next up previous
Next: First results of the Up: IRAM Newsletter 54 (December 2002) Previous: News from the 30m

VESPA is operational

The thorough upgrade of the 4096 channel autocorrelator at the 30m telescope is now complete. The new correlator backend named VESPA (VErsatile SPectrometer Assembly) combines hardware from the old 30m and interferometer correlators in a powerful new design. The control and data acquisition software is working, and VESPA is available for general use since May giving the 30m a much needed boost in acquisition power for high resolution data. The hardware is organized in 6 units (Fig. 3). Each digital chassis contains 12 correlation boards of 256 delay channels.

Figure 3: VESPA in the backend room at the 30m telescope. This correlation spectrometer consists of 6 units, each having a digital chassis with 12 correlator boards and an analog chassis with 15 RF modules.
\mbox{\psfig{,width=8.5cm} }

A total of 18432 time domain channels are now available. With two synthesizers installed in each unit, a large range of spectral resolutions and bandwidths become possible, increasing the correlator capabilities by a factor 3 in most configurations. In addition, new user modes, not available in the old 30m correlator, have been installed, increasing performance further. Table 1 summarises the capabilities of VESPA in its 4 user modes.

Table: VESPA user modes. Bandwidths are nominal values, the spectral resolution column lists channel separation.
  basic parallel multi--beam polarimetric
resolution sections $\times$ bandwidth receiver pairs sections $\times$ bandwidth receiver pairs
kHz $N \times$ MHz one two $N \times$ MHz one two
3.3 $6\times10$ -- $1\times60$              
6.6 $6\times20$ -- $1\times120$              
10 $6\times20$ -- $1\times120$              
20 $12\times20$ -- $1\times240$ 180 60, 120 $ 2\times20$ -- $1\times40$    
40 $12\times40$ -- $1\times480$ 360 120, 240 $ 4\times20$ -- $1\times80$ 120 40, 80
80 $12\times40$ -- $1\times480$ 480 160, 320 $ 4\times40$ -- $1\times160$ 240 80, 160
320 $12\times80$ -- $1\times640$ 640 320, 640 $ 4\times80$ -- $1\times320$    
625                 480 160, 320
1250 $12\times160$ -- $1\times640$ 640 640, 640 $ 4\times160$ -- $1\times640$    
2500                 640 240, 480

A full representation of its complexity is not possible here. Interested readers are referred to the VESPA users guide1 and to a more technical summary at A very detailed graphical configurator program (written in python) can be downloaded from

VESPA is used in its basic mode to connect to the (single pixel) SIS receivers. This is the most flexible mode, since any user defined spectral band can be connected to any of the 4 receivers in a valid configuration. Each spectral band can be individually moved to any IF frequency in the range 100 to 600 MHz. Table 1 shows that up to 12 spectral bands are available for the common spectral resolutions ($\ge20$ kHz). In particular, individual bands can be joined to form one contiguous spectral band connected to one receiver. A maximum bandwidth of 480 MHz can thus be covered at the often useful spectral resolution of 40 kHz.

VESPA extends this basic mode towards spectral resolutions of 10 kHz and smaller. In these extra high resolution modes an unprecedented velocity resolution of 10 m/s is reached at 100 GHz. Observations of the coldest molecular clouds profit from this high spectral resolution (Fig. 4) and small differential motions between parts of such clouds can now be studied. The corresponding requirement on the frequency stability due to the various Doppler corrections is, however, challenging and needs more study. In sensitivity-limited observations it is advantageous to use two (orthogonally polarized) receivers tuned to the same frequency. The parallel mode of VESPA supports this frequent observing strategy. Inasmuch as possible, VESPA employs a single synthesizer for parallel bands, leaving the other synthesizer for additional bands or a larger contiguous band. Table 1 shows the maximum contiguous band available at a given resolution for each receiver in such a pair of parallel receivers. VESPA can also be configured (not shown in the Table) in several narrower parallel bands whose combined width does not exceed the maximum listed in the table. Even two pairs of parallel receivers can be handled in this mode. The table gives maximum bandwidths for each pair.

Figure 4: HCN Spectrum (88 GHz) of a cold dark cloud showing the combined effects of hyperfine splitting, optical depth, and velocity structure. The spectrum was obtained with VESPA at a resolution of 3.3 kHz (11 m/s) and 30 MHz bandwidth in frequency switching mode. The insets display the 3 velocity-resolved hyperfine components.
\mbox{\psfig{,width=8.6cm,angle=270} }

The main use of VESPA will probably be with the multi-beam 1.3mm SIS receiver HERA which has currently a center-filled square array of 9 beams. In its multi-beam mode VESPA constitutes an efficient spectrometer offering resolutions from 20 kHz to 1.25 MHz with bandwidths shown in the table. At the frequently used resolution of 80 kHz, VESPA provides one band of 160 MHz or 2(4) independent bands of 80(40) MHz for each of the 9 beams. These bandwidths per beam will be halved however, when HERA is upgraded to 18 beams.

Fortunately, during the assembly of VESPA, not all the memory of the origin of most of its hardware was erased. VESPA has retained the capability of doing cross correlations which was its main function at the IRAM interferometer. In its polarimetric mode, VESPA cross correlates the IF powers from two orthogonally polarized receivers tuned to the same frequency. Stokes U and V spectra can thus be derived when linearly polarized receivers are used. Apart from the superiority of fully digital correlations over analog correlations, VESPA polarimetry also offers much more bandwidth than the 40 MHz available with the IF polarimeter (see Tab. 1). Several questions of calibration have still to be studied, however, before the polarimetric mode can be made available. VESPA was designed by G. Paubert in collaboration with the Grenoble backend group headed by M. Torres. Ph. Chavatte, A. Lapinov, J.Y. Mayvial, Th. Merrien, and A. Sievers, and M. Vidal were actively contributing at various stages of this 1.5 year project.

G. Paubert & C. Thum

next up previous
Next: First results of the Up: IRAM Newsletter 54 (December 2002) Previous: News from the 30m