Draft High Freq

From CASA Guides
Jump to: navigation, search


This document is intended for observers planning EVLA observations at "high" frequencies. For this purpose we will define "high" frequencies as those above 18 GHz, in the K band (18- 26.5 GHz), Ka band (26.5-40 GHz), and band Q (40-50 GHz). All these receiver bands share at least some of the same problems and solutions, as compared to lower frequency bands, in particular antenna pointing, atmospheric phase coherence and opacity/flux calibration. For an overview, general performance, and some specifics of receiver band (e.g. sensitivity, etc.) of the EVLA, consult the Observational_Status_Summary.

In particular, the calibration overheads for high frequency observing are typically considerably larger than for lower frequency observations, significantly impacting the overall time request. As described in more detail below, the overheads grow with increasing frequency and maximum baseline length. For this reason proposers must consider the overall observing strategy in detail at the proposal stage.

Proposing for High Frequency Observations

In this section we will discuss several issues which should be considered when proposing for EVLA high frequency observations. All proposals should be prepared using the EVLA Proposal Submission Tool (PST).

K, Ka, and Q band Receiver Status

As of 2011, all EVLA antennas have been equipped with K, Ka and Q band receivers. Information on the sensitivity of the receivers over their tunable frequency ranges can be found at Observational_Status_Summary#Sensitivity.

Technical Justification Considerations

There are several things to consider that will influence your overall time request directly like configuration (and hence angular resolution) and correlator setup, and others that fall into the category of overheads, such as the extra time needed for calibration. These are discussed in detail below.

Configuration: The EVLA is reconfigurable; with the D-configuration providing the shortest baselines (poorest angular resolution, but highest surface brightness sensitivity) and the A-configuration providing the longest baselines (highest angular resolution, but poorest surface brightness sensitivity). See the configuration schedule for details for each call for proposals. It is generally important to consider (1) what angular resolution is required for your science at the desired observing frequency; (2) for resolved sources, how does the desired angular resolution compare to the required surface brightness sensitivity -- particularly important for thermal emission; and (3) how much flux will be resolved out by the array configuration that gives the desired angular resolution. Observational_Status_Summary#Resolution in conjunction with the Exposure Calculator can help answer these questions. Additionally, as described in more detail below calibration at high frequency is more challenging with longer baselines. It is a good idea to weigh the cost-to-benefit of going to higher frequency/longer baselines in terms of the required observing time.

Coorelator Setup:

Overall Observation Set-up: The overall setup of your proposed observations can significantly impact the overheads and should be investigated early in the proposal process. For example, where are your targets in the sky? If they cannot be observed in a reasonably short observing block, you will want to break them up into multiple groupings. Each will require their own calibration


During the proposal submission process the EVLA exposure calculator is an essential tool. Use the exposure calculator to estimate the time request for your target sources and for your calibration. Generally you use all antennas and polarizations, robustness and line bandwidth to calculate the target image sensitivity. For gain (phase) calibrators you would want to calculate the scan time needed for robustly detecting your calibrator with 10 sigma on a single baseline (2 antennas) in a single polarization in the total (subband) bandwidth. Depending on the array and frequency, coherence times may require a rapid cycling between gain calibrator and target source(s). Note that typical "fast-switching" or "loop" overheads may be 30 to 50 percent of the total observing time. Key is that your gain calibrator scans are separated in time no longer than the atmospheric phase coherence time, and that you detect your calibrators on every baseline in each of the calibration scans. Remember that spending slightly more observing time on your calibrators may make the difference between detecting them or not, and thus between being able to calibrate your data or not! The loss of the extra time spent on calibration typically, for non-snapshot observations, decreases your sensitivity with only a few percent. If you can't afford these few percent flexibility on your RMS, then you're not being realistic expecting to detect sources at only a few sigma. For blind detection experiments you need to aim for at least 8 sigma in your calculations.

As soon as you are moving a large angular distance on the sky (about 20 degrees) you need to perform a new pointing scan. That is, when moving between flux/bandpass calibrator and region of your target sources, or after one hour, when the sky has rotated 15 degrees. Also large temperature changes (such as near sunset) may require new pointing scans. Count on at least 3 minutes per pointing scan.

Add extra time for bandpass and flux calibration scans (typically a few minutes each, plus the extra pointing), slew, and 10 minutes startup time for each scheduling block. For example, if you need good uv-coverage you'd typically want to do a long track in an observing run of several hours. But if you are observing in popular LST ranges (or anticipate a relatively low priority), you may be better off in cutting the observing run in several separate hour-long blocks to benefit from short gaps left by higher-priority projects and accept the less homogeneous uv-coverage. Each observing run comes with 10 minutes startup overhead so if instead of an 8-hour track you decide to ask for four 2-hour observations, you need to ask for an extra half-hour. The best way to figure out how much you need exactly for all the overhead for your science observing is to make the observing schedules before submitting your proposal, but of course we're not requiring you to do that. Just make sure the total time request is realistic and reasonable.

Observing Preparation for High Frequency Observations

In this section we will discuss several items which should be considered when preparing for high frequency observations using the Observation Preparation Tool (OPT). The OPT manual is at https://science.nrao.edu/facilities/evla/observing/opt .

Observing Strategies

Flux Calibration Absolute gain calibration is difficult in the largest arrays at the highest frequencies, principally because both 3C48 and 3C286 are resolved. Currently the recommended primary flux calibrators are: 1331+305 (3C286) and 0713+438. If you cannot get to either of those, 0410+769 and 0319+415 (3C84) are also acceptable, although 0319+415 may be variable, and 0410+769 has uv restrictions (see the EVLA Calibrator Manual) and may also be variable. The most recent models are currently available at: https://science.nrao.edu/facilities/evla/data-processing/data-processing/flux-calibrator-models-for-new-evla-bands Once finalized (following the 2011 A configuration), these will be incorporated into the existing data reduction packages (CASA/AIPS). For most observing projects, the effects of atmospheric extinction will automatically be accounted for by regular calibration when using a nearby point source whose flux density has been determined by an observation of a flux density standard taken at a similar elevation. However, at high frequencies (most notably K-band, Ka-band, and Q-band), both the antenna gain and the atmospheric absorption may be strong enough to make `simple'.

Phase Calibration

Adequate phase calibration is a complicated function of source-calibrator separation, frequency, array scale, and weather. And, since what defines adequate for some experiments is completely inadequate for others, it is impossible to define any simple guidelines to ensure adequate phase calibration in general. However, some general statements remain valid most of the time. These are given below. Tropospheric effects dominate at wavelengths shorter than 20 cm, ionospheric effects dominate at wavelengths longer than 20 cm. Atmospheric (troposphere and ionosphere) effects are nearly always unimportant in the C and D configurations at L and P bands, and in the D configuration at X and C bands. Hence, for these cases, calibration need only be done to track instrumental changes - once per hour is generally sufficient. If your target object has sufficient flux density to permit self-calibration, there is no need to calibrate more than once hourly. The smaller the source-calibrator angular separation, the better. In deciding between a nearby `S' calibrator, and a more distant `P' calibrator, the nearby calibrator is usually the better choice. At high frequencies, and longer configurations, rapid switching between the source and nearby calibrator is often helpful.

For K band,

For Ka band

For Q band, tracking phase is particularly difficult, requiring phase calibration scans as often as every ten minutes or using the Fast Switching schemes described below.

Reference Pointing

Because the systematic pointing errors at the VLA generally are 10-20 arc seconds (but can be as bad as one arc minute), they are a significant fraction of the primary beam at 43 and 22 GHz (FWHP about an arc minute at 43 GHz and about two arc minutes at 22 GHz). Therefore, high frequency observers will want to make frequent pointing measurements in their observations. The basic procedure is as follows:

Determine pointing offset by pointing up on a nearby, bright point like source. This is done at C or X-band and full bandwidth to give maximum sensitivity. Tell the on line system to apply those corrections to any number of subsequent scans. Repeat (1) when the pointing has changed significantly, for instance the source may have moved to a different AZ and EL. New pointing corrections should also be done when there is a change of temperature which could alter the shape of the dish (e.g. sunset and sunrise). As a role of thumb, point up every hour or so (except when temperature is changing rapidly). The dwell time should be long enough to allow a five-point cycle, which means it should be at least 2.5 minutes. The integration time should be 1 second.

In the OPT, for Scan Mode on the Observing Preparation Tab specify "Interferometric Pointing". Use the NRAO Default Instrument Configuration "Primary C band pointing" or "Primary X band pointing". The results of the reference pointing scans will be provided with your observing log (request it from observe@nrao.edu if you did not receive it).

How often to point:

As a rough guide, pointing up every hour or so seems adequate, except when the temperature is changing rapidly. The most predictable instances of the latter are sunrise and sunset, around which times it is advisable to point up every half hour.

Flux density required for pointing calibrator:

For successful pointing calibration at X band using the standard 10 second integrations, the pointing calibrator should be at least 0.3 Jy. Stronger is better for flux densities up to about 1 Jy; for even stronger sources, thermal noise no longer dominates the errors in the solutions. Note that the source must be this strong on individual baselines! The VLA determines pointing corrections using amplitudes calculated from self-calibration of the interferometric visibilities, using a point-source model. A resolved source will work less well than one which is truly point-like.

Distance from the pointing calibrator:

A pointing scan should not be applied to sources more than 30 degrees in AZ or EL from the pointing calibrator - often the phase calibrator is suitable also for a pointing calibrator. Avoid observing sources within 10 degrees of the zenith where changes in AZ are too rapid to calibrate.

Size required for pointing calibrator:

The source used for pointing calibration should obviously be small compared to the primary beam, to avoid confusing pointing errors with source structure. A more stringent limit is that there be enough flux on individual baselines to allow ANTSOL to find a solution; and that any structure larger than the synthesized beam should not cause significant variations in the gains reported by ANTSOL on the timescale of the pointing scan (a few minutes). These considerations imply that EVLA phase calibrators suitable for the array and wavelength in question will be the most reliable pointing calibrators as well.

Pointing at high elevations:

The pointing model for the EVLA is particularly bad at elevations above 80 degrees (Yun 1997), and the residual pointing errors change rapidly with sky position there as well. Furthermore diurnal source motion near zenith can lead to rapid changes in azimuth as the source transits, again giving large differences in the residual pointing errors from one moment to the next. Both the a priori and referenced pointing therefore do a very poor job at high elevations.

What happens during a pointing scan:

The pointing scan itself consists of a series of five-point pointing cycles. The phase center of the array is kept constant at the nominal position of the calibrator. For each of the five positions, ANTSOL derives the antenna gains via self-calibration of the interferometric visibilities using a point-source model. The amplitude gains from the five ANTSOLs are collected, and the beam width and pointing offset in Az and El derived for each antenna. The offsets are considered valid and recorded only if three conditions are met: There must be enough flux at each pointing, as reported for each antenna by ANTSOL, to be able to determine a statistically significant solution. The offset and beam width that are estimated must be reasonable. There must be valid solutions for both polarizations, since the offsets stored are the average of the offsets determined for each polarization.

Doppler Setting

The EVLA provides the capability of calculating the appropriate sky frequencies for a specified sky position, rest frequency, velocity, rest frame and velocity type. This is done only at the start of a scan (and currently is only available for OSRO modes; for RSRO modes, one must still use the Dopset program to calculate the sky frequencies for a particular date range - given the dynamic scheduling of the EVLA, these should be updated on a time period to avoid shifts in the line for multi-epoch observations.

Fast Switching Due to variations in the water vapor content of the troposphere, there will be an additional source of phase noise when observing at high frequencies. Tropospheric phase fluctuations can be characterised by a "root phase structure function", corresponding to the rms phase variations as a function of baseline length:


where is in cm and beff is in km. The value of K varies according to the weather, with K = 20 under good weather conditions at the VLA site (typical fall and winter nights), K = 30 under average weather conditions at the VLA (fall and winter days, summer nights), and K > 60 under poor weather conditions (summer days). The exponent n = 0.7 for baselines shorter than about 1 km, and n = 0.33 for baselines longer than 1 km. The characteristic timescale for these variations is simply the baseline length divided by the wind speed in the turbulent layer, typically about 10 m/s.

There are currently three ways to deal with those phase fluctuations, and on this page we will discuss method number 3:

If your source is stronger than 100 mJy, you can apply self-calibration. If your target source contains a maser source (e.g. SiO), you may want to exploit the technique of self-calibration by monitoring the atmospheric phase fluctuations using the maser in one IF and applying the solution to data taken in the other IF. If your source is weak and you observe in A or B array, you can use the method of fast switching. This is just calibration of your phases using an external calibrator, however the cycle time is short enough to 'stop' the phase fluctuations. Normally, there is a 20 seconds overhead each time the VLA slews to a new source. The fast switching mode at the VLA has been introduced to avoid those extra 20 seconds, and thus allows a much shorter cycle time. It has been shown that fast switching phase calibration at the VLA will stop tropospheric phase fluctuations at an effective baseline:


where tcyc is the calibration cycle time and va is the velocity of the winds aloft (typically 10 m/s at the VLA site). Hence, fast switching will be effective for cycle times shorter than the baseline crossing time of the troposphere b/va. Currently fast switching is considered useful in A and B arrays, but may also be useful at more compact configurations if your sources are at low elevation.

How to determine a suitable cycle-time

In order to determine your cycle time, you need to decide how large phase fluctuations are acceptable in your observations - note that a 90 degrees phase RMS will easily wipe out a source. The table below shows a typical RMS phase in degrees for the VLA at 7 and 13 mm after calibration. These values are derived with the Site Test Interferometer. Measurements of phase were made by observing a geostationary satellite beacon (at 50 deg elevation) at 11.3 GHz with two 1.8 m satellite dishes separated by 300 m. The results were then scaled to 7 mm and 1.3 cm. (To estimate the RMS phase noise at 1.3 cm, simply double the cycle time between phase calibrator and source from Q band.)

RMS phase (in degrees) for the VLA at 7mm and 1.3cm after calibration under median conditions

7 mm tcycle=2 min tcycle=5 min tcycle=10 min 1.3 cm tcycle=4 min tcycle=10 min tcycle=20 min Month Day .. Night Day .. Night Day .. Night Day .. Night Day .. Night Day .. Night Jan 25.5 .. 18.8 47.1 .. 34.6 74.7 .. 54.9 25.5 .. 18.8 47.1 .. 34.6 74.7 .. 54.9 Feb 25.0 .. 15.1 46.1 .. 27.8 73.2 .. 44.2 25.0 .. 15.1 46.1 .. 27.8 73.2 .. 44.2 Mar 30.2 .. 20.9 55.7 .. 38.4 88.4 .. 61.0 30.2 .. 20.9 55.7 .. 38.4 88.4 .. 61.0 Apr 45.4 .. 24.5 83.5 .. 45.1 132.6 .. 71.6 45.4 .. 24.5 83.5 .. 45.1 132.6 .. 71.6 May 38.1 .. 21.9 70.1 .. 40.3 111.3 .. 64.0 38.1 .. 21.9 70.1 .. 40.3 111.3 .. 64.0 Jun 39.1 .. 20.3 72.0 .. 37.5 114.3 .. 59.5 39.1 .. 20.3 72.0 .. 37.5 114.3 .. 59.5 Jul 41.7 .. 27.6 76.8 .. 59.0 122.0 .. 80.8 39.1 .. 20.3 72.0 .. 37.5 114.3 .. 59.5 Aug 51.1 .. 29.7 94.1 .. 54.7 149.4 .. 86.9 51.1 .. 29.7 94.1 .. 54.7 149.4 .. 86.9 Sep 62.0 .. 34.9 114.3 .. 64.3 181.4 .. 102.1 62.0 .. 34.9 114.3 .. 64.3 181.4 .. 102.1 Oct 51.6 .. 30.8 95.1 .. 56.7 150.9 .. 89.9 51.6 .. 30.8 95.1 .. 56.7 150.9 .. 89.9 Nov 32.3 .. 29.7 59.5 .. 54.7 94.5 .. 86.9 32.3 .. 29.7 59.5 .. 54.7 94.5 .. 86.9 Dec 22.4 .. 15.6 41.3 .. 28.8 65.5 .. 45.7 22.4 .. 15.6 41.3 .. 28.8 65.5 .. 45.7 Thus decreasing your fast-switching cycle time from 5 minutes to 2 minutes under median conditions will improve the final, measured rms phase noise in your map by about a factor of 2. The minimum allowed time per source is 20 seconds, implying a minimum total cycle time of 40 seconds.

Winter observations are clearly best for high frequencies. Try to schedule observations during the night when phase stability is likely to be better. If you have to observe during the day, then choose a short cycle time to keep the phase rms below 30 degrees (or at worse 40 degrees). If you are doing a detection experiment, try for the best conditions possible.

Note that the amount of effective on-source time depends on the distance, and orientation, of the slew between the source and calibrator. Ensure that you get enough time on source (see below). As an example, in a recent test observation of a source and calibrator separated by 1.8 degrees, a minimum cycle (20 sec for source/20 sec for calibrator) yielded 7 seconds of on-source time for each source. It has been found that a cycle time of about 100 seconds (70 sec on source, 30 sec on calibrator) is adequate under average observing conditions at the VLA, for source-calibrator separations less than 4 degrees or so. In this case the expected residual tropospheric phase fluctuations under good weather conditions at the VLA will have an RMS of about 18 deg at 7mm after calibration for all baselines longer than about 500m, and less than this on shorter baselines.

How to implement fast switching in OPT=

For fast switching phase calibration a minimum calibrator flux density of 0.3 Jy is recommended. It is also recommended that every hour or so an observation be made of a stronger calibrator (1 Jy or greater), to determine pointing corrections and to track amplitude variations. In the OPT, fast switching is achieved by setting scan loops.

Atmospheric Phase Interferometer

The NRAO has recently installed a Atmospheric Phase Interferometer (API) at the VLA site, and software has been installed for real time monitoring of the phase stability. The API measures continuously the tropospheric contribution to the interferometric phase using an interferometer comprised of two 1.5 m dishes separated by 300 m observing an 11.3 GHz beacon from a geostationary satellite. The primary purpose of the API is to serve as a guide to high frequency observers at the VLA. In particular, the API data are meant to be used to estimate the required calibration cycle times when using fast switching phase calibration, and in the worst case, to indicate to the observer that high frequency observing may not be possible given the weather conditions. Detailed instructions as to the use of the API data can be found on the Atmospheric Phase Interferometer home page.


Observer's Checklist Choose observing strategy Are your sources strong (at least 0.1 Jy, preferably stronger)? Then you can apply self-calibration to the source, and it is sufficient to observe the calibrator every 30 minutes at 22 and 43 GHz.

Are your sources weak? Apply either the fast-switching technique or, if your source contains a maser you can use the maser for self-calibration. Select calibrators

Primary flux calibrators: Try to observe 3C286 (or 3C48 as a secondary choice) to achieve absolute gain calibration. Note however, that especially in A- and B-array you will need to use clean component models since both 3C286 and 3C48 are resolved at high frequencies.

Normal calibrators: from the List of EVLA Calibrators, select compact calibrators that lie as close as possible to your target source. A flux density of around 0.5 Jy is recommended at full bandwidth, and preferably stronger at narrower bandwidths.

Fast Switching calibrators: These calibrators need to be at least 0.3 Jy, and to have a maximum distance from your target source of around 3 degrees.

Bandpass calibrators: if you are observing spectral lines, ensure you have also a stronger (a few Jy if possible) calibrator in order to perform bandpass calibration.

Suitable pointing sources: referenced pointing scans have to be done at sources which are at least 0.5 Jy, and should be as close as possible to the target.

Setup Instrument Configuration:

Continuum mode: use available NRAO standard setups but consider: Time averaging: make sure you have short enough integration time in order to reduce time averaging losses. Spectral line modes see table.

Velocity resolution: make sure the spectral resolution is enough to resolve your line.

Hanning smoothing on line is possible, with a cost of a factor of 2 in spectral resolution.

Make sure you have enough line free bandwidth to measure the continuum flux.

Time averaging: make sure you have short enough integration time in order to reduce time averaging losses.

Integration time: the minimum possible integration time depends on the total number of channels produced in the correlator.

Check your scheduling block:

General: Are your sources up? Check in the SummaryReport.

Have you included referenced pointing scans at C or X-band?

Is the integration time for those scans 1 second?

Is the total dwell time for the pointing scans at least 2.5 minutes?

Are the pointing corrections applied to the correct scans?

Are you using Scan Timing=Duration(LST) and not On Source?

Are your source coordinates entered correctly and in the same epoch (J2000 or B1950)?

Have you integrated long enough to achieve your desired sensitivity?


Are your AC and BD IFs at different frequencies?

Is your integration time short enough to reduce time averaging losses?

Spectral line considerations:

Are the rest frequencies of your lines correct?

Is the velocity definition correctly specified?

Is the rest frame correctly specified?

Fast switching:

Did you check that the slew time between your sources is shorter than the cycle time? In normal mode (not fast switching mode) the slew time is the move time minus 20sec.

Is the integration time as short as possible for the configuration?