next up previous contents
Next: Special Concerns for Specific Up: INFORMATION AND ADVICE ON Previous: Automatic Insertion of Geodetic   Contents


Scheduling the VLA

For much detailed information on VLBI at the VLA, follow the links to the VLBI at the VLA guide.

The WIDAR correlator is capable of providing phased array output from the VLA in a VDIF format to be recorded on a Mark5C disk recording unit. Such data can be played back on the DiFX correlators, at least. Single dish recording is a subset of phased array recording. Since recorder ready data are provided by WIDAR, none of the normal VLBI backends are used. But the properties of the correlator are rather similar to the RDBE so, in fact, scheduling the VLA has become more similar to scheduling other antennas than it was with the old system.

The SCHED output used by the VLA is the VEX file. In many ways, the setups are similar to those for the VLBA. The user does not need to worry about the details of the external VLA LO setup, although the front end name (FE) does need to be provided as there can be ambiguities. The software involved in translating the VEX file and making the VLA scheduling blocks (VEX2OPT) takes care of most details. The important factor is that the LO and baseband frequency specifications, taking into account the IF sideband, add up to the RF frequency of the bottom edge of the desired baseband. The bottom edge is required because the EVLA phased output is always net upper sideband.

As for setups, the VLA is fairly similar to a VLBA antenna. But there are several crucial differences. There are a maximum of 4 baseband channels in the initial version (2012), which only allows a maximum of 512 Mbps when used with the 32 MHz bandwidths of the RDBE PFB personality on the VLBA. The VLBA DDC personality can have up to 128 MHz bandwidths so it can provide 2 Gbps compatibility with the VLA with the 4 channels. All of the channels at the VLA must be net upper sideband. They must come in polarization pairs -- having four separate frequencies is not possible. Each baseband must come from a separate IF with RCP in A or B and LCP in C or D. There are some cases above 18 GHz where the BD pair must be at a lower frequency than AC, but that mainly affects Ka band ( 30 GHz), which is not available on the VLBA. SCHED will not try to protect against that although the hooks are in the code in case of future need.

A new mode has been implemented at the end of 2012 that allows up to 4 basebands to be extracted from each 1 GHz-wide IF. That will allow an 8 pair, 32 MHz baseband configuration that matches the RDBE PFB personality on the VLBA at 2 Gbps, at least for dual polarization. The SCHED checking will assume this mode. All channels must still be upper sideband.

The restriction to net upper sideband needs to be considered when using the RDBE with the PFB personality on the VLBA. Either the VLBA LO needs to be above the RF frequency, or the sideband inverting capability of DiFX needs to be invoked. For some bands, a high VLBA LO is not possible so the DiFX mode will be needed. If the LO/sideband combination from a station for a channel spans the same RF range as the LO/sideband combination at another station, despite opposite sidebands, SCHED will detect the overlap and schedule for sideband inversion.

Sample schedules that include the VLA have been provided. One using the PFB personality of the RDBE at the VLBA is jvla.key (modify for 16 channels!). The WIDAR baseband channels can have much wider bandwidths than the PFB 32 MHz channels. They can be up to 128 MHz wide so, with 4, there is enough bandwidth to feed 2 Gbps, matching the output of the VLBA stations when they use the DDC personality of the RDBE. For an example of using 128 MHz bandwidths with WIDAR and the RDBE/DDC, see vladdc.key. The other option is to activate a mixed bandwidth mode in DiFX so that one WIDAR channel can be correlated against 4 PFB channels, but that mode has not yet been implemented.

For scheduling, the user should be aware that the VLA slews at 40 deg/min in azimuth and 20 deg/min in elevation, which is about half the speed of the VLBA. This can be an issue for widely separated sources. It is, however, faster than some other HSA or Global VLBI antennas. The use of DWELL for scheduling should insure that adequate on-source time is obtained with the slower antennas.

Another issue is that, each time a frequency is seen for the first time, there needs to be a one minute dummy scan during which the levels of the analog signals into the samplers are set. SCHED can handle this when using DWELL scheduling by inserting the appropriate amount of time as a gap before the first scan with each frequency. The time required is given by the antenna catalog parameter TLEVSET. VEX2OPT will insert the required dummy scan in this gap. Alternatively, the user can explicitly insert DUMMY scans. These are scans that are not recording, not doing pointing, and not doing phasing and are at least as long as TLEVSET (1 minute for the VLA). SCHED will not try to insert a gap, or claim a late on-source time, based on TLEVSET before such a scan. It is traditional to put a series of such DUMMY scans at the start of the observing block, stepping through all the frequencies that will be used in the observing block. Note that an enhancement needed for the DUMMY scans on the VLA is to avoid needing DUMMY scans even for small frequency changes such as those associated with different Doppler frequencies for different sources.

ARRAY PHASING:

An important concern unique to the VLA and other interferometers used in phased array mode to generate a single data stream for VLBI is the need to adjust the individual antenna phases so that the signals add coherently when summed. The instructions to actively phase during a scan, to hold phase from a previous active scan, or not apply phase offsets are given in the VEX file in ``intents''. The relevant intents are AUTOPHASE_DETERMINE, AUTOPHASE_APPLY, and AUTOPHASE_OFF. Each can be preceded by VLA: (eg VLA:AUTOPHASE_DETERMINE if there is a reason to make them specific to the VLA. They can be provided directly using the INTENTs input to SCHED, but SCHED will also generate them as needed based on VLAMODE parameter if that is used, which is recommended. The use of VLAMODE is preferred because the defaulting behavior between scans is cleaner -- it does not get tangled with other uses of INTENTs. SCHED will not allow the use of both phasing INTENTs and VLAMODES as then can step on each other.

Phasing scans can be added simply as additional VLBI scans, or the VLA can be sent to a source not observed by the rest of the antennas and for which no recording is made.

For successful phasing of the array, a source must be greater than about 0.1 Jy (That is an old VLA number but is probably still reasonable. See the guide referenced above for details) and have a position that is good to a fraction of the VLA synthesized beam (enhanced sensitivity is only obtained over this area). It must have small structure phases and not have other sources of comparable strength in the primary beam that might confuse the phasing algorithm. The position accuracy is especially important if a calibrator is being used to phase the array for observations of another source.

Adding phasing sources is tricky, because it is desirable to spend a minimum amount of time on them, but if they are missed, the rest of the data will be bad. When phasing, the SCHED scan during which the phasing occurs is broken into subscans by the VLA as only one solution is done per subscan. The SCHED scan should be long enough for at least 4 such subscans. The length of the subscans is set by the scan-dependent parameter VLAPTIME in seconds. The default is 10 seconds which is reasonable for most good phasing sources and is the minimum SCHED will allow. If it is necessary to phase on a weak source, this could be extended. With the default subscan length, the SCHED scan for phasing should be at least 40 seconds on-source.

The interval between phasing scans depends on the observing frequency, the VLA configuration, and the weather. Intervals between 15 and 30 minutes are typical.

REFERENCE POINTING

At Q band (7mm, 43 GHz), the VLA a priori pointing is not good enough so reference pointing must be used. That is also controlled through use of INTENTs.

****** Document how to insert pointing scans, and the use of automatic insertion of pointing scans.

The old parameter VLAPSRC allowed automatic insertion of VLA phasing scans in the observe files. That scheme no longer works. But an item on the to-do list is to make that parameter insert VLA phasing scans. That has not yet been done.

OTHER CONSIDERATIONS

If doing DELZN segments to measure the tropospheric delay, it is best to use a single dish mode as the phasing can take time and can confuse the troposphere solution algorithm.

It is still early days for VLBI observations using the WIDAR correlator to generate the phased array output data. This section will likely need more information eventually.

Note that normal output from the WIDAR correlator will be obtained during VLBI observations. Such observations may be useful for determination of large scale structure, total flux density, polarization, or absolute amplitude calibration. The user may need some VLA specific scans, such as on a flux calibrator, to make full use of such data.

This section has changed drastically with the advent of the upgrade to the VLA and the use of the WIDAR correlator. If you have a perverse interest in the old system, see the obsolete sections.


next up previous contents
Next: Special Concerns for Specific Up: INFORMATION AND ADVICE ON Previous: Automatic Insertion of Geodetic   Contents
Craig Walker 2014-04-14