Previous
abstract 
Next
abstract
The U.S. Naval Observatory has conducted a program of speckle observations of double stars beginning in 1990. Over the past twelve months, several significant improvements have been made in the data acquisition system and in offline reduction methods, and two systematic errors have been detected and corrected. The observations employ an ITT image-intensified CCD camera attached to the USNO 66 cm refractor. The CCD is a TI3PCCD operating in the frame-transfer mode. Frames are acquired at the standard video rate of 33 frames per second. The autocorrelation function of a double-star image is computed in real time using the directed vector autocorrelation (DVA) algorithm introduced by Bagnuolo et al. (1992, AJ 103 , 1399). Pairs separated by 200~mas can be resolved with this system, but resolution is a function of both magnitude and the magnitude difference between the two stars. Originally, only 16~ms could be spent on the DVA computation for a frame without losing subsequent frames. This time has been doubled by operating the frame grabber asynchronously with the host computer. About twice as many frames can now be processed per unit of time for most double stars in the program. The signal-to-noise ratio has been enhanced by utilizing independent photometric information. Allowing the secondary star to occupy a fixed pass band on the intensity scale, pixels which are not bright enough to correspond to the associated primary star are filtered out in software. This not only increases the signal-to-noise ratio of the autocorrelation function, but increases the frame acquisition rate for pairs having a large difference in magnitude. The limiting magnitude difference has been thus increased from 2 to 3 for pairs separated by two arcseconds or more. Position errors on the order of one-half pixel (25~mas) in the East-West direction occur as a result of charge-transfer inefficiency in the CCD readout. This is corrected by symmetrizing the DVA after quadrant information has been extracted. More seriously, periodic white/black balance variations originating in the video output electronics can produce shifts of up to 80~mas, again in the East-West direction. Fortunately, only pairs having intensities close to the threshold are affected. Selected observations have been recalibrated by dividing the autocorrelation functions by those of flat fields exhibiting the white/black balance oscillation. A procedure has been established to accurately null these oscillations in the video output.
