The Noise Bias Problem In Optical Speckle Imaging Experience With A Real Detector

We have obtained diffraction limited images for bright objects using various large astronomical telescopes, including the MMT, but the results have not been convincing for fainter objects. All speckle interferometric techniques utilizing image power spectrum or autocorrelation methods to obtain calibrated image amplitudes are subject to noise bias induced by the Poisson statistics of photon detection. This noise bias, independent of image frequency for ideal detectors with delta-function like point spread functions, must be removed to obtain calibrated image amplitudes. The noise bias characteristic of our intensified TV detector, both as a raster event localizer (faint object mode) and as a conventional image amplitude detector (bright object mode), is not only image frequency dependent but signal rate and signal distribution dependent as well. This precludes signal independent calibration of the characteristic bias functions and sets limits on object faintness for successful image recovery. Implications for detector development are noted.