This section contains subsections with detailed discussions and advice on various aspects of scheduling. It surely is not complete, but much basic advice is available.
A tremendous amount could be said about observing strategy. See the article by Joan Wrobel in the book “Very Long Baseline Interferometry and the VLBA” (1995: ASP) for a good discussion. That book also contains a lot of information about VLBI and is a good reference for all observers to have. Other important sources of information are the “VLBA Observational Status Summary” and “General Instructions on Observation Preparation” (which is sent to scheduled users). These and many other useful documents can be accessed on the Internet from the “VLBA Information for Astronomers” page. This section of the SCHED Manual will be limited to a few important concepts to keep in mind while scheduling. It is assumed that the observer already has a reasonable idea of how much time is needed per source, what frequencies to observe, etc. The suggestions here are more oriented toward smooth observing, processing, and calibration.
Fringe finders: In order for the staff at a VLBI correlator to tell if the correlation has gone well, it is necessary to have occasional scans on sources that will provide strong fringes. A minimum of two such scans should be provided and, for longer projects, there should be one every 4-6 hours at a minimum. The “VLBA Observational Status Summary” and “General Instructions on How to Prepare for Observing” documents mentioned above have lists of appropriate sources.
The fringe finders can also be used for “manual phase cal”. This is the process of aligning the phases and delays on all individual IFs. This is simply done by running the fringe fitting program on the calibrator and applying the results to all data. For this purpose, it helps to have a scan with all antennas at reasonably high elevations simultaneously.
Finally the fringe finders typically make good bandpass calibrators because then have good SNR in each spectral channel.
Setting the flux scale: Setting the flux scale of VLBI images is not trivial. Ideally, one would have a compact source of known flux density that could be used, as 3C286 and 3C48 are on the VLA, to calibrate the flux densities. However, by their nature, sources that are compact to VLBI (essentially none are unresolved) are variable and, therefore, do not make good flux calibrators. There are two ways to deal with this.
One is to try to obtain the flux density of a compact calibrator at a close enough time to the VLBI observations that variations will not be significant. At the time of the VLBI observations, some of the larger antennas can be asked to measure flux densities of sources being observed (see the TANT command. Better yet, if you have a phased array, such as the VLA, or can get an hour or so of VLA test time, accurate flux densities of your VLBI calibrators can be measured. In order to do this, be sure to include a VLA flux calibrator (usually 3C286 or 3C48) in the VLA schedule. You don’t need to record VLBI data on it. This method can go awry if the calibrator has intermediate scale structure that is compact to the interferometer but resolved out by VLBI.
It is also possible, and perhaps preferable, to rely on the apriori calibration of the VLBI antennas. For this, it is best to look carefully at your data and determine which subset of antennas is giving consistent calibration. With the VLBA at intermediate frequencies, it is possible to get the flux scale right to within a few percent this way. You just have to be sure not to let any antennas with bad weather or other problems contribute to the flux scale. In the AIPS calibration task CALIB, use ANTUSE to limit the gain normalization to the antennas whose gains you trust. Also specify a generous minimum elevation, such as 30 degrees, for the gain normalization.
When using the apriori gains for absolute flux calibration, care must be taken with bandpass calibration and channel selection. The gains are measured using the full bandwidth of the baseband channel - they are based on total power measurements from the baseband converters. Such gains apply to the average across the full baseband. If you attempt to apply such gains to a data set with the edge channels removed, there will be an error of a few percent because those edge channels typically have lower gain (which is probably why you removed them) and the average for the remaining channels will be higher. You can deal with this either by doing the calibration including all edge channels, or by doing a bandpass calibration based on all channels. Bandpass calibration brings all channels to the average level.
If you are going to rely on apriori calibration, be sure to obtain some data at close to the frequencies at which antenna gains have been measured. These frequencies are given in the vlba_gains.key available by anonymous ftp to ftp.aoc.nrao.edu. This is the file that you will need for VLBA calibration. At those frequencies, the Tsys used in calibration and the gain are based on the same values of Tcal so the Tcal value cancels out. If you observe at other frequencies, you will be depending on the Tcal values being correct (or at least having the right ratio) at your frequency and the frequency where the gains are measured. That is not assured.
For any of these methods to work, it is best to include a strong, compact calibrator and to observe it several times to check consistency.
There appears to be an offset of on the order of 6.5% between Tsys values measured with the legacy VLBA system’s baseband converters and Tsys measured with the new, wide-band RDBE. The cause of that offset is not yet understood, but is thought to be some issue with the old system. During the transition, the gains distributed are for the old system. An adjustment appears to be required to for data data calibrated using Tsys measured either by the RDBE or by DiFX and gains derived from the old system (all gains distributed so far as of July 2012). For a discussion of this effect, see /htmladdnormallinkVLBA Senesitivity Upgrade Memo 34 https://science.nrao.edu/facilities/vlba/publications/memos/sensitivity-upgrade/index.sensimemo34.pdf. Stay tuned on developments in this area.
“VLBA Information for Astronomers”
Setting the relative gains of antennas: Another invaluable use of a strong, compact calibrator is to help ensure that all antennas and all IFs have consistent calibration. After apriori calibration and setting of the flux scale has been done, a model of the calibrator can be used to derive whatever additional gain adjustments are needed. These can be applied to the whole observation to obtain the best possible calibration, short of self-cal.
Scan Gaps, “Readback tests”, and module-swap gaps: Before the Mark5C era, one had to be very careful about scan gaps to avoid data loss due to resync times, but also not to put the data at risk through excessively long record scans (period of no media stoppage). With the Mark5C, most of these issues go away. The media are stopped from the end of one scan to the expected time of good data at that station in the next scan. Resyncs don’t take time with the software correlator.
With the older media (including currently (2012) in use Mark5B and VLBA disk recording), the recording media should be stopped occasionally to help prevent major amounts of data loss if something should go wrong and, on non-VLBA controlled stations, to allow disk bank changes. Two minute gaps every hour or two used to be used for readback tests with the tape systems. Tape is no longer used, but an equivalent data integrity test is likely to be built into the disk based systems in the future. Also, if the disks are not stopped, all the data ends up in one file. If there are problems closing the file, the data can become unreadable. As of early 2008, it is recommended that a gap of 30s or more be inserted every hour or two. In fact, any pause in recording will start a new recording scan, but beware that SCHED prevents gaps of less than MINPAUSE seconds. For field system controlled stations, (most non-VLBA stations) gaps should also be inserted to allow bank changes. The VLBA can switch between mounted disk banks (modules) on the fly, but the field systems need a pause in the data recording. Such gaps should be inserted every 22 minutes for recordings at 1 Gbps and proportionally less often at lower bit rates. These gaps need to be more than 10s long.
Phased VLA: The phased VLA has special calibration needs. If the target source is either too weak or too resolved to be used for phasing up the array, special scans on a calibrator must be inserted for this purpose. This can be done with SCHED . Or it might be possible to do it with the VLA scheduling software, but it is probably better to explicitly include the calibrators in the SCHED files. (see the section on VLBI at the VLA for details).
Also, the effective beam of the phased VLA (or any other phased interferometer such as Westerbork) is the synthesized beam. This can be less than an arcsecond in the worst cases. This places far greater demands on source positions required for observing than is typical at single antennas. Be sure to provide positions of sufficient accuracy, which can mean 0.1 arcsecond.
Anyone using the VLA for VLBI should consult the VLBI at the VLA guide.
Calibrator scans: Except for phase or fringe reference calibrators, there will typically be only 2 or a few more scans on each calibrator of the types discussed above. It is tempting to put these scans at the beginning and end of the scheduled observing time. However this is quite risky as those are the time periods most likely to experience problems. The beginning is especially risky because, if there are any problems at a site, they may not have been fixed yet. The end is ok, except at sites where the main target source has been down for a while and the antenna has been idle. It may not wake up again cleanly for the calibration scans. It is best to take time (a few minutes on each calibrator) out of the middle of the schedule a couple of times to get the calibration data.
Protected scans: Starting in Oct. 2011, ongoing VLBA observations will be preempted at the Pie Town and Mauna Kea stations once per day for up to 1.5 hours for Earth Orientation Observations. This is in return for operations funding from the USNO. There is some flexibility in the exact times of the preemption. SCHED parameter PREEMPT can be used to protect key scans such as special calibrators or geodetic segments. See the writeup for that parameter for more details.
24 hour schedules: With dynamic scheduling, it is useful if 24 hour schedules can be “wrapped” by starting part way through, then picking up the first scans at the end. If you have a 24 hour schedule, it would help the process if you insert a unique COMMENT at the logical places for wrapping the schedule. The scheduler will use the WRAP24 switch to cause SCHED to copy the scans onto the end of the schedule to create a 48 hour schedule. He/she will then use DOSCANS to select the desired 24 hour block. Unique comments help find the same place in the two copies of the schedule. Example eg24.key demonstrates such a schedule.
Set-and-remember: Both the VLBA (RDBE) and the VLA new systems require the power levels for the analog signals to be set before the wide band samplers. Changing the attenuators can cause phase errors, but there is a lot of dynamic range in the allowed levels. Therefore the attenuators are only set once for each frequency setup per observation. Time needs to be left for that operation. The RDBE needs 5 seconds. The VLA needs 60 seconds. SCHED will warn when inadequate time is provided. Also, SCHED will schedule in the required time automatically using the numbers in the station catalog parameter TLEVSET when using dwell time scheduling when the scan itself is a recording scan, a pointing scan, or a phasing scan (VLA) or is not long enough. After the first scan with a setup, the RDBE/PFB needs about 2 seconds at the start of each scan to set the levels for the final down-sampler to 2 bits. That time will be provided automatically for if the station file parameter TSETTLE is higher, which it normally is for other reasons. For the DDC personality, the level setting will be adjusted regularly during the scan. The VLA doesn’t need extra time for that, although that operation is done.
Scans have a nominal start time and a stop time that may be set explicitly by the user or derived by SCHED based on a variety of criteria. Scans have a separate start time, often offset from the nominal start time, that SCHED assigns for the start of recording. SCHED also has a reasonably good idea of when good data will start to be available after any required slews and setup times are completed. Throughout most of SCHED, the scan times presented to the user are the nominal start and stop times. On the other hand, the recording start time is what will be given in the files sent to the telescopes and correlators. The VEX file, which is becoming dominant as the telescope and correlator control file, uses the recorder start time as the scan time, but contains the offset from that to the good data start. Some systems, including the VLBA and VLA now use the good data start as the time to start recording or correlation. This section describes how all this is controlled using a number of SCHED input parameters.
There are a variety of ways to set the nominal times of a scan. A START time must always be set for the first scan of a schedule — SCHED obviously has to know when to start. After that, the ways to set the nominal scan times are:
Set the START time and the STOP time. This is uncommon except with SCHED inputs derived automatically from other sorts of schedules.
Set just the STOP time. SCHED will use the previous scan’s stop time plus any gap as this scan’s start time.
Set the START time and the DURation. Sched will set the start time as requested and the stop time to the start time plus the duration.
Set a duration using DURation. SCHED will set the start time to the previous scan’s stop time pus any GAP and the stop time to the start time plus the duration.
Set a duration using DWELL. SCHED will set the start time to the previous scan’s stop time plus the time it takes to get the antennas on source and observing. Then it will set the stop time to the start time plus the requested dwell time. Note that this option does not yet take into account the time to switch between frequency bands (say 7mm to 2cm). There is an option, with a second argument to DWELL, to not wait for the last, or last few antennas. This can be useful when there is a slow antenna involved or when some antenna is too close to the zenith. A third argument can be used to insure that the slow antennas get some time on source. GAP can be used to set the minimum interval between the scans. This is the most recommended scan scheduling scheme.
Use any of the above to set the start and stop times, and then offset the resulting start time with PRESCAN. Only the start time, not the stop time, is affected by PRESCAN. If PRESCAN is positive, the start time is delayed. If PRESCAN is negative, the start time is moved earlier by the requested amount, although it will not be moved into the previous scan. This used to be the primary scheme for controlling time provided for media acceleration and correlator synchronization, but has been superceded by GAP, PRESTART, and MINPAUSE. A still-useful function is to delay the start of scans scheduled with DWELL somewhat to be more sure that the first recorded data will be good.
The calculation of the time required for an antenna to start delivering good data is based on on information in the station catalog. See the section on that catalog for details of the parameters mentioned here. The slew time is calculated based on the pointing direction at the previous and next sources along with the slew rates and accelerations for the antenna. An additional time TSETTLE is added to the slew time for the antenna to settle and the electronics to get ready. Additional time might be added for digital systems to set power levels based on TLEVSET for the first time a setup is seen (VLBA and VLA). The total time to be ready will not be less than MINSETUP, even if the slew is fast and the settling time (TSETTLE) is short.
The slew time calculator tries to take into account the need for cable wraps. The algorithm is simply — the shortest path to the next source will be taken even if that turns out not to be optimal later. Early VLBA experience suggested that a simple, predictable algorithm was preferred over a more optimal, but hard to predict, algorithm.
After the nominal times are determined, the recording start time is adjusted as described in the Adjusting the Scan or Recording Start Time section below using PRESTART, and MINPAUSE. Such adjustments were used to allow time for the recording media to get up to speed, formatters to reconfigure between scans with different setups, and for playback to synchronize on correlation. These were significant issues with the old tape systems, and to a lesser extent with the MARK5A system. The newer MARK5C recording system and DiFX correlators, and perhaps others, are able to record and correlate data beginning right at the assigned start time so such adjustments are not required.
The date specification for a scan is for the scan stop time, regardless of how the scan times are specified. If there is a scan that crosses midnight, this can cause some confusion, especially if it is the first scan of the experiment and the date is being specified along with START. If a schedule crosses a day boundary and START or STOP times are being specified, the new day should be specified. However, if midnight is crossed during any form of duration scheduling, the day will be incremented automatically.
All scans for a given station must be specified in time order. However, it is not necessary for scans for different stations to be in time order. This allows, for example, for the scans for one antenna to be specified separately from the scans for another antenna. While this works, it is not recommended because SCHED does not try to identify scans that can be correlated together. Anything that depends on knowing what the whole array is doing is likely to fail. DWELL time scheduling is one such item because SCHED must know how long the slews are for all antennas in a scan. Plotting of u-v coverage is another because SCHED only plots u-v points for baselines between antennas in the same scan. The estimates of data volumes and rates from the correlator, given in the summary file, are yet another because they depend on counts of baselines. Finally, any VEX file produced in such a way is likely to cause problems at a correlator that depends on it, such as DiFX (VLBA and many others).
SCHED allows sidereal time scheduling by means of the LST parameter. For VLBI, the concept of LST needs a bit more specification since it is different for each element of the array. LST can take an argument which is the station whose LST is to be used. If there is no argument, that station is assumed to be the VLA since LST scheduling was most commonly used for VLA schedules when the capability was first added to SCHED. Now that dynamic scheduling is being used on the VLBA, many users will want to set LST=VLBA_PT to conform to the style of scheduling requested for such projects.
If LST is specified, there are two ways to set the date. With the original method, the year and month are ignored and the day is assumed to be the (modified?) julian sidereal day number. Finding the LST day can be a bit painful, but a rough estimate can be had from the fact that LST day 60501 was 2006 Feb. 16. The estimate can be put in SCHED to narrow down to the right date. The second method is much easier and the scheme normally used. It is to specify the UT day in the usual manner. SCHED will attempt to figure out the LST day number, taking into account the fact that 0 hours LST is sometime in the middle of the UT day. It will also check if your start time is in the approximately 3.9 minutes where the result is ambiguous (LST days are shorter than UT days) and request specification of the LST day number — giving you the options.
If sidereal time scheduling is requested, most times and durations are assumed to be in sidereal units. Some exceptions are PRESCAN and MINPAUSE.
It is a very good idea, when using LST scheduling, to check the final output schedules, which are all in UT, to be sure that they are for the right day.
When dynamic scheduling, it is often difficult to mesh projects together perfectly. This can lead to gaps in scientific observing which causes an under-utilization of the array. SCHED supports two features that are meant to help with this situation. The first is that the user can specify optional scans on the ends of the project by setting PREEMPT = EXTRA for those scans. The scans will appear in the summary file so their properties are clear, but will not be passed to the machine readable files used to control the antennas. To activate them, so they are on the control files, parameter DOSCANS can be used to select the scan range to be used. The second option applies to 24 hour schedules which could be started at times other than in the original schedule and wrapped around the day. Parameter WRAP24 causes SCHED to copy the scans onto the end of the original scans to make a 48 hour schedule. DOSCANS is then used to select the desired 24 hour block. It is expected that DOSCANS and WRAP24 will be used mainly by operations personnel.
NOTE: Concerns about formatter reconfigurations will disappear when the digital backends take over. About half of VLBA projects use the new system as of the start of 2013 and that should go to all during the year, so soon this section will be obsolete.
A factor to consider when planning scan times is legacy formatter reconfigurations. These happen when the internal switching and setup of the formatter at the station has to be changed. Such changes happen when any of a number of parameters, including the number of channels, the sample rate, the BBC assignments, the BBC sidebands, and the pulse cal detector setup, are modified. A formatter reconfiguration takes about 8 seconds on the VLBA and during that time the formatter is not sending valid time codes to the recorder. If this happens during recording, it knocks the correlator out of sync for both the duration of the reconfigure plus the time to resync. In practice, it also seemed to confuse the old VLBA hardware correlator somehow for maybe one or two stations and they can take over a minute to resync. This is not likely to be the case for the software correlator.
It is best to avoid reconfigurations if possible. In the rare cases where that is not possible, try to provide a gap between scans of sufficient length to do the reconfiguration. The formatter configuration requested during a gap is the same as that during the following scan, so this only requires using the GAP command (or any other mechanism for having one scan start a while after another ends). Common reasons that reconfigurations occur in schedules are changes in the sample rate, changes in the BBC sideband (remember for net upper sideband, the 20cm and 13cm systems on the VLBA use lower sideband at the BBC), and changes in the kHz part of the frequency which changes the pcal detector frequency (changes in the MHz part will not change the pcal setup and will not trigger a reconfigure).
The nominal scan start time, if set using DWELL, is the time good data is expected to be available. If recording and correlation could start instantly, that would also be a good time to start the recorders. The more modern systems, especially the Mark5C recorders and the software correlators, approach this ideal and the recordings are started at the time specified in the VEX file for the start of good data. That time is calculated based on the expected slew rates and on any extra time, specified using the TSETTLE, MINSETUP and TLEVSET parameters in the station catalog. The start can also be postponed with GAP or PRESCAN. With fixed scheduling (DUR, START, etc), the nominal scan starts are forced by SCHED , but the time of good data, if later than the scheduled scan start, is still set by the actual expected slews and additions and that is believed by the Mark5C control software and the correlator. Parameters MINPAUSE and PRESTART, discussed below have no actual effect if the derived start time is before the expected start of good data.
With older systems, including the MARK5A systems that are still in use at least part time at the beginning of 2013, it can take a small amount of time to get the the recording going at the stations and the correlation fully synced up. With previous generation correlators and tape recorders, this was a serious issue and it was advisable to start the recordings at least 20 seconds before good data. Modern (2010) correlators are faster so the total time needed for both starting recording and synchronizing is only a few seconds, or less. Another reason for possibly starting recordings early, or even recording continuously between scans is that the Mark5A disk system can only handle up to 1024 recording scans. A recording scan is the time between stops of the recording media. There might be several source scans in a recording scan. Fast-switching, phase-referencing observations with the older systems could run into this issue so the default MINPAUSE (see below) has been set to prevent recording gaps during the fast-switching in appropriate cases. This section describes the tools in SCHED to allow the recordings to be started early.
The schemes described in the scan times section (parent of this section) are used to set the times of the scans as reported in the output files meant for human consumption. But the telescope control files actually give the times for the recording to start and stop. For the legacy (Mark5A) systems, those times are in the crd.xx files for VLBA systems and as the uncommented scan-wide start time in the $SCHED section of the VEX file. For the RDBE (VLBA etc) and WIDAR (VLA) systems, it is the data good time in the VEX file, shown for each telescope as an offset from the start time.
There are two primary parameters that can affect the recording start time used for the legacy systems. They are PRESTART, and MINPAUSE. PRESTART is used to request that the recording be started the requested amount of time (record time) before the scan start time. If that time is earlier than the previous stop time, the recorder will be left running.
The extreme, and often useful, case of a pre-start is to not stop recording between scans. This is especially useful if you have many short scans with short intervals between them, such as when phase referencing. MINPAUSE sets the smallest gap between scans for which the recording will be stopped. If the gap is smaller, the recorder will be left running. MINPAUSE used to be in units of playback time, so it was multiplied by the speed up factor to get the effect at record time. The speedup factor is no longer a simple concept so that adjustment has been removed and now MINPAUSE applies to the record time.
The default value for PRESTART is 5 seconds when the DAR is the legacy VLBA system or the MKIV system. Otherwise it is zero, which will be the case with the modern digital systems. The default value for MINPAUSE is 10 seconds when using the Field System or the legacy VLBA DAR. Otherwise it is zero which is the case for the new VLBA digital system. The Field System with the DBBC may be switched to zero in the future.
PRESTART is applied before MINPAUSE. First the recording start time is shifted earlier, then the interval from the last stop time is examined to determine if the recording should be left running. The defaults of PRESTART=5, MINPAUSE=10 should be ok in most situations (they are also in a state of flux as of the end of 2010, so it is possible they have changed since this was written). Users should not need to worry about these parameters most of the time. The offset of the recording start time from the scan start time can be displayed in the summary file by adding PTSTART to the arguments of SUMITEM. The recording start time is also available in the sch. files.
Another way to shift the recording start time is to use the parameter PRESCAN. This parameter was the original way of introducing gaps between scans, but was considered obsolete in recent years. However, it may have a use as a way of increasing the time allowed between scans beyond what would be given with the station parameter TSETTLE, although a negative PRESTART can do the same thing. PRESCAN shifts the start time in either direction (positive shifts to later times) after the other parameters discussed above set the time. If a positive PRESCAN is given, and the DURation or DWELL is increased by the same amount, the whole scan can be shifted farther from the period of the slew, allowing somewhat increased confidence that the recording will start with good data.
All of projects on the VLBA that do not involve other antennas or special constraints are scheduled dynamically. That means that they are put into a special queue, along with information about their minimum requirements, and then are run at an appropriate time given the weather and condition of the equipment. This increases the overall quality of VLBA output by avoiding observations at times when it is clear that the results will be poor, even if it also introduces inefficiencies in scheduling that mean that there is some idle time (projects don’t mesh together perfectly). For example, there is not much point in observing at 43 GHz when there is bad weather at many sites. But a 1.6 GHz observation at the same time might be fine. Dynamic scheduling also allows more flexible response to targets of opportunity. Galactic sources, in particular, tend to have short periods of enhanced activity so it is best to be able to observe when they are high. Of course, it is possible that some projects in the dynamic scheduling queue will never be scheduled.
The PI for a dynamically scheduled project will be given a window in LST at the VLBA_PT (Pie Town) that will be scheduled, if the observation is done. It is useful slotting projects together if the PI allows some flexibility in the actual start time. The PI should prepare a schedule using the LST parameter with Pie Town as an argument (LST=VLBA_PT).
When a program is selected for a dynamically scheduled time slot, VLBA operations personnel will modify the date (calendar or lst day number) to match the scheduled day. This last minute modification is necessary because the stations do not have an LST concept and the machine readable files delivered by SCHED must be in UT. The files will then be loaded to the sites with instructions about the time window to use, which may well be a subset of what is actually in the schedules. Because of this last minute modification of the date, it is best to minimize the use of dates in the SCHED input. Use durations instead.
To provide flexibility in the start time, it is useful to schedule using DWELL to set the scan lengths. Then SCHED can adjust the gaps between scans to allow for the slew times that will be experienced by the actual observations. If doing an astrometric project with “DELZN” segments (geodetic type observing all over the sky to solve for the zenith atmospheric delay), it is useful to use the ability of SCHED to construct such segments automatically. Since such segments depend on observing sources near rise and set, they cannot be moved around, so without the automatic construction of the segments, the start time cannot be changed.
For most VLBI observations, the data are recorded at the stations and played back later at the correlator. It is not yet possible for the recording process to be totally transparent to the user, but dealing with it is now much less of a burden than it was in the tape era. A top level issue is that a user is typically allocated a total amount of recording media that he/she is allowed to use. This number may be less than the total that could be recorded during the observation. It is required because the overall media supply may not be large enough to allow full time recording at the maximum bit rate and the scheduling committees need to apportion the resources according to the project needs and priorities.
Beyond the overall amount of data recorded, the disk-based systems still require the user to pay attention to the need for data quality checks, to maximum reasonable recording scan lengths, to the need for time at non-VLBA stations for module swaps, and to dealing with the fact that the correlatable data is not provided starting at the exact time specified for starting the recording. These topics are discussed in the Observing Strategy and Scan Times section and will not be discussed in detail here.