Method of estimating a reflection profile of an optical channel
Field of the invention
The invention relates to a method of and a device for estimating a reflection profile of an
optical channel.
Background
These days, access networks, also called passive optical networks (PON), are used for
connecting a plurality of customers to a core network of data transportation.
In such an access network, the interconnection between the core network and the access
network is given at a so-called central office, which contains an optical line terminal (OLT).
The OLT is connected via at least one optical fiber, preferably called optical feeder fiber, to a
so-called remote node. At this remote node, an optical downlink signal transmitted by the OLT
is split onto different optical branches, to which one or more customers are connected by
means of optical network units (ONU).
The customers send via the optical branches optical uplink signals towards the remote node,
which combines these optical uplink signals to a combined uplink signal and transmits the
combined signal via the optical feeder fiber to the central office, which contains the OLT.
In order to determine the transmission properties of the transmission channel into which a
device transmits an optical signal, a measurement technique of optical time domain
reflectometry (OTDR) can be applied. In such an OTDR measurement, a reflection profile of
the transmission channel is estimated. Preferably, the technique of OTDR is carried out at the
OLT.
For the purpose of OTDR, an optical measurement signal in the form of an optical pulse may
be transmitted into the optical channel, which comprises one or more optical fibers. Such
Optical fibers are usually made of inhomogeneous material, which causes backscattering of
the optical measurement signal. The backscattered optical signal, preferably called received
response signal, may then be recorded over time as the reflection profile. Knowing the
propagation speed of the optical signals within the optical fibers, the received response signal
can be converted from the time domain to a distance.
Different imperfections of the optical channel, such as e.g. an open connector or a dirty fiberconnector,
may cause characteristic increased or decreased backscattering of the
measurement signal, which in turn may be observed as reflection peaks within the reflection
profile. By examining the reflection profile and the reflection peaks contained in the profile,
one may derive, at which distance an imperfection is present within the optical channel.
Instead of using a single optical pulse as the measurement signal, a more advanced technique
of OTDR may be employed. This advanced technique makes use of an optical signal that is
modulated in its amplitude in dependence on a correlation sequence. The received response
signal is first recorded and then used for determining the reflection profile. This is achieved, by
correlating a time-discrete version of the received response signal with the initial correlation
sequence itself. In the case, that the auto-correlation function of the correlation sequence is
equal to or approximated by the dirac delta function, the result of the correlation yields an
estimate of the impulse response of the optical channel in the time domain, which is an
approximation of the reflection profile.
When transmitting an optical transmission signal carrying transmission data into the optical
channel using a transmission device, it is one possibility to carry out the technique of OTDR, by
using a further, separate device. The transmission device and the separate OTDR device are in
this case both coupled to the same optical channel, preferably via an optical coupler.
A more advanced technique is that of embedded OTDR (eOTDR), in which the transmission
device itself contains the hardware for generating the optical transmission signal as well as the
hardware that is necessary for carrying out an OTDR measurement. Preferably, the optical
transmission signal is directly modulated in dependence on a correlation sequence, wherein
the frequency of this direct modulation is chosen, such that it does not disturb data reception
at a receiving side. After transmitting into the optical channel the optical transmission signal,
which carries the directly modulated measurement signal, the received response signal of the
optical channel can be obtained, by filtering out that frequency, at which the optical
transmission signal was modulated. This received response signal can then be used for
determining a reflection profile via the technique of signal correlation as it has been described
previously above.
Summary
Proposed is a method of estimating a reflection profile of an optical channel. The method
comprises different steps.
A measured reflection profile of the optical channel is provided. One or more reflection peaks
are estimated within the measured reflection profile.
A residual reflection profile is determined, by removing the estimated reflection peaks from the
measured reflection profile. Furthermore, a modified residual reflection profile is determined,
by modifying one or more estimated crosstalk frequency components within the residual
reflection profile.
Finally, the estimated reflection profile is determined, by superposing the estimated reflection
peaks and the modified residual reflection profile.
In order to grasp the advantages of the proposed method, the following aspects may be taken
into consideration.
When carrying out a method of embedded OTDR, the hardware components carrying out the
measurement of the reflection profile may experience an impact of crosstalk noise caused by
other hardware components present within one same device. Such crosstalk noise may
degrade the measured response signal and thus also the measured reflection profile.
Crosstalk noise may be present especially, in the case that the integration of the OTDR devices
into the transmission device is given as a Small Form-Factor Pluggable (SFP).
For obtaining a reliable estimated reflection profile, the impact of crosstalk noise needs to be
reduced. One countermeasure may be, to shield the hardware devices, in order to reduce the
amount crosstalk noise caused by electromagnetic fields generated by the hardware devices.
Such a countermeasure is on the one hand cumbersome and on the other hand may also be
expensive, due to space restrictions given for a device carrying out eOTDR.
In order to remove the crosstalk noise from the measured reflection profile, filtering techniques
may be applied. Such filtering techniques may not only have an impact on the crosstalk noise,
but also on reflection peaks present within the measured reflection profile. As it has been
previously described, such reflection peaks should remain present within a filtered reflection
profile, in order to detect imperfections of the optical channel.
The proposed method first estimates the reflection peaks and then removes these peaks, in
order to yield a residual reflection profile. Next, the residual reflection profile is filtered, such
that estimated crosstalk frequency components are modified. Due to the fact, that crosstalk
usually causes one or more spectral peaks within the spectrum of a reflection profile, the
crosstalk noise may be reduced, by modifying the estimated crosstalk frequency components.
After modifying the crosstalk frequency components, the yielded modified residual reflection
profile is superposed with the previously estimated reflection peaks.
Thus, the proposed method of estimating a reflection applies a filtering technique, in which
estimated reflection peaks are first separated from the measured reflection profile, before
separately filtering the crosstalk noise in the frequency domain, and then finally superposing
the filtering result with the preserved reflection peaks again. Thus, this filtering technique for
reducing crosstalk impact avoids a severe degradation of the reflection peaks present within
the finally estimated reflection profile. Therefore, the estimated reflection profile contains well
preserved reflection peaks. Thus, the estimated reflection profile may be used for detecting
imperfections of the optical channel via the reflection peaks, as well as reliably estimating the
attenuation profile of the optical channel due to reduced crosstalk noise.
Brief description of the Figures
Figure 1 shows a flowchart of the proposed method according to an embodiment.
Figures 2a and 2b show a measured reflection profile in the linear and the logarithmic
domain.
Figure 3 shows a second derivative of the measured reflection profile in the linear domain.
Figure 4a shows estimated reflection peaks.
Figure 4 b shows a residual reflection profile.
Figure 5 shows a magnitude of a spectrum of the residual reflection profile.
Figure 6 shows a first derivative of the magnitude.
Figure 7 shows a modified residual reflection profile.
Figures 8a and 8b show an estimated reflection profile in the linear and the logarithmic
domain, respectively, plotted over a time scale on the abcissa.
Figures 9a and 9b show the estimated reflection profile in the linear and the logarithmic
domain, respectively, plotted over a distance scale on the abcissa.
Figure 10 shows a proposed device according to an embodiment.
Description of embodiments
Figure 1 shows a flowchart of the proposed method according to a preferred embodiment. In
the step SI , an OTDR trace as a measured reflection profile of an optical channel is provided.
This OTDR trace may be measured by a device carrying out also the proposed method.
According to an alternative solution, the OTDR trace may be provided by a first device to a
second device via a data interface of the second device, which receives the data of the OTDR
trace, and which then provides the OTDR trace to further sub-devices of the second device
carrying out the further steps of the proposed method.
The OTDR trace is preferably a measured reflection profile in the form of a time discrete
sampled signal. Preferably, the sampling frequency of the sampled measured reflection profile
lies within the range of several megahertz, preferably 40 MHz. Such a sampling frequency of
40 MHz corresponds to a sampling interval of 25 ns. Taking into consideration typical
propagation speeds of optical signals within optical fibers, the sampling interval of 25 ns
corresponds to a resolution in distance of 2.5 m.
The measured reflection profile RP is provided to a next step S2 of the proposed method.
Figure 2a shows an example of a measured reflection profile RP in the linear domain, wherein
a timescale corresponding to a time resolution in microseconds is shown along the abcissa.
The reflection profile RP contains in this example at least three reflection peaks PI , P2, P3, as
well as crosstalk noise, which is clearly visible as a noise signal overlaying the reflection profile
RP.
The reflection profile RP of Figure 2a is once more shown in Figure 2b as a reflection profile
RPL in the logarithmic domain. Both refection profiles RP and RPL of the Figures 2a and 2b are
plotted here as time continuous signals over a time scale of microseconds, wherein a person
skilled in the art will acknowledge, that such time continuous plots may represent time discrete
and sampled reflection profiles. Thus, the plotted values of the reflection profiles RP and RPL
can be considered as time discrete values plotted over discrete indices corresponding to the
timescales shown in the Figures 2a and 2b.
Coming back to Figure 1, reflection peaks EP present within the reflection profile RP are
estimated within the step S2. These estimated reflection peaks EP may then be stored in a
separate step S21 . The measured reflection profile RP is provided in the time domain, while
the estimated reflection peaks EP are preferably also estimated in the time domain. For
estimating the reflection peaks, a second derivative of the measured reflection profile is
determined. This may be obtained, by determining for the time discrete measured reflection
profile the signal of second differential order in the time domain. Having yielded this second
derivative, a reflection peak is estimated to be present for those time discrete indices, for which
the second derivative exceeds a defined threshold.
Figure 3 shows the second derivative SD together with the threshold Tl . The threshold Tl may
be determined, by determining the median absolute value of the second derivative SD.
Preferably, the threshold Tl is chosen to this median absolute value multiplied by a fixed
factor. The fixed factor is preferably chosen within the range between the values 1 and 10. In
a preferred embodiment, the fixed factor is chosen to the value 6.
As it was previously outlined, for estimating the presence of one or more reflection peaks,
those time-discrete indices of the second derivative are determined, for which the second
derivative exceeds the threshold. The estimated reflection peaks are then determined as those
values of the measured reflection profile RP, whose time-discrete indices correspond to or are
equal to the determined time-discrete indices.
Preferably, not only these determined values of the measured reflection profile RP are used as
the estimated the reflection peaks and then removed, but also further values of the measured
reflection profile RP are determined used as the estimated reflection peaks. These further
values are those, whose corresponding further time-discrete indices lie within predefined time
windows, which are centered around the previously determined time-discrete indices. Thus,
preferably, a combined set of values corresponding to a combined set of time-discrete indices
is used for estimating the reflection peaks, wherein the combined set of time-discrete indices is
a combination of the previously determined set of time-discrete indices and the further timediscrete
indices.
For a sampling frequency of 40 MHz for the measured reflection profile RP, a time window
has preferably a width of 8 time discrete indices, which corresponds to 200 ns or a distance
resolution of 20 m. Furthermore, the timely width of the time window corresponds to a fullwidth-
at-half-maximum of an optical pulse, when using only a single optical pulse for OTDR
measurement instead of a correlation sequence. Even furthermore, the timely width of the time
window corresponds to a timely width of a dirac-delta function resulting from OTDR
measurement using a correlation sequence.
The time window is preferably a weighting window, which defines weighting factors for the
values of the measured reflection profiles falling within the time window. According to a first
solution, the weighting window is a simple boxcar window applying a constant factor of 1 to
those value of the measured reflection profile, which fall within the time window. According to
another solution, the weighting window is a raised cosine window, which applies different
values within the range 0 to 1 to those values of the measured reflection profile, which fall
within the time window. By using such a raised cosine window, a more smooth transition of
values representing a reflection peak and values representing no reflection peak is achieved.
Figure 4a shows a number of estimated reflection peaks EP in the linear domain.
Coming back to Figure 1, the estimated reflection peaks are removed from the measured
reflection profile RP within a step S3. This yields a residual reflection profile RRP. Figure 4 b
shows the residual reflection profile RRP in the linear domain. The reflection peaks are
removed from the residual reflection profile RRP, by replacing the initial values of the
measured reflection profile at those indices, at which reflection peaks were detected, by the
mean value of the reflection profile of those indices, which are adjacent to the indices at which
reflection peaks were detected.
Coming back to Figure 1, the obtained residual reflection profile RRP is then used for filtering
out the crosstalk noise.
Figure 5 shows a magnitude S of a spectrum of a residual reflection profile. Frequency
components CT of the crosstalk noise are also indicated in Figure 5. A modified residual
reflection profile is determined, by modifying estimated crosstalk frequency components within
the residual reflection profile's spectrum.
Such modification is carried out in the step S4 of the Figure 1. The crosstalk frequency
components CT shown in Figure 5 are estimated in the frequency domain and also modified
in the frequency domain. Such crosstalk frequency components CT can be expected as
spectral peaks within the lower frequency domain, while normally no discontinuities are
expected for higher frequencies within the residual reflection profile's spectrum.
For estimating the crosstalk frequency components CT, the spectrum of the reflection profile is
determined, by transforming the residual reflection profile from the time domain to the
frequency domain. This is preferably carried out by a frequency transformation, which is
discrete in the time domain and discrete in the frequency domain. Preferably, the Fast Fourier
Transform (FFT) is used. This yields a complex discrete frequency spectrum. The magnitude S
of such a discrete frequency spectrum is the one shown in Figure 5, wherein the frequency
scale as a continuous scale is plotted along the abscissa. In this example, the number of
discrete values plotted is 4096.
For finally detecting the crosstalk frequency components CT, the complex discrete frequency
spectrum is differentiated once, preferably by determining the first order differential signal.
This yields a complex first derivative. Then, the magnitude of this first derivative is determined,
which yields real values.
Figure 6 shows the magnitude FD of the first derivative of the residual reflection profile's
spectrum. Those spectral indices of the first derivative's magnitude FD are determined, for
which the first derivative's magnitude FD exceeds a spectral threshold T2. The complex
spectrum is then modified in its complex values at those indices, which correspond to the
determined spectral indices.
The threshold T2 may be determined, by determining a median value of the first derivative's
magnitude FD. Preferably, the threshold T2 is chosen to this median value multiplied by a
fixed factor. The fixed factor is preferably chosen within the range between the values 1 and
10. In a preferred embodiment, the fixed factor is chosen to the value 6.
Preferably, the threshold T2 is determined, by determining a median value of the first
derivative's magnitude FD within a predefined spectral range FR. This spectral range FR has a
lower frequency limit LL and an upper frequency limit UL. This median value may then be
multiplied by a fixed factor. The fixed factor is preferably chosen within the range between the
values 1 and 10. In a preferred embodiment, the fixed factor is chosen to the value 6.
Preferably, the modification of the complex spectral values of the spectrum is carried out not
only for the spectral values at those spectral indices, for which the first derivative's magnitude
FD exceeds the spectral threshold T2, but also for further values of the spectrum, whose further
spectral indices lie within one or more predefined spectral windows, which are centered
around the previously determined spectral indices.
Thus, preferably, a combined set of values corresponding to a combined set of discrete
spectral indices is used for estimating the crosstalk frequency components and then modifying
these components, wherein the combined set of discrete spectral indices is a combination of
the previously determined set of discrete spectral indices and the further discrete spectral
indices.
Preferably, the spectral window has a predefined width, which is preferably of a width of eight
spectral indices. Preferably, this spectral window is a weighting window, which may in one
alternative solution be a boxcar window applying weighting factors of 1 or 0. According to an
alternative solution, the spectral window is a raised cosine window applying varying weighting
factors within the range between 0 and 1.
The corresponding complex spectral values, which correspond to the spectral indices used for
estimating the crosstalk frequency components, are modified to a respective mean value. For
one of these corresponding complex spectral values, the respective complex mean value is
determined as the mean of the adjacent spectral values, which are adjacent to the
corresponding values to be modified.
Preferably, the estimation of the crosstalk frequency components and the modification of these
components is limited to the predefined frequency range FR.
The modified complex spectrum is then transformed back from the frequency domain to the
time domain, using preferably a time discrete and discrete frequency transform as an inverse
transform. This inverse transform is preferably the inversed Fast Fourier Transform (IFFT) .
An obtained modified residual reflection profile is then provided from the step S4 to the step
S5, as shown in Figure 1. Such a n example of a modified residual reflection profile MRRP is
shown in Figure 7 in the linear domain. As shown in Figure 1, the yielded modified residual
reflection profile MRRP is then superposed with the previously estimated reflection peaks EP
within a step S5. This yields an estimated reflection profile ERP.
Figures 8a and 8 b show an estimated reflection profile in the linear domain ERP as well as in
the logarithmic domain ERPL, wherein these profiles are plotted over a time scale on the
abcissa. By comparing the initially measured reflection profile RP of Figure 2a with the
estimated reflection profile ERP of Figure 8a in the linear domain, it has to be noted, that the
amount of crosstalk noise is reduced, while the impact of the step of reducing the crosstalk
noise onto the reflection peaks is kept to a minimised amount. Looking at Figure 8a, one can
clearly see that not only the initially visible reflection peaks PI , P2 and P3 can be observed,
but also the reflection peak P4, which was previously not easily visible within the reflection
profile RP of Figure 2a.
The proposed method is of advantage, since it first separates the estimated reflection peaks
from the measured reflection profile, before then filtering the residual reflection profile, for
reducing the amount of crosstalk noise. Furthermore, by later on superposing the estimated
reflection peaks with the modified residual reflection profile, a n estimated reflection profile is
obtained, in which both aims, reducing crosstalk noise and preserving reflection peaks, is
achieved.
Figures 9a and 9 b show an estimated reflection profile in the linear domain ERPD and in the
logarithmic domain ERPDL, respectively, wherein these profiles are plotted over a distance
scale o n the abcissa.
Figure 10 shows the proposed device for estimating a reflection profile of a n optical channel
according to a preferred embodiment.
The device D contains a data interface Dl, over which data representing a measured reflection
profile can be received.
Preferably, the data interface Dl is connected via a data bus DB to a processing device P as
well as a memory device M .
The memory device M and the processing device P are operable, such that they jointly carry
out the different steps of the proposed method described in detail above.
Thus, the memory device M and the processing device P are operable, to jointly estimate
reflection peaks within a measured reflection profile, and furthermore to determine a residual
reflection profile, by removing the estimated reflection peaks from the measured reflection
profile.
Furthermore, the devices P and M are operable to jointly modify estimated crosstalk frequency
components within the residual reflection profile, for determining a modified residual reflection
profile. Finally, the devices P and M are operable, to jointly superpose the estimated reflection
peaks and the modified residual reflection profile, for determining the estimated reflection
profile.
The functions of the various elements shown in Figure 10, including any functional block
labelled as 'processor', may be provided through the use of dedicated hardware as well as
hardware capable of executing software in association with appropriate software. When
provided by a processor, the functions may be provided by a single dedicated processor, by a
single shared processor, or by a plurality of individual processors, some of which may be
shared. Moreover, explicit use of the term 'processor' or should not be construed to refer
exclusively to hardware capable of executing software, and may implicitly include, without
limitation, digital signal processor (DSP) hardware, network processor, application specific
integrated circuit (ASIC), field programmable gate array (FPGA), read only memory (ROM) for
storing software, random access memory (RAM), and non volatile storage. Other hardware,
conventional and/or custom, may also be included.
It should be appreciated by those skilled in the art that any block diagrams herein represent
conceptual views of illustrative circuitry embodying the principles of the invention. Similarly, it
will be appreciated that any flow charts may be substantially represented in computer
readable medium and so executed by a computer or processor, whether or not such computer
or processor is explicitly shown.

CLAIMS
1) Method of estimating a reflection profile of a n optical channel,
comprising
providing a measured reflection profile (RP) of said optical channel,
estimating one o r more estimated reflection peaks (PI , P2, P3) within said measured
reflection profile (RP),
determining a residual reflection profile (RRP), by removing the estimated reflection
peaks (PI , P2, P3) from said measured reflection profile (RP),
determining a modified residual reflection profile (MRRP), by modifying one o r more
estimated crosstalk frequency components (CT) within said residual reflection profile
(RRP),
determining the estimated reflection profile (ERP), by superposing said estimated
reflection peaks (PI , P2, P3) and said modified residual reflection profile (MRRP).
2) Method according to claim 1,
wherein said measured reflection profile (RP) is provided in the time domain,
comprising furthermore
estimating said reflection peaks (PI , P2, P3) in the time domain,
removing said estimated reflection peaks (PI , P2, P3) in the time domain.
3) Method according to claim 2,
wherein the step of estimating said reflection peaks (PI , P2, P3) in the time domain includes
determining a second derivative (SD) of said measured reflection profile (RP),
determining those indices of said second derivative (SD), for which said second
derivative (SD) exceeds a threshold (Tl ) .
4 ) Method according to claim 3,
wherein said step of estimating said reflection peaks (PI , P2, P3) in the time domain includes
using those values of said measured reflection profile (RP), whose indices correspond to
the determined indices.
5) Method according to claim 4 ,
wherein said step of estimating said reflection peaks (PI , P2, P3) in the time domain includes
using furthermore those values of said measured reflection profile (RP), whose indices
lie within one o r more predefined time-windows centered around said determined
indices.
6) Method according to claim 3,
wherein said threshold (Tl ) is determined, using a median absolute value of said second
derivative.
7) Method according to claim 1,
wherein said measured reflection profile (RP) is provided in the time domain,
comprising furthermore
estimating said crosstalk frequency components (CT) in the frequency domain,
modifying said crosstalk frequency components (CT) in the frequency domain.
8) Method according to claim 7,
wherein the step of estimating said crosstalk frequency components (CT) in the frequency
domain includes furthermore
determining a spectrum (S) of said residual reflection profile (RRP), by transforming
said residual reflection profile (RRP) from the time domain to the frequency domain,
determining a first derivative (FD) of a magnitude of said spectrum (S),
determining those spectral indices of said first derivative (FD), for which said first
derivative (FD) exceeds a spectral threshold (T2).
9) Method according to claim 8,
wherein the step of modifying said crosstalk frequency components (CT) in the frequency
domain includes furthermore
modifying those values of said spectrum (S), whose spectral indices correspond to the
determined spectral indices.
10) Method according to claim 9,
wherein the step of modifying said crosstalk frequency components (CT) in the frequency
domain includes furthermore
modifying also those values of said spectrum (S), whose spectral indices lie within a
one o r more predefined spectral-windows centered around said determined spectral
indices.
11) Method according to claim 10 ,
wherein the step of modifying said crosstalk frequency components (CT) in the frequency
domain includes furthermore
setting spectral values within a respective spectral window to a respective mean value,
wherein said mean value is determined as the mean of those spectral values, whose
indices are adjacent to said respective spectral window.
12) Method according to claim 8,
wherein the step of determining said modified residual reflection profile (MRRP) includes
furthermore
determining said modified residual reflection profile (MRRP), by transforming the
modified spectrum back to the time domain.
13) Device for estimating a reflection profile of an optical channel,
comprising
at least one data interface (Dl),
at least one memory device (M),
at least one processing device (P),
wherein said data interface (Dl) is operable to receive a measured reflection profile (RP) of
said optical channel,
and wherein said memory device (M) and said processing device (P) are operable, to jointly
estimate one o r more estimated reflection peaks (PI , P2, P3) within said measured
reflection profile (RP),
determine a residual reflection profile (RRP), by removing the estimated reflection peaks
(PI , P2, P3) from said measured reflection profile (RP),
determine a modified residual reflection profile (MRRP), by modifying one o r more
estimated crosstalk frequency components (CT) within said residual reflection profile
(RRP), and
determine the estimated reflection profile (ERP), by superposing said estimated
reflection peaks (PI , P2, P3) and said modified residual reflection profile (MRRP).