This section studies the principal
parameters that intervene in the reconstruction of FBG by layer-peeling. The
limits of the reconstruction methods are presented. The influence of the layer
thickness and the number of used spectral points is evaluated. The
reconstruction from a starting complex reflection response or impulse response
is analyzed from the point of view of the available dynamic range and the
influence of noise.
We have shown in section 3.3.1 that the
spectral reflectivity of homogeneous gratings can saturate when the grating
length and refractive index modulation amplitude are sufficiently important.
This effect is explained by the fact that the light components, for which the
wavelength is in the saturation bandwidth, are completely reflected before the
grating output. This effect is not limited to homogeneous FBGs.
For such gratings that have a
saturation bandwidth in reflection, the reconstruction by layer-peeling reaches
its limits as a part of the grating is not probed by all possible wavelengths.
This can be observed in the reconstruction of homogeneous gratings with
different lengths and identical refractive index modulation amplitude (Dnac
of 2×10-4), as presented in Fig. 3-24. The reconstruction of the
1 mm long grating is complete, but for the 10 mm grating, the last
2 mm show a small coupling amplitude decrease and also a small remaining
coupling amplitude after the grating output position, indicating reconstruction
errors. The layer-peeling algorithm clearly fails to reconstruct the 20 mm
long FBG as we observe that the amplitude and phase information of the coupling
coefficient are not reconstructed for more than half the grating length. We
have performed the reconstruction of the 20 mm long FBG with two spectral
resolution values and if a better reconstruction is observed for the higher
spectral resolution, the complete reconstruction also failed in this case.
Fig. 3-24 Layer-peeling reconstruction with 5 mm layer thickness of the
coupling coefficient amplitude (top) and local Bragg wavelength calculated from
the coupling coefficient phase (bottom) for homogeneous gratings of refractive
index modulation of 2×10-4
for different lengths L and number of spectral points M (N represents the
number of layers)
Fig. 3-25 Layer-peeling reconstruction of the coupling coefficient amplitude
(top) and local Bragg wavelength calculated from the coupling coefficient phase
(bottom) for the FBG2 (left) and a 10 mm long homogeneous FBG with a Dnac
of 5×10-4; the reconstruction parameters are : D = 3 mm,
M = 30 N for solid lines and D = 20 mm,
M = 10 N for dashed lines; the dashed lines are shifted for
clarity
Another illustration of this spectral
depletion effect that limits the reconstruction by layer-peeling is presented
in Fig. 3-25 where the reconstruction of a 10 mm long and period
chirped FBG with constant Dnac (FBG2) is compared to the reconstruction of a FBG
with the same parameters but without the chirp. The homogenous grating
reconstruction falls down after 3 to 4 mm, while the complete chirped
grating can be reconstructed. The chirped grating has a smaller maximal
reflection amplitude and a broader bandwidth that prevents the complete
depletion of a particular spectral bandwidth.
Recently, it has been demonstrated that
the worst-case error amplification factor in reconstructing a grating from its
complex reflection spectrum by layer-peeling is of the order of 1/Tmin,
where Tmin is the minimum transmission amplitude through the grating
[3-20].
The limitations of the reconstruction
for not too strong FBGs can be partially overcome in two ways. The first
possibility is to reconstruct the grating from both sides and to combine only
the first half of the reconstruction distributions. The second alternative is
an experimental method that is a consequence of the results presented in Fig. 3-25,
where a predefined chirp function is applied to the grating by a temperature
ramp or a strain ramp. The applied chirp needs to be sufficient to reduce the
maximal reflectivity of the grating.
|