A 93% contrast is obtained with the He-Ne laser. The main source of
contrast variations with time comes from temperature gradient and
mechanical constraints on the input fibers. When special care is taken to
avoid fiber bending and twists, the laser contrast variation is lower than
7 rms over a week. Using high-birefringent fibers which are far less
sensitive to mechanical stresses will improve the contrast stability.
With an halogen white-light source, the contrast obtained is of the order
of 7 with a potassium ion beam combiner connected with low-birefringent
fibers (Fig. 2b) and 78
with a silver ion beam combiner
connected with high-birefringent fibers (Fig. 2c). Two main
sources of interferometric contrast drops between the two components have
been identified: chromatic dispersion and polarization mismatch.
The consequence of residual differential dispersion between the two
arms is to spread out the fringe envelope and decrease the contrast.
Since the delay line translation is not perfectly linear, the Fourier
relation between space and time is affected and an accurate estimate
of the dispersion is difficult. Only the number of fringes and the
shape of the interferogram gives an idea of the existing differential
dispersion. The theoretical number of fringes is given by the formula
and the interferogram
contains about 13 fringes. Such a spread is not sufficient to
explain the contrast drop between the laser- and white-light
interferograms. More detailed studies of residual chromatic dispersion
are in progress.
In the present case the contrast decay is mainly explained by
differential birefringence. Low-birefringent fibers are known to be
highly sensitive to mechanical constraints and temperature changes,
leading to unpredictable birefringence. Coupling between polarisation
modes can occur leading to a contrast loss which can be worse for
unpolarized incident light (case of the thermal white-light source).
This is confirmed by the preliminary results obtained with
high-birefringent fibers and the incident light polarized along the
neutral axes: the contrast reaches 78 (Fig. 2c).
The apparent asymmetry of the interferogram could be due to residual
differential polarization and/or dispersion. Full characterizations
are in progress and will be reported in Paper III [Haguenauer et al. (1999)].
Figure 3 summarizes the photon losses in the two components. We express the losses in terms of remaining photons when 100 incoherent photons are injected at each waveguide input. For the component made from potassium ion exchange, we obtain 20 and 14 photons on each photometric channel and less than 20 photons in the interferometric channel leading to a total of 54 photons for 200 photons injected, hence a total throughput of 27%. For silver ion exchange, respectively 30, 31 and 25 photons have been measured leading to a throughput of 43%. The main difference between the two results comes from the coupling efficiency between the fiber and the waveguides and the propagation losses.
|
Table 1 summarizes estimation of losses coming from different
origins. The propagation losses and the coupling losses have been measured
with a straight waveguide manufactured in the same conditions. The Fresnel
losses have been theoretically estimated to 4. Any function causes
additional losses which cannot be evaluated separately but have been
estimated to 10
. One should notice that the reverse Y-junction acts as
only one of the two outputs of an optical beamsplitter (see Paper I).
Therefore 50
of the light is radiated outside the waveguide. The first
two columns of Table 1 show that our measurements are
consistent with the theoretical performances computed from the different
optical losses reported in the table.
Last column of Table 1 gives an order of magnitude of expected improvement in the future. The main progress concerns the beam combination function. We should be able to retrieve the second half of the combined photons thanks to new combination schemes like X-couplers, multiaxial beam combiners or multimode interferometric (MMI) multiplexers (see Paper I) at the cost of a slight chromaticity of the function. Some components including these new functions are being manufactured and will be soon tested. The ultimate optical throughput would be around 70-80%, twice more than our current results.
Copyright The European Southern Observatory (ESO)