International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B-YF. Istanbul 2004
5) apodization to avoid "Gibbs effect" (ringing);
6) inverse cosine transform application to retrieve at-sensor
radiance spectrum.
An ideal interferogram, therefore, is symmetric with respect to
the zero path difference and contains cosine contributions only.
However, a misaligned sampling grid, which doesn't match the
region of zero path difference, results in an evident asymmetry
in a real interferogram profile as it happens in Figures 7-8.
The spectrum of the employed laser source has been retrieved
(result of inverse cosine transform), and is plotted in Figure 9,
where we have not adopted any apodization technique.
0.045
—0— Laser HeNe spectrum
0.035
0.025
0.015
Intensity (u.a.)
0.005
-0.005400 500 600 " 700 800 900 1000
Wavelenght (nm)
Figure 9. Uncalibrated spectrum of radiance retrieved from the
interferogram depicted in Figure 7.
As already stated this type of interferometer needs neither a slit
nor an aperture, leading to much larger flux of light available
than that from traditional dispersive spectrometer. However, the
resolving power of this system R=KÄ/max/Sk& is determined
only by the number N of photosensitive elements of the detector
array. This is the main reason that has led us to employ a 1024-
elements array.
Moreover, for a continuous spectrum, most of the radiation
energy is concentrated in the central fringes of the interferogram
but the physical information more relevant is distributed on the
wings of the interferogram. Therefore, the detector needs a large
dynamic range, which, in turns, determines the maximum
variability of signal over the image.
The spectral resolution seems to take great advantage from this
experimental set-up. Due to the circumstance that the employed
spectral source may be approximated to an impulse-like
radiation field, the measurement above described is also a good
test to estimate the instrument spectral resolution &k , which
obeys the following law (Persky, 1995):
ök = — | (7)
20PD max
The obtained spectral resolution is about 175 cm (which
corresponds to 7 nm at 632 nm) across the entire bandwidth.
According to (4) the related maximum optical path difference is
about 30 um.
However, we believe this result could be improved because
spectral resolution has been lost due to two main troubles which
affected the shown data. The first one concerns with the
circumstance that the Nyquist cut-off frequency of the input
spectrum as filtered from the detector spectral sensitivity did not
134
match the available sampling frequency, thus generating
aliasing in the blue side of the spectrum. The second trouble
reason is related to the adopted optical configuration. We have
verified that a different distance between the two folding
mirrors can improve the obtained spectral resolution.
In Figure 10 six frames of a brief sequence of measurements
(images of the sky and landscape around Florence) are shown.
Let us note how the scene details seem to move from left to
right while the pattern of interference fringes remains fixed.
i
: IT p) m
Figure 10. Six frames of a brief sequence of measurements
(images of the sky and landscape around Florence city).
As can be seen, the number of interference fringes is smaller
than that obtained in previous laboratory measurements due to
the poor coherence degree of the solar radiation.
Data processing is currently under evaluation in order to
retrieve physical properties from these images.
6. CONCLUDING REMARKS
In this paper a new stationary imaging interferometer operating
in the Sagnac configuration (developed at our laboratory) has
been discussed. Experimental activity has been carried out in
order to calibrate the instrument response and to measure its
spectral resolution. Some problems connected to data
processing methodology have been discussed regarding in
particular dark-signal removal, DC offset compensation, inverse
cosine transform routine implementation, and noise filtering.
