High-Performance FX Correlator System for enhanced ALMA

We have proposed a FX correlator system for an enhanced ALMA. Maximum bandwidth per IF is 4096MHz and spectral resolving points per antenna per IF is 128 x 1024. The number of correlation is 3160 for 80 antennas per IF and the number of output frequency bins for each correlation is 16 x 1024. We will prepare 4 sets of this correlator for 4 IF bands. This FX correlator system always realizes both high spectral-resolution ( < 0.1km/s at 100GHz ) and wideband ( > 1000km/s at 850GHz ) observations simultaneously for 4 IF bands up to 850GHz. Main changing point from the very large correlator system previously proposed is to apply flexible frequency-channel smoothing for the correlated data. It reduces the output frequency data from 128 x 1024 to 16 x 1024 per IF and eliminates the fear that the large amount of the frequency channels might increase the costs of post-detection computing and archiving. We estimate that the total cost of this FX correlator becomes half due to this change compared to the previously proposed one. 1. REQUESTED PERFORMANCES FOR FUTURE CORRELATOR In order to realize all the scientific goals with ALMA, -the correlator should not role out plausible experiments which would otherwise be allowed by the design.-, and ALMA future correlator should also support serendipitous discoveries. 1.1 REQUIREMENTS FROM MM AND SUB-MM SCIENCE Much more molecular and ion line emission and absorption will be observed in the sub-millimeter wavelength than in the millimeter wavelength, and they are sometimes blended each other. Especially at star forming region like Orion, emission line forest is detected shown in Figure 1. With the sensitivity of the enhanced ALMA, these line forests can be observed at more distant massive star forming regions. Thus the spectral resolution is one of the most important factors in order to analyze the target line emission correctly. We need the spectral resolution higher than 1MHz for the mapping observations of massive star forming regions like Orion. More than a few thousands of spectral channels are needed over the 4-GHz bandwidth for the success of not only line survey but also usual mapping observations in the sub-millimeter wavelength. The proto-planetary disks are one of the most interesting objects to study using ALMA, and different kinds of physical conditions will be observed in one field of view of ALMA. In the outer edge of the disk at about 500 AU from the proto-star, excitation temperature is relatively low ( about 10 K ) and the Kepler velocity is 1 km/s. But in the inner edge of the disk at 0.1 AU from the star, excitation temperature and gas density become high and the Kepler velocity is about 100 km/s. At the middle radius of about 10 AU from the star there might exist gap formation by a proto-planet. In order to investigate the molecular gas in the outer region and the gap formation, we need the velocity resolution of 0.1km/s. On the other hand, wide velocity coverage more than 100km/s is necessary for the gas in the inner edge of the disk. Further multi-line observations will be essential to study the physical and chemical evolution of such ploto-planetary systems, because different lines will be excited with such different kinds of physical and chemical conditions. We can also detect the continuum emission from the disk, and have to destingwish continuum and line emission precisely because we are able to obtain sensitive continuum data and emission line data simultaniously at one time. Thus we should make wideband( >> 100km/s ) and high-resolution( 0.1km/s ) observations simultaneously to go on the research for the physical and chemical evolution of the proto-planetary disks. Multi-line imaging study for nearby low-mass star forming regions become popular at millimeter wavelength, e.g., CO(1-0), HCO(1-0), and CO(1-0). Sub-millimeter multi-line and continuum imaging study will be more powerful tool * Correspondance: Email: [email protected] to investigate the formation and evolution of not only the proto-planetary system but also massive star-forming regions. For the molecular envelope of late-type stars, similar type of observations will be important and fruitful. Figure 1. Orion KL wideband spectra at 850micron. Extra-galactic millimeter and sub-millimeter observations require wide spectral window more than 1 % of the observational frequencies, and wider continuous bandwidth more than 2GHz is necessary to observe the velocity field of distant objects at sub-millimeter wavelength up to 850GHz. In case of molecular line observations of radio-loud objects, subtraction of continuum from line emission is essential, and much more line-free channels will be needed for precise continuum subtraction shown in Figure 2. Precise continuum subtraction depends on both the accuracy of passband calibration and the determination of continuum baseline level in a band. In order to obtain ten times better singal-to-noise ratio at continuum baseline ( line-free regions ) than that of the line channels, we need a hundred times more frequency channels, which corresponds to wide velocity coverage typically more than 1000 km/s ( 2.8 GHz at 850GHz ). We should support exciting serendipity like H2O maser observations of NGC4258. Using ALMA, we will be able to observe sub-pc regions of the nuclear molecular disks of nearby AGN at the distance less than 10Mpc. High excitation molecular and ion lines will be detected on the nuclear disks. Wide velocity coverage( > 1000 km/s ) and relatively highresolution( < 10 km/s ) observations will be important up to 850GHz. Much improvement of sensitivity with ALMA will allow us to detect weak emission, of course, and absorption line in distant galaxies. Absorption line forest observations such as Dumped Ly alpha forest might be interesting to investigate the formation and evolution of galaxies. In this case we also need wider bandwidth( > 1000 km/s ) including ambiguity in redshift z and high resolution( < 1 km/s ) simultaniously. Figure 2. CO from the central region ( < 1kpc ) of radio-loud galaxies 1.2 REQUIREMENTS FROM ENHANCED ALMA SYSTEM The enhanced ALMA means that ALMA will increase its sensitivity and obtain new capabilities at higher submillimeter wavelength due to the contribution of Japan( see ALG report ). It is now considered to consist of maximum about 80 antennas. ALMA SIS receivers will have their instantenious bandwidth of about 4GHz with better Trx. ALMA analog IF system will transmit both two sideband signals with both polarizations simultaneously from one receiver band. Thus the correlator system for the enhanced ALMA should process all the above analog data at one time to obtain better sensitivity and to save observing time. If we proceed two polarization data at one time, we can always obtain two times better signal-to-noise ratio. In mm and sub-mm regions, several important molecular lines are existed in both upper and lower 4GHz bands ( e.g., see Fig. 1 ), and for the higher sub-mm frequencies ( > 600 GHz ), good atmospheric condition ( opacity < 1.0 ) will be occurred in the limited fraction of time( less than 40 % ; m\Matsushita et al. ). Thus simultaneous observations of both sidebands are effective to save observing time, especially for higher sub-mm observations. Other than the sensitivity and saving observing time, more simple analog IF system will help us to reduce the analog parts and to save maintenance man-power and cost in case of the enhancement of the array. 2. HIGH-PERFORMANCE FX CORRELATOR SYSTEM FOR ENHANCED ALMA 2.1 COMPARISON OF FX AND XF ARCHITECTURE First we will briefly review the comparison of the performance of FX and XF architecture. Originally Chikada et al. described the cost-performance of FX and XF correlators. They calculated the total gate number of the FX-type correlator NG(FX) to that of the traditional XF correlator NG(XF) as a function of frequency channel number( n ) and antenna number( Figure 8 in No.8 reference ). They defined complexity of the operation Cg ; how many gates are needed to make one operation at unit clock frequency. Its unit is gate operation MHz . It is heavily depend on the technology. M is the total number of spectra obtained at one time, NO is the number of operation at unit time per one spectrum, and B is the bandwidth. Cg and NO are different between FX and XF. NG = Cg x M x NO x B/n. Cg in Nobeyama FX and XF type correlators( 2-bit direct multiplier ) were 400( F in FX type ), 1100( X in FX type ), and 30( XF type ) in 1980s technologies. Here we again calculate Cg using the parameters of VSOP FX and UWBC. Cg in 1990s' technologies are 570( F in FX type ), 3900( X in FX type ), and 21( XF type ). Cg in X part of FX is larger than that in 1980s. However, rough trend about the ratio of NG(FX) to NG(XF) for n and M does not change between 1980s and 1990s. In case of 64 antennas with the same bandwidth, the ratios of NG(FX) and NF(XF) are calculated using both sets of Cg. They are shown in Figure 4. In n > 1000, NG(FX)/NG(XF) < 0.1 0.0001 0.001 0.01 0.1 1 10 100 10 10