Nod & Shuffle

Introduction

• Improved sky subtraction
• Potential increased density of slits

• Increased observing overheads
• Potential decreased field of view

Further details about the Nod & Shuffle mode are given below, including:

• Basic concepts
• Band and micro shuffling
• Nodding
• Observation definition
• Overheads
• Performance
• Focal plane masks
• Example masks and data
• Data reduction
• Commissioning team

The Nod & Shuffle Concept

Nod & Shuffle addresses the problem of subtracting night sky emission by frequently nodding the telescope pointing between an object position and a sky position while simultaneously shuffling the charge on the CCD detectors between science and storage (unilluminated) regions. This is similar to the common practice in near-infrared astronomy of nodding or beam-switching with the exception that the detector is not read out after each nod as is the case for background-limited near-infrared observations which do not suffer from read noise. The resulting image produced by the Nod & Shuffle technique contains two spectra obtained quasi-simultaneously through each slit in the focal plane mask - one of the object and the one of the sky. Though these spectra are stored in different regions of the CCD detector, they were imaged with exactly the same pixels through identical optical paths. The effects of pixel response (flat-field), fringing, irregularities in the slit, and temporal variations in the sky cancel out when one subtracts the sky spectrum from the object spectrum. For long exposures, one can realize a factor of 10 times improvement in the systematic uncertainties associated with subtraction of bright sky lines, especially in the red (600-1000 nm) where such errors typically dominate over photon or read noise. There is a trade-off, however; the noise in sky-subtracted Nod & Shuffle spectra is higher (up to a factor of sqrt(2) compared to classically sky-subtracted spectra using a very long slit.)

Because with Nod & Shuffle one no longer derives the sky from regions adjacent to the object, one can use significantly shorter slits than in classical MOS spectroscopy which typically uses 5-10 arcsec long slitlets. In the limit where the slits have the same length as the object size (or the seeing disk) one can have a density of these ``micro-slits'' which is therefore 5-10 times higher than in classic MOS mode. This can be particularly advantageous when attempting MOS spectroscopy of crowded fields. However, because part of the detector is used for charge storage and therefore can not be illuminated, one necessarily loses between 50 and 66% of the GMOS field of view with Nod & Shuffle, depending on the Nod & Shuffle layout (shuffling mode) selected. A further disadvantage of Nod & Shuffle observing is the increased overheads described below. The spectral images produced by Nod & Shuffle are, naturally, different from those produced by classical MOS spectroscopy and the data analysis must be adjusted accordingly. However, additional tasks to handle the reduction of Nod & Shuffle spectral data will be included in the upcoming release of Gemini IRAF scripts so this should not be considered a disadvantage.

Choosing between Band- or Micro-Shuffling

Below we show three schematic layouts for Nod & Shuffle (click on each image for a larger view). The large, inset red square represents the GMOS imaging field of view (this is the area within which one is allowed to place slits in classical MOS observing). The science regions are shown as white with black filled rectangles representing the slits. The red and blue shaded regions represent storage regions. Normally the red shaded regions store the science data when the sky position is being observed, and the blue shaded regions store the sky data when the science target position is being observed. Because the sky and target are not observed simultaneously, the red and blue storage regions are allowed to overlap on the detector. These examples are not to scale and are for illustrative purposes only.

	[a] The simplest possible layout is band-shuffling with a single science band. Here the middle third of the detector is the only science region, and the top and bottom thirds of the detector are used for storage exclusively. This band layout allows one to place a very high density of slits of variable length in a centralized region; however, one loses 67% of the field of view for slit placement.
	[b] This example has two science bands and three storage bands. Because storage regions between science regions can be shared, one starts to win back field of view as the number of bands is increased. Here we are using 40% of the GMOS detector for science (available for slit placement) and 60% for charge storage. Once more each of the science bands can contain a high density and number of slits of variable length. Each band must have the same height and there must be two storage regions immediately adjacent to each science region; however, they do not need to be regularly spaced or fill the entire detector as shown here. This can be seen better in the next example.
	[c] The limiting case of many bands where each science region contains exactly one slit and, therefore, each slit has the same length. This special case is knows as micro-shuffling. As the number of micro-shuffled bands increases and the size of the slits decreases one can use nearly 50% of the GMOS field of view for slit placement.

Nod Definition with Nod & Shuffle

• How much additional overhead can the observations tolerate? Both the size and direction of the telescope nod will affect the amount of extra overhead incurred by the Nod & Shuffle observations compared to classical spectroscopic observations. While minimizing these extra overheads shouldn't be the primary consideration when selecting a telescope nod configuration, they can have an important impact on the observing efficiency.
• Does one wish to nod along the slit and observe the target in both the sky and object positions? This is a useful mechanism by which one can double the exposure time on the target and greatly reduce the overheads associated with Nod & Shuffle observing. However, this may not be possible or desirable for certain applications, e.g. observing crowded fields or extended objects. When nodding along the slit there are certain additional considerations to take into account when designing the Nod & Shuffle focal plane masks.
• If one is not nodding along the slit, one must define a sky position which does not place any objects in the slitlets. This can be tricky especially if one is using a large nod to a position for which deep imaging does not exist. It may be beneficial to request pre-imaging for both the target and sky positions if one anticipates using a very large (several arcmin) nod with Nod & Shuffle. Or perhaps more useful would be to request a short exposure (5min) at the sky position through the mask once it has been designed - this is the best way to make sure there are no bright objects contaminating the spectral image of the sky.
• Can one still guide in the sky position? For large nods (several arcmin or more) it is unlikely that the OIWFS will still be able to reach the guide star (the OIWFS patrol field is 3.54 arcmin x 4.15 arcmin). At this time, guiding with two separate guide stars in the target and sky positions is not a supported GMOS operational mode. It is still possible to take the sky spectra unguided (specify OIWFS in FREEZE mode for the sky position) but one must make doubly sure there are no bright objects that can wander into the slits during the unguided sky exposure. Note that observing with the sky position unguided will also help to reduce the Nod & Shuffle overheads.
• One is restricted to a single nod definition for each Nod & Shuffle observation. This means that the telescope will continuously nod between one target position and one sky position; this is known as the standard Nod & Shuffle mode. There are plans to support a pattern mode in a future implementation where the telescope nods between the target position and any number of user-specified sky positions in the same Nod & Shuffle observation.

Defining a Nod & Shuffle Observation

• cycle: A Nod & Shuffle cycle is defined as the smallest useful quantity in the Nod & Shuffle exposure. In theory this is one exposure in the target position and one exposure in the sky position.
• A = B: The exposure time in the target position during one cycle is defined as A. The exposure time in the sky position during one cycle is defined as B. For Nod & Shuffle observations with GMOS the exposure time A will always be equal to B.
• A is even: Because of the slightly uneven performance of the GMOS blade-type shutter, Nod & Shuffle exposures are always taken in pairs in order to preserve photometry. Therefore the exposure time A must be an even integer number of seconds.
• B/2, A/2, A/2, B/2: In order to obtain the best sky subtraction, the sky exposure brackets the target exposure. Therefore each Nod & Shuffle cycle with exposure time A is composed of 4 equal length sub-exposures taken in the following order - sky, object, object, sky.
• observation: A Nod & Shuffle observation is defined to have a set number of cycles, each with observing time A. Therefore the total observing time (total open shutter time) for a Nod & Shuffle observation with a number of cycles C is equal to 2 × A × C.
• shuffle distance: The shuffle distance defines the height of the bands (or the length of the slits for micro-shuffling). Normally the shuffle distance is a positive number, the sign determines whether the charge is shuffled up or down for the sky exposures. The object is always observed with the charge in the normal (no shuffling) configuration. For a positive shuffle, the resulting spectral images will have the sky spectrum located beneath the object spectrum with a separation equal to the shuffle distance. The shuffle distance is always specified in unbinned pixels, regardless of whether or not the spectral data is binned in the y-direction. For example, for a single-band Nod & Shuffle observation the shuffle distance is +1536 pixels and for a micro-shuffled image with 1 arcsec slits the shuffle distance is 14 pixels. However if the spectral data is binned in the y-direction, one must make sure that the shuffle distance is a multiple of the binning (e.g. one cannot have a shuffle distance of 21 pixels if binning by 2 in the y-direction, but a shuffle distance of 20 or 22 pixels would be OK).
• charge traps: The presence of local defects (charge traps) in GMOS detectors #2 and #3 lead to low-level horizontal stripes in Nod & Shuffle spectral images. Efforts to correct these charge traps are underway. Meanwhile there are two approaches to minimizing the effects of these defects. Since semester 2003B we have providing a special kind of dark exposure for Nod & Shuffle. Taking these dark exposures is very time consuming, and they are therefore only guaranteed to be made available for the following Nod & Shuffle configurations:
  • A=60, 15 cycles, shuffle distance=31 pixels, CCD binned 2x1
  • A=60, 15 cycles, shuffle distance=70 pixels, CCD binned 2x1
  • A=60, 15 cycles, shuffle distance=1536 pixels, CCD binned 2x1
  Any Nod & Shuffle darks defined by PIs in their Phase II proposals other than the above configurations will be taken on a best-effort basis. The Nod & Shuffle darks are effective for correcting for most of the effect of the defects. However, for very deep observations it is recommended to also translate the detector between Nod & Shuffle observations by ± a few pixels in the y-direction. This allows the defects to be removed in the data reduction stage via a suitable rejection algorithm. Nod & Shuffle programs that aim to go very deep should have their observations split into at least three dithered exposures. The translation of the detector must be defined by the PI in the OT using a GMOS sequence to change the DTA-X offset. 
• baseline calibrations: For Nod & Shuffle programs all baseline calibrations are taken in exactly the same manner as for classical long-slit and MOS spectroscopy programs. Additional calibrations or calibrations taken in Nod & Shuffle mode must be requested explicitly in Phase I and Phase II proposals and Nod & Shuffle Darks must be defined by the PI in their Phase II OT program.

Additional Observing Overheads with Nod & Shuffle

• Rule of thumb:If you are nodding along the slit and use nods of less than 10arcsec but greater than 2arcsec, the approximate time required to get 1800sec open shutter time (A-position plus B-position) is 2200sec, plus time for acquisition and readout. For nods with a total distance of 2arcsec or less one can elect to use electronic offsetting of the OIWFS, in which case this time is reduced to 2000sec.
• GMOS Nod & Shuffle overheads are defined as the time per cycle that the shutter is closed. This excludes readout and target acquisition. This also assumes nodding along the slit, meaning all open-shutter time is spent collecting data on the target.
  • Example 1: Exposure time A = 60sec, Overhead = 24sec, nodding along the slit.
    Effective overhead (time spent not observing target) per cycle is 24sec, or 24/(60+60) = 20%
  • Example 2: Exposure time A = 30sec, Overhead = 24sec, nodding along the slit.
    Effective overhead (time spent not observing target) per cycle is 24sec, or 24/(30+30) = 40%
  • Example 3: Exposure time A = 60sec, Overhead = 24sec, nodding off to sky.
    Effective overhead (time spent not observing target) per cycle is 84sec, or (24+60)/60 = 140%
• Additional overheads above the 24 sec depend on the size and the direction of the telescope nod, and appear to be limited by the speed at which the GMOS OIWFS moves into position. Nods in the q-direction appear to be much more efficient (less overhead) than nods in the p-direction (perpendicular to the slit).
• Nods in the q-direction can be approximated as adding another additional 1sec of overhead per cycle for every 15 arcsec of distance, eg. a nod of q=30 arcsec gives an overhead of 26 sec per cycle.
• Nods in the p-direction can be approximated as adding another additional 0.6sec of overhead per cycle for every 1 arcsec of distance, eg. a nod of p=30 arcsec gives an overhead of 42 sec per cycle. It is strongly recommended to nod in the q-direction (parallel to the slit).
• Very small absolute nod sizes (< 2.0arcsec) can take advantage of electronic offsetting of the OIWFS. In this case the GMOS OIWFS does not actually move, only the position of the spots on the wavefront sensor is electronically shifted. The effective overhead per cycle is reduced to 13.5 seconds, so observations which can tolerate such small nods are encouraged to use this feature.

Performance Delivered by Nod & Shuffle

The recommendations and performance numbers given in this section are based on observations done during the commissioning and System Verification in August and September 2002.

• Recommended Exposure Times: Choosing the best exposure time for Nod & Shuffle observations is highly dependent on the stability of the current cloud cover conditions. There is a trade-off because shorter exposure times, while yielding more accurate sky-subtraction also lead to significantly increased overheads. During Commissioning and System Verification for Nod & Shuffle it was determined that an exposure time of A = 60sec gave acceptably low sky-subtraction residuals for photometric or nearly photometric (CC = 70%) and relatively stable sky conditions. For highly variable sky conditions or heavier cloud cover (CC = 90% or Any) an exposure time of A = 60sec gave unacceptably high residuals and we recommend a shorter exposure time, e.g. A = 30sec.
• Sky subtraction and Nod & Shuffle -vs- Classic GMOS Spectroscopy: Nod & Shuffle gives the possibility of eliminating systematics from the sky subtraction, even in the precense of very strong sky lines. The advantage of Nod & Shuffle is largest in the red, due to the many strong sky lines. By eliminating the systematics from the sky subtraction, Nod & Shuffle makes possible to combine spectroscopic observations to reach total exposure times of many tenths of hours without being affected by systematic sky residuals. It has been demonstrated that with total exposure times of 28 hours it is possible to determine redshifts from absorption lines of objects as faint as i'(AB)=24.7 mag. For exposure times of a few hours and fairly bright objects, typically i' brighter than 20-21 mag, the advantage of Nod & Shuffle is smaller. However, if you are concerned about getting high signal-to-noise spectra and are studying faint absorption features, you may still want to consider Nod & Shuffle. Example longslit spectra taken with Nod & Shuffle illustrate these points.
• GMOS Integration Time Calculator (North, South): The GMOS ITC does not account for systematic uncertainties in the subtraction of bright skylines, and as such is quite easily adapted for estimating realistic sensitivities using Nod & Shuffle. All one needs to do is to set the sky aperture to 1 times the target aperture in the Analysis Method under Details of Observation instead of using the default sky aperture of 5 times the target aperture.

Focal Plane Masks Used with Nod & Shuffle

• Long-slits: There are standard long-slits available for Nod & Shuffle observing which utilize the central third of the detector for science, the top and bottom thirds are used for charge storage. These longslits are listed on the GMOS long-slits page. Requests for long-slit configurations other than these will be handled as multi-object spectroscopy and the user must design the masks using the procedures described below.
• Multi-Object Spectroscopy (MOS) Masks and the Maskmaking Software GMMPS: The mask making software supports the design of Nod & Shuffle MOS Masks for both micro-shuffling and band shuffling. Details are available on the MOS observer pages.
  

Example Nod & Shuffle MOS Masks and Spectral Images

• Example gifs of MOS masks - Still to come.
• Example gifs of data - Still to come.
  • Point out locations of the object and the sky data - Still to come.
  • Point out the charge traps - Still to come.

Reducing Nod & Shuffle Data

The Gemini IRAF package contains a task "gnsskysub" that performs the basic shifting and substraction of a Nod & Shuffle exposure. The example longslit spectra show a raw Nod & Shuffle exposure, and the result after processing with "gnsskysub" and combining several exposures to clean the cosmic-ray-events. The recommended steps for the full reduction are as follows:

• Bias subtraction and trimming, using gsreduce
• Processing with gnsskysub
• Further processing as for normal longslit or MOS spectra, but without the sky subtraction since that has already been taken care of by gnsskysub

The Nod & Shuffle Commissioning Team

Last update April 30, 2004; Inger Jørgensen
Previous version June 25, 2003; Kathy Roth and Inger Jørgensen