ATUC Meeting
Triennium Funding - Science Case

at the ATNF in Epping, Sydney

- March 23, 1999 -

2GHz bandwidth upgrade to the ATCA

1. Overview

The current ATCA correlator provides ample width and resolution for Galactic work up to the maximum planned operating frequency of 115 GHz (where a FULL_16_512 correlator configuration gives a 37 km/s usable velocity width with a channel spacing of 0.1 km/s), and it is expected that a great deal of exciting Galactic science will be done as a result. However, the maximum velocity width is currently 600 km/s which is inadequate for extragalactic work, and this proposal is for an upgrade to make it suitable for extragalactic work. This upgrade will also have the effect of increasing the achievable continuum sensitivity of the ATCA by a factor of four. The following sections discuss the science that would be achievable with the bandwidth upgrade. Little or none of this could be done with the present correlator.

2. Science background - Extragalactic 3mm molecular lines

Star-forming galaxies are much more luminous in the 3mm CO line than in the 21cm HI line. For example, NGC4945 has a single-dish HI flux of about 270 Jy km/s, but a single-dish CO flux about 25 times stronger, whilst the CO flux of NGC253 is about 100 times stronger than its HI flux. Even after allowing for the increased system temperature and lower antenna efficiency at 3mm compared to 21 cm, such galaxies are much more easily detectable in CO than in HI. For example, 80% of the star-forming galaxies in a recent project (COLA - Norris et al., in preparation) were detected in CO in a survey with a Tsys of 300K on the 15m SEST telescope, but only 20% of them were detected in HI in a survey with a Tsys of 30K on the 54m-equivalent ATCA, in a similar integration time.

The effect of this, together with the large collecting area of the upgraded ATCA, is that the ATCA at mm wavelengths will be amongst the most sensitive radio telescopes in existence for detecting distant galaxies. For example, a typical star-forming galaxy (such as a middle-of-the range COLA galaxy) at a redshift of 15000 km/s (and thus at a CO frequency of 109 GHz) would be undetectable (i.e. require several days integration) with the HIPASS survey, would be detectable by SEST after about 8 hours integration, but would be detectable at 5-sigma in about 40 minutes in CO on the upgraded ATCA. An 8-hour synthesis image would remain > 5-sigma even if the source were extended over several synthesised beamwidths.

All these figures are for "normal" galaxies. For ultraluminous infrared galaxies the redshift may be extended to cosmological distances. For example Arp220 would be detectable to z=4 (at which redshift the ATCA at 100 GHz would be observing a higher intrinsic CO transition), and the lensed galaxy 10214+4724 would be observable to z=10 or greater. For such galaxies, CO observations will rival optical observations for detecting high redshift galaxies. More importantly, the ATCA data will tell us about the molecular material rather than the stellar material, and so such studies will be very important for studying the evolution of galaxies.

Many key projects could be envisaged in this area, such as:

  • Studying high-redshift starburst galaxies, such as those in the Phoenix survey and HDF-S, to try to establish their evolutionary state.
  • Imaging of interacting galaxies, to watch the kinematics and evolution of molecular material
  • Imaging of starburst galaxies, to see how the molecular material is distributed
  • Tracing kinematics in galaxies with high extinction
  • Measurements of the CO mass of a high-redshift sample of galaxies, to establish the evolution of molecular mass with redshift.
  • A survey of the Hubble Deep Field South at the redshift of major groupings such as the STIS quasar Lyman absorption lines

I note in passing that the bandwidth upgrade will also be important for some lower-frequency projects, such as observations of water megamasers at 22 GHz, but the case for these is not as compelling as the 3mm case and so will not be discussed further.

3. Science background - continuum

Increasing the bandwidth of each IF by a factor of 16 (from 128 MHz to 2 GHz) gives a factor of four increase in sensitivity. Whilst interference will probably preclude use of this at 13 and 20cm, and perhaps at 6 cm, it is likely to be useful at 3 cm and shorter wavelengths, provided there is sufficient spectral resolution to enable flagging of interfering signals. This factor of four will make the ATCA sensitivity similar to that of the current VLA.

This gain increase will be especially useful at the shortest wavelengths (12, 7, and 3mm) where interference is unlikely to be a problem, but where sources tend to be weaker so selfcal becomes more difficult. In particular, it means that we can selfcal on source four times weaker, or, equivalently, integrate for one-sixteenth of the time for the same flux level on a given source. This is likely to be especially important at short wavelengths ( 3 and 7 mm), as the atmospheric coherence timescale for selfcal becomes shorter. It also means we can take advantage of the higher resolution of the ATCA at 7 mm to image sources at greater effective sensitivity (i.e. taking into account the spectral index) than we currently enjoy at 3 cm.

This increased sensitivity will not just allow us to do what we currently do at 3 cm in one-sixteenth of the observing time. It will make an enormous difference in the ability of the ATCA to detect and image weak millimetre sources such as the thermal dust emission in protostars and circumstellar disks, and in starburst, active, and infrared-luminous galaxies.

4. Technical Background

At present, the maximum bandwidth available on the ATCA is 128 MHz in each of two IF's, with full polarisation information. Furthermore, only 33 spectral channels are available in each of those IF's. By overlapping the two bands (by 32 MHz) to provide continuity, and rejecting the outer 10% of each band, the highest usable bandwidth available for spectral line use is therefore 202 MHz, with 51 spectral channels across it, giving a channel spacing of 4 MHz, but a resolution of ~8 MHz. The following table shows the resulting velocity width and resolution.

Table 1: Maximum usable velocity width and resolution of the ATCA

Band (GHz)

Total width (km/s)

Resolution (km/s)













The typical CO width of a galaxy is of the order of 300-600 km/s, and so if a galaxy were observed in this configuration, the CO line would be indistinguishable from a bandpass calibration error. Denser molecules such as HCN typically have similar widths (e.g. Henkel et al 1994, A&A 284, 17). Experience therefore suggests that a usable velocity width of ~1500 km/s is the minimum for adequate bandpass calibration. At present, therefore, the ATCA is not suitable for extragalactic work at 3 mm.

This problem is alleviated somewhat at high redshifts, where the observing bandwidth necessary for a given intrinsic velocity width is reduced by a factor of (1+z), but even at z=2 the current bandwidth is marginal. Much of the science discussed above is at lower redshifts than this, and so the 1+z is not sufficient to fix the problem.

Upgrade option 1 (cheap and dirty)

Since the correlator is capable of running in a one-bit 256 MHz bandwidth mode, the array could be doubled in bandwidth by relatively minor modifications to the LO and signal distribution system. A relatively cheap system would double the bandwidth for one IF (full polarisation) to 256 MHz on 5 antennas, at an estimated cost (W.E.Wilson, private communication) of $90k + 2 man-years, giving a velocity width of 1200 km/s, at a resolution of 24 km/s. This would enable some niche extragalactic work, particularly at high redshift, but would not provide a general extragalactic capability needed for all the science listed above. This would also provide no improvement in continuum sensitivity, as the extra sensitivity from the increased bandwidth is compensated for by the loss in sensitivity caused by going to 1-bit sampling.

Upgrade option 2 (preferred)

This proposal is to upgrade the signal path and correlator to a bandwidth of 2 GHz, with 1024 channels per product. This will give a velocity width of 6000 km/s, which is ample for all galaxies, and also allows a sufficient velocity range to make feasible searches for galaxies at uncertain redshifts (e.g. a search for CO in the HDF-S, associated with the STIS quasar redshift absorption lines).

WEW has made some rough estimates of resources required to achieve a 2GHz BW on the ATCA, taking into account the changes that would need to be made to the following systems:

  • IF/LO,
  • samplers,
  • data transmission,
  • delay units,
  • correlator

Specifications and costs:

Antenna-based components: four 2GHz BW IFs from each antenna, costing $1.5M.

Correlator component: for 4 products per baseline, each giving 1024 channels across 2GHz, costing $1.2M

(NB: This assumes that there will be a correlator chip available  - developed by someone else ( i.e. as was the case for the MB correlator ) - which we could use.  If we need to develop our own chip, add another $0.5M.)

Manpower:12 man-years (in addition to the above costs)

TOTAL COST: $2.7M plus 12 man-years