|
Beware! This document is very out of date and almost certainly only of use for those people attempting a MkIII experiment and/or one scheduled using PC-SCHED. Most spectral line users are directed towards the following guide to scheduling spectral line experiments that pays particular regard to the capabilities of the EVN MkIV Data Processor and the EVN.
As there are a number of things that require extra attention, due to the present scheduling software and correlator, when you are planning/doing spectral line VLBI on the EVN that we advise that you make particular use of those JIVE support scientists who are experienced in this area e.g. Huib van Langevelde.
In the planning phase you need to be aware of the spectral line capabilities of the correlator, particularly if it is in Bonn. For hardware details concerning the this correlator consult the Correlator User Manual. This constrains what you can do in two ways:
This implies that if you are applying for observing time, you should be aware of the correlator resources you are asking for. If you want 112 complex spectral channels and 112 real autocorrelations for LL and RR, you are asking for all 12 crates on a single baseline (where a single crate can correlate a bandwidth of 28 MHz). Hence it has to be correlated in almost as many passes as you have baselines i.e. it takes a long time.
Furthermore, for many OH line experiments, 2 MHz/112 is probably not giving you sufficient resolution. Getting more channels is not an option (see above) so you want to go to smaller bandwidth. But, because of fixed data rate, this means that the tapes will need to be recorded at lower tape speeds to accomodate the same playback data-rate at the correlator. You should be aware that this causes a speed-up factor at the correlator, where, for instance, your 500 kHz bandwidth experiment will be correlated 4 times faster than real time. This helps to reduce the burden on the correlator, as not so many passes on a tape are required to fill up a reasonable amount of observing time. But another effect is that sync-up times can become quite long in speed-up mode. Therefore, you may want to have continous tape motion during recording (except of course at the end of a scan).
The recording modes of MkIII are limited to a fixed set. This implies that if you want more than a single polarization of one transition (mode E) you will need to select mode E or C and record much more data than you are going to use. You can build in some redundancy by observing a line twice (or seven times), but remember that not all stations have enough narrow filters for all their BBC's. To not make correlation any more complicated than necessary you should indicate which channels you plan to correlate, in order to have at least the right filters in the same channels at all stations. Consider whether you want to use the autocorrelation amplitude calibration scheme. Realize that it will only work if all antennas see a strong and identical spectrum (watch out for phased arrays). You probably want Tsys data taken during your experiment anyway.
There are some special considerations for calibrating your line experiment. First, as line sources have a finite spectral signature, there is not much to calibrate delay on. So, unless you are doing an absorption experiment, you will have to observe a delay calibrator every hour, or more often. The line sources will only allow tracking of rate and phase. You may also want to spend some extra time on a strong calibrator to determine the bandpass. For both cases it might be best to use the exact same frequency setting of your target source. Even when there is no physical reason why you cannot acquire this a few MHz off, it is wise to have the same setup in order to avoid extra complications in AIPS. For the same reason rename your source for every new frequency setting you are going to use.
Spectral line scheduling in PC-SCHED can be a nightmare, even though recently some accomodations have been made (version 3.6). You start out by calculating the proper setting of the band-edge for your source. This involves calculating the Doppler components for the day of observing and your sources position, shifting the rest-frequency by the proper convention (radio or optical) and subtracting (for USB) half of the bandwidth. Note that all these can be set to 10 kHz accuracy only, which is usually sufficient.
Next you make your own version of the codes.sch file for PC-SCHED which should have a new block for each of the frequency setting (for each line source) that you are going to use. If you change the bandwidth code in the F card of the configuration, PC-SCHED will attempt to change the tape speed in the output SNAP files and ask for the appropriate filters in the PRC files. Continuous tape motion can be achieved by setting CAL to a negative (or zero) value in the PC-SCHED parameter section. But check the results carefully, PC-SCHED has been known to insert whole days in order to optimize the start and stop times. Don't worry too much about warnings that equipment is not ready for the next scan, this is the consequence of continous tape motion. Also Tsys} measurements will have to be done during scans rather than before integration. A known bug is that in the PRC files PC-SCHED writes the track assignment procedure has an ET command in it after all. This will stop the tape at every call of the procedure, usually in SNA files after each source change. You have the edit the PRC files to avoid the ET commands. Furthermore, the negative CAL times give rise to non-consecutive time tags in the SNA files; these look ugly, but are not a problem and will be skipped by the telescopes if that time has passed at the instant they are read.
In the cover letter you should indicate clearly what it is you want to do. Realize you are creating trouble for the friends at the stations, who have to set up your non-standard frequency assignments and switch (or physically put) the right filters where you want them. With the standard LO settings in the CODES.SCH they might not be able to reach your sky-frequency and some iterating might be necessary. Finally remind them in the cover letter that they have to switch the phase cal tone off. Otherwise it will create spikes in your spectra.
Spectral line correlation is not done absentee in Bonn. You will have to make arrangments to go there. It may be a good idea to ask people in Bonn to check there are fringes on your calibrators before you set out on your travel. The correlator needs a special setup file, tailor-made for your experiment, so you need to get involved in that. Furthermore, after correlation of a few passes you need to inspect your data to make sure everything is right. Another thing to do is to make notes of which correlations were producing valid data. The correlator software fringe detector will generally no be able to find any fringes for line sources, so there is no clear way to discriminate between good and bad data on the A-tapes.
The AIPS package is completely spectral line proof. But several problems with the interface, MK3IN, for spectral line data have been known, especially for the autocorrelation data. Like with things mentioned above, you may need help from somebody who has experience with spectral line VLBI.
Last updated: Thu Aug 31 09:32:30 MET DST 2000
 |