PMD Tolerance Testing
of Optical Interfaces
Harald Rosenfeldt, Agilent Technologies,
Harburger Schlossstr. 6-12, 21079 Hamburg, Germany
harald_rosenfeldt@agilent.com
Abstract
Improving the tolerance to polarization
mode dispersion (PMD) is considered
to be one of the major prerequisites
for the success of modern high bit
rate optical communication systems.
Various approaches such as optical
compensation, electrical mitigation,
multi-level modulation formats promise
to increase the PMD tolerance of optical
systems, whereas the question of how
to experimentally characterize these
solutions needs to be answered before
commercial deployment. This is not an
easy task since these systems need to
be characterized with respect to fi rst and
higher order PMD but also with respect
to their dynamic behavior. We show that
deterministic polarization controllers
combined with in-situ measurement
of PMD can help to explore the PMD
tolerance of an optical communication
system and to generate reliable and
repeatable results by avoiding statis-
tical elements such as polarization
scramblers. These elements can be
combined to form a PMD testbed which
allows to stress a system by applying a
deterministic amount of PMD including a
well-defi ned rate of change. Such a PMD
testbed can be used during development
of adaptive mitigators as well as for
compliance testing. Finding an agree-
ment on standard test procedures for
such a testbed will make the evaluation
of PMD tolerant receivers easier and
more comparable.
Keywords: PMD, PMD Measurement,
PMD Tolerance, PMD Testbed,
Polarization Synthesizer, In-situ PMD
1. Introduction
The main objective of determining the
PMD tolerance of an optical transmission
system is to fi nd out on which fi ber link
it can be operated with the desired
availability. Although the aim seems very
obvious, it turns out to be a rather com-
plicated task to guarantee a predefi ned
outage probability for a transmission
system on a specifi c fi ber link. Main
reason for that is the statistical nature
of PMD which always leaves a risk that
the fi ber link moves to a state exceeding
the PMD tolerance of the system.
The established way of tackling this
problem is to determine the tolerance of
the system with respect to differential
group delay, DGD. A typical value for the
DGD tolerance of a system would be
approximately 30% of the bit period, i.e.
8 ps for a 40 Gbit/s system. Considering
that the instantaneous DGD of a fi ber
link can be approximately 3 times
higher than the PMD value (due to the
Maxwellian distribution of the DGD),
the maximum tolerable PMD would
be 2.7 ps. Although being very simple,
this rule-of-thumb leads to design rules
which can easily be applied in the fi eld.
However, this practical rule works be-
cause the estimated tolerance is rather
conservative. At a bit rate of 10 Gbit/s,
this might be acceptable, but a maximum
tolerable PMD of 2.7 ps will disqualify
most long haul fi ber links for the use of
40 Gbit/s data rate. Therefore, network
operators consider a PMD tolerance of
around 8 ps to be desirable. This would
allow a suffi cient number of links to be
upgraded to 40 Gbit/s.
To achieve this tolerance at these data
rates adaptive systems will apparently
be part of the receiver to improve the
quality of the received signal. These
systems can be optical systems (e.g.
optical PMD compensators) but also
electronic systems (e.g. Viterbi decoders).
In any case, an adaptive system will
add dynamic properties to the receiver
and thus raises the need for character-
izing the dynamic performance.
Characterizing the dynamic perfor-
mance of a PMD tolerant receiver
means to examine the ability of the
receiver to follow different sorts of
perturbations, i.e. stress patterns.
Choosing certain stress patterns is
closely linked to the question about the
required speed of a PMD compensator.
Published results show that PMD fl uc-
tuations can be in the order of sub-mil-
liseconds [1, 2]. However, most of the
time, the PMD varies much slower and
tends to have time constants of min-
utes to hours [3, 4]. It is obvious that an
aerial fi ber is likely to exhibit a different
fl uctuation pattern than a buried fi ber.
Therefore, a method for characterizing
dynamic properties of a receiver which
does not act on assumptions about the
fi ber link would be helpful. This will lead
to an objective characterization compa-
rable to the frequency-response of an
amplifi er. Having gathered information
on the dynamic receiver properties, the
individual dynamic properties of a fi ber
link can be compared and a qualifi ed
decision can be made whether the
receiver can be used on this particular
link or not.