Space Telescope Science Institute
COS Data Handbook 2.00
Table of Contents Previous Next Print

COS Data Handbook > Chapter 1: COS Overview > 1.1 Instrument Capabilities and Design

The Cosmic Origins Spectrograph (COS) is an HST fourth generation spectrometer, designed to enhance the spectroscopic capabilities of HST at ultraviolet (UV) wavelengths. COS was built by Ball Aerospace Corporation to the specifications of Dr. James Green, the Principal Investigator (PI), at the University of Colorado at Boulder in conjunction with the COS Instrument Definition Team (IDT). Designed to primarily observe faint point sources, COS is optimized for maximum throughput, and provides moderate and low resolution spectroscopy in the UV and limited imaging in the NUV.
COS is a slitless spectrograph that employs two circular 2.5 arcsec diameter science apertures, the Primary Science Aperture (PSA) and the Bright Object Aperture (BOA). The PSA is an open aperture and the BOA contains a neutral density filter to attenuate the flux of bright objects. COS also contains two calibration apertures, the Wavelength Calibration Aperture (WCA) and the Flat-Field Calibration Aperture (FCA). Light from external sources does not reach these apertures. Instead they are illuminated by internal calibration lamps. The FCA is not available for observers, but the WCA can be used by observers to obtain wavelength calibration spectra. The WCA can be illuminated by one of two Pt-Ne wavelength calibration lamps. Similarly, the FCA can be illuminated by one of two deuterium flat-field calibration lamps; however, this is restricted to observatory calibration programs.
The instrument has two channels: a far-ultraviolet (FUV) channel that is sensitive across the 900-2150 wavelength range and a near-ultraviolet (NUV) channel that provides wavelength coverage from 1650-3200 . The COS optical design achieves its high performance, particularly in the FUV, by minimizing the number of reflections in the optical path and the use of large format detectors which maximize the wavelength coverage per exposure. Each channel has its own photon-counting detector and a selection of gratings (Table 1.1). The NUV channel also has a mirror that can be used in two modes for imaging. The FUV channel uses a single reflection system where a high-efficiency, first-order, aspheric holographic grating completely corrects the beam in the dispersion direction but has low spatial resolution perpendicular to dispersion. Only one channel may be used at a time.
Table 1.1: COS Spectroscopic Modes
Normal wavelength range ()1
Bandpass per exposure ()
Resolving Power R = λ/Δλ
(m pixel–1)
9002 – 1450
3 35
3 35
3 41

Normal wavelength ranges are for the primary (default) central wavelength setting.
2. Modes covering wavelengths below 1130 are expected to have a much reduced resolving power of about 2000.
3. Second order contamination possible longward of 2150 .

FUV Spectroscopy
The FUV channel employs a large format cross delay line (XDL) detector consisting of two 16384 x 1024 pixel segments, referred to as FUV segments A and B. The segments are separated by a physical gap of 9 mm, which makes it impossible to obtain a continuous spectrum across the two segments with a single setting. The supported central wavelength positions were selected to enable full wavelength coverage of the gap. Table 1.2 shows the wavelength ranges of both segments for all possible FUV grating and central wavelength combinations.
Table 1.2: Wavelength Ranges for FUV Gratings for FPPOS = 3
Central wavelength setting ()

These are new modes that are currently being characterized, so these numbers are preliminary. 1055 and 1096 central wavelength settings have been offered starting from Cycle 19 and 1222 central wavelength mode starting from Cycle 20.

It is not yet clear how much of the G140L segment B short wavelength (< 900 ) ranges will be available due to uncertainties in the HST OTA throughput. This is currently being investigated.

The G140L grating and 1105 central wavelength setting moves the zero-order image onto segment B. Therefore, only segment A is available for this setting.

Beginning in Cycle 18, the G140L 1230 setting was replaced by the 1280 setting.

NUV Spectroscopy
To retain efficiency utilizing the square format of the NUV detector, three mirrors simultaneously image three, fully aberration-corrected, spectra onto a single 1024 x 1024 Multi-Anode Micro-channel Array (MAMA) detector. Consequently, three separate regions of the spectrum are imaged onto the detector. These spectral regions, referred to as stripes A, B, and C, each span the physical length of the detector in the dispersion direction - but are not contiguous in wavelength space. The allowable grating positions were defined with two objectives: the capability of obtaining full spectral coverage over the NUV bandpass and maximizing scientific return with a minimum number of grating positions. As a result, several of the supported central wavelength positions were selected to maximize the number of diagnostic lines on the detector in a single exposure. Table 1.3 shows the wavelength ranges of the three stripes for all possible NUV grating and central wavelength combinations
Table 1.3: Wavelength Ranges for NUV Gratings
Central wavelength setting ()
1334 – 17331
1768 – 19672
3161 – 35603

The wavelengths listed for central wavelength 2635 in stripe A are for completeness only and also in case a bright emission line falls onto the detector. Note that the NUV detector’s sensitivity at these wavelengths is extremely low. To obtain a low-resolution spectrum at wavelengths below about 1700 we recommend G140L and the FUV channel.

The values in shaded cells are wavelength ranges as seen in second-order light. In these cases the achieved dispersion is twice that for first-order mode. Note, however, that some first order light may contaminate the spectrum depending on the SED of the target.

The spectrum longward of 3400 may be contaminated by second order spectrum, depending on the SED of the target.

Grating Offset Positions (FPPOS)
For each NUV and FUV central wavelength setting there are four grating offset positions (FPPOS=1-4) available to move the spectrum slightly in the dispersion direction. This allows the spectrum to fall on different areas of the detector to minimize the effects of small scale fixed pattern noise in the detector. Figure 1.1 shows the shifts in uncalibrated x pixel coordinates of the stripe B spectra for all four FPPOS positions.
Figure 1.1: Grating Offset Positions (FPPOS)
This figure shows spectra obtained at all four FPPOS positions using the G185M grating with a central wavelength setting of 1850. The individual plots show the collapsed counts from the stripe B spectra versus the uncalibrated x pixel coordinates. Note that the three features marked 1, 2, and 3, shift slightly for each FPPOS position.
NUV Imaging
COS imaging may only be done with the NUV channel and the spectral coverage includes the entire NUV bandpass from ~1650-3200 . This mode utilizes a flat mirror with two available mirror settings, MIRRORA and MIRRORB. The first setting uses a primary reflection off the mirror surface, and the second setting provides an attenuated reflection. MIRRORB and/or the BOA may be used to obtain images of brighter objects, but MIRRORB produces a secondary image and the BOA produces an image with coma that degrades the spatial resolution (Figure 5.2 and Figure 5.3). While the spatial resolution of COS NUV MIRRORA (Section 1.2) images can be quite good, the field of view is very small. Furthermore, because the optics image the sky onto the detector – not the aperture – the image includes some light from sources out to a radius of about 2 arcsec. However, only point sources within about 0.5 arcsec of the aperture center have essentially all their light imaged, and so the photometric interpretation of a COS image can be inherently complex.
Data Collection Modes
COS has two modes of data collection, TIME-TAG and ACCUM, and only one mode can be used for a given observation. In TIME-TAG mode the position, time, and for FUV, pulse height of each detected photon are tabulated into an events list, while in ACCUM mode the photon events are integrated onboard into an image. TIME-TAG data have a time resolution of 32 ms, and can be screened as a function of time during the post-observation pipeline processing to modify temporal sampling and exclude poor quality data. COS is optimized to perform in TIME-TAG mode, although ACCUM mode is fully supported in the pipeline processing. ACCUM mode should be used primarily for UV bright targets that can not be observed in TIME-TAG mode due to high count rates. Users should note that FUV data taken in ACCUM mode use sub-arrays since the 18MB of onboard memory cannot hold a complete FUV image (containing both detector segments). ACCUM mode omits only the wavecal region and unused detector space, therefore the FUV ACCUM subarrays contain all of any external spectrum. The FUV ACCUM subarrays, whose sizes are 16384 x 128, are shown in Figure 2.2.

Table of Contents Previous Next Print