Operational concepts definition document

Instrument description

Integral field unit

Near-infrared spectrograph

Filters

Hardware suppression

Detector

Observing considerations

The near-infrared sky

The near-infrared sky in J and H is dominated by airglow emission from molecules, principally OH and O2. This emission is spatially and temporally variable and has to be accurately subtracted from near-infrared imaging and low resolution spectroscopice observations. This necessitates observing the sky on timescales shorter than the variability timescale, and integration times are typically < 1min. However for medium and high-resolution spectroscopy there are considerable line free regions. Observing in these line free regions enables considerably fainter targets to be observed, and also does not require such accurate sky subtraction. The integration times can therefore be longer, which places the observations in the background limited regime.

In high resolution (~17000) spectroscopy of the J and H bands the line widths of 1Å are consistent with their being unresolved (Maihara et al 1993). The expected intrinsic width is around 0.1Å. The total integrated line intensities in the J and H bands are 17000 photons/s/m2/arcsec2/micron and 26000 photons/s/m2/arcsec2/micron respectively. This is considerably greater than the intensity level in between the lines of 590 photons/s/m2/arcsec2/micron (Maihara et al 1993).

Array sampling

There are various techniques for sampling the counts at a pixel in a near-infrared array. These are summarized below:
Reset-Read
Pixels are reset and read out once at the end of the integration.
Reset-Read-Read or Double correlated sampling
Array is read immediately after the initial reset and before the final reset at the end of the integration.
Multiple end-point sampling or Fowler sampling
The array is read multiple times after the initial reset and before the final reset
Sample up the ramp
The array is read out at several equally spaced times between the initial and final reset
Which of these sampling techniques gives the best S/N value depends on whether the observation is background or read-noise limited. Garnett and Forrest (1993) calculate the best technique to use in each situation. This is summarized below:
Background limited: Double correlated sampling > sample up the ramp and Fowler sampling
Read noise limited: Sample up the ramp > Fowler sampling > Double correlated sampling

The optimum strategy for CIRPASS therefore depends upon whether our observations are read-noise or background limited.
Defining:
O = object photons/s/pixel
S = sky photons/s/pixel
B = instrumental background photons/s/pixel
D = dark photons/s/pixel
rn = read noise
t = integration time
An observation is read noise limited if
(O+S+B+D)*t < rn2

and background limited otherwise.

For given values of S,B,D and rn, the maximum integration time for which and observation can be read-noise limited occurs when the object flux is zero. Observations where the object flux > 0 will always become background limited at shorter times because the object flux is also contributing to the LHS of the equation.

So for typical values for CIRPASS of
S = 605 photons/s/m2/micron/arcsec2 -> 0.0384 photons/s/pixel
B = 0.05 photons/s/pixel
D = 0.05 photons/s/pixel
rn = 7 electrons - readnoise on a single read
tmax = 49/(0.0384+0.05+0.05) = 354 s

So if the integration time is > 354s then the observation will be background limited, and the optimum sampling technique is correlated double sampling. The noise on the final value after correlated double sampling is
sqrt((O+S+B+D)t + 2rn2)

The maximum integration time on an object is driven by various factors including the variability in the inter-line sky and in the instrumental background, the need to take calibration observations and detector non-linearity. More work will be done to assess the effect of these various factors. For now, a conservative maximum integration time of 15 mins is assumed. Note that during maximum integration we may want to sample up the ramp in order to easily reject cosmic rays. The cosmic ray rejection software will fit a line to the resulting measurements and reject outliers. However this line fitting will not be used to give the final count rate value, which will be found by subtracting the initial from the final read (the double correlated sampling method).

Bad pixels

Persistence

Non-linearity

The HAWAII-1 (1024 sqr) detectors are linear up to values of ~100000 electrons.

Observing modes

Beam switching

This technique is used for point sources. The source is placed at one end of the IFU (position A) for the first integration. The telescope is then moved to place the source at the other end of the IFU (position B) and the integration is repeated. Subtracting the exposure at A from that at B then removes the background counts from both object spectra and leaves negative and positive object spectra. These are then extracted and combined to give the final spectrum. Note that the row order for object and surrounding sky spectra may be different at the two offset positions. The order of the beamswitched observations should be ABBAABBA..... An order of ABABAB...would give a spurious positive or negative object signal in the presence of steadily increasing or decreasing background counts.

In the case where more than one observation is taken at each position, the data reduction produces several 2D A-B images, combines them, then extracts the 1D spectra and proceeds as above.

The above strategy will not remove any bad pixels that occur in the same spatial resolution element along a row in the matching spectra at the two offset positions. The bad pixel removal can be improved by moving the instrument so that the spectra move on the detector after each pair of observations, e.g. AB move BA move AB move..... The 2D A-B images are then aligned before combining (which will remove the bad pixels).

Offset sky

Stepping source along IFU

For point sources (or small extended sources) a possible alternative to beam switching would be to step the source along the long axis of the IFU. The source is placed at one end of the IFU for the first integration. After the first and subsequent integrations , the telescope is then moved to place the source N lenses away on the IFU (where N*lens diameter > source size) and the integration is repeated. Subtracting one data frame from the following data frame would then remove the background from the object counts, as in beam switching. The spectra are then extracted and spectra from the same parts of the object are combined.

This technique has the advantage of placing the spectra at a different position after each offset without instrumental dithering. However, because the spectra are only combined after extraction a bad pixel in a spectrum lead to all the pixels in that spatial resolution element being rejected in the combination and hence good data being thrown away. Data reduction will be slightly more complicated than in beam switching as the rows containing the same parts of the objects will have to be worked out at several offset positions and not just at one offset position.

Observation examples

Outline of an observing run

The required frequency of calibration observations will be assessed during upcoming lab testing. Here we assume that the flat field and wavelength stability and repeatabilty are such that the required calibration frames for an observing night can be taken during the preceding and/or following day. (The preceding day means that the calibrations are available for data reduction during the night observing, but the following day will be required if the observing schedule is not rigidly defined in advance).

An overview of a typical observing run is presented here. Initially CIRPASS observations will be carried out in service mode i.e. the observer will not be present at the telescope.

When CIRPASS is put on the telescope at the beginning of a run the following initial checks and calibrations will be carried out:

At the beginning of every night:

After every change of wavelength range:

After every change of lens scale:

Daytime calibrations required are: Lamp flat fields for each wavelength range used.

Observation: Standard stars
Object parameters: J=14
Mode: beam switch
Required S/N: 100 Integrate for 300s at each IFU position, gives S/N=76 (summing over all lenses with emission). Combining these two exposure gives a S/N ~ 100.

Observation: Star formation in damped Lyman alpha galaxies.
Object parameters: The expected line flux from damped Lyman alpha galaxies depends upon their assumed size and star formation rate. Two cases are considered here, both having a star formation rate of 1 Msol/yr and with the DLA having a half light radius of 0.9 and 3.25 h100-1 kpc (the minimum and maximum observed sizes of Lyman break galaxies). Assuming a flat universe with a zero cosmological constant, the DLA angular sizes are 0.22 and 0.78 arcsec.

Assuming H0=65, a star formation rate of 1 Msol/yr gives F(Lya)=0.42 photons/s/m2. Assuming F(Ha)/F(Lya)=1 and F(Hb)/F(Lya)=0.04 this implies F(Ha)=6.43x10-21 W/m2 and F(Hb)=3.48x10-21 W/m2. The ETC requires W/m2/Å, so assuming that this flux falls onto 2 pixels=4.4Å and /2 to give the flux within the half-mass radius gives F(Ha)=7.31x10-22 W/m2/Å and F(Hb)=3.96x10-22 W/m2/Å.

The first scenario (r0.5=0.22 arcsec) implies F(Ha)=4.81x10-21 W/m2/Å/arcsec2 and F(Hb)=2.60x10-21 W/m2/Å/arcsec2. The object light falls on 2 lenses. A S/N of 3 in one lens requires 2000s in Ha and 12000s in Hb. A S/N of 3 summing over the two lenses requires 1000s in Ha and 6000s in Hb.

In the second scenario (r0.5=0.78 arcsec), F(Ha)=3.82x10-22 W/m2/Å/arcsec2 and F(Hb)=2.07x10-22 W/m2/Å/arcsec2. The object light falls on 23 lenses. For S/N=3 summed over 23 lenses requires 3.7 hrs in Ha and 21.6 hrs in Hb.

Mode: Although the source in this case is extended, the number of lenses filled by the source is small compared to the total number of lenses in the array and so beam switching is possible.

Bibliography

Garnett and Forrest 1993, IRAC (not sure what this is)
Garnett and Forrest 1993, SPIE, 1946, 395 (haven't read this but abstract is the same as above)
Maihara et al 1993, PASP, 105, 940
Last modified: Tue Jan 9 16:07:26 2001