JCMT Observing modes, Durations, and Noise

Contents

• Jiggle Chop
• Jiggle patterns
• Loop control
• Shared/Separate offs
• Duration
• Noise
• Jiggle Position-Switch
• Jiggle Patterns
• Loop Control
• Shared/Separate offs
• Duration
• Noise
• Grids Postion-Switch
• Grids patterns
• Loop Control
• Shared/Separate offs
• Duration
• Noise
• Raster maps (not included yet)
• Comparing observing modes

Jiggle-Chop observations

General description:

The advantage of JCHOP observations is that the patterns can be executed fast, with minimal telescope motions, and that the frequent sky observations result in relatively flat baselines. JCHOP observations are limited to targets that fit within the maximum throw of the secondary mirror of 180". Both the size of the throw and the direction can be set by the observer in the MSB for the observation, but a customary choice is to chop in AZ i.e. at a constant airmass. If there are no sky positions free of emission within 180" of the target, the observer will have to select a position-switched observing mode instead.

Jiggle patterns:

• A receptor-centered jiggle pattern places the target in the centre of the region mapped by one of the receptors and it will also be on one of the pixels in the output map. All NxN jiggle patterns such as 3x3, 4x4, or 5x5, are receptor-centered and primarily intended for use with the single-receptor receivers or dual-pol receivers with two receptors on the same position on the sky.

HARP does not have a central receptor. When used with a receptor-centered jiggle pattern it will align one of the inner 4 receptors with the target instead. Hence, when a NxN jiggle pattern is selected with HARP the target will be far off-center in the resulting map, or, stated differently, the map will be very assymetric around the target position. The orientation of the map is determined by position of the K-mirror, which can choose any of 4 angles, differing by 90 degrees, depending on the elevation of observation. In short, only a region of 1.5x1.5 arcmin will be mapped predictably. Combining observations from different elevations may result in a windmill pattern around that region:

The four special patterns intended for use with HARP that fully sample the footprint of the array are harp4 and harp5, coming in two variations each. The receptors in HARP are spaced 30" apart and these patterns fill the 30"x30" post-stamp of each receptor with '4 by 4' or '5 by 5' points, spaced 7.5" or 6" respectively. Compared to a Nyquist sampling of the nominal beam at 345 GHz, the harp4 patterns slightly under-sample, whereas the harp5 patterns over-sample. Harp4 is the recommended pattern for general map making and the cube resulting from the 4x4 receptors of HARP will have 16x16 pixels in e.g. RA, Dec and cover a region of 2x2 arcmins. The harp5 pattern will have 20x20 pixels in RA, Dec covering the same region.

Harp4 and Harp5 can be pixel-centered or map-centered. To understand the distinction one needs to realize that the spatial dimensions of a HARP cube will always have an even number of pixels i.e. won't have a central pixel. This presents two choices of where to put the target in the final map:

• A pixel-centered jiggle pattern will place the target on one of the central 4 pixels of the output map, slightly offset by half a pixel in each direction from the center of the map. As is the case for the receptor-centered map, discussed above, the map will thus be asymmetric around the target but only by one map pixel. Due to the possible orientations of the K-mirror the edge of the map will this be uncertain by 7.5" or 6", respectively. The pixel-centered harp jiggle patterns are named harp4 and harp5 in the JCMTOT.

• A map-centered jiggle pattern will place the target exactly in the center of the output map, precisely between the central 4 pixels. The resulting map will be symmetric around the target position and independent of the K-mirror orientation. However, none of the map pixels will align exactly with the target position, which should not be an issue for an extended source, but might be a concern for a source with a point-like core. The map-centered harp jiggle patterns are named harp4_mc and harp5_mc. Map-centered patterns are also referred to as array-centered since the target is placed on the center of the array between the 4 receptors.

The four harp patterns are illustrated below (click on each pattern to show them in a pop-up window). The grey (cyan?) dots in each pattern are the jiggle positions and the centers of the pixels in the output map. In each figure the pointing target is indicated by the small circle and either aligns with a pixel or the center of the map. (For completeness: A receptor-centered pattern will select a configuation in which the blue cross coincides with a map pixel, but with the target circle on one of blue crosses instead of close to the center.)

Loop control:

The dwell time per jiggle position is calculated to be as long as possible to allow the whole pattern to be completed within the max_time_between_chops, but not to exceed 2s per point. For a 16-pt jiggle this means about 0.56s/point and for a 25-pt jiggle 0.36s/point. Note: while the time between chops of 9s is shorter than the 30s typically used with position-switching, it is much longer than what is required for an accurate determination of continuum emission. For this observers can select continuum mode in the their MSB which changes the max_time_between_chops to 0.1s and forces a sky reference for every jiggle position: the observing mode changes to chop-jiggle (chop-before-jiggle) instead of a jiggle-chop (jiggle-before-chop). However, the frequent chops significantly increase the overheads: limited testing so far suggests that the typical penalty for using continuum mode is an increase of the observing time by a factor of 1.7CHECK!.

Step_time, max_time_between_chops, and max_time_between_nods have been set to defaults that can not be changed by observers directly. Since their optimal settings may depend on e.g. the weather, opacity, and seeing, they are however under the control of the operators, although this is not yet being actively used at present. PIs of science programs that would significantly benefit from settings different from the defaults should contact their assigned support scientist.

Shared vs. Separate offs:

Observations benefit from shared offs in two ways and the more so the more points between the offs (the larger the raster-map). To illustrate, consider 1s observations of 16 positions (which would be a very small raster map). Separate offs would require 32s on the sky equally split between ons and offs. Using shared offs the observation needs only 16+4=20s. But also for each point individually the noise is reduced because of the longer observation of the off. Expressed in a time required to reach a particular rms the combined effect makes shared off observations 4 times faster than separate observations in the limit of many points. More precisely, the factor is ~4/(1+2/sqrt(np)) or ~2.7 for the relatively small HARP jiggle patterns and ignoring fixed overheads (which reduce the factor to closer to ~2). Nevertheless, one night instead of two nights is a huge gain, which is why shared offs are the default for JCHOP observations.

Of course, the above gain comes at a price: all 16 or 25 jiggle-positions from every receptor are correlated because they have the same sky reference (off). As long as the pixels are kept separate this is not an issue because noise remains dominated by the on measurement and the points can be regarded to be only weakly correlated. However, any operation in which pixels are combined will 'suffer' from this correlation in the sense that it will combine pixels that are not fully independent and the noise will hence not drop with the customary sqrt(np) factor. Note that this both applies to operations that use a spatial convolution (smoothing) as well as total flux measurements that sum or average over pixels such as aperture photometry. In the extreme, when averaging data over the whole footprint of a harp4 observation, the noise will be 1/4th (16 independent receptors) instead of 1/16th (196 pixels) of the noise per pixel. Finally: note that JCHOP continuum mode observations always will have separate offs because there will be only a single jiggle position per sky reference.

In conclusion: if imaging or velocity/frequency data are the primary aim of the observations, shared offs are hugely beneficial and recommended; if accurate flux measurements over extended regions are the primary aim or the final maps will be spatially smoothed significantly, separate offs are recommended instead.

Duration:

Noise:

The expected noise can thus be calculated using:

    rms = 1.04 * Tsys*1.23/sqrt(df*t_on) * sqrt(1+1/sqrt(np))

    1.04: 4% excess noise, presently unexplained
    1.23: Correlator digitization factor
    df:   channel-spacing in Hz
    t_on: 'on'-only integration time in sec
    np:   1 for separate offs else the number of point 
          sharing an off. For patterns not exceeding 9x9
          this is the number of points in the jiggle pattern.

Jiggle-Position Switch observations

General description:

This mode is recommended for mapping ≤ 2 arcmin fields where an emission free position is relatively far away from the target. Jiggle-PSSW (JPSSW) observations utilize the secondary to observe a regular grid-pattern of points but, contrary to JCHOP, use a position-switch of the whole telescope to observe sky positions free of emission.

The advantage of JPSSW over JCHOP is that sky reference positions are not restricted to a distance within 180" from the target and can be as far as degrees away. The main disadvantages are the less accurate sky subtraction due to the larger switch and the longer time between sky references. Another disadvantage is the additional overhead resulting from moving the whole telescope to the reference position. The position can be specified as either a relative position or a fixed position. The switch will be to this single position only rather than two positions on diametrically opposite sides of the target, as is the case for JCHOP. (However, note that the reference position can be defined in a coordinate system other than RA, Dec resulting in it rotating from observation to observation in that frame). In order to minimize the number of moves, the telescope executes an OFF-ONs-ONs-OFF etc. pattern, with the switches pairs-wise associated.

Jiggle patterns:

Loop control:

The dwell time per jiggle position is calculated such that it is as long as possible to allow the pattern to be completed within the max_time_between_refs, but not to exceed 2s per point. As for JCHOP there is a continuum mode which changes the max_time_between_refs to 5s at the cost of more overhead. Note that position-switch observations are not well suited for measuring the continuum level, both because of their typically larger switch, as well as the relatively long time between sky references even in continuum mode.

Shared vs. Separate offs:

Duration:

Noise:

See Jiggle-Chop for a more detailed discussion

Grid Position-Switch

Grid patterns:

Loop control:

Similar to JPSSW, there is continuum mode for GPSSW, which changes the max_time_between_refs to 5s at the cost of more overhead. Note that position-switch observations are not well suited for measuring the continuum level both, because of their typically larger switch, as well as the relatively long time between sky references even in continuum mode.

Shared vs. Separate offs:

Finally, as soon as the integration time per grid-point becomes such that only a single grid-point can be observed before having to go the sky reference (i.e. 15sec by default) there will be only one point per reference and the distinction between separate and shared offs disappears: the observations will use separate offs.

In conclusion: for any observations with integration times ≥15s per grid-point the choice of shared or separate offs is irrelevant. For shorter integration times a choice for separate offs comes with a heavy penalty in the form of excessive overheads.

Duration:

Noise:

    rms = 1.04 * Tsys*1.23/sqrt(df*t_on) * sqrt(1+1/sqrt(np))

     np:  number of points with shared offs or 1 if none:
          max(1,int(30/t_on)) i.e. the number of t_ons that fit 
          in 30sec, but no less than 1. (30secs = time_between_refs.)

See Jiggle-Chop for more details.

Comparing Observing modes

• Most efficient: Jiggles with shared offs, either beam-switched (JCHOP) or position-switched (JPSSW). Compared to GPSSW or separate jiggles, these observations take 30% less time.
• Single-position, GPSSW, and JCHOP/JPSSW with separate offs: these observations don't get the benefit from the shared offs or accrue additional overheads due to e.g. a position switch to the sky reference.
• GPSSW observations with separate offs and integrations time < 15s per grid point: these have excessive overheads due to increased number of position switches to the sky reference.

In addition to taking less time, JCHOP and JPSSW with shared offs will have a systematically lower noise as well. This is shown in the next two, side-by-side figures. The figure on the right zooms on the lower-left region of the figure top-left. Using the equations above we find that compared to t_on_nonsh = 30s, shared off observations with 16 points will need t_on_sh = 18.75s to reach the same rms. Using the equations for duration of JCHOP observations at the top which include the fixed overheads, one then finds that the shared offs JCHOPS will take about half the time as the separate offs JCHOPS: 15min instead of 33min as is illustrated in the final figure.

The figure below shows the benefits in terms of the duration of the HARP4 jiggle observation, including realistic overheads, needed to reach a certain rms (assumed Tsys = 250 K and frequency channels of 1 MHz). Note that shared/separate offs are only relevant within a single observation, which typically takes 10-20 minutes: the noise in coadds of successive observations will scale as usual with the sqrt(nr of observations).