with the power meter, we measure loss “L”, we measure connection loss
[A], the loss of the fiber and any intermediate connections or splices
in the cable to test [CTT] and connection [B] but our “0 dB
reference” is now “0-[1]” and the loss measured is
L = (0- [1]) – ([A]+CTT+[B]) = measured loss of the cable
Note
the measured loss is the same as the 1 cable reference decreased by the loss of the
connection between the reference cables [1] when the “0 dB reference”
was set, which is not necessarily the same as saying the loss measured
does not include one of the connectors, since the loss of [1] is not
necessarily the same as [A] or [B]. That unknown factor causes this
method to be more uncertain (many would say "less accurate") than the 1 cable reference.
Is it possible to compensate for the included connections?
Someone making this test who understands this issue could compensate for the difference by calibrating [1] this way:
Set reference by 1 cable reference :
[LS]-------ref cable-------0-[PM]
Add the receive reference cable and measure again:
2 cable reference : [LS]-------ref cable-----0-[1]-------ref cable------[1][PM]
This
test gives you a direct measurement of [1] which can now be added back
in when testing the cable, giving you the same result as method B. But
in order to do so requires the ability of the instruments to measure by the 1 cable reference , so it just adds another step to the process – why not just
use 1 cable reference in the first place. And the loss measured [1] will be
slightly different each time the two connectors are mated, adding to
the uncertainty of the measurement.
The 3 Cable Reference, Formerly Method C
Set the reference with method A but do not change the output of the LS.
Method C: [LS]-----ref cable----0-[1]------ref cable------[2]------ref cable------(0-[1]-[2])[PM]
where
“0” equals the output of the launch reference cable measured by [PM] as
in the 1 cable reference, decreased by the loss of the connections [1] and [2]
between the reference cables, so the [PM] now measures (0-[1]-[2]).
OFSTP-14/7: [LS]------ref cable------[A]------cable to test------[B]------ref cable------[PM]
Now
with the power meter, when we measure loss “L”, we measure connection
loss [A], the loss of the fiber and any intermediate connections or
splices in the cable to test [CTT] and connection [B] but our “0 dB
reference” is now “0-[1]-[2]” and the loss measured is
L = (0- [1]-[2]) – ([A]+CTT+[B]) = measured loss of the cable
The
measured loss now is the same as 1 cable reference decreased by the loss of the
connection between the reference cables [1] and [2] when the “0 dB
reference” was set, which is not necessarily the same as saying the
loss measured does not include one of the connectors, since having two unknown connectors causes this method to be more uncertain (many would say "less accurate") than either of the other methods.
Putting some numbers on this methodology
Consider that the test is being made by someone who is careful with all
the procedures and has very good reference cables, which should have
mating losses of 0.2 dB or better. If we test a cable plant with the 1 cable reference at 3.0 dB, with the 2 cable reference it will measure 2.8 dB and with the 3 cable reference,
2.6 dB.
We’ve done actual tests like this
under strictly controlled conditions and got 2.96 dB, 2.66 dB and
2.48 dB. See the table below.
The
measurement uncertainty, called “standard deviation” in statistics, can
be determined by making multiple measurements, averaging and
calculating standard deviation. What was found is this test here was 1 cable reference had a standard deviation of 0.02 dB, 2 cable reference had a
standard deviation of 0.20 dB and 3 cable reference had a standard deviation of
0.24 dB. This is expected because the 1 cable reference has no variation caused by
connections included in setting the reference, while the 2 cable reference has one
and 3 cable reference has two and every fiber optic connection has some variation each time
it is connected.
| Test Method | Results
Loss in dB, Standard Deviation | | 1 Cable Reference | 2.96 dB, +/-0.02 dB | | 2 Cable Reference | 2.66 dB, +/-0.20 dB | | 3 Cable Reference | 2.48 dB, +/-0.24 dB |
Another unknown creates a problem with these tests on multimode cable –
mode power distribution (MPD.) Even if we control the MPD in the launch
cable using a source of known MPD and a mandrel wrap as called for in
some standards, as soon as we add another connector, we change the MPD
in a fairly uncontrolled fashion. Connectors act as mode modifiers,
stripping off higher order modes which are the most lossy in
connections but converting low order modes to higher order modes also.
So setting the reference with connections adds to the uncertainty of
mode power distribution in the testing.
Yet
another issue is the cleanliness of connectors. If one of the
connectors is dirty when the reference is set by the 2 or 3 cable reference methods, but
the connectors are subsequently cleaned or the dirt falls off, the loss
will change because of the extra loss included in the reference setting
process. This has even been known to cause a LSPM test to show a “gain”
which sounds impossible.
“Fudging” the Methodology
Manufacturers
of test equipment with fixed connector interfaces have invented some
creative methods of "fudging" a 1 cable reference test. Some of them do
a 2 cable reference, add a hybrid adapter cable or two to allow
connection of
the cable to test, then do some behind the scenes software calculations
and “kazam!” a claim it is a 1 cable reference test because it includes
2 more connections in
the test than in the reference! Looking at the info above regarding the
uncertainty of measurements anyway, think about how uncertain a
measurement is when you try to guess at the loss of several newly added
connections!
"0" dB Reference: [LS]-------ref cable------0-[1]-------ref cable-------(0-[1])[PM]
Test: [LS]------ref cable------[X]--adapter---[A]------cable to test------[B]------ref cable------[PM]
L = (0- [1]) – ([X]+[A]+CTT+[B]) = measured loss of the cable
The uncertainty of the measurement is primarily determined by the difference in the losses [1] and [X].
Which method to use?
Most cabling standards today recognize the need for the 3 cable reference for some
types of connectors but prefer the 1 cable reference. Network standards for link
loss have referenced all methods, so it's important to check the
network specifications. Many telcos still use the 2 cable reference. Whatever
method you use, the loss data you document must include the method used
to be valid and be comparable to other tests.
|
