Overcoming the Legacy of Galileo

Ever since the invention of the first telescope by Galileo in 1609, the need to "point" it in the desired direction has always been assumed to be a fundamental requirement. But this assumption was based only on the technology used until now to build all almost all radiotelescopes. By marrying modern computing with an array of many small antennas, a new generation of telescopes can be built which look in all directions at the same time, just as if there were thousands or millions of conventional telescopes all in the same spot, each pointed in a different direction. The new telescope is called Argus, named after the mythological guard being that had 100 eyes and could watch in all directions at the same time.

A second assumed fundamental requirement that started with Galileo's telescope is that only one person can look thru it at a time. This has evolved without conceptual change into today's method of assigning telescope time to individual observers. Existing radiotelescopes can typically only be used by one research project at a time. There is great competition to use the available telescope time, with some prospective users being turned away. Argus can conduct all desired research programs at the same time, with no mutual interference or compromise needed, making its scientific productivity much greater than comparable conventional telescopes. This capability is particularly important when seeking international funding and participation, since it makes wider usage possible.

A third assumption from Galileo's telescope is that precisely-made, moving components such as lenses or mirrors must be used. That has evolved into the the large dish radiotelescopes of today, which are ultimately limited by gravity, wind and differential temperature. Argus has none of these limitations. And as telescopes grew larger than Galileo's, they required massive yet accurate machinery to support and steer them. Argus has no machinery, no moving parts, no tight tolerances and little need for mechanical maintenance.

The radiotelescopes of today are typically large steel structures, one-of-a-kind with no mass production, with construction costs dominated by labor, which increases with time. An Argus telescope consists of a large number of very small antennas, connected together with a large number of small computers. Mass production of the pieces is automatic, and construction cost is dominated by the cost of computing, which decreases with time. Hence the cost of an Argus telescope must become less than a Galilean telescope at some point, even if its priceless all-seeing advantage is ignored.

Since Argus is made primarily of computers, other aspects of computing technology are automatically available. The raw telescope data can be recorded for permanent storage, so that retroactive observations of phenomena discovered later can be made. Data can be sent continuously to researchers around the world via the Internet, making it unnecessary for them to come to the telescope.

The time has come to seriously consider a fundamentally different approach for radiotelescopes. Instead of large steel dish structures, a large number of small omnidirectional antennas can be used in an array to obtain much greater performance at ultimately lower cost. Such arrays are commonly called "phased" arrays, but that implies narrow bandwidth, so a more correct term for what is discussed here is a "timed" array.

Advantages of Argus Over a Dish-type Antenna

Compared to a conventional dish, an Argus timed array provides many advantages, including simultaneous high-gain omnidirectional sky coverage (no scanning), high sensitivity (arbitrarily long integration time), high resolution, variable beam size and shape, low and movable sidelobes, detection and tracking of transient and moving sources, adaptive and retroactive observations, interference rejection, higher aperture efficiency and fault tolerance.

Information and energy are falling on any radio telescope from all directions all the time, and the vast majority of it is ignored; that is in one sense considered "good." The larger a dish antenna is, the worse it becomes in terms of using all the energy and information that falls on it. Figure 1 below illustrates the extremely low total efficiencies of some well-known dish-type antennas, in comparison to the Argus approach. The sensitivity of an Argus array is the same as that of a dish having the same total effective collecting area and the same sensitivity receiver.

Figure 1. Efficiencies of various antennas.

In terms of flexibility, an Argus array has a number of advantages. It can be easily expanded or changed in shape; its resolution can be chosen independently of its collecting area; and its resolution, beamshape, and sidelobes can be changed at will by software.

In terms of capability, an Argus array can do many things a dish cannot do, including observe multiple (all) objects simultaneously, track rapidly moving objects, detect transient events in unknown directions, survey the entire sky in a single integration period, observe adaptively in response to current results, and reobserve retroactively objects or events not recognized initially. The retroactive observations can be done by playing back the recorded data from the array elements, and if desired the beam and processing equipment can be reoptimized for the reobservation.

Argus has many advantages over a dish in terms of its ability to deal with radio frequency interference (RFI). The elements can be designed to have nulls at the horizon for rejection of terrestrial signals. The elements are on the ground, in contrast to the elevated feed of a dish, hence the signal strength of terrestrial signals is less. Small shield fences can be used around the elements or array if necessary for further rejection of terrestrial signals. The direction of any RFI signal is immediately known to Argus since one of its beams always points toward the RFI source, and it will be strongest in that beam. That beam will also provide a nearly noise-free version of the RFI which can be used to characterize and identify it and to blank it or cancel it in the rest of the beams. Diagnosis of RFI is immediate with no need to steer the telescope "off-source" to see if it goes away. Since each beam can be separately optimized, permanent nulls can be generated by each beam in the direction of known fixed RFI sources. Adaptive nulls can be generated in real time as needed to deal with transient RFI. Moving RFI sources such as aircraft or spacecraft can be immediately identified as such by their movement among the beams, and henceforth tracked, predicted, and removed from the telescope output. Argus can also identify RFI sources by their distance, since it can simultaneously focus itself at all distances. A modest 64-element Argus can resolve distances out to about three kilometers, whereas an Arecibo-sized Argus can do so out to 500 kilometers. These distances would allow discrimination against almost all manmade signals.

We have constructed and operated a prototype eight-element circular Argus array at 162 MHz (1).

Argus Design Criteria

The elements of a general-purpose Argus array should have hemispherical coverage, aimed straight up. They should have nulls at the horizon for rejection of terrestrial interference, have dual circular polarization, be broadband, and mass producible. The best candidates are from the helix family. A multifilar contrawound conical helix can achieve these requirements.

The Argus array geometry should have approximately circular symmetry (for uniform beams), and not have uniform spacings (to avoid grating lobes). Placing the elements logarithmically spaced along the arms of a multiarm logarithmic spiral (see Figure 2 below) achieves these requirements. To calibrate the array occasionally, small remote-controlled omnidirectional transmitters are placed inside and near the array.

Figure 2. Element locations along the arms of the multiarm logarithmic spiral. The open squares are calibration transmitters.

The number of elements required in an Argus array to have performance comparable to a dish telescope depends upon the desired application and frequency. For example, if it is desired to observe only a single direction with maximum effective collecting area, then the number of elements for comparison with some existing dishes is:

But note that if an Arecibo-size Argus array were operated at 1500 MHz, it would also be looking in every other direction at the same time, and so would be equivalent to 1,600,000 Arecibos.

If it is desired to do an all-sky survey in a given time, then the equivalent number of elements required is:

Argus Computing Architecture

The performance of an Argus array (as measured by its number of elements, number of beams, and bandwidth) is limited primarily by its computing power. Hence this is the most critical portion of the design. Fortunately, available computing power is rapidly increasing and its price is falling. A small computer is used at each of the n elements, which does all computations that can be done on the data coming from that element. A different set of m small computers is used to perform the calculations for each of the m beams. In general, m is much greater than n, since the array is sparse (2).

All the element and beam computers may communicate via a token ring network. All of the element and beam computers are dedicated, programmed, and optimized to do just one set of fixed calculations, so they can be made very fast. The element weightings used for beamforming are kept in lookup tables that are separate for each beam and can be rapidly changed as desired.

In addition to the element and beam computers, there is another much smaller group of small computers attached to the network, each dedicated to some special project. Examples of such projects include monitoring a pulsar, tracking a spacecraft, lunar occultation, identifying RFI, calibrating the system, etc. Each special project computer is free to use whatever data it wishes and make whatever calculations it wishes, with no interference with the main computers or with each other. Hence there is no limit to the number of special projects that can occur simultaneously. One particularly important special project is to record all the element data in a compressed form for later analysis. This makes it possible to reobserve an event that occurred long ago, but was not recognized at the time. The special projects computers can also be attached to the worldwide Internet, making it possible for anyone anywhere to control them and to obtain data from them.

The computational power required for an Argus array of equivalent size to a large dish is greater than can be reasonably achieved today in the microwave region. But future developments in computing will make this possible, and today modest arrays at lower frequencies are possible.

Once an essentially noise-free image of the sky is obtained by long integration, a differential mode of operation can be used. In this mode, the telescope output displays only the differences between the "normal" sky and the current sky. This drastically reduces the amount of data to be displayed, and allows for immediate discovery of anything which has changed, moved, appeared, or disappeared. Such discoveries could automatically be announced immediately by one of the special projects computers to everyone around the world who chose to receive such announcements, via an Internet newsgroup or mailing list.

Converging Technologies

Many technologies are now advancing rapidly in directions relevant to Argus. The Institute of Electrical and Electronics Engineers (IEEE) just published a special Communications issue (May 1995) on Software Radios. It describes how the cellular telephone industry expansion is forcing the development of radio technology that uses computers to process multiple complex communications signals instead of using traditional analog filters and amplifiers. Q-Dot corporation in Colorado is now manufacturing CCD chips for dedicated electronic beamforming applications. Optical beamforming of microwave signals is an active area research area for military radar applications.

IEEE sponsored a conference on Mass Storage Systems in the Fall of 1995. The development of mass storage systems is being forced by the entertainment industry, who want to provide video-on-demand home access to every movie and television program ever made. The amount of the data produced by an Argus array having 10**3 elements, 10**5 beams, 10' resolution, 100s integration and 10**4 spectral points is about 1 terabyte/day. This is less than the data rate produced by a number of other scientific efforts today: NASA Earth Observation Satellite (8 TB/day), CERN (10 TB/day), Fermilab (3 TB/day). By the year 2000, optical tape cartridges will hold 1 TB each.

Holographic storage systems are being developed by Optitek and GTE (3) that are predicted to store a trillion bits in a cubic centimeter crystal of lithium niobate in three to five years. Holographic storage also allows massively parallel computations to take place within the storage itself, such as the addition or subtraction of entire images in a single step.

Protein-based computers are being developed at Syracuse University (4) that can in principle be 1/50 the size and 1000X the speed of semiconductor- based computers. This computing technology was first recognized by the Russians, who have used it to build a yet-secret military radar processor. (Note that Argus requires computations similar to a radar.) Protein memory cubes can attain the same trillion bits per cubic centimeter as the crystals mentioned above. They must have very uniform composition, which is aided by low gravity manufacturing. Two space shuttle flights have already carried experiments to do this. Birge predicts that this technology will dominate computing within eight years.

Perhaps the ultimate in computation may emerge from the field of quantum computation. Seth Lloyd of MIT (5) is pursuing the use of individual atoms as storage devices. A single electron energy level is used to represent a bit, which is flipped by laser radiation. A single grain of salt contains about 10**18 atoms and could hence store that many bits. Because of internal instabilities and imperfections among the atoms, a quantum processor would have to spend about 99.9 percent of its time doing error detection and correction. Nevertheless the remaining speed is about 10**8 times faster than a Pentium processor. This technology is about 20 years from practical application.

The Big Picture

It is commonly believed that humankind is basically aware of everything that goes on around us in the universe. This may seem logical, given all the telescopes in operation around the earth. But the fact is that all telescopes combined see only a tiny fraction of the universe and frequency spectrum at any one time, and as larger telescopes are built, they see even less. In our quest for ever greater detail about the trees, we are ignoring the forest. There are undoubtedly transient events occurring all the time of which we are unaware; previous examples include pulsars and supernovae. We have no global view of our electromagnetic environment, encompassing both natural and manmade signals. We have an obligation to open our eyes widely and be aware of our surroundings so we can learn more about the universe and understand the big picture. Argus will make this possible.

References

1. Bolinger, James. A Simultaneous Multi-beam Phased Array Using Digital Processing Techniques. Master's thesis. The Ohio State University (1988).

2. Brown, Stephen B. Radio Camera Arrays for Radio Astronomy. Master's thesis. The Ohio State University (1993).

3. Philip E. Ross, Forbes, Sept 26, 1994, p 170

4. Robert R. Birge, Scientific American, March 1995, p 90

5. Wired, March 1995, p124

This paper was presented during the "High-Sensitivity Radio Astronomy" conference at Jodrell Bank, Nuffield Radio Astronomy Laboratories, University of Manchester, England, on January 22 - 26, 1996.