The FUSE Observer's Guide Appendix A: Details of the Instrument Version 2.0, May 10, 2000

WARNING!!! Some information in this document may be dated!!

This appendix was formed from material cut out of the FUSE Observer's Guide, ver. 1.2. It contains extra discussion that was deemed "optional" for Cycle 2 observers and beyond, but may still be of interest. Some of this information has been duplicated and updated in the FUSE Observer's Guide, Version 2.0. For Any material shown here and in Version 2.0, the information in the main document is more current and should be used.

Outdated material in this appendix will be updated only as time permits.

Table of Contents

Appendix A. The Instrument

A.1 Optical Design

A.2 Telescope Mirrors

A.3 Focal Plane Assemblies

A.4 Spectrograph

A.4.1 Wavelength Coverage and Dispersion

A.4.2 Effective Area

A.4.3 Spectroscopic Resolving Power

A.4.4 Spectral Astigmatism and PSF

A.5 Detectors

A.5.1 Detector Background

A.5.2 Detector Flat Field

A.6 Fine Error Sensors

A.7 Instrument Data System

This multi-channel design allowed the coatings on the mirrors and gratings to be selected to maximize reflectivity in the wavelength ranges above and below ~1020 Å. Two mirrors and two gratings are coated with SiC to provide wavelength coverage below 1020 Å, while the other two mirrors and gratings are coated with aluminum and a LiF overcoat. The Al+LiF coating provides about twice the reflectivity of SiC at wavelengths > 1050 Å, but has very little reflectivity below 1020 Å. We will thus refer to "the SiC channels" and the "LiF channels" below.

The four channels can be thought of as comprising two nearly identical "sides" of the instrument. A side consists of one LiF and one SiC channel, each of which produces a spectrum that falls onto a single detector. Each channel has a bandpass of about 200 Å. Thus, at least two channels are required to cover the entire ~290 Å wavelength range of the instrument. All four channels cover the 1015-1075 Å region.

Figure A.1-1 also shows the orientation of the instrument prime coordinate system (X,Y,Z). The two LiF channels are on the +X side of the instrument, which is always kept in the shade (i.e., the -X side always points toward the sun). This orientation minimizes the amount of sunlight that can make its way down the baffles surrounding the LiF channels. Minimizing stray light in the LiF channels is crucial to the operation of the Fine Error Sensor (FES) guidance camera which operates at visible wavelengths. The orientation of the satellite is actually biased by 2.5° (in roll around the Z axis) in order to keep the radiator of the operational FES in the shade.

A.2 Telescope Mirrors

The four telescope mirrors are identical off-axis paraboloids. The blanks are made of Zerodur, with a triangular rib structure on the back to provide a lightweight but very stiff substrate. Each mirror is attached to the front of a honeycomb-sandwich intermediate plate by means of tangential-blade flexures, which minimize mounting-induced distortions. Three stepper-motor actuators are attached to the rear of the intermediate plate, which allow tip-tilt-focus adjustment of the mirror. Refocusing may be required periodically as the structure shrinks due to desorption of water from graphite epoxy. The tip-tilt mechanism is used to provide rough alignment of the mirrors to the FPA entrance apertures.

The off-axis angle of the SiC mirrors and the LiF mirrors differ slightly. This angle is defined by aperture stops placed over the surfaces of the mirrors. Otherwise, the mirrors are identical except for the coatings. Some of the specifications of the mirror assemblies are given in Table A.2-1. The point spread function (PSF) at the focal plane places 90% of the light within a diameter of 1.5 arcsec.

Thermal control is maintained by heaters attached to the intermediate plates, which radiatively couple to the rear of the mirror substrates.

A.3 Focal Plane Assemblies

The four apertures include:

• The high resolution aperture (HIRS: 1.25 × 20 arcsec) to ensure that maximum resolution (24,000 - 30,000) is maintainable even if the telescope imaging or pointing stability degrade below specifications. (This is not a problem to date.) This aperture does not allow all of the light from a point source in to the spectrograph, however.

• The medium resolution aperture (MDRS: 4 × 20 arcsec) that provides maximum throughput while minimizing airglow contamination. In the limit where telescope PSF and pointing stability are at the specs, this aperture also provides a spectral resolution comparable to that obtained in the HIRS aperture. However, thermal drifts of the separate channels, and in particular of the SiC channels with respect to the LiF channels, has made the general use of this aperture problemmatic. (See section __.)

• A large square aperture (LWRS: 30 × 30 arcsec) that can be used for observations of faint extended objects, producing a filled-aperture resolution of about (~100 km/s), or for point source observing. In the limit where telescope PSF and pointing stability are at the specs, this aperture also provides a spectral resolution comparable to that obtained in the HIRS and MDRS apertures, except for any airglow lines. The thermal motions the channels need to be modelled and removed, however, for best results. (See section __.) For nominal operations, we consider this to be the default aperture.

• A pinhole aperture (PINH: approximately 0.5 arcsec in diameter) that (in principle) can be used for observations of extremely bright objects. However, operational and calibration uncertainties will not permit use of this aperture for normal science operations during Cycle 2.

Note to Table A.3-1: The aperture throughputs are computed assuming nominal PSF (90% encircled energy in 1.5 arcsec diameter), pointing jitter of <0.5 arcsec (1 sigma), and pointing accuracy of <0.2 arcsec.

Figure A.3-1. The locations of the FUSE apertures in the sky for a slit position angle of 0° with North in the +Y direction. Positive aperture position angles correspond to a counter-clockwise rotation of the apertures on the sky.

NOTE: The pinhole was designed for possible use in observations of very bright targets, for which FUSE has not been optimized. Its flatfield and wavelength calibration will be different than for the other apertures, and it will not be photometrically calibrated, since its throughput depends strongly on the pointing stability. Its use will not be supported in Cycle 2.

Each FPA can be moved independently in two directions. Motion tangential to the Rowland circle, which is roughly in the dispersion direction and perpendicular to the apertures, allows co-alignment of the channels and in principle will permit "focal plane splits" for high signal/noise ratio observations of bright targets. Motion in "Z" (perpendicular to the plane of the FPA mirror) enables focusing of the apertures with respect to the spectrograph grating and detector.

The front surface of the FPAs on the two LiF channels are reflective in visible light. Light not passing through the apertures is reflected into a visible light CCD camera. Images of stars in the field of view (FOV) around the apertures are used for acquisition and guiding by this camera system, called the Fine Error Sensor (FES). Further information on the FES is contained in Section A.6.

A.4 Spectrograph

The spectrograph has a Rowland circle design, with a diameter of 1652 mm. Light entering the spectrograph through an FPA aperture illuminates one of four diffraction gratings. The gratings are holographically ruled on spherical substrates made of fused silica. Table A.4-1 lists some of the design parameters of the spectrograph and diffraction gratings.

FUSE can obtain spectra from about 905 Å to 1187 Å. The spectra from the four channels are imaged onto two microchannel plate detectors, with a SiC spectrum and LiF spectrum on each (from each aperture). Therefore, each detector covers the entire wavelength range. The two channels are offset on the detector perpendicular to the dispersion direction to prevent the spectra from overlapping. Each detector is divided into two functionally independent segments (A and B) separated by a small gap. To ensure that the gaps do not fall at the same wavelength region in both detectors, they are offset slightly with respect to each other. Table A.4.1-1 lists the wavelength coverage of each of the eight detector segment/channel combinations. Nearly the entire wavelength range is covered by more than one channel, and the important 1015-1075 Å range is covered by all four.

Although the holographic rulings allow for partial correction of astigmatic optical aberrations, the image of a point source still has a vertical extent of 150 to 1100 µm (~ 14 to 100 arcsec) on the detector, depending on wavelength. This means that virtually no spatial imaging capability is available, since the HIRS and MDRS slits are only 20 arcsec in length (which projects to 220 µm on the detector). Only in the limited spectral regions near the minimum astigmatic heights in each spectrum will marginal spatial information be derivable.

The SiC channels have a dispersive plate scale of 1.03 Å/mm while the LiF channels have a scale of 1.12 Å/mm. Coupled with the size of the detector pixels, this results in a scale of 6.2 mÅ/pixel in the LiF channel and 6.7 mÅ/pixel in the SiC channel (in the X or dispersion direction).

Note to Table A.4.1-1: The dispersion direction of the two SiC channels is opposite that of the LiF channels. Both SiC 1 A and LiF 1 A refer to the same physical detector segment.

Figure A.4.2-1 The predicted beginning-of-life effective area as a function of wavelength (all 4 channels summed). The prediction assumes a loss of 20% per year (from the time of manufacture to launch) in reflectivity of optical surfaces. The effective area will change as instrument components are integrated and characterized.

Figure A.4.3-1. The measured spectral resolution as a function of wavelength (after correcting for astigmatism).

This image illustrates a number of interesting features of the spectra (some of which are due to the testing setup). First, note that the spectral resolution degrades with the width of the slit. This is due to the fact that the slit was fully illuminated during the test, which would more closely match what would be expected in orbit for emission from a large extended source (such as a supernova remnant or planetary nebula). Under conditions of nominal pointing accuracy and jitter, a point source spectrum would have roughly the same spectral resolution in all slits. Second, notice the astigmatic height and curvature of the spectra in the LiF channel (the upper 3 spectra). This is a natural consequence of the optical design. The narrow slit is on-axis, and the spectrum through this slit is symmetric about the dispersion axis. Note that the other two LiF spectra are asymmetric since they are off-axis. The spectrum through the MDRS slit (the top spectrum) is tilted slightly to the left while the spectrum through the LWRS (the 3rd spectrum down from the top of the image) is tilted to the right. The SiC spectra all have much smaller astigmatic heights because the spectra are near a holographic correction point at these wavelengths (see Figure A.5.1-1). The fact that the spectral resolution is lower at a spatial focus point is also obvious from these spectra (compare the widths of the LiF and SiC spectra).

Figure A.4.4-1. Molecular hydrogen emission spectrum recorded by detector 1B. The top 3 spectra are from the LiF channel and the bottom 3 are from the SiC channel.

The spectra discussed above are from side 1 of the instrument. For side 2, the order of the slits is reversed from what is shown in Figure A.4.4-1. The LWRS slit is at the top of the image and the MDRS slit is below the HIRS slit. The LiF spectra are still imaged in the upper half of the segment, with the SiC spectra in the lower half. During I&T, when the gratings were being tightened, the LiF grating was twisted slightly, resulting in a slight tilt to the "on axis" spectra. Consequently, the side 2 LiF spectra are not symmetric about the dispersion axis.

A.5 Detectors

Two microchannel plate (MCP), double-delay line detectors are used in the FUSE instrument. The MCPs and delay line anodes are curved to match the Rowland circle. In such a detector an incident photon strikes a potassium bromide (KBr) photocathode, which covers the top surface of the high-voltage biased Z-stack of MCPs. The MCPs provide electronic amplification of this initial photoelectron, resulting in a cloud of ~2 × 10⁷ electrons at the bottom of the stack. This electron cloud then impinges on a helical delay line anode below the MCP stack. The events are located in the spectral direction on the anode by measuring the time required for the electron cloud pulse to propagate in the x-direction. Events are located in the spatial direction by measuring how the charge in the cloud is split by the interleaving wedges of the anode in the y-direction.

The total charge that can be extracted from the MCPs is a limited resource that is managed carefully. High S/N observations of any target, bright or faint, deplete this resource and may cause gain sags in the illuminated region of the detector. FOG Section 3.5.2 describes the bright object limitations in greater detail.

The active area of a detector must be approximately 170 × 10 mm in order to capture the full spectral range from a pair of gratings. This requires the use of two MCP stacks ("segments") in each detector. Each segment is 88.5 × 10 mm, and they are placed end-to-end in the long (dispersion) direction, with a ~7 mm gap between the stacks. Due to edge effects, the effective gap size is ~10 mm. The two detectors are displaced slightly in the dispersion direction so that no wavelength interval falls in the gaps of both detectors. Throughout this document the detector segments are labeled 1A and 1B (segments A and B on side 1), 2A and 2B (segments A and B on side 2).

The photon event locations in each detector MCP segment are digitized into 16,384 × 1024 pixels. On all segments, the X pixel size is ~6 µm. However, the Y pixel size is different for the different detector segments. For segments 1A and 1B, the Y pixel size is ~10 µm. For segments 2A and 2B, the Y pixel size is 14.7 and 16.1 µm, respectively. The detector resolution, however, is ~25 × 50 µm (i.e. the detector resolution is oversampled by ~4). The full instrumental resolution, including the optics and pointing jitter varies from ~33 to 50 µm (corresponding to R = 20,000-30,000; see Section A.4.3 for plots of spectral resolution versus wavelength).

Information for every photon event is packaged into a 4-byte packet and sent to the Instrument Data System (IDS). Information in this packet includes the detector and segment number of the event, the X,Y location on the segment, and the pulse height of the event.

If the total count rate from the two detectors exceeds 32,000 counts/second, the bus between the detectors and the IDS cannot keep up, and events are randomly discarded at the detector. Spectral integrity is maintained, but photometric accuracy is somewhat compromised. The total number of photons detected (before discarding photons above the 32,000 count rate limit) is telemetered to the IDS from the detector. For count rates higher than ~40,000 per segment, the dead-time correction begins to become important (> 20%).

Two sets of wire grids lie above the MCPs. A "QE grid", located several millimeters above the surface of the MCPs, creates an electric field above the plates. Photons which strike the MCP between the pores create photoelectrons which escape in the direction perpendicular to the MCP. The purpose of the QE grid is to deflect these photoelectrons back to the MCP, and thereby prevent a loss in the quantum efficiency of the plates. Deflected photoelectrons return to the grid within ± several pores of where the photon originally hit. The latest I&T data indicate that the QE grid causes < ~10% degradation in the spectral resolution (an option to observe with the QE grid off to improve resolution may be supported in later observing cycles). Above this QE grid is a "plasma grid", whose job is to keep plasma in the earth's atmosphere from impinging on the MCP. This grid will reject particles with energies up to several volts in energy. Highly energetic particles will not be stopped by this grid and will result in either normal energy background events in the MCP or high energy events which are rejected by the pulse height discriminators in the detector.

A stim lamp is located just below the internal spectrograph baffles, and about 1 meter above each detector. This stim lamp will be used in orbit as an aid in calibration. The count rate from the lamps are high enough that they can be used to provide some information on the flatfield and distortion properties in the detectors. The shadows of the 2 wire grids will be visible in exposures taken with the stim lamps.

The temperature of each detector electronics package is regulated by a connection to a baseplate which is strapped to one of two external radiators. During a long observation, the baseplate is expected to be stable in temperature to < ± 2°, resulting in the electronics being stable to < ± 1°. Large changes in satellite attitude (and solar illumination of the radiators) are likely to result in large absolute changes in the temperature of the baseplate, and will therefore affect the temperature of the electronics. The performance of the electronics are sensitive to changes in temperature, resulting in a "drift" of the measured photons events in detector coordinate space, and a loss of spectral resolution. This effect may result in our having to schedule observations to minimize beta angle changes between targets, and has the potential to impact the scheduling of time-critical observations.

A.5.1 Detector Background

The pulse height distribution (PHD) of background events is different than photon-induced events. Background PHDs have a negative exponential shape, while photon-induced PHDs have a gaussian-like shape. Therefore, it is possible to reject low energy background events with proper settings of the pulse height discriminators. However, within the normal energy range of photon induced events, background events are indistinguishable from photon induced events. The background rate stated above is the value after rejecting the low energy events.

The background count rate is especially important for faint targets where the astigmatism in the spectra is large. The astigmatic height of the spectra in each channel is shown below in Figure A.5.1-1. Based on this figure, one can see that the detector background per spectral resolution element will be lowest at the wavelengths where the astigmatism is minimized (i.e., 930 and 1040 Å).

Figure A.5.1-1. The predicted astigmatic spectral height as a function of wavelength. The given height encloses 90% of the events in an isolated, unresolved emission line from a point source as seen through the HIRS or MDRS aperture.

The detector background rate is the ultimate limiting factor in the ability to detect faint sources. The contribution of the background to spectra varies with wavelength in proportion to the astigmatic height of the dispersed spectra. Figure A.5.1-2 is a plot of a 1 count cm^-2 sec^-1 background level converted into an equivalent flux as a function of wavelength. The peak at 985 Å is due to the low effective area and high astigmatic height of the LiF channels at that wavelength. For sources with fluxes at or below the background level, it will be advantageous to discard segments with low effective areas and large astigmatic heights. The background is reduced to a minimum if only the SiC segments are used below 1030 Å, and only the LiF segments are used above 1030 Å. The dashed line in Figure A.5.1-2 shows this minimal background flux level.

Figure A.5.1-2. The solid curve shows the flux equivalent to a background counting rate of 1 count cm^-2 sec^-1 as a function of wavelength. Scattered light and airglow is not included. The variation with wavelength reflects both the astigmatic height of the extracted spectrum and the sensitivity of the FUSE LiF and SiC channels. Background in the extracted spectrum is minimized if the LiF channels are ignored below 1030 Å and the SiC channels are ignored above 1030 Å, as shown by the dashed curve.

The FUSE Project currently has little more than a global knowledge of the detector background properties because the level is so low. (It would take an exposure of almost 1.7 million seconds to obtain 100 counts per pixel.) However, we expect that the background will show high frequency spatial variations similar to those seen in the flat-field images, depending upon where in the MCP stack an event originates. Events induced on-orbit by high energy charged particles will also vary with time depending on the FUSE satellite position. Due to these effects, the FUSE team initially projects that the background can be established to an approximate accuracy of 10% on all wavelength scales, predominantly limited by the unknown level of spatial variations and their stability with time. This implies that a source can be detected at the 3 sigma level when its flux is 30% of the background, and a source can be detected at the 10 sigma level when its flux is equal to the background. In this regime, higher S/N ratios cannot be obtained by increasing the integration time nor by reducing the resolution, since the observation is limited by the uncertainty in the background and not by the Poisson noise of the source! Figure A.5.1-3 shows the S/N=3 floor for FUSE. The listed exposure times show how long it takes to overcome the Poisson noise to reach the S/N=3 floor. See Appendix C for an example.

Figure A.5.1-3. The solid curve shows the S/N=3 sensitivity floor for FUSE. This assumes that the background is only known to an accuracy of 10% of the value shown by the dashed curve in Figure A.5.1-2. The integration time required to reach the 3-sigma Poisson noise limit in a 1 Å bin is shown at three wavelengths for the flux levels given by the curve.

As the FUSE team accumulates observing time through the various apertures, we expect to obtain a better knowledge of the background properties. If it proves to be stable, and once sufficient background observations have been obtained to produce a high S/N model, we hope to be able to predict the background on all wavelength scales to an accuracy comparable to that obtained in our flat-field images, on the order of a few percent. The S/N=3 floor for FUSE shown in Figure A.5.1-3 would then be proportionately lower. Even then, however, the S/N in an observation of a faint target will be limited by the background uncertainties, not the Poisson noise of the observation.

A.5.2 Detector Flat Field

A portion of a flat field image for segment B of the FUSE flight detector on side 1 of the instrument (i.e., segment 1B) is shown in Figure A.5.2-1. Several things should be noticed in this image: (1) there is a hexagonal ("chicken wire") fixed pattern in the flat field, and (2) a small region of low response (located near the right hand side of the image). The chicken wire pattern is caused by the fiber bundles in the MCP, and have a diameter (at the anode) of ~0.6mm (~0.6 Å).

Figure A.5.2-1. Flat field image of a region on segment 1B. This image is 3000 × 3333 µm (500 × 333 pixels) in size. Darker colored areas have higher intensity. The contrast has been enhanced to show the "chicken wire" pattern, which is typically a 10-20% effect across the fiber bundle boundaries.

A region from the flat field for segment 2B is shown in Figure A.5.2-2. This flat field exhibits a number of properties not evident in the flatfield for segment 1B above. First, notice that there are spots of low sensitivity which range in size from several pixels up to many resolution elements across. These regions of lower sensitivity also appear to be "edge brightened". It is currently believed that the low sensitivity spots are due to a blocked pore in the MCP, and the electric field behind the MCP stack is disorted in a way that redistributes the charge to a region around the pore. Detector 1 has a few large dead spots in it (mostly between the expected positions of the spectra), but detector 2B has many more smaller low sensitivity regions. The effect of these spots on the spectral resolution will be determined during I&T.

Another feature of the flatfield in segment 2B is a Moire pattern, which has an amplitude of 5-10% and a period of about 8 pixels in X (slightly larger than one spectral resolution element). The amplitude of the Moire pattern varies across the detector segment, being quite visible in some locations (10% effect) while almost invisible at other wavelengths. The Moire pattern is much less evident on segments 1A, 1B, and 2A. The pattern itself is apparently an interference effect caused by the stacking of glass plates in the MCP stack. The stability of the Moire pattern and the low sensitivity spots is currently unknown, but will be studied prior to launch with the flight spare detector.

The hexagonal and Moire patterns represent redistributions of light, and are not true changes in the response function. Hence, the effect of these fixed pattern distortions is to degrade the spectral resolution. It is not possible to correct for the redistribution of light (this being a deconvolution). However, on most of detector segments, the distortions are at the few percent level, and are "averaged" out by the astigmatic nature of the spectra.

Figure A.5.2-2. Flat field image of a region (500 × 333 pixels) on segment 2B. Darker colored areas have higher intensity. The image contrast has been enhanced to show the Moire pattern which is a ~10% effect.

A.6 Fine Error Sensors

The FES limiting magnitude is V ~ 13.5 (for centroiding to 0.2 arcsec accuracy), sufficient to find at least one guide star in 85% of the fields at the Galactic Poles and virtually 100% of the fields at lower Galactic latitudes. In some fields near the galactic poles it may be necessary to schedule observations at a particular time to maximize availability of guide stars.

Each FES consists of a pair of mirrors that re-image the FPA onto a 1024 × 1024 pixel frame transfer SITe CCD, which is masked to a 512 × 512 pixel image area. The pixels are 24 µm square. The CCD is cooled to -60C with a 2-stage thermoelectric cooler coupled to a radiator on the side of the satellite. Only one of the two FESs is used at a time, and the satellite is rolled (around the Z axis) by 2.5° to keep the active radiator in the shade. The other FES will not be used unless dictated by a decline in performance of the primary unit.

Each FES has a three-position filter wheel. In the first position is a nearly zero-loss glass blank, which serves only to maintain the proper optical path length in the system. In the second position is a neutral density filter with an attenuation of 99.5% (6 magnitudes), for use when acquiring bright targets or when guiding on bright stars near the science target. In the third position one FES contains a V-band filter and the other FES contains an R-band filter. There is no shutter on the FES, and consequently the CCD is flushed continuously (when not exposing) to avoid charge build up.

The imaging properties of the FES are modest. The stellar point spread functions are typically 5 arcsec FWHM, depending on field angle and color. Nevertheless, an image of the field (with the target either in or out of the slit as requested by the investigator) will routinely be provided to users to confirm that the proper target was acquired. We expect the photometric calibration of the FES to be good to < 0.1 mag.

A.7 Instrument Data System

The IDS communicates simultaneously with all instrument subsystems, and is responsible for controlling all instrument functions, including thermal control, actuators on the mirror assemblies and FPAs, and detector and FES operations. It accepts data from the FUV detectors and FES, and packetizes the data for transmission to the Spacecraft solid-state recorder. The IDS also collects "housekeeping" telemetry (temperatures, voltages, etc) from these subsystems, packetizes them, and sends them to the recorder.

The IDS plays a crucial role in the pointing performance of the instrument. After a slew to a new target field, the IDS processes the FES image of this field, and determines the boresight pointing based on comparison with a star table uplinked from the ground. This measured pointing is sent to the spacecraft Attitude Control System (ACS) to update the current pointing. Once the FES begins sending centroid information to the IDS for a set of guide stars, the IDS computes the measured quaternion (pointing vector) once every second and sends it to the ACS to maintain pointing stability.

Please send comments or questions to

Return to the FUSE home page.