How Are Radio Telescope Images Produced and Why Does This Require a Supercomputer?

Question: I’m curious about the computation aspect of radio astronomy. How do you transform radio signals over some period of time and from 27 dishes into a 2 dimensional image? Why does this require a supercomputer?  — Bill

Answer:  Let me try to paraphrase the description of how a radio interferometer works from NRAO’s description of this technique.  A collection of two or more radio telescopes can collect the information necessary to create a two-dimensional image of a radio source by first observing the same radio source all at the same time. If we also then know (or measure) the relative distances between each of the antennas in our array to a very high level of accuracy, then we can calculate what the arrival time of the radio waves from a radio source will be at each antenna in our telescope array.  Note that these arrival times will be slightly different dependent upon the position of the radio source in the sky. This difference is a time delay in the phase of the radio wave coming from the radio source.  When we combine these two offset waves, they will not overlap perfectly due to their phase shift, creating what are called interference fringes. As the Earth turns and the telescopes tilt to keep tracking the radio source, the angles of the observations change. This tracking movement of the telescopes changes the distances the radio signals travel from the source to each of the telescopes, in the same way that shadows are longer when the Sun is lower. This translates to different phase delays between the waves reaching each telescope. The longer we observe, the more variations we get. The more variations we get, the more different pieces to the collection of information needed to tell what the object looks like in two-dimensions we get. Note also that the farther apart we separate the telescopes, the sharper their view of the sky becomes.  A special-purpose computer (somewhat of a “super computer”) called a correlator synchronizes the incoming data from the different antennas to within a few millionths of a second of each other. It pairs up each antenna to every other antenna in the array, creating hundreds of unique pieces of information on the object that is being observed. To keep up with this constant and complex data stream, our correlators are among the fastest supercomputers in the world, performing their calculations at femtosecond speeds – up to 16 quadrillion operations every second.  The information that this correlator is then fed to some special-purpose software that assembles the information the corrlator gathered into a two-dimensional image of the radio source in the sky.

Jeff Mangum