1.2 COS Physical Configuration

Figure 1.3: The COS Optical Path and the Locations of the Mechanisms.

Scaled with all elements shown in their correct relative locations.
The COS optical design includes an external shutter, two science apertures, two calibration apertures, two Optics Select Mechanisms (OSM1 and OSM2), and separate NUV and FUV detectors. COS also has an independent calibration lamp assembly containing two Pt-Ne and two deuterium lamps, which can illuminate the detectors with an emission line or a continuum spectrum, respectively. The COS optical design and elements are displayed in Figure 1.3.

External light enters the aperture mechanism through either the PSA or the BOA and illuminates OSM1, which contains the three FUV gratings and a mirror. Each grating can be set to one of several positions, to obtain different wavelength ranges. The positioning of the OSM1 mechanism is not precisely repeatable, and this can cause small, but significant, variations in how the spectrum or image is projected onto the detector. This non-repeatability can be corrected in post-observation data processing using separate or concurrent (TAGFLASH) calibration lamp exposures (wavecals). The COS FUV channel optical path is illustrated in Figure 1.4.

Figure 1.4: The COS FUV Optical Path.

If the OSM1 is set to the mirror position, incoming light is directed to a collimating mirror, and then to OSM2, which contains a mirror for imaging and the four NUV gratings. Each grating offers multiple positions. As is the case with OSM1, the positioning of OSM2 does not repeat exactly, and the data need to be corrected in post-observation data processing via either separate or concurrent wavecals. If a grating is in place on OSM2, the dispersed light is imaged onto the NUV detector by three separate parallel camera mirrors (NCM3a, b, c). This results in three spectra, or stripes, covering different wavelength ranges. Full wavelength coverage may be obtained through multiple observations with different grating positions. Alternatively, if the plane mirror is in place on OSM2, the undispersed light is sent to the middle camera mirror (NCM3b) and then imaged onto the NUV detector. The plane mirror on OSM2 may be used in either of two settings, designated as MIRRORA and MIRRORB. The MIRRORA setting employs a direct reflection from the plane mirror. For the MIRRORB setting, the plane mirror is slightly offset to provide primary reflection off the order-sorting filter and hence an attenuation factor of approximately 25 compared to the MIRRORA setting. The COS NUV channel optical path is illustrated in Figure 1.5.

Figure 1.5: The COS NUV Optical Path (spectroscopic only).

A series of beam-splitters and fold mirrors direct light from the calibration lamp assembly (see Figure 1.3), through either the WCA or FCA and into the optical path. The calibration lamp assembly can provide continuum illumination with its deuterium lamps and emission line illumination with its Pt-Ne lamps to both the NUV and FUV spectrographs. The Pt-Ne lamps may be operated during TIME-TAG science exposures in order to produce concurrent wavelength calibrations (TAGFLASH mode).

1.2.1 The COS Detectors

COS uses two detectors, an FUV XDL and an NUV MAMA. Table 1.4 gives an overview of their characteristics.

Table 1.4: COS Detector Characteristics.

Detector Characteristic




CsI (opaque)

Cs2Te (semi-transparent)



MgF2 (re-entrant)

Wavelength range

<900–2150 Å

1650–3200 Å

Active area

85 × 10 mm1

25.6 × 25.6 mm

Pixel format (full detector)

16384 × 10241

1024 × 1024

Image size recorded per spectrum

16384 × 128 (ACCUM)1
16384 × 1024 (TIME-TAG)1

1024 × 1024

Pixel size

6 × 24 μm
0.023 × 0.092 arcsec

25 × 25 μm
0.0235 × 0.0235 arcsec

Spectral resolution element size (= "resel")

6 × 10 pix2

3 × 3 pix

Plate scale: Along dispersion (per resel)

0.13 arcsec

0.075 arcsec

Plate scale: Cross dispersion (per resel)

0.92 arcsec

0.075 arcsec

Plate scale: Imaging (per resel)


0.075 arcsec

Quantum efficiency

~26% at 1335 Å
~12% at 1560 Å

~10% at 2200 Å
~8% at 2800 Å

Typical Dark count rate (away from SAA)3

1.13 cnt s–1 cm–2
1.63×10–6 cnt s–1 pix–1
9.77×10–5 cnt s–1 resel–1

140 cnt s–1 cm–2
8.77×10–4 cnt s–1 pix–1
7.89×10–3 cnt s–1 resel–1

1 Sizes given are for an individual FUV segment.
2 For most modes.
3 NUV dark rate is time dependent. NUV and FUV dark rates are for 2017. For updated information on NUV and FUV dark rates please see the COS performance monitoring pages.

FUV Channel

The FUV channel uses a large-format, windowless solar-blind cross delay line (XDL) detector. This is a two-segment photon-counting detector with microchannel plates feeding an XDL anode. The data are digitized to a 16384 × 1024 pixel format for each segment; however the active area is only 14191 × 439 for Segment A (FUVA) and 14374 × 426 for Segment B (FUVB). Because there are no physical pixels, fiducial electronic pulses are recorded at specific times throughout an observation to permit alignment of data to a standard reference frame. These electronic pulses are referred to as "stim pulses." Figure 1.6 schematically shows the COS FUV XDL segments with the locations of the active areas and stim pulses. The stim pulses emulate counts located near the edges of the anode, beyond the illuminated portions of the detector. A zoomed-in image of one of the FUV stim pulses on Segment B is shown in Figure 1.7. An example of an FUV external science spectrum taken with Segment B is shown in Figure 1.8, with a simultaneous wavelength calibration spectrum.

Figure 1.6: The FUV XDL Detector.

Layout of the FUV detectors (drawn to scale). The slight curvature at the corners of the active areas is also present on the flight detectors. The red and blue dots show the approximate locations of the stim pulses. The numbers in parentheses are the pixel coordinates at the corners of the segment's digitized area.
Figure 1.7: COS FUV Stim Pulse.

Left: A portion of an image in the FUV detector with a typical stim pulse is shown. Right: A histogram of the stim pulse profile in the x and y direction. The electronic stim pulses are used to remove thermal distortions and to map the XDL detector elements to a standard reference frame.
Figure 1.8: Example of a COS FUV Spectrum.

Wavelength calibration spectra for FUV Segment B with G160M/1600 obtained during ground testing. The upper spectrum is from the internal wavelength calibration lamp obtained through the WCA. The lower spectrum is from an external lamp obtained through the PSA. The bright streak at the bottom is due to an area of enhanced background on the detector segment. Note that the size of the active area is smaller than the overall digitized area. The stim pulses are also visible in the upper left and lower right corners.
With each recorded event on the XDL detector, the total charge in the associated electron cloud incident on the anode is recorded. For FUV TIME-TAG data this pulse height amplitude (PHA) is sent to the ground along with the position of the event and can be used during data analysis to help identify non-photon events, see Figure 1.9. For FUV XDL ACCUM mode data, only an integrated pulse height distribution (a histogram of the PHA data) for the entire segment is available.

A photon landing on an FUV detector segment creates an event (a cascade of electrons) at the backside of the detector which is characterized by a PHA that is detected by the electronics. The detector electronics distinguishes between real and background noise events by the value of the PHA, with background events having low PHAs and real events large PHAs. However, as a region of the detector is exposed to more and more light, the PHAs that it produces become smaller, an effect called "gain sag." Gain sag can cause two effects: the mis-registration of event positions and localized sensitivity loss.

Mis-registration of event positions as a function of PHA is termed "walk." Walk has been identified in both the dispersion (X) and cross-dispersion (Y) directions. A walk correction is made by the COS calibration pipeline for TIME-TAG data (see Section 3.4.5).

Localized sensitivity loss occurs when the PHAs for some pixels become too small to be distinguished from background events, causing events to be missed or filtered out. This results in a localized region of low sensitivity. Eventually, the gain of all of the pixels in a region becomes so small that photons landing on that location no longer create events with valid PHAs, see Figure 1.9. In that case, no events are registered and the region is termed a "dead spot." When this occurs, it is necessary to either increase the high voltage applied to the detector (which increases the PHAs of all the pixels), or to move the aperture so that the science spectra land on a different portion of the detector (which has not been exposed to as much light). The COS FUV detectors have already experienced localized gain sag on regions of the FUVB and FUVA detectors exposed to the bright Ly-α airglow line when the G130M and G140L are used respectively. As a result, the detector high voltage has been raised on several occasions (see Appendix B.1). Furthermore, the default location for the science spectra and target acquisitions is periodically moved to an unsagged region of the detector. For more information on lifetime positions consult Appendix A.1.

Figure 1.9: Example of a COS FUV Pulse Height Distribution at different times.

The PHAs for photon events falling in a given region shifts to lower values as charge is extracted from the detector. The normalized distributions from a single area are shown over 4 different periods from 2009.6 to 2010.9. Over this period, the peak in the pulse height distribution (modal gain), as measured by a gaussian fit, has decreased from nearly 13 to just above 3. The red region on the left shows the calcos cutoff value. Any events falling in this area are screened out by calcos processing as noise, and the events will be lost to the final spectrum.

NUV Channel

The NUV channel uses a 1024 × 1024 pixel Multi-Anode Micro-channel Array (MAMA) detector. This detector has a semi-transparent cesium telluride photocathode on a magnesium fluoride window, which allows detection of photons with wavelengths from ~1150 to ~3200 Å. The NUV MAMA provides no pulse-height information, but may be used in both ACCUM and TIME-TAG mode. The NUV channel creates three spectrum stripes on the MAMA detector, resulting in three separate stripes for the science data and three for wavelength calibration data as shown in Figure 1.10.

Figure 1.10: Example of a COS NUV Spectrum.

Wavelength calibration spectra obtained from the internal source through the WCA (upper three stripes) and an external source through the PSA (lower three stripes). The stripes are designated A, B, and C, in going from bottom to top for each source. Wavelength increases from left to right in each stripe and from bottom to top (hence the SHORT, MEDIUM, and LONG designations).