Advice for designing high precision photometry observations (exoplanets, brown dwarf weather)
Updated 28 July 2016
Observations to obtain high precision relative photometry with Spitzer need to be carefully planned to minimize the effect of instrument and observatory systematics that add correlated noise to the time series. This memo suggests some best practices that should be followed to ensure the cleanest raw data. While these recommendations will mitigate sources of systematic noise, they will not remove it entirely and decorrelation methods will need to be applied to the input time series before analysis of the light curve. In addition, the history of the telescope pitch angle with respect to the Sun, target attitude in the spacecraft-sun reference frame and other currently not well characterized parameters will cause the suggested observing strategy to be more or less effective on a case by case basis. The basic strategy is to obtain a measurement of existing residual images at the position that the source will be observed. This is done in a period of telescope attitude settling after slewing to the source. The source is then placed on a position of minimal photometric variation of a well characterized pixel (the "sweet spot") for the observations. This precise pointing is enabled by use of the PCRS peakup imager. Depending on the length of the observation, re-pointings may be necessary to keep the source from drifting off the region of minimal variation. Finally, an observation is executed to measure the residual image on that pixel at the end of the time series.
An Example 40-hour Observation with PCRS Peak-Up
For a 40-hour staring mode observation you need to specify in Spot:
- A 30 minute dithered AOR with target not on sweet spot
- a 12-hour PCRS peak-up AOR with target on sweet spot
- a second 12-hour PCRS peak-up AOR with target on sweet spot
- a third 12-hour PCRS peak-up AOR with target on sweet spot
- a 4-hour PCRS peak-up AOR with target on sweet spot
- A 10 minute dithered AOR with target not on the sweet spot
Chain all the AORs together so that they will be observed back to back (you can safely ignore any warnings that Spot may give about a chain being longer than 24 hours)
The initial 30 minute and final 10 minute AOR
The goal of the first AOR is twofold. First, the AOR is long enough in most cases to remove the initial slew settling that occurs when acquiring a new target. To allow the telescope position to settle, we recommend this 30 minute AOR be chained in front of the long scientific staring mode AOR to eliminate the initial drift. Traditionally, this pre-AOR has been observed in staring mode to be consistent with the science observations.
However, if observers will not be using these 30 minutes in their science analysis, we recommend dithering and not using PCRS peak-up for that AOR . This allows observers to generate their own local-in-time dark frame for the science observations which can be used to remove persistent images, thereby increasing final photometric precision. Persistent images have been shown to be potentially a 0.1% effect in photometry in comparing multiple epochs depending on the channel, frametime, and previous observations in the schedule. Additionally, we recommend that observers insert a short dithering AOR (~10 minutes) after their science AOR, to capture the most accurate dark image for their dataset. This post-AOR should have the same frametime and dither pattern as the pre-AOR, it just does not need to be 30 minutes long to accomodate the telescope position settling. The reason to use the same frametime and dither pattern is that the dark pattern will change if these things change, and we only want to measure a background pattern change due to latents. A combination of cluster targets and dithers may be used to build the necessary redundancy to produce a localized dark frame. The source should be dithered off the frame if possible to produce a representative background. PCRS peak-up is not necessary for the dithered AORs since the goal is to not have the source on the sweet spot. Additionally, not using PCRS peak-up will save a few minutes of observing time.
If using the subarray, the following two options are available for generating dithering AORs. Generating dithered AORs for full array includes these options as well as a host of other dither patterns, and it is much easier to achieve a well dithered final local-dark since 64 frames are not taken at one position.
- Use one of the available dither patterns - select "dither pattern" = yes , choose either a 4 position Gaussian or a 9 position Reuleaux pattern. The challenge with this option will be in generating AORs that are 30 minutes in duration, since the patterns are fixed in number. The pre-AOR may be longer than 30 minutes, but this will be charged to your total requested time. Each BCD file of 64 frames will then be observed at a different position.
- Use a Fixed Cluster target (not position) - do not select a dither pattern, instead when generating the target for the observations, select it as "Fixed Cluster" and then enter in as many offset pointings as you wish to observe. You will likely want to choose "Array" offset coordinates and enter the offsets in units of pixels. Similarly, select "Observe the offsets only" so that there is never a target at the sweet spot. Although this allows observers freedom in the total duration of the AOR, it will require two different targets for the observation, since the target for the science observation needs to be a different "Fixed Cluster" to get the correct offsets for PCRS peak-up as described above.
The PCRS Peak-Up Science AORs
For PCRS peak-up to succeed, the following must be true:
- The peak-up target must be in the magnitude range 7.0 mag < V < 12.1 mag.
- There should be no stars brighter than about 13.0 V mag within 40 arcseconds of the peak-up target.
When adding PCRS peak-up to your observation, you have the choice of peaking up on either the science target itself or an offset star. If the visual magnitude of your target is between 7 and 12.1, then we recommend you use the target itself. Otherwise you can either use the "PCRS selection" button in Spot to choose a nearby star from the PCRS Guide Star Catalog, which contains a carefully selected subset of stars from the Tycho catalog, or specify your own offset star. It is possible that there are stars within the 7 - 12.1 V mag limit of PCRS that are not in the catalog because the catalog has a different brightness limit. If Spot does not find a catalog star, we recommend that the user perform a search for stars in SIMBAD (Coordinate query) within a 30 arcminute radius of the target's coordinates and select a suitable star from the list of found objects. An overlay of the Guide Star II catalog (GSC-II) in Aladin is recommended to weed out candidate peak-up stars that have bright companions around them. For proper motions we recommend the use of the Tycho-2 catalog (available in VizieR). Enter the peak-up star information manually in Spot by clicking on the "PCRS peak-up" tab and then select "Position" in the "Type" box. The coordinates, epoch of the coordinates and the proper motions (in arcsec/year; for Right Ascension this is the R.A. proper motion times cosine Dec, as directly given in, for example, the Tycho-2 and SIMBAD listings). Finally, a V-Magnitude must be entered in the "Specifications-Position" box. Keep in mind the requirements above in choosing your peak-up star (both of which are automatically met by the PCRS Guide Star Catalog).
We have a code available that helps with selecting the best available peak-up star. Choosing the best star requires a rough optimization of position error, visual magnitude, and angular distance. Brighter stars are more desirable because the peakup integration time will be shorter, although all brightnesses are feasible. Only catalog stars with angular distances less than ~half a degree are shown as viable catalog stars. If this tool shows a star which is not listed in Spot, and it is your choice for the best catalog star, you can always enter it manually as a PCRS peak-up star. The catalog used here is exactly the same as that used in Spot.
Proper motion errors are the largest source of position errors. Observers should consider both the position errors of their targets as well as the PCRS catalog star in use. The goal in choosing a catalog star is to have a combined error on the position of the catalog star plus the position of the target star which is small compared to the width of the sweet spot (0.35 arcsec). If there are equally good candidate peak-up stars with respect to positional errors, the star that has the best combination of being closest to the science target and being the brightest within the limits specified above should be selected.
If the proper motion errors on either your target or available catalog stars are too large, we recommend doing PCRS peak-up observations of the catalog star and the target to derive individual offsets for that particular star pair. These pre-science observations are necessary to determine the relative positions of the peakup star and target. They must be performed at least 6 weeks prior to the scheduling of the science observations. We recommend using six sample AORs that perform 10 repeat exposures (no dithering) on the science target each, to build up statistics on those offsets. Each AOR should be a peak-up AOR on the exact catalog star planned for use in the science AORs. You will then be able to transform the actual measured positions of where the target falls on the pixel in the pre-observation frames to a cluster offset in the target definition for future AORs which will shift the target onto the sweet spot. Contact the SSC Helpdesk for the transformation equations.
See the Spot Version 19 Release Notes for the mechanics on how to use Spot to add PCRS peak-up to an AOR.
Putting Your Science Target on the "Sweet Spot"
The position of minimal gain variation is not exactly at the default pointing center for the subarray and small offsets to the source position need to be specified. After peak-up, you can ensure that your target is placed in the "sweet spot" region by specifying your target as a Fixed Cluster target in Spot, with Array (Row,Col) offsets given in the table below. (MODE = readout mode)
|Ch||Mode||FOV||Array Row (")||Array Col (")|
Note the following:
- You should select Offset Coordinates "Array" (Row/Perp, Col/Para) when specifying the cluster target.
- Make sure to check "Observe the offsets only" when specifying the cluster target.
- The IRAC Instrument Support team is only characterizing the Channel 1 and Channel 2 Subarray sweet spots, so to use the available gain maps to correct your photometry, you should use the offsets given for FOV=Sub for your given readout mode (Full or Sub). Also, the majority of warm high precision photometry observations use these well-characterized subarray pixels; therefore, there are ample datasets to use to characterize the behavior of these pixels if you wish to do so.
- If you choose Mode=Full and FOV=Sub, you will receive one full array image per exposure ("Number of Frames"), with the target placed on the Subarray sweet spot located on pixel [24,232] on the full array. Note that we use the convention that the center of the bottom left BCD pixel is labeled [1.0,1.0].
- To get the best results, you should use staring mode ("Mapping Mode" = No and "Dither Pattern" = No) and use only one FOV per AOR. Offsetting to another FOV or to multiple positions within a map will be less accurate than the initial PCRS to Science Instrument offset and will eliminate the benefit of using PCRS peak-up.
Breaking Up Long Stares Into Multiple AORs
One important scientific benefit of the PCRS peak-up is the ability to reposition a target onto the same spot on a pixel during a long stare, thus enhancing the precision of time-domain measurements. Typically, the Spitzer pointing system drifts systematically by approximately 0.35" per day. This means that in about 12 hours a target placed halfway between the center and the edge of the well-characterized 0.5-pixel "sweet spot" of the Channel 1 or Channel 2 subarray has a good chance of drifting out of the calibrated region. To eliminate this problem, we recommend breaking up all staring observations that are longer than 12 hrs into separate AORs lasting 12 hrs or less, all of which should have PCRS peak-up (see below for revisions to this for guide star peak-up). The AORs should be identical copies of each other, with the exception being that the Number of Frames should be changed to make each AOR shorter than 12 hrs and obtain the total duration as desired. They should also be linked together using Chain constraints, to ensure that they are scheduled as a single unit. We do not recommend peak-ups more often than about 10 hours, however, as the reacquisition error using peak-up is of order 0.1 arcsecond, which is comparable to or in excess of the expected pointing drift for observations shorter than about 10 hours. These minimum and maximum duration recommendations were changed to be more conservative regarding the error in re-acquiring the source.
After examining the statistics of PCRS Peak-Up over the past 3 years, we are revising slightly our recommendation that states observers should break up long AORs into 10-12 hr pieces, with a Peak-Up at the beginning of each (October 2015).
Guide Star Peak-Up
- Ch 1 (512 measurements) 40% in sweet spot, 68% within 0.4 px of center of sweet spot, 68% repeat visit RMS < 0.09 px
- Ch 2 (531 measurements) 61% in sweet spot, 68% within 0.3 px of center of sweet spot, 68% repeat visit RMS < 0.10 px
When the Peak-Up is on a different star than the science target (“guide star Peak-Up”), the repeatability is somewhat worse than originally expected from early statistics. More than 40% of all guide star PU measurements do not place a target within the half pixel region surrounding the peak of response (the “Sweet Spot”). Part of this is due to errors in relative astrometry between the Peak-Up star and the target. However, even repeat visits to the same target (using the same Peak-Up star) have an RMS spread in initial position worse than about 0.1 pixels 32% of the time, which risks positional discontinuities between AORs. Therefore we are doubling the recommended duration of AORs with guide star Peak-Up, to 20-24 hr.
- Ch1 (208 measurements) 90% in sweet spot, 68% within 0.1px of center of sweet spot, 68% repeat visit RMS < 0.06 px
- Ch2 (520 measurements) 98% in sweet spot, 68% within 0.08px of center of sweet spot, 68% repeat visit RMS < 0.06 px
Our analysis of Self Peak-Up shows that more than 90% of AORs make it inside the sweet spot, the region of peak response. Furthermore, repeat visits to the same target result in a spread of initial AOR positions smaller than 0.06 pixels 68% of the time. This is comparable to (or less than) the size of pointing wobble plus jitter (about 0.08 px), ensuring significant positional overlap from one AOR to the next. If the Peak-Up is on the science target, we continue to recommend peaking up every 10-12 hr.
In a small fraction (about 10%) of self Peak-Up cases, repeated visits have resulted in greater than 0.1 pixel differences in initial positions. Those observers using a data analysis technique that is indifferent to a larger drift in position away from the sweet spot, and who wish to avoid a discontinuity in position between AORs, should consider longer AOR durations.
In summary, you should make sure your AORs fulfill the following conditions:
- Long stares are broken up into chained 10-12 hr AORs (or 20-24 hr AORs for guide star Peak-Up), each containing a PCRS peak-up.
- A 1/2 hour long AOR with dithering should be added to the beginning of each chain with the same channel and frametime as the subsequent AOR in the chain. Additionally, we recommend that a short (~10 minutes) dithering AOR be added to the end of the science AORs.
- All science AORs in a chain have the same target position and proper motion, and their peak-up target coordinates are the same.
- The target should be specified as a "Fixed Cluster", and the offsets should be specified as given in the table above for the current channel, readout mode, and FOV. (The SSC is only building an intrapixel gain map for the subarray sweet spot, so we strongly recommend using the offsets corresponding to the Subarray FOV.)
- "Array Row/Perp, Col/Para" should be selected for the "Offset Coordinates", and "Observe the offsets only" should be selected.
- The peak-up V magnitude is within the acceptable range (7 to 12.1).
- All AORs with the same peak-up target have the same peak-up V Magnitude.
- All AORs in a chain have the same readout modes (subarray/full array) and frametimes, per channel.
- Only one channel is selected under "Field of View".
- Please turn on "Data Collection" in both channels unless you are using 0.4 or 2 second full array frames or 0.02 second subarray frames. In general, you will obtain better results using the subarray versions of the 0.4 and 2 second frametimes.
- If you have specified a proper motion for your main or peak-up target, the given position should be consistent with the given Epoch (i.e., the current equinox J2000 position of the target may be derived from the given position by multiplying the proper motion by the time between the given Epoch and the current epoch).
- If your target is within 20 pc of the Sun, please contact the SSC Helpdesk so that we can adjust the coordinates to account for annual parallax.
Perl Script to Check AORs
We have made available a perl script, check_peakup_aors.pl, that may help you review your AORs before submitting to the SSC.