Data reduction

 A malfunction in the camera control computer caused the loss of 80% to 90% of the short exposures. As a consequence, the speckle noise remains 3 times more intense than expected and dominates the photon noise.
The dark-speckle algorithm is described in [Labeyrie 1995] and [Boccaletti et al. 1998]. The data reduction software provides two images. The first is equivalent to a long exposure, computed by co-adding all the short exposures containing the position of the photon-events. The second image, called the ``cleaned map'', displays the result of the dark-speckle algorithm. The number of zero-photon events occuring in binned pixels containing 2x2 camera pixels, is cumulated over successive frames. The cleaned map, thus obtained in negative form, can be inverted with a suitable gamma exponent. In principle, the cleaned image is more sensitive to the faint structures than the conventional long exposure.
Photon-counting data are affected by the so-called photon-centroiding ([Tiébaut 1994]) resulting from the limited resolution of the electronics in space and time. The counting behaviour is expected to improve with forthcoming detectors.

 

Figure: a. Negative cleaned image of the binary star $\delta$ Per, obtained in H$\alpha$ ($\lambda=653.6nm$, $\Delta\lambda=10nm$) with the $0.45\arcsec$ mask. The intensity of the image represents the number of zero photon collected with 9503 short exposures of 20ms. The hypothetic companion (arrow) is visible at the edge of the mask ($PA=202\degr$, $\rho=0.293\arcsec$) as a ``hole'' in the cleaned map, amidst the fixed speckles generated by the spider structure. At this location, the zero photon-count is lower than elsewhere.

The signal to noise ratio (SNR) is measured in the cleaned map according to Eq.(6) from [Labeyrie 1995]. The magnitude difference is derived from the CP20 long exposure by comparing the flux of the star outside the mask, and the flux of the companion when the star is under the mask.
To assess the performance of the system, one can compute the expected maximum brightness ratio theoretically reachable (Eq.(7) [Boccaletti et al. 1998]). This value only depends on the sampling (150 pixels/speckle area) and on the adaptive optics (AO) gain which is the brightness ratio between the peak of the corrected image and the halo level. The average AO gain obtained in the present run is about 15 at $0.5\arcsec$ from the star. The total integration time is assumed to be very large compared to the speckle lifetime. This leads to a 8.8 magnitude difference. However, as described in sections 3.2 and 3.3, this goal, although modest with respect to the $\Delta m=22$ goal attainable in principle with future adaptive optics, was not reached owing to practical problems. Among these was the loss of 90% of the observing data, due to a computer failure.