A control system is described that aligns the 96 beams of the Los Alamos National Laboratory KrF laser system to within a pointing accuracy of 5 µrad within 5 min and maintains the alignment in real time. This performance is made possible through a novel use of random noise. The 96 beams, together with optical benchmarks, are imaged on a single television (TV) camera. The pointing angles of those beams are estimated from the arithmetic means of the pixel coordinates within the beam images. The pixel intensities of each TV frame are mapped into a binary decision array based on whether or not the pixel intensity is above or below a threshold criterion. Existing, or introduced, random noise in the TV signal causes the contents of this array to vary from frame to frame, even when the actual beam is stationary. The beam positions are estimated from the pixel coordinates and their associated elements within this array. Finally, the beam angle estimates are updated from these position estimates, each TV frame, in combination with all previous estimates. This finds the contributions of the beam edges to the beam position by directly using pixels with intensities both above and below the beam threshold criteria, eliminating the need (possibly unrealizable) to do so by software interpolation algorithms. It does this very quickly, resulting in great data compression without use of computer time.