SD Forum #5 - minutes

SD - Forum #5 - minutes

Tuesday, November 19

  1. Build the basic model, the essential desk replacement. This replaces one LSI mini computer and two serial lines with a PC running pSOS and LAN communications. All the current control software is unchanged; one shared library is revised to change the communications from serial line to ethernet socket.

    This implementation allows us to drive the antenna using either Bunyip or Yowie. Bunyip remains critical for the A/D users (eg, SPOT).

    Switching from old to new requires two steps:

    • Move the Coord and Servo serial lines from the LSI to the PC.
    • Change the environment variable which points to the desk library.

    Operational as of November 7.

  2. Incorporate the "Tel_xx" routines into Desk; this eliminates one further process in the control chain. It would give all platforms (bunyip, yowie, wirrun, the suns and yeti) equal access to the control system. All platforms would use the control software developed for SPECTRA

    Switching from the old to the new requires two steps:

    • Move the serial lines.
    • Change a host name and port number in an ascii file.

    Design underway. Trials expected late november.

  3. Incorporate the Coord functionality into Desk. This would remove a further LSI from the system, as well as one serial line. The control system overheads would be reduced by several seconds.

    Trials expected late january.

  4. Incorporate TrkFcc into Desk. This can happen independently of item 3, but its benefits won't be realised until the Coord step is completed.

    Trials expected late december.

----------------------------------------------------------------

Additional points:

Along with the benefits of the ethernet interconnectivity come the problems associated with the network. Since the Parkes network has at times experienced problems (there is a wealth of anecdotal data), we need to keep a close eye on the on-line portion of the network, possibly also taking steps to make pavo exclusively an on-line machine.

Platform Independent Antenna Control and Monitor Facility (mjk)

Stage one of this operation is complete: a library of control routines has been prepared allowing tasks running on the unix systems to control the antenna.

These libraries will remain operational through the evolution of Desk.

A simulation antenna has also been provided, to allow painless testing of the control software.

Stage two is partially complete: the unix software is in prototype form, but requires the new desk to be in place.

Evolution of the Parkes Utilities (P. te L-H)

Over the coming months the utilities will evolve to adapt to the new environment (to take advantage of the new Desk, and to provide the calibration services needed now that the new focus cabin is in place). The list below reflects the order and priority of the effort.

  • Functionality as designed by ATOMS -> done
  • Libraries with function calls to the hardware managers (eg. Antenna/DESK) -> on-going
  • Inter-process communication/GUI -> under debate
  • User Interface -> Glish /TK ? Perl/TK ?
  • Smaller Utilities (eg, Coord; coco) -> done
  • Antenna control (sun-based 64m) -> evolving
  • Calibration (SPOT) -> next.

-------------------------------------------------

Additional notes:

  • The correlator software is firmly vms-based at present. W.Wilson envisages a move to unix, sometime in the next few years - the exercise will involve several man-years of effort (software and hardware interface).
  • The hardware interfaces will increasingly lean towards PCs and the pSOS operating system. This is new territory for the parkes environment, but not for the correlators, where we have now had several years of satisfactory results .

Multi-Beam Software and the aips++ connection (LSS)

The software architecture is largely complete.

Lister identified several areas requiring further effort:

  • Interface to the control task (MULTI). The present design has a simple command line interface. We would like to offer some form of GUI. We are exploring the option of a Glish-based system. (The difficulty is that Glish has not yet been ported to a vms system, whereas MULTI has to run on wirrun (vms); a cooperating pair of processes may answer the problem).
  • Scheduling is still modest.
  • The NFS link between the RPFITS archive data and PKSLoad, the real-time diagnostic processor is very slow (50 Kbytes/sec). (Some questions were raised here about the data rates - if we record autocorrelations only we can operate RPFITS in "real" mode; if we wish to record complex numbers, - if we record cross-correlations between diferent beams, then we may need to operate in "complex" mode for all the data, thus doubling the file sizes).
  • The calibration machinery is incomplete.
The New Parkes Conversion System (GM)

This was discussed at length in july ( sd-forum #2; see also the polarimetry discussion in sd-forum #4). The general design is unchanged. Some questions remain which need clarification:

  • Do we want 4 IFs ? Some arguments advanced:

    - multiple transitions within a band (eg, HI and OH).

    - dual beam operation at high frequencies.

    (4 IFs seems a natural consequence of adopting the standard 4-IF AT rack).

  • Do we need to provide bandwidths > 500 MHz for the Bonn polarimeter ?

    This is relevant only for frequencies above 4 GHz.

  • Could we satisfy the 500 MHz bandwidth with contiguous 256 MHz bands and 4 IFs?

    This is unlikely to be easy, as the Bonn polarimeter expects an analogue signal centred on 300 MHz, and up to 500 MHz bandwidth. The up/down conversion operations required to merge the two 256 MHz passbands is probably no simpler than providing the additional conversion and filtering of the original design.

    The real question is thus the original one: how important is a 500 MHz bandwidth?

Interference and Baseline Ripple Assoicated with the new Focus Cabin

This topic was introduced here to outline the problem - it will be discussed at greater length at the next forum, when the picture should be a bit clearer.

Significant problems have been observed recently - in the multi-beam tests in late october; with the H-OH receiver last week; during the pulsar runs. It is clear that any weakness in the RF cable integrity will produce serious baseline problems; the path between the LNA output and the feeds is not yet clear - W.Wilson will investigate this next week.

next meeting : Tuesday, 14 January 1997

Staff space
Public