Differential Mode Delay (DMD) for multimode fiber types and its relationship to measured performance

Differential Mode Delay (DMD) for multimode fiber types and its relationship to measured performance
DIFFERENTIAL MODE DELAY (DMD) FOR MULTIMODE FIBER TYPES AND ITS
RELATIONSHIP TO MEASURED PERFORMANCE
Rick Pimpinella and Al Brunsting
Fiber Research Department
Panduit Corp., Orland Park, IL 60427
determined by the fiber’s bit error rate (BER). Optical
pulses are launched into the MMF core at precisely
controlled offsets from the fiber axis and have fast rise and
fall times, referred to as restricted launch conditions,
compared to overfilled launched conditions using an LED.
Emerging pulses from the MMF are measured and saved in
the time domain (voltage vs. time) by a fast received and
data storage device. We will show that the DMD
measurements correlate with MMF length, fiber type, and
the corresponding BER.
Abstract – With the introduction of Vertical
Cavity Surface Emitting Lasers (VCSEL’s) used
with multimode fiber (MMF), Differential Mode
Delay (DMD), modal interference effects and the
influence of connector interfaces are found to
reduce the observed bandwidth of multimode
fibers (MMF’s). In this study the performance of
three example MMF’s is characterized by
measuring DMD and is compared to Bit Error
Rate (BER) measurements on the same three
fibers.
Fiber length is also considered.
Controlling the spatial locations of the launch
pulses, we develop methods aimed to certify the
performance of optical fiber for 10Gb/s Ethernet.
The source for these measurements was an 850nm
VCSEL.
2. PERFORMANCE OF MULTIMODE FIBERS (MMF)
2.1 Refractive index profile
The refractive index of the MMF core varies as a function
of distance from the core axis (refractive index profile).
The specific shape of the refractive index profile in the core
has a significant effect on the distribution of the guided
optical power in the fiber. But more importantly, the index
profile profoundly influences the velocities of the various
propagating modes, which results in dispersion.
A
parabolic index distribution nearly equalizes the group
velocities of the propagating modes and fiber
manufacturers strive to optimize the specific index
distribution to minimize dispersion.
©2005 Optical Society of America
OCIS codes: (060.0060) Fiber optics and optical communications,
(060.2270) Fiber characterization, (060.2300) Fiber measurements.
1. INTRODUCTION
It is now clear that of the bandwidth of a significant
fraction of legacy MMF and laser-based transceivers (such
as VCSELs) does not support data rates up to 10-Gb/s and
lengths of 300-m. This is due mostly to the design of these
fibers which were optimized for lower bit rates and light
emitting diode (LED) transmitters. The fiber cores were
designed to be overfilled and thereby the LEDs couple
optical power into all, or nearly all, of the fiber modes.
However, when the small spot from a VCSEL illuminates a
fiber core, optical power is coupled into a small fraction of
the available modes (several hundred).
The actual
supported bandwidth of installed legacy MMF, using laserbased transceivers, can be (and usually is) unpredictable.
A general refractive-index profile is given by Eq. (1).
n(r ) =
{n
clad
]
α 1/ 2
r < rcore
r ≥ rcore
(1)
where ∆ is the relative refractive index difference between
core and cladding and given by ∆ = 0.5.(ncore2 - nclad2)/
ncore2. The power law parameter is α.
A set of perturbed refractive index profiles, relative to
Equation (1), has been proposed to describe MMF that is
representative of installed fiber [2]. Even though these
perturbations were used for 1300nm and a core diameter of
62.5µm a similar set of perturbations can be generated for
850nm and a core diameter of 50.0µm to describe those
corresponding applications. In ref. 1 the effects of
perturbed refractive index profiles on DMD-type results
were simulated and shown to be significantly dependent on
the refractive index profile and amount of offset launch [2].
New laser optimized MMF supports these higher data rates
up to 300-m, specified as OM-3 by ISO 11801 and 850-nm
Laser Optimized 50 micron (850 LO 50) by TIA. The only
accepted way to characterize these high bandwidth MMF’s
is with DMD measurements. See, for example, ref. [1].
This paper will describe such DMD measurements and
relate the results to measured MMF performance as
OFC/NFOEC Conference
[
ncore ⋅ 1 − 2 ⋅ ∆ ⋅ (r rcore )
1
2005 Technical Proceedings
DMD measurements, described in Section 3, are similar to
these offset launch conditions.
differences in modal output in combination with the optical
properties of MMF result in an observed modified behavior
for VCSEL signal propagation. For 10GbE systems, the
dispersion and interference effects acting on the
propagating modes can significantly degrade signal
performance, thereby limiting the maximum transmission
distances to a few hundred meters. The dispersion and
interference effects result in several optical power penalties
that reduce the system’s overall optical power budget.
2.2 Vertical Cavity Surface Emitting lasers (VCSELs).
In MMF systems the optical source is traditionally an LED
and depending on the material system, operates at either
long or short wavelengths (1310nm or 850nm respectively).
Unlike Fabry-Perot (FP) and Distributed Feedback (DFB)
lasers, LED’s are mainly surface emitting devices, having a
broad spectral output with large numerical aperture (NA).
The large core diameters in MMF (50µm, 62.5µm and
larger) are designed to efficiently couple the LED output
power. The LED output overfills the MMF core resulting
in an Overfilled Launch (OFL) condition. For an OFL
condition the performance of MMF is limited to relatively
low bit rates (< 650Mb/s) and, modal dispersion is
relatively well behaved.
2.3 Bit error rate (BER)
Flawless propagation of bits of data through the fiber link is
the ultimate design goal of the MMF in that optical link.
For our purposes there are two basic measures of this
transport performance: (1) the number of bit errors
compared to the total number of bits transmitted (BER) per
unit time and (2) the speed of the bits passing through the
fiber (data rate in bits/sec). Here we will focus on BER.
State-of-the-art high-speed multimode systems such as
10Gigabit Ethernet (10GbE) use newer devices known as
VCSEL’s. They, like LED’s, are surface emitting devices;
however, the internal structure of the VCSEL comprises a
layer of multiple quantum wells between Bragg Reflectors
forming a laser cavity. Hence, although the device is a
surface emitter, the output of the VSCEL is similar to that
of the more traditional edge-emitting laser. The light
emission of the VCSEL is from a single longitudinal mode,
although multiple transverse modes with slightly different
wavelengths are supported due to the lateral cavity
diameter.
In Fig. 1 consider digital pulses (bits) being launched into a
MMF, represented by “initial pulses”. The mean optical
power of the (zero, one) bit is (p0, p1), W. There is noise on
these pulses, pnoise. After the pulses have propagated
through the MMF they suffer distortions due to differential
mode delay and other effects, represented by “final pulses”.
Additional noise is present in the final pulses.
The difference between p1 and p0 for the (initial, final)
pulses is (OMAinitial, OMAfinal) or optical modulation
amplitude. There are two bimodal distributions of optical
powers for both the p1 and p0 pulses, shown on the right
side of Fig. 1. The logic of the associated circuitry is
programmed to decide if a given propagated pulse is a 1 or
a 0. A comparison is made between the detected optical
power of a given pulse within a given time window with a
given power threshold, pTH. If a given pulse (bit) starts as a
pinitial(t)
pnoise
optical power, p(t)
optical power, p(t)
Whereas edge-emitting lasers couple few longitudinal
modes into single-mode fiber (SMF) and LED’s couple
large numbers of modes uniformly across the core of the
MMF, VCSEL’s couple a few transverse modes into a large
core MMF in an under-filled launch condition. These
pfinal(t)
p1
t
OMAinitial
In general
width of
distribution is
reduced.
p1
pTH
OLfinal
OMAfinal
OLinitial
p0
initial
pulses
p0
final
pulses
No. of readings
No. of readings
Fig. 1. Explanation of bit error rate (BER). Digital pulses are launched into the multimode fiber (MMF) under test
and undergo distortions that result in bit errors. See the text for an explanation of these symbols.
OFC/NFOEC Conference
2
2005 Technical Proceedings
0 and falls above pTH, that bit will be erroneously recorded
as a 1. A similar error occurs for a starting 1 bit which falls
below pTH. Those bit errors are presented as the gray
overlap areas, OLinital and OLfinal.
the pulse maximum of the leading edge of the fastest pulse
and 25% of the pulse maximum of the trailing edge of the
slowest pulse in this sequence (see Fig. 2).
If all the pulses in the DMD sequence are temporally
aligned; then we expect that arrival time of each pulse,
propagated by a distribution of mode groups, will be nearly
the same from pulse to pulse. This condition translates into
lower noise (lower temporal jitter) and a smaller BER
(improves). See Fig. 1. Conversely, as the temporal
misalignment increases in the DMD sequence, the temporal
jitter increases, and the BER increases (becomes worse).
The final bit error rate is then the ratio of the final overlap
area, OLfinal, to the total area of the bimodal distribution on
the right side of Fig. 1. To reduce the final BER it can be
seen that OMAfinal must be increased and/or pnoise must be
decreased.
If the data sequence contains consecutive identical digits
(case A), then the BER may change compared to
alternating digits (0’s and 1’s, case B). This is because
case A may contain significant low-frequency components
compared to case B.
To measure DMD consider the difference between tfast and
tslow, which is comprised of DMD (due only to the fiber),
the launch optical pulse, and the chromatic dispersion [in
ps/(m*km)]. See ref. [1].
3. DIFFERENTIAL MODE DELAY (DMD)
4. MEASUREMENTS OF DMD AND MMF
PERFORMANCE
In addition to BER another way to characterize MMF is to
measure DMD. A spatially small (compared to the MMF
core) and temporally short optical pulse is launched in the
core of the MMF end face that is under test. At the output
end face the resulting signal is measured [3 - 5]. This
measurement is repeated, starting at the axis of the MMF
core and moving outward to the core/cladding interface [1].
See Fig. 2 (not scaled and illustrates only the principle).
Due to the cylindrical symmetry of the fiber this linear scan
responds to many of the MMF modal structures. The
launching spot can originate from a single mode fiber (or
equivalent). There are additional requirements for the
detection system and methods [1].
4.1 Materials and methods for DMD measurements
The layout of Fig. 3 was used to make DMD
measurements. A VCSEL source is embedded in a
commercially available XFP transceiver, powered by an
evaluation board. Voltage pulses from the pulse generator
are differentially supplied to the evaluation board and in
response the VCSEL generates optical pulses. The pulse
width, period, delay, amplitude, and voltage offsets are all
controlled from the pulse generator.
Optical pulses at 850nm are launched into SMF from the
VCSEL, resulting in an approximately 18dB signal
attenuation, compared to MMF. The launch SMF is
positioned to better than 0.5µm accuracy and repeatability
by the x-y-z precision location control and the bare fiber
For a given offset launch only a weighted subset of all the
possible mode groups is excited in the MMF. At the next
offset launch location in the DMD measurement sequence a
different weighted subset of mode groups is excited. In the
full sequence of a DMD measurement all the mode groups
are excited, according to the weights for each offset
location. The different mode groups will, in general, have
different propagation times as discussed in Section 2.1,
illustrated in Fig. 2 (right-hand plot). Two times are
measured, tfast and tslow, determined by the time of 25% of
Fiber
core
Fiber
cladding
Evaluation PC
board
x-y-z
precision
location
control
Pulse generator
Trace due to launching spot 1
Trace due to launching spot 2
Trace due to launching spot 3
Optical table
MMF under test
+ +
+
stationar
y fiber
holder
trigger input
High frequency
scope
+
SMF = single mode fiber
MMF = multimode fiber
optical input
+
tfast
+
tslow
primary path
alternate path
time
Optical power
meter
Fig. 2. End face of a MMF, showing three idealized launching spots into
the core and an idealized and resulting DMD plot. Leading and trailing
edge times (25%) threshold are identified with a “+”.
OFC/NFOEC Conference
bare fiber holders
gap
VCSEL source
MMF end face
Launching
spot 2
Launching
spot 3
optical power with offsets
Launching
spot 1
launch SMF
Fig. 3. Summary of the experimental layout.
3
2005 Technical Proceedings
the elements in this array (based on optical power, µW) to
determine an accurate measurement of x = 0 and y = 0.
The raw signal from the optical input of the high frequency
scope was noisy. Mainly this was due to the low coupling
efficiency of the VCSEL into the SMF as mentioned above.
Digital filtering was used to reduce much of this noise. An
example is shown in Fig. 4. For each interior point of 4096
raw points the optical power was averaged over a window
of 200 consecutive points, centered on the raw point. Near
the boundaries the averaging was asymmetric and included
only those points within the overall time range. For that
portion of the filtered optical power that was beyond the
range of each pulse, the background structure was deleted
and set to zero. The results were thus uncontaminated by
background.
Fig. 4. An example of the digital filtering used for the DMD
measurements.
holder. The MMF under test is located with a bare fiber
holder mounted on a stationary fiber holder. The gap (less
than 10µm) between the two fiber end faces is where the
DMD offset distances are defined. The axis of the SMF
output beam is perpendicular to the end face of the MMF to
within 1.0°.
Methods from ref. [1] were used to determine DMD in the
time domain and are summarized here. The launch SMF
was positioned at offset launch locations of x = 0, 2, …,
24µm and y = 0µm. At each location raw optical pulses
from the MMF under test were captured by the scope (see
Fig. 3). After digital filtering (Fig. 4) the time locations of
25% of the maximum optical power of each pulse were
determined, marked by +’s in Fig. 5. From the complete
set of +’s the left most “+” and the right-most “+” were
determined and are indicated by thicker dashed lines in Fig.
5. This difference defines the time between the leading
edge of the fastest resultant pulse and the training edge of
the slowest resultant pulse. The pulse width due only to the
Fiber type 2
Fiber type 3
24
26
26
22
24
24
20
22
22
20
20
8
6
4
2
12
10
8
6
14
12
10
8
6
4
4
2
2
time, ps
time, ps
1000
800
600
400
200
1000
800
600
400
0
-2
200
0
-2
0
0
-2
time, ps
1000
10
14
16
800
12
16
18
600
14
18
0
16
0
offset launch distance, µ m
18
offset launch distance, µ m
28
offset launch distance, µ m
28
400
Fiber type 1
26
200
Positioning the axis of the SMF congruently with the axis
of the MMF requires a detailed procedure (see “Launching
spot 1” in Fig. 2). A coarse location is first determined by
finding the (horizontal, vertical) edges (x, y) using the
optical power meter and the x-y-z precision location
control. Based on this estimate of x = 0 and y = 0, a matrix
of measurements is taken with a step size of 5.0µm for both
x and y directions. Numerical methods are used to weight
Fig. 5. DMD measurements for the three types of fiber. See the text for details.
OFC/NFOEC Conference
4
2005 Technical Proceedings
fiber can be determined by subtracting the width of the
launch pulse from this resultant difference. The pulse
width due only to the fiber per unit length, DMD in units of
ps/m, is given as follows: [(time distance between these
two dashed lines) – (time between corresponding launch
pulse points)] / (fiber length). For example, for Fiber type
1 (left-most “+”) – (right-most “+”) = 525ps. The time
between the 25% maximum points of the launch pulse is
229ps. The fiber length is 2438m so that DMD = (525ps229ps)/2438m = 0.121ps/m.
Here discrete p0(fk) values are the N Fourier transform
elements of the P0 values, given in Eq. (2), where k = 1, 2,
…, N and where fk are the N frequencies. The Fourier
transform function, FT{…}, can be computed in
Microsoft® Excel with the Fourier Analysis tool, for
example. In Eq. (3) the subscript k was deleted from the
time, t, dependency to indicate that all the P0(t) values are
used to calculate each p0(fk) value. The magnitude of each
complex element within the array of N Fourier transform
values is squared to give each p0(fk) value to convert from
electrical to optical power (e.g., squaring the IMABS(…)
function in Excel where the arguments are the elements of
the FT{…} array).
Longer fiber lengths are typically used for these types of
DMD measurements. The lengths used in this study were
chosen to balance our requirements for sensitivity to the
various DMD pulse delays and our corresponding BER
measurements.
The relationship between time and frequency is
∆f = 1 ( N ⋅ ∆t )
4.2 Methods for MMF bandwidth measurements
where ∆t is the time step size (in sec) between the
successive N times, tk, and ∆f is the corresponding
frequency step size (in Hz) between the successive N
frequencies, fk.
Following the notation of ref. [1] (Sections 6.2.1 and 6.2.2),
let U(rj,tk) be the digitally filtered output pulse where U is
optical power (in units of Watts), rj (j = 1, 2, …, M) are the
M offset distances (µm) between the axis of the launch
SMF and the axis of the MMF under test (see Fig. 3), and tk
(k = 1, 2, …, N) are the N discrete times (sec). Note that
U(rj,tk) is in the time domain. For simplicity here let W(rj)
be the mean of the 10 VCSEL weighting factors, given in
Annex D of ref. [1], simulating a typical VCSEL. This
method can be readily expanded to include the full set of
weighting factors.
Similarly the reference pulse in the time domain is
converted to the frequency domain as given in Eq. (5).
r ( f k ) = FT {R ( t )}
M
p0 ( f k ) = r ( f k ) ⋅ h ( f k )
(2)
From Eqs. (2) and (3) and the associated measurements the
p0(fk) values are determined. Using Eq. (5) and its
associated measurements the r(fk) values are determined so
that the h(fk) values can found from Eq. (6):
h( f k ) = p0 ( f k ) r ( f k )
To represent our VCSEL we took the mean of these matrix
elements at each offset distance to weight our results, using
W(rj), (j = 1, 2, …, M). Summing over all M offsets as
described in Eq. 2, gave a resultant pulse in the time
domain (N = 4096 values in units of Watts vs. ps) for each
of the three MMF types. Convert P0(tk) in the time domain
to the frequency domain using Eq. (3):
OFC/NFOEC Conference
(6)
where h(fk) is the fiber frequency response, also called the
fiber transfer function. This implies that the resultant
function, p0, is comprised only of the reference pulse, r,
and the fiber frequency response, h. The effects due to the
connectors and other sources that may affect p0 are ignored
here.
Also in the time domain let R(tk) be the reference pulse that
is launched into the MMF under test. This is measured and
filtered in the same manner as the U(rj,tk) values. In Fig. 3
note the alternate path directly from the VCSEL source to
the optical input of the scope to allow an unfiltered
measurement of R(tk).
p0 ( f k ) = FT {P0 (t )}
(5)
From the Convolution theorem [6, 7] we have
A resultant output temporal response function, P0(tk),
simulates all significant mode groups that are simulated by
the typical VCSEL (through the W values) and the MMF
under test (through the U values).
P0 (tk ) = ∑ j =1W (rj ) ⋅ U (rj , t k )
(4)
(7)
The normalized functions for p0(fk), r(fk), and h(fk)
[p0(fk)/p0(f1), r(fk)/r(f1), and h(fk)/h(f1)] are plotted in Fig. 6.
The effective modal bandwidth (EMBc) is found by
locating that frequency, fb, where h(fk) is 1.5 dB down from
the zero frequency, f1. The EMBc frequency, fc, is given by
fc = 1.414. fb. In practice fb must be found by interpolation
within the h(fk) array, e.g., a linear interpolation. EMBc
typically has units of MHz.km. This means that fc, in Hz, is
to be multiplied by the length of the MMF under test in km
and divided by 106 Hz/MHz.
(3)
5
2005 Technical Proceedings
specified BER for a minimum receiver power. We
characterize the performance of the fiber under test by
measuring the BER as a function of receiver average
power. The input power to the optical receiver is
incrementally reduced by means of an external variable
optical attenuator (VOA). For compliance we have
confirmed that our 50/125µm MMF under test performs to
a maximum operating range. For reference, a baseline
BER curve is routinely generated for a reference MMF.
1.1
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
r(fk)
p0(fk)
For the reference MMF we achieve a specified BER for an
average optical receiver power that exceeds the minimum
system requirement.
It is important to note that
performance is highly dependant on the characteristics of
the transmitter output pulse, which has been adjusted to
meet the requirements. It was observed that seemingly
small changes in output launch conditions have large
measurable effects on BER performance.
4.5E+09
4.0E+09
3.5E+09
3.0E+09
2.5E+09
2.0E+09
1.5E+09
1.0E+09
5.0E+08
h(fk)
0.0E+00
Normalized power, W
Fibe r type 3
frequency, Hz
Fig. 6. An example of the three kinds of frequency responses,
described in Section 4.2
4.3 Materials and methods for BER measurements
4.4 Results
To characterize the optical fibers under test for 10Gb/s
Ethernet performance we configured a test and procedure to
emulate a 10GBASE-S Ethernet link. The method employs
a Bit Error Rate Test (BERT) system to simulate the optical
transmitter and receiver and to measure the link
performance in terms of bit error rate. The BERT is
programmed to take repetitive measurements over extended
periods of time. Modal interference effects in the optical
link are minimized by eliminating unnecessary modal
losses.
Using these methods to determine DMD in the time
domain, bandwidth, and BER; three distinct fiber types
were evaluated. All had core diameters of 50µm. Type 1
was laser optimized for 10GbE data rates. An optical time
domain reflectometer (OTDR) measurement of fiber length
yielded 2438.4m for the DMD measurements. Type 2 was
not laser optimized for 10GbE. Its OTDR length was
1019.1m. And Type 3 was laser optimized for 10GbE and
had an OTDR length of 1016.1m. A summary of those
measurements, including digital filtering, is shown in Fig.
5.
The fiber under test is connected to the BERT using two
patch cords with their cladding modes suppressed. The Bit
Error Rate Test configuration is shown in Fig. 7 where TP
refers to test point and (Tx, Rx) refer to optical (transmit,
receive) functions. A pattern generator drives the VCSEL
transmitter. The output power levels of the transmitted
signal were compliant to IEEE 802.3ae [8].
5. DISCUSSION AND CONCLUSIONS
From Fig. 4 it is clear that digital filtering is required to
significantly reduce the high frequency component of the
noise in the raw measurements. Smooth curves are
required to measure DMD where the leading and trailing
edges of the slowest and fastest points are analyzed to
determine tfast and tslow.
The required performance for a 10GbE system is a
Pattern
Generator
VCSEL
Tx
In Fig. 5 we see that the noise is greater for Fiber type 1
than for either type 2 or type 3. This is due to the length of
type 1, 2438.4m for the DMD measurements while type 2
was 1019.1m and type 3 was 1016.1m. Comparing Fiber
type 2 with Fiber type 3 in Fig. 5, we see that the pulses for
Patch Cord
TP1
Patch Cord
Fiber type 3 line-up nearly perfectly, corresponding to a
small DMD for Fiber type 3. The pulses for Fiber type 2
are clearly displaced from each other as a function of offset
launch distance, corresponding to clearly not having 10GbE
performance at 300m as shown in Fig. 8 (discussed below).
Variable
Optical
Attenuator
TP2
Each of the three fiber types was terminated with
connectors such that the total fiber length was 300m. BER
measurements were made on all three fiber types and on
control and baseline samples. Under automatic control all
Optical
Rx
`Fig. 7. Bit Error Rate Test Configuration.
OFC/NFOEC Conference
6
2005 Technical Proceedings
Average BER for 3 fiber types
which satisfy the 10GbE requirements, we see that the
correlation with measured DMD and measured bandwidth
is not perfect. From our experience BER results are
sensitive to connector types and blemishes at the connector
interfaces, which translate into mode selective losses within
the fiber link. We suspect this cause for this type of
disagreement.
1.0E+00
1.0E-01
1.0E-02
1.0E-03
1.0E-04
Fiber type 1
BER
1.0E-05
1.0E-06
Fiber type 2
1.0E-07
Fiber type 3
1.0E-08
Generally, we can conclude that there is a less than perfect
correlation between BER and measured DMD and
bandwidth. DMD measurements mostly depend on modal
dispersion but are not so responsive to other causes that
limit bandwidth (e.g., inter-symbol interferences). On the
other hand, BER is a realistic measurement of bandwidth
that includes most, if not all, causes of bandwidth
limitations. For the end customer the performance of the
fiber link is most importantly determined by BER and data
rate.
1.0E-09
1.0E-10
1.0E-11
1.0E-12
-7.0
-7.5
-8.0
-8.5
-9.0
-9.5
-10.0
-10.5
-11.0
-11.5
-12.0
-12.5
-13.0
-13.5
-14.0
-14.5
1.0E-13
Received power, dBm
Fig. 8. BER measurements for 3 fiber types. The two curves with solid
markers meet the requirements for 10GbE.
6. ACKNOWLEDGMENTS
results were acquired and subsequently analyzed and
plotted (see Fig. 8).
The contributions of Bo Wang are acknowledged who
prepared many of the samples and took the BER
measurements. Manho Chung also is acknowledged for
preparing other samples for the DMD measurements. Ken
Reeder arranged for the availability of several of the
samples. A discussion with John S. Abbott, Corning Inc.,
was very helpful.
The results are summarized in Table 1 where it is seen that
for smaller DMD results (ps/m) we have higher bandwidths
(MHz.km). Within the MMF under test if the pulses that
propagate within the available modal groups all arrive at
the scope at nearly the same time are indicated by small
DMD values which translate into higher bandwidth
capacity. Our results are consistent with this interpretation.
The last column compares our measured results with the
standard [6] which requires that the MMF transmit data at
10Gbits/s with BER ≤ 10-12 and received power ≤ -9.9dBm.
Fiber types 1 and 3 pass this requirement but Fiber type 2
does not.
7. REFERENCES
[1] “FOTP-220: Differential Mode Delay Measurement of Multimode
Fiber in the Time Domain,” TIA/EIA Standards Document, TIA/EIA 455220-A, (January 2003).
[2] M. Webster, L. Raddatz, I. H. White, and D. G. Cunningham, “A
statistical Analysis of Conditioned Launch for Gigabit Ethernet Links
Using Multimode Fiber”, J. Lightwave Technol., 17, 1532 – 1541 (1999).
For Fiber type 2 the BER measurement was much different
that the BER measurement for the other two fiber types
(see Fig. 8). This is consistent with the measured DMD
result for Fiber type 2 that is larger than the other two fiber
types and the measured bandwidth for Fiber type 2 that is
smaller than the other two fiber types.
[3] L. Raddatz, I. H. White, D. G. Cunningham, M. C. Nowell, “An
experimental and theoretical study of the offset launch technique for the
enhancement of the bandwidth of multimode fiber links”, J. Lightwave
Technol., 16, 324 – 331 (1998).
[4] S. E. Golowich, W. A. Reed, “Technique for measuring modal power
distribution between an optical source and a multimode fiber”, U.S. Pat.
6,788,397, (2004).
Comparing the BER results for Fiber types 1 and 3, both of
[5] J. B. Schlager and D. L. Franzen, NIST Symposium on Optical Fiber
Measurements, PP. 127 – 130, (1998).
Table 1. Comparisons of measured DMD,
measured bandwidth, and BER for three types of
MMF’s.
[6] J. B. Thomas, “An Introduction to Statistical Communication Theory,”
New York: Wiley, 1969, p. 143.
At BER =
DMD meas.
meas.
10-12,
received
MMF length, DMD bandwidth BER
.
MHz
km
type
m
ps/m
length, m power, dBm
1
2438 0.121
2830
300
-10.2
2
1019
0.254
1404
300
No signal
detected.
3
1619
0.114
2451
300
-11.3
OFC/NFOEC Conference
[7] W. H. Press, et al., “Numerical Recipes: The Art of Scientific
Computing,” Cambridge, UK: Cambridge University Press, 1986, p. 383.
[8] IEEE 802.3ae – 2002, Part 3: Carrier Sense Multiple Access with
Collision Detection (CSMA/CD) Access Method and Physical Layer
Specifications (2002).
7
2005 Technical Proceedings
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF

advertisement