Characterization of High Speed Optical Detectors for Purpose of
OM4 Fibre Qualification: Selective Mode Detection
F. J. Achten
1
and D. Molin
2
1
Prysmian Group, Zwaanstraat 1, 5651CA Eindhoven, Netherlands
2
Prysmian Group, Parc des Industries Artois Flandres, 62092 Haisnes Cedex, France
Keywords: High Speed Optical Fibre, High Speed Detector, Optical Fibre Qualification.
Abstract: The most important characterization of OM4 fibre is the ‘Differential Mode Delay’ (DMD) measurement.
The measured DMD profile is some kind of ‘roadmap’ of the OM4 fibre; it contains all relevant data from
which the main optical parameters are computed (for instance ‘Effective Modal Bandwidth’). Most
requirements for the measurement equipment are well defined within standardization documents. However
the requirements for the detector are still under discussion. This paper shows the state of the art of
commercially available detectors (high speed optical electrical converters, fibre coupled) from different
manufacturers. A method to characterize these detectors is suggested, and it shows the ‘ideal’ detector is not
yet commercially available.
1 INTRODUCTION
In today’s world the need for high speed data
transmission is growing. In high performance data
centers multimode fibres are commonly used. Today
the operational wavelength is 850 nm, but ‘wide
band multimode fibres’, covering the range 850 nm
to 950 nm (likely to be OM5) are being developed
and tested in systems (CommScope, 2015, Molin
2014). Some publications exist to discuss the
specifications that wide band fibre should reach 950
nm (Pimpinella, 2014, Bigot, 2015)). These fibres
are qualified by a tightly standardized measurement
method called ‘Differential Mode Delay’ (DMD)
(Oh, 2012, TIA, 2003). The main parameter
considered is the ‘Effective Modal Bandwidth’
(EMB). The EMB is computed from the ‘un-
normalized DMD profile’ (TIA, 2003, IEC, 2017).
The specification for the EMB value becomes more
relaxed when the wavelength increases from 850 to
950 nm. This is caused by the fact that the ‘penalty
caused by Chromatic Dispersion goes down. At 850
nm the EMB specification is 4700 MHz.km, at 950
nm EMB is specified at 2450 MHz.km (note: precise
value are still under discussion at IEC and TIA).
When the fibre is optimized at 850 nm, the EMB
will be much higher than the specified value.
However, when the range 850 nm to 950 nm is
considered, the EMB values at 850 nm and 950 nm
will be closer to the specified values. This means the
precision of the DMD measurement method
becomes more relevant. On the launching side (side
of the fibre where the laser pulses are launched into),
the laser pulse characteristics and the exact position
and size of the launch spot are very relevant. On the
detector side, two properties are relevant. First of all
the detector must be sufficiently fast to detect small
changes of the laser pulse shape after travelling
through the fibre under test. If fibres are measured
close to the fibre length used in systems, the laser
pulse must be very narrow in time domain (and also
in wavelength domain). A typical value for the Full
Width Quarter Max (FWQM) of the laser pulse is 10
ps. A typical value for the detector bandwidth to
proper detect this pulse is 25 GHz. The FWQM of
the pulse detected by such a detector will be typical
35 ps (referred to as ∆T
REF
).
Secondly, the detector must detect all the modes
guided by the fibre under test. The detector response
should be close related to the actual spatial and
angular power distribution leaving the fibre under
test. High speed detectors often tend to have a
sensitive area smaller than the field leaving the fibre
under test. The speed of the detector drastically
reduces when the diameter of the sensitive area gets
larger (Hui, 2012). To catch all the light that leaves
the fibre under test, the internal fibre pigtail of the
Achten, F. and Molin, D.
Characterization of High Speed Optical Detectors for Purpose of OM4 Fibre Qualification: Selective Mode Detection.
DOI: 10.5220/0006541201410147
In Proceedings of the 6th International Conference on Photonics, Optics and Laser Technology (PHOTOPTICS 2018), pages 141-147
ISBN: 978-989-758-286-8
Copyright © 2018 by SCITEPRESS – Science and Technology Publications, Lda. All rights reserved
141
detector may consist of a tapered or lensed core. The
consequence of such a structure is some light will
leave the internal fibre pigtail from the side, and so it
will not reach the sensitive area of the detector. This
causes selective mode detection’. Sometimes the
internal fibre pigtail is bended to fit the small
housing. This will introduce macrobend losses. As
far as we know, no commercial available detector
exists that uses a ‘bend insensitive’ fibre design as
the internal fibre pigtail. Another cause of selective
mode detection may be imperfections of the
sensitive area. If the detector suffers from selective
mode detection, the measured DMD profile of the
OM4 fibre under test is not theoretically correct, and
as a consequence the computed EMB value isn’t
correct either.
This paper describes a method to characterise a
detector, so it can be used to qualify OM4 fibre. The
method is tested on five different detectors from
different manufacturers. These detectors are all
specified for multimode fibre, with a very high
optical bandwidth of at least 10 GHz. A very high
bandwidth is mandatory to qualify short length of
OM4 fibre, for instance close to the maximum
system length of 400 m. All detectors are fibre
pigtailed with a 50 or 62.5 µm core Graded Index
fibre. Also a low bandwidth detector is used as a
reference. This detector cannot detect the individual
pulses (too slow), but can measure the power
distribution on a near DC level. Two of the used
detectors were developed for 850 nm pulse
measurement (so are not suited to qualify wide band
multimode fibres up to 950 nm). For now the
detectors are investigated at 850 nm, but the same
method (for the 950 nm sensitive detectors) can be
used for other wavelengths in the range 850 - 950 nm.
2 CHARACTERIZATION
METHOD
To characterize a detector, a special designed optical
fibre is used. Referred to as ‘mode separating fibre’.
Regular multimode fibres are designed with an
‘Alpha Profile’ refractive index core (Oh, 2012). If
the Alpha value (α) is optimized for a particular
wavelength (OM4: 850 nm), and the refractive index
profile is very accurate, the EMB reaches very high
values at that particular wavelength. This means all
launched modes reach the detector at the same time
after travelling through the fibre. The pulses are
shaped exactly equal as launched into the fibre
(assuming a short fibre length of max 1 km, so
Chromatic Dispersion effects can be neglected). A
typical value for α is 2.0. The special designed
optical fibre (‘mode separating fibre’) has an α of
1.6. The profile is shown in Figure 1.
Figure 1: refractive index profile of the special designed
fibre (’mode separating fibre’).
Because of the low α, the mode groups experience
different times of flight through the fibre. This can
be clearly seen on the resulting (normalized)
measured DMD profile of Figure 2b (used is a
Titanium Sapphire Mode Locked laser, 10 ps pulses,
850 nm). Figure 2a shows a simulated DMD profile
on an ‘ideal’ α = 1.6 fibre. The simulation model is
described in (Gholami, 2011). The method (and
equipment requirements) to get a fibre DMD profile
is described in detail in TIA and IEC documents
(TIA, 2003, IEC, 2017). The 18 mode groups reach
the fibre end at clear different moments in time. So
one can derive the positions in time of the mode
groups leaving the fibre, and the power of the pulses
present within each mode group launched at
different radial offset positions. If the α = 1.6 fibre is
‘perfect’ (equal to the input to the simulation
model), the difference between the detected DMD
profile and simulated DMD profile is a measure for
detector performance. The closer both DMD
profiles, the better detector performance (it then
detects all mode groups leaving the α = 1.6 fibre).
The used length of the α = 1.6 fibre (mode
separating fibre) is 550 m.
a b
Figure 2: simulated and measured DMD profile after 550
m of the ’α = 1.6 fibre’ (mode separating fibre). 18 Mode
Groups are visible and clearly separated in time.
PHOTOPTICS 2018 - 6th International Conference on Photonics, Optics and Laser Technology
142
3 MODE SEPARATING FIBRE
In Graded Index (GI) multimode fibres, the modes
are sorted into mode groups. Modes of the same
mode group exhibit nearly the same group index and
the group index difference with its neighbour. Thus
the time of flight differences are nearly the same for
all mode groups. This difference of time of flight
between consecutive mode groups is at first order a
function of the alpha value (α), the numerical
aperture (NA) or Delta (Oh, 2012), the core diameter
and the wavelength of operation (λ). And at second
order a function of the dopant content (full
Germanium, full fluorine or a Germanium &
Fluorine co-doping).
For the mode separating fibre, the α value is
adapted so that the mode groups experience a clear
different time of flight when travelling through the
fibre. Visible by performing a DMD measurements
at a specific minimal fibre length. This condition
can be expressed as follows:
|
Δ𝜏
|
∙ 𝐿
ΔT
𝑅𝐸𝐹
> 𝑋
Where Δτ is the time delay difference between
consecutive mode groups in ps/m, L is the minimum
fibre length to use in the DMD measurements in m,
ΔT
REF
is the Full Width Quarter Max (FWQM) of
the ‘reference’ pulse used in the DMD measurement
in ps (the detected pulse launched to the fibre), and
X is a ‘threshold’ that is larger than 4.
For a 50 µm core GI multimode fibre with an
NA = 0.200, one can approximate Δτ as follows:
Δ𝜏
(
𝜆, 𝛼
)
= 𝑃
00
+ 𝑃
10
∙ 𝜆 + 𝑃
01
∙ 𝛼 + 𝑃
11
∙ 𝜆 ∙ 𝛼
+ 𝑃
02
∙ 𝛼
2
The constants P
xy
are computed using a simulation
model (Gholami, 2011), the result is shown in
Figure 3a. Figure 3b shows the pulses leaving the
fibre, referring also to Figure 1. When α = 1.6 → Δτ
= 0.36 ps/m, so the delay between 2 mode groups
after 550 m fibre is 200 ps.
The total power of all mode groups leaving the
fibre depends on the launch radial offset. While
approaching the edge (25 μm) of the core, the power
will go down (some pulse energy is launched in the
cladding, and power is lost because of core-cladding
interface artefacts and slight fibre bending). The
distribution of the power over the core radius
depends also on the length of the fibre.
Figure 3: (a) 3-D plot showing which α suits best for the
mode separating fibre = 1.6), used at a specific
wavelength, to reach a clear separation of the mode groups
after travelling through the fibre (b).
Figure 4 shows the theoretical power distribution
over a Graded Index 50 μm core fibre. From now on
referred to as ‘Shape’.
Figure 4: Theoretical power distribution (’Shape’) over a
Graded Index 50 μm core fibre. The distribution changes
with fibre length.
Note: the shape of the trace shows a dip in the
centre as the fibre gets longer. This is because of the
higher Ge content near the centre of the core, so
attenuation increases faster with length for light
travelling through the centre region of the core.
The mode separating fibre is a useful tool to
characterize a detector on selective mode detection,
to verify whether the detector is suited to qualify
OM4 fibre; detect DMD profiles with maximum
accuracy.
4 THE DETECTORS
According to the DMD measurement standard (TIA,
2003, IEC, 2017), the detected launch pulse should
have a FWQM below a certain value. This value
(∆T
pulse
) depends a.o. on the spectral width of the
laser (δλ), the chromatic dispersion of the fibre
(D(λ)) and the length of the fibre. The lowest DMD
Characterization of High Speed Optical Detectors for Purpose of OM4 Fibre Qualification: Selective Mode Detection
143
values to be measured with sufficient accuracy
depends on T
pulse
. For OM4 fibre the specification
is 0.10 ps/m. Figure 5 shows the relation between
fibre length and the minimum required value for
T
pulse
(for the used laser and fibre: δλ = 0.1 nm,
D(λ)= 95 ps/nm.km). The experiments are done on
the mode separating fibre with a length of 550 m.
Figure 5: laser pulse and detector requirements to qualify
OM4 fibre at different lengths.
A total of five commercially available detectors is
studied. It is not the intention to review these
detectors, and promote the best performing detector.
The intention is to review commercial availability of
very high speed multimode detectors that are
mandatory for OM4 fibre qualifications when short
fibre length are measured. For instance 400 meter,
the maximum system length. The five detectors are
all commercially available (high speed optical
electrical converters, fibre coupled, some less easy
to find), and will be left anonymous. The exact
internal structure of some of these detectors is
unknown.
Detector #1 is an integrated optical-electrical
module (responsivity in A/W is not specified). The
detectors #2 and #5 use an internal tapered pigtail;
the sensitive detection area is near 30 µm diameter.
Detector #3 is special because it is very fast (45
GHz), despite using a ‘direct’ coupling of the 50 µm
internal pigtail to the sensitive detection area
(without internal fibre lensing or tapering). The
relatively large area must be thin to reach the high
detection speed, and as a consequence the
responsivity is very low (Hui, 2012).
For instance, from Figure 5, it shows detector #1
is too slow (10 GHz), case an OM4 fibre is
measured at maximum system length (400 m), the
specified detector bandwidth should then at least be
20-25 GHz. Only four of such detectors (multimode,
20 GHz) are commercially available to date and to
our knowledge.
5 MEASUREMENT RESULTS
5.1 Reference
First objective is to verify the theoretical Shape
using 550 m of the α = 1.6 test fibre. The only way
to verify this is using an optical detector with an
homogeneous sensitive area significantly larger than
the modes field diameter of the light leaving the α =
1.6 fibre. The output of the fibre must be radiated
directly on the sensitive area of the detector, no
optics (or fibre) in between. This measurement is
done by performing a ‘DMD scan’ in 4 directions
(4-Quadrant scan: 4Q). The way to do this is well
described in the standardization documents (TIA,
2003, IEC, 2017). The laser source we use is a 10 ps
Titanium Sapphire Picosecond laser. The 4Q scan
enables accurate alignment of the fibre, so the
launch at 0 µm radial position indeed is at the optical
centre of the fibre. After the DMD scan, the 4
quadrants are combined, for instance by folding the
pulses at each launch radius. The pulses leaving the
fibre normally go to the high speed detectors under
test, connected to a sampling module and signal
analyser. However the large area detector is too slow
to follow the fast Ti:Sapp laser pulses, so the
incoming beam is modulated with a chopper at 160
Hz, connected to a Lock In Amplifier. At each radial
launch position the signal is measured. Finally the 4
quadrants are averaged, resulting in the results
shown in Figure 6.
Figure 6: theoretical and verified ‘DC ’power distribution
(Shape) after 550 m α = 1.6 fibre.
PHOTOPTICS 2018 - 6th International Conference on Photonics, Optics and Laser Technology
144
(a)
(b)
(c)
Figure 7: (a) reproducibility of detectors #3 and #4.
(b) Shape of all five detectors.
At the outer launch positions (23 & 25 µm from the
centre), the experiment shows higher power, we
believe this difference is caused by inaccuracies in
the fibre index profile close to the cladding
(scattering, leaky modes). Up to 23 µm theory and
experiment are almost in accordance.
5.2 Detectors under Test
The coupling of laser pulses into the α = 1.6 fibre is
realized by a direct coupling of a HP780 launch fibre
(single mode at 850 nm, launch spot diameter 5 μm)
to the α = 1.6 fibre. For both fibres, the cleave quality
is checked by an interferometric technique, and is far
below an angle of 1 degree (to avoid angular coupling
to the α = 1.6 fibre). The 4Q DMD scan (including
alignment, 850 nm) is executed five times per detector
to visualize the reproducibility of the measurement.
For detectors #3 and #4 the five power distributions
(‘Shapes’) are shown in Figure 7a. The responsivity
of detector #3 is very low, causing this detector to
have the poorest reproducibility. Figure 7b shows the
power traces (Shape) of all five detectors.
It is clear that none of these detectors reach the
theoretical Shape. Since the exact internal structure of
the detectors in unknown, it is hard to explain the
observed differences.
5.3 Discussion
The internal fibre pigtails of detectors #2 and #5 are
‘tapered’. So the pigtail output field to the detector
sensitive area is reduced to a smaller area (from 62.5
to 30 μm diameter). In some way modes are lost, and
do not reach the sensitive area (or reach the area, but
do not generate current). It makes sense the lost
modes are the modes that travel closest to the
cladding. The Shape for both detectors is nearly
equal, detector 2 is slightly closer to the theoretical
Shape (detector 2 has a slightly lower bandwidth, so
maybe the sensitive area is just a little larger).
Detector #1 shows a Shape equivalent to #2 and
#5, which might mean it is also equipped with an
internal tapered (or lensed) fibre pigtail.
Detector #3 is the oddball of the selection. It has
an internal 50 µm pigtail that is not tapered or lensed,
and -as a consequence-, uses a larger and thinner
sensitive area. So it causes a significantly lower
responsivity (Hui, 2012). Still, the Shape does not
reach theory. For the outer part (narrow shape) two
explanations might be considered: bend losses of the
internal pigtail (50 µm shows more bend loss
compared to 62.5 µm), as the pigtail is bended over
few cm inside the detector housing. We also noticed a
poor coupling of the pigtail to the sensitive area
(slight mismatch). An interesting solution might be to
use ‘bend-insensitive’ 50 µm fibre as the internal
pigtail. The ‘dip’ in the middle is more difficult to
explain. Possibly the homogeneity of the sensitive
detector area is poor because it needs to be thin to
reach the high speed (45 GHz).
Characterization of High Speed Optical Detectors for Purpose of OM4 Fibre Qualification: Selective Mode Detection
145
(a)
(b)
(c)
(d)
Figure 8: Detected mode group powers of the #3 detectors
considered at launch radii of 0 μm (centre), 9 μm (most
mode groups detected), 18 μm (relevant for ‘inner’ DMD),
and 23 µm (‘outer’ DMD).
Finally detector #4. The Shape is ‘roundish’. The
size of this detector is very small, and it is
completely sealed. The internal structure is
unknown. This detector proves to be the least suited
to qualify OM4 fibre; it shows the narrowest shape
of all five detectors.
5.4 A Closer Look
The previous results will show as well by using a
regular OM4 fibre rather than the α = 1.6 fibre; the
differences in Shape are almost equivalent. Using
the α = 1.6 fibre however generates the ability to
verify the detector response not only per launch
offset radius, but also per mode group. To simplify
analysis, we consider detectors #2 (widest Shape),
#3 (centre dip) and #4 (narrowest Shape).
In Figure 8 the detected powers per mode group
at 4 launch offset radii are plotted together with
theoretical values derived from the simulated DMD
profiles (detector #2 shown in Figure 2a).
To scale the power levels equal, it is assumed at
9 μm radial offset, all mode groups are detected.
From Figure 2 it shows the agreement between the
simulated and measured DMD profile (detector #2)
is fair, however when approaching the cladding, the
differences increase. This is probably caused by
core-cladding interface artefacts of the α = 1.6 fibre.
Further, to optimize the model, one must know the
exact Alpha value, core-size and Delta (Oh, 2012),
and these must be very constant over full length (550
m) of the α = 1.6 fibre (here’s another challenge).
From Figure 8a, theory, ‘odd’ mode groups are
symmetric whereas ‘even’ mode groups are anti-
symmetric (Gholami, 2011). So no power at fibre
output by the even mode groups. This is well
confirmed by the experimental data.
Detector #3 fails to detect full power of the first
mode group, while the third and fifth mode group
approach theory. Probably caused by an artefact in
the centre of the detector sensitive area, which is a
reason to reject detector #3 for OM4 fibre
qualification.
Figure 8b and 8c show a typical but unexpected
result. The measured mode group power
distributions of the three detectors shift to higher
order modes compared to theory. One might expect
the opposite as higher mode groups are more
sensitive to selective mode detection. We expect this
is caused by local imperfections of the index profile
of the mode separating fibre (α = 1.6 fibre, figure 1).
This observation needs further research.
Figure 8d, when launching close to the cladding,
it clearly shows the loss of power for all three
PHOTOPTICS 2018 - 6th International Conference on Photonics, Optics and Laser Technology
146
detectors compared to theory. Main cause of the
narrow experimental Shapes shown in Figure 7b.
6 CONCLUSIONS
To date, to our knowledge, no commercial available
high speed detector (≥ 20 GHz) exists that can detect
all modes leaving the OM4 fibre under test.
We made a specially designed ‘mode separating
fibre’, and we used this fibre to characterize
performance of five commercially available high
speed detectors. The method clearly shows the
limitations of these detectors to qualify OM4 fibre.
Which does not mean these fibres will fail in
systems, it just shows the ideal detector for
qualifying OM4 fibre is not yet commercially
available. The next challenge is to explain the
observed mode selective detection by considering
the internal structure of these detectors in detail.
Another challenge is to bring simulated and
measured data closer by improving the quality
(accuracy of index profile and homogeneity over
length) of the mode separating fibre. From there
improvement to the detector design may lead to the
‘perfect’ detector for purpose of OM4 fibre
qualification. A first improvement to the internal
structure of the detector might be to use a ‘bend-
insensitive’ Graded Index multimode fibre to serve
as internal pigtail. This will cause less modes to
leave the pigtail from the side.
ACKNOWLEDGEMENTS
The authors would like to thank J.G.A. Achten,
retired teacher technical English, for reviewing this
paper.
REFERENCES
CommScope, 2015. Wideband Multimode Fiber - What is
it and why does it makes sense? (White paper)
http://www.commscope.com/docs/wideband_multimo
de_fiber_what_why_wp-109042.pdf
Molin D., Achten F., 2014. WideBand OM4 Multi-Mode
Fiber for Next-Generation 400Gbps Data
Communications, IWCS 2014.
Oh, K., Paek, U., 2012. Silica Optical Fiber Technology
For Devices And Components (Chapter Five), A John
Wiley & Sons, Inc, 2012.
Pimpinella, R., Kose, B., Castro, J., 2014. Wavelength
Dependence of Effective Modal Bandwidth in OM3
and OM4 Fiber and Optimizing Multimode Fiber for
Multi-Wavelength Transmission, IWCS 2014.
Bigot, M., Molin D., 2015. Wide-Band OM4 Multimode
Fibers for Future 400Gbps and 1.6Tbps WDM
Systems, IWCS 2015.
TIA, 2003. FOTP-220 - Differential Mode Delay
Measurement of Multimode Fiber in the Time Domain
TIA-455-22-A, January 2003.
IEC, 2017. IEC 60793-1-49 ED3 - Optical fibres - Part 1-
49 (Draft): Measurement methods and test procedures
- Differential mode delay, February 2017.
Hui, R., O’Sullivan, M., 2012. Fiber Optic Measurement
Techniques (Chapter 1, pg 35). Elsevier Academic
Press, 2009.
Gholami, A., Molin, D., Sillard, P., 2011. Physical
Modeling of 10 GbE Optical Communication Systems.
Journal of Lightwave Technology, Vol. 29, No. 1,
January 2011.
Characterization of High Speed Optical Detectors for Purpose of OM4 Fibre Qualification: Selective Mode Detection
147