Special-Purpose Computer for Radio Astronomy

Abstract

Radio astronomers developed a digital spectro-correlator to obtain the power spectrum of astronomical objects for the studies of their spectral lines. The degital spectro-correlator has functions of calculation of auto-or cross-correlation function, integration, and Fourier transformation. Characteristics of the digital spectro-correlator for radio astromony is 1) calculation with small bit numbers and 2) real-time spectral estimation.

Performance of the specro-correlator is characterized with its total bandwidth (full width of the spectrum), the number of spectral resolving points and the total number of the spectra obtained at one time. They are directly related to the data processing speed and the physical size (gate number of the correlator system. In the recent millimeter-wave observations, we need the total bandwidth of a few hundred MHz and the number of the resolving points of a few hundreds. Thus the processing speed and the gate number of one spectro-correlator unit is a few giga operation per sec and a few mega gates.

Now radio astronomers start to discuss the next generation spectrocorrelator system for sub-millimeter observations. It requires the total bandwidth of a few CHz with a few thousand spectral channels. The largest cor-relator systems among planned projects is for large array system. In such large array system, we have to calculate about one thousand sub-millimeter spectra at one time. Thus we need the processing speed of a few tera operation per second and the total gate number of about a few giga gates at the system clock of 100 MHz. We will introduce the key points and key technologies to realize the above huge spectro-correlator system.

the spectro-correlators. Necessary technologies to realize the LMSA correlator until the beginning of the 21 Century are shown in Table 3. We expect that the progress of those technologies is smooth.

Assuming that the above technologies are available, we can make a block diagram of LMSA correlator system (Figure 2). This system includes almost all the requests of the users. However, we have to resolve some difficulties to construct the real hardware of this system, e.g., large number of signal fanout (50) and connection (2bit x 1000) in front of the X part of the correlator. Thus we start to discuss the specialization of the above system with our wisdom; we have to understand the users’ requests more correctly (and eliminate unnecessary functions or bits) in order to realize the correlator and LMSA itself having the power cutting open the radio astronomy in the 21 Century.