Final Report 1

By Jim Chaffee

The contents of this report are classified TOP SECRET

Introduction

This is an interim report on recent developments in the Eye on the Sky program, conducted for the U. S. Air Force Research Laboratory by Sandia Laboratory for Alien Artifacts under contract AF-08-875903612. The author is Dr. Maurice Derreter.

The developments considered herein relate to application of software developed for code search. Some of the events involving personnel have been reported separately but will be included for completeness sake.

The Eye in the Sky program has been in existence for a little more than eight years, gathering and analyzing data from antennae watching the sky for alien transmissions. Recent improvements in software tools have led to new discoveries, to be discussed below.

The author has been attached to this project for three years. He developed and tested the software based on original concepts and algorithms stemming from his PhD work and a subsequent two years postdoctoral research at the Institute for Advanced Mathematical Applications under Air Force Research Laboratory and University of Ohio joint sponsorship.

Review of Tools

Hardware in the field consists of 2,048 miniaturized antennae with embedded spread spectrum uplinks. These are distributed all over the globe. In U. S. held territory they are placed in sparsely populated locations with unobstructed visibility of the sky, military bases, and on buildings dominating cityscapes. In foreign countries they are placed in locations negotiated with the local government, as well as on all embassies, consulates, and U. S. military bases. Some have been secretly placed in hostile nations. There are a number of these antennae on satellites.

Called a Spider Eye, the classified antenna design constructed with molecular manipulation technology consolidates thousands of antennules (miniaturized active elements) on an antenna the size of a pinhead acting as a phased array. The antenna operates to low elevations without multipath interference, listening to a broad swatch of sky. (Note: The author is aware the name Spider Eye is incorrect. More appropriate would be the name Arthropod Eye, but the author was told it didn’t have the same sex appeal, and Bug Eye was out of the question.)

Data is transmitted directly to satellites by a jamming and interference resistant wireless spread-spectrum chirp uplink. Each antenna transmits a unique code to identify its input. Time-tagged using GPS time for synchronization, all raw data input from each antenna retain their code identification and are kept as individual units at a central clearing house in New Mexico, with back-up in Cheyenne Mountain. Data sets from single antenna units are termed frames, though the image that comes to mind is more a window.

Because of the synchronized, precise time-tagging, the antennae provide a highly coordinated overlapping view of the entire sky around the earth. Since the project searches for codes, the term listening is preferred to viewing.

The software innovation, also described in detail in earlier reports and detailed again here in an appendix, consists of parallel signal processing algorithms in a single program called FODA-SE, short for Foreign Origin Detection and Analysis-Software Examiner. The program runs on a massively parallel MIMD integrated-memory hypercube controlled by a dedicated mainframe.

FODA-SE initially passes over the data with a neural net, digitizing the analog data. The net is based on the Protzel-Jeffries memory model refined to detect bits and associate intensities. Requiring gigabytes of memory, the net takes two days to process a week of data in batch mode, though it now runs recursively in real-time. Tested with noise-smeared simulated data, the net detected over ninety-nine percent of the bits, even in extremely low signal-to-noise environments, while conventional methods found at best fifty percent. The bits come out shaded, not simply black and white but in shades of gray indicating noise intensity.

The output bit stream passes into the multidimensional search software, a specialized type of hybrid phase-locked-loop behind a modified Hinkley detector and Grenander pattern recognition algorithm. The Hinkley detector gives an initial heads up, the pattern recognition algorithm ties in with the loop, and when a code is detected, the loop chases it wherever it leads, pulling along the paired detector and pattern recognizer. The software organizes the conglomeration of disparate antennae, with its hodge-podge of apertures, into a tuned array. The Hinkley detector aids lock even with dynamics, though we don’t expect much motion. Nonetheless, it gives the loop a kind of infinite bandwidth on the front end, while the pattern recognizer squeezes down, forcing the loop to follow the pattern. When a pattern is detected, we term the resulting process focusing.

The software searches for patterns, patterns within patterns, and so on. It looks for patterns in time and out of time, patterns at differential bit rates, patterns of differing intensity, patterns of different frequency, all using a wavelet basis devised specifically for multidimensional scaling, zooming in and backing away. Multidimensional lattice coordinates defined by shades of gray, time, and individual frame are the natural space in which the software searches. A pattern is a periodic sequence in this space, repeating with regularity in some parameter, and once detected is learned and retained by a Lauritzen-Kiiveri Markov neural net.

Individual time-tagging of data from each antenna using a precision synchronized network is essential. Time provides one of the code-dimensions. Without it, phasing the disparate antennae into an array would be impossible. But most important, without individual time tags detecting the alien code would have been impossible. The emphasis on this point is due to earlier discussions of saving money by timing overlapping frames (sky views) jointly. If this had been done we would not have detected the alien code. This striking fact seems impossible, and indeed there is as yet no explanation. However, the case will be presented below.

• << TEX Home
• 1
• Page 2
• 3
• 4
• Next >>