Overheads and observing technique

This page provides information on how COMICS observations are performed and how to estimate the true amount of observing time required for a particular observation. The following equation may be used:

[Important equation; please use a graphics-capable browser]
  • telapsed is the total elapsed time in seconds,
  • S/N is the required signal-to-noise ratio,
  • NEFD is the Noise Equivalent Flux Density, i.e., the 1-sigma-1-second sensitivites listed for imaging and spectroscopy (the NEFD for chopping only should be used, regardless of your actual mode of observing),
  • η is the observing efficiency (integration time divided by elapsed time; see tables on this page),
  • Nbeam is the number of beams used in the observation (4 for chop-and-nod, 2 for chopping only), and
  • Non-source is the number of these beams for which the source is on the array.

N.B.: The above expression takes into account the sqrt(2) worsening of the sensitivity for off-chip chopping and nodding. Please be sure to use the NEFD for chopping only (i.e., those listed for relevant modes) observations.

Chopping and nodding

All observations require chopping with the secondary mirror to perform background subtraction. In addition, nodding is also required unless the target is brighter than about 5e-14 W/m2/um/arcsec2 (about 8 Jy/arcsec2). Both chopping and nodding should also be used whenever faint spatial structures are being studied. Note that chopping and nodding each reduce the sensitivity of observations by a factor of sqrt(2) compared to staring, because the noise in the target aperture is increased by this amount when subtracting frames. Please contact the COMICS Support Astronomer for more detailed information.

Chopping only

If you are observing with chopping only (e.g., imaging of bright sources or spectroscopy), then you can get both beams on the array (i.e., Non-source = 2 in the above equation) only if the source is small enough for it to fit twice on the array.


If you are chopping and nodding, then Nbeam = 4. Depending on the geometry of the source, 2, 3, or 4 of these chop/nod positions can have the target on the array. Here are some examples (in each case, the red and black shapes represent the positive and negative chop beams):

Source type Nod position 1 Nod position 2 Non-source
Compact (can fit onto array 4 times) 4
Elongated (can fit onto array 3 times) 4
Extended (can fit onto array 2 times) 3
Very extended (can fit onto array only once) 2

Chopping modes

The chopping throw can be up to 60'' - however, the AG system (Auto-Guide using guide stars) cannot be used if the throw is more than 30''. The chopping direction can be specified in terms of Azimuth-Elevation or Right Ascension-Declination.

Since the chopping profile is not a perfect square wave, the object is not stationary immediately after the secondary mirror makes a chopping throw. In imaging mode, the first few frames taken after a chop throw are automatically discarded; this loss of time is accounted for in the observing efficiency numbers given below. In medium-resolution spectroscopy mode, it is recommended to keep all frames, although the first will suffer from this effect. Please contact the COMICS Support Astronomer for more information.

Image quality: RAW and ADD modes

COMICS can achieve diffraction-limited performance at 10 um. However, tracking and guiding errors produce poorer image quality (about 0.5'' FWHM) for exposures longer than about 50 msec. In RAW mode, every frame is saved to disk, allowing a diffraction-limited image to be obtained via "shift-and-add". However, please note that shift-and-add is only feasible for relatively bright sources (brighter than a few Jy) with distinct features (such as a peak) to apply it off-line. In ADD mode, data is co-added before being saved, resulting in degraded image quality but much higher observing efficiency (see below). This effect is unimportant for Q-band observations since the diffraction limit is >0.5''.

N.B.: The default mode is “ADD”. Please read the following link carefully and contact Support Scientist if you would like to use “RAW”. [A correction has been made (2014 Oct). Please visit the link.]

Observing efficiency

Overheads are large for COMICS due to the high mid-IR background and rapid data rate. Since chopping is necessary even at sites such as Mauna Kea, observations in mid-infrared can only ever be 50% efficient on-source. However, it is possible to recover the 50%-loss by chopping on-chip (so that the "sky" beam also contains an image of the source). In reality, time to chop, readout, etc. will reduce the overall efficiency somewhat.


Although all 320 columns of the detector must be read out, a subset of the rows can be read out (making the final image a thinner rectangle) to reduce the readout time. This is the recommended mode in the Q-band as the high background quickly fills up the wells and may saturate the detector. However, it has now been confirmed that it is possible to readout the entire array and still keep the efficiency at about 20-30% in the N-band (ADD mode with only one beam on source). If on-chip chop/nod is possible (with compact sources smaller than about 15 arcsec) then the efficiency will simply double and up to ~70% efficiency has been achieved.

In the Q-band, the efficiency can be as high as 30% with a partial readout of about half of the 240 rows (again ADD mode with only one beam on source), and up to about 70% with on-chip chop/nod. However, if the full-array readout is required (e.g. for an extended source), the efficiency will be considerably reduced (see table below). The following table gives the observing efficiency (i.e., integration time divided by elapsed time) when default readout is used, together with the maximum number of rows which can be read out to maintain the best observing efficiency. As mentioned above, note that reading out full array for filters which usually use partial readout reduces the efficiency. Note also that the efficiency and maximum width will decrease in poor weather when the mid-IR background increases.

All on-chip chopping.
†: Default readout (FR=Full Readout,
PR=Partial Readout/width)
Filter Default readout Full readout
70% (FR) --
N9.7 73% (PR/160) 48%
72% (FR) --
ArIII 58% (FR) --
SIV 68% (FR) --
NeII 69% (FR) --
Q17.7 74% (PR/80) 24%
74% (PR/120) 36%

The above figures do not include the overheads for object acquisition and standard stars. Remember to account for these additional overheads when estimating the amount of time your observations require (see below).


Spectroscopy requires chopping only. The observing efficiencies are listed in the following table.

Chopping along the slit (i.e., both positive and negative beams on the slit).
Grating Efficiency
NL 67%
NH 85%

In addition, you must add 5 minutes to introduce your target or into the slit.

Flatfields are required for spectroscopy. For the NL grating, there is a slight, but negligible, variation in the pixel-wavelength relation with telescope position. Assuming the grating setting is not changed during the night, a single flatfield (5 minutes) may be used to calibrate your data. For other gratings, strong fringing makes this variation important, and precise flatfield frames must be taken at the same elevation and grating setting (i.e., before slewing the telesocpe) in order to properly calibrate your data. A flatfield takes about 5 minutes, and one will be required for your target, and one for your standard star.

Standard stars

Add about 5 minutes to take images of a photometric standard star. For a spectroscopic standard, add 15 minutes (5 for acquisition, 10 for actual observations), plus an additional 5 minutes for a flatfield (medium and high-resolution spectroscopy only).


Here are a couple of worked examples to help you understand the factors to consider in determining the actual observing time required. See also the examples on the COMICS front page.

Example 1. Q-band imaging

We wish to observe a source approximately 5'' in size in the Q-band to a limiting (5-sigma) point source depth of 50 mJy.

The imaging sensitivity table indicates that the Q17.7 filter is the most sensitive (its NEFD is 320 mJy for point sources) and we would therefore like to use this. From this page, we learn that maximum observing efficiency is obtained by reading out 60 rows (7.8'') of the array, which is large enough for our small target. Because we maintain the full length of the array (320 pixels = 42''), we can nod our target along the array and have all 4 chop-and-nod beams on-chip. According to the equation at the top of this page, the elapsed time will be:

telapsed = (5 × 320/50)2 / 0.74 × 42/(2 × 4) = 2568 seconds (46 minutes).
The actual observing procedure will be:
  1. Observation of standard (5 mins)
  2. Observation of target (46 mins)
  3. Observation of standard (5 mins)
The total elapsed time is therefore 56 minutes.

Example 2. N-band medium-resolution spectroscopy

We wish to take a medium-resolution spectrum covering the entire N-band (8-13 um) of a point source with a flux of 500 mJy, achieving a S/N per pixel of about 10.

According to the spectroscopic sensitivty table, the NEFD for the NM grism is 2500 mJy. Further details on the NM grating reveal that this sensitivity is obtained over most of the wavelength range, and that 2 grating positions are required to fully cover the N-band. We will chop along the slit, and both beams will be on the array since the source is compact. The elapsed time at each grating position is:

telapsed = (10 × 2500/500)2 / 0.69 × (22 / 2 × 2) = 3623 seconds (60 minutes).
The actual observing procedure will be:
  1. Flatfield at standard star (5 mins)
  2. Standard star acquisition (5 mins)
  3. Observation of standard at grating position 1 (10 mins)
  4. Flatfield at target (5 mins)
  5. Target acquisition (5 mins)
  6. Observation of target at grating position 1 (60 mins)
  7. Change grating position (5 mins)
  8. Observation of target at grating position 2 (60 mins)
  9. Flatfield at target (5 mins)
  10. Flatfield at standard star (5 mins)
  11. Standard star acquisition (5 mins)
  12. Observation of standard at grating position 2 (10 mins)
The total elapsed time is therefore 180 minutes.

Copyright © 2000-2019 Subaru Telescope, NAOJ. All rights reserved.