The Physics of the Accelerating Universe Camera

PAUCam is attached to the prime focus of the WHT. The main parabolic mirror of the WHT has a diameter of 4 m, a focal length of 10.5 m, and a focal ratio of f/2.5.

To correct the image in the prime focus, an optical corrector of four elements²¹ plus an atmospheric dispersion corrector is installed close to the prime focus, providing an effective focal length of 11.79 m (f/2.8) and an unvignetted FoV of 40' (see Figure 2).

The distance between the last lens face of the corrector and the PAUCam focal plane is 118 mm. This is a small gap to place the optical elements of the camera. There are two optical elements: (a) the entrance window and (b) the filters. The light path transverses these elements in a converging and tilted beam.

Using the optical design provided by the ING, the company Fractal S.L.N.E.²² designed the entrance window and the filters, analyzing their effects in the final image.

4.1.1. Entrance Window

As the camera operates in a vacuum, an entrance window is needed to allow the light in while maintaining the camera vessel in the vacuum. The dimensions of the focal plane demand a window diameter of 280 mm to avoid vignetting. The window produces degradation of the image quality. The spot diagram of the telescope with the prime focus corrector gives an rms radius of 7.3 μm in the field center. Using a flat optical window, this value would increase up to 14.0 μm. To try to minimize this value, we designed a window of fused silica with two different curvature radii: 1567 mm for the external face and 1588 mm for the internal one, with a thickness in the center of 20 mm necessary to avoid deformations due to the pressure difference between the two window faces. With this configuration, the spot rms radius in the field center is marginally improved to a value of 13.6 μm. However, the improvement at larger distances from the center is larger with respect to the flat-window option. The best value for the spot diagram is achieved outside the field center at the central position of the four central CCDs (see Figure 5) with an rms radius of 12.3 μm. The outer CCDs installed at the edges of the FoV, used for guiding and calibration, were not used to optimize the spot diagram.

Zoom In Zoom Out Reset image size

The entrance window was manufactured by the Mexican Instituto Nacional de Astrofísica, Óptica y Electrónica (INAOE),²³ with long experience in astronomical instrumentation. It was coated with BBAR-580, delivering a reflection not larger than 2% in the wavelength range of 400–1100 nm.

4.1.2. Filters

PAUCam has two sets of filters installed in two different jukeboxes (see Section 4.5.1). When observing, these filters are placed within a few millimeters of the active area of the CCDs in a vacuum environment, at low temperature (∼250 K), and in a focalized and nontelecentric tilted beam. Since the filters are under vacuum conditions, a common manufacturing requirement from all filters is that the coating cannot produce any significant outgassing that can be deposited onto the CCDs. An overview of the filter set is given in Table 1. Pictures of the filters installed in the filter exchange system can be seen in Section 4.5.1.

Table 1.  General Characteristics of the PAUCam Filter Sets

Type	Size (mm)	Width (nm)	Range (nm)	Avg. Transmission
External broad	203.4 × 208.8 × 3.0	∼90–150	300–1080	98%
PAUS tray broad	65.7 × 33.8 × 3.0	∼90–150	300–1080	96%
PAUS tray narrow	65.7 × 33.8 × 3.0	13	450–850	97%

Download table as:  ASCIITypeset image

BB filter set. The BB filter set was manufactured by Asahi Spectra.²⁴ The filters cover the whole focal plane of the camera, 203.4 mm × 208.8 mm in size. To minimize the distortion of the focalized beam, the thickness of the filters is 3.0 mm. This filter set provides PAUCam with the capability of enlarging its science reach and fostering the interest of other potential users. Their photometric transmission bands have the same design as the filters used in the Dark Energy Camera (DECam; Flaugher et al. 2015) that were manufactured by the same company. There are six BB filters—u, g, r, i, z, and Y—and they were calibrated at the ICE-CSIC/IEEC/IFAE labs.

The setup to calibrate the filters was composed of a light source, a collimation system, a monochromator, an integration sphere, and a calibrated photodiode. To avoid dust contamination during the tests, the filters were installed inside a hermetic box in the clean room. The transmission of the window glass was calibrated with the same system described above prior to the filter transmission measurements.

Filter transmission measurements were done at different positions. These values were compared to check for good uniformity across the filter area and averaged to obtain the transmission curves shown in Figure 6.

Zoom In Zoom Out Reset image size

Any filter designed with multilayer coatings changes its transmission profile depending on the incidence angle of the light beam. The effect of the focalized and tilted beam compared to a parallel incidence beam in the final filter transmission profiles is practically negligible for the BB filter set, as the effect is much smaller than the wavelength transmission width of the filters.

PAUS filter set. The small-sized filter set was manufactured by the Canadian company Iridian Spectral Technologies.²⁵ The PAUS uses this set. It consists of 40 NB filters covering the eight central CCDs in five trays and several small BB filters covering the 10 external CCDs of each tray. Each of the eight central CCDs, for which the PAUS has been optimized, has five different NB filters associated with it, one for each tray.

The optical properties of the incoming beam and the environmental conditions influence the effective transmission profiles, which are more critical in the NB filters and can have an impact on the photometric redshift performance of the PAUS. To understand the variation of the transmission differences, we modeled their transmission profiles in different parts of the filters. The focus position will also shift with wavelength, but this has been measured to be negligibly small.

The BB set is composed of griz filters that share the same characteristics as the external filter set described above, with a somewhat lower transmission for these smaller sets in the g band, especially at bluer wavelengths (85% versus 95%). The NB filters display a "top-hat" transmission curve (Figure 7) with a nominal FWHM of 13 nm and a central wavelength distance between two consecutive filters of 10 nm. This leads to some overlap between filters adjacent in wavelength. The central wavelengths of these NB filters span a wavelength range from 455 to 845 nm. They were measured at a calibration setup at CIEMAT (Casas et al. 2016). To test the homogeneity, the transmission was measured in a matrix of 15 points obtaining a very low dispersion. The mean transmission measured for each filter was averaged and turned out to be above 95%.

Zoom In Zoom Out Reset image size

The camera focal plane is composed of 18 scientific-grade Hamamatsu Photonics²⁶ CCDs, type S10892-04, of 2k × 4k pixels. The CCDs are back-illuminated and fully depleted with 15 μm sized pixels (see specifications in Table 2). In addition, these CCDs showed very low dark current (<1 e⁻/pixel/h) at the operational temperature of ∼–100°C, low readout noise (<5 e⁻ rms), and a low number of cosmetic defects. The CCDs are distributed on the focal plane (Figure 8) according to their performance during characterization. The CCDs that showed the poorest performance (two units) were installed at the corners of the focal plane and used for guiding, while the best CCDs were installed populating the focal plane from the center to the borders, ensuring that the eight central CCDs are the ones with the best performance. The planarity of the assembled chips in the focal plane assembly (FPA) was checked with a Stil OP10000²⁷ confocal sensor with a precision of 0.9 μm mounted in a specially manufactured x–y movable table to scan the complete focal plane. The peak-to-valley planarity of the FPA was measured to be 26 μm.

Zoom In Zoom Out Reset image size

Two test benches at IFAE/ICE-IEEC (Figure 9) and CIEMAT were developed in order to characterize the CCDs (Jiménez et al. 2012). The system is composed of a cryostat that allows one to read out one CCD at a time via a MONSOON (NOAO)²⁸ controller. The CCDs are cooled down to 173 K while being pumped below 1 × 10⁻⁶ mbar. The test bench is composed of a closed optical path illuminated by a 150 W xenon arc lamp. The beam coming from the lamp source, whose intensity on the focal plane can be controlled by a neutral density filter, crosses a set of lenses in order to collimate the light as much as possible before entering a filter wheel that holds three long-pass filters. A monochromator with a resolution of 1 nm is placed after the filter wheel, and the resulting monochromatic light is sent to an integrating sphere holding a monitor photodiode characterized versus an NIST-calibrated reference photodiode head in absolute terms of W cm⁻². The final result is that the light coming from the integrating sphere is a flat-field monochromatic light with a known absolute light power and intensity.

Zoom In Zoom Out Reset image size

Due to the importance of controlling these complex test-bench instruments and synchronizing them with the CCD readout electronics (Section 4.3) to get the desirable CCD characterization data, we developed several LabVIEW²⁹ and TCL/TK applications that allow us to manage all of the systems and collect and save the characterization data for further analysis. The setup is remotely controlled in fully automatic mode by suitable software applications that can perform wavelength scans from 300 to 1100 nm. Figure 10 shows the quantum efficiency averaged for all CCDs determined at operational temperature.

Zoom In Zoom Out Reset image size

The PAUCam readout electronics (Castilla et al. 2012; de Vicente et al. 2014) are based on a MONSOON front end and a panVIEW (Honscheid et al. 2010) application as the readout electronics controller. Both systems have been extensively used in several world-class instruments, and there is enough know-how around this tandem to be considered a reliable solution. Developed at CTIO (La Serena, Chile),³⁰ panVIEW is a specific application of a pixel acquisition code (PAN) that is in charge of acquiring the pixel data from the MONSOON front end. In addition to that, it is able to control all of the memory registers and variables used by the front-end electronics to perform the CCD readout, generate suitable flexible image transport system (FITS)³¹ headers, and define and log temperature test points, voltages, and many other user-defined variables. In PAUCam, there are four panVIEW applications, one for each MONSOON crate, which run in an individual Linux server equipped with a Four Input Links for Atlas Readout board.³² Finally, each panVIEW is commanded by the PAUCam data acquisition interface (DAQi; see Section 6.2) in order to start up and configure all of the required settings.

Concerning the MONSOON front end, PAUCam uses the DECam (Flaugher et al. 2015) physical implementation, which, roughly speaking, is composed of a master control board supporting a High-speed Optical Link for Atlas interface board³³ to use an S-LINK protocol for data transfer, a 12 channel acquisition board (ACQ) used to generate bias and Capture detector's video is the correct one, and a clock board that generates the clocks and is able to drive up to 36 independent clocks and replicate them up to 135 times.

The readout electronics configuration can be changed on the fly by software to use the whole focal plane for science images (normal mode) or the top and bottom CCDs as guiders (guiding mode) while the rest remain on science mode. The readout speed for the science mode is 150 kHz and can be increased up to 300 kHz while reading one region of interest (ROI) for the guiding mode.

In order to comply with the specifications for the camera readout system, it was necessary to use four MONSOON crates: two of them reading out eight CCDs each, and the other two reading out just one CCD each in order to reach different speeds and exposure times for guiding tasks. In order to get the whole readout synchronized, it was necessary to use a master–slave configuration with the dependency of the readout mode selected. There are three readout modes.

• 1.  
  The full mode is the default operation mode of the camera. It reads out all of the CCDs at the same time. For this mode, there is just one master MONSOON that commands the others, set as slaves.
• 2.  
  The guider mode is able to read out the 16 central CCDs while using two of the CCDs at the corners for guiding purposes. This mode requires two MONSOONs as masters, one for the 16 CCD configuration and the other for the guider CCDs.
• 3.  
  The focus iteration mode is used for the focusing procedure in combination with the focus of the telescope. For this configuration, the same image is exposed several times. Between each Exposure, a number of rows are displaced, and the telescope focus is changed. The number of focus iterations and the focus increment between exposure and CCD rows displaced are user-defined. At the end of the configured expositions, the CCDs are read out. Finally, a software algorithm (Section 7.1) searches for the best focus value reached on every focus change. Figure 11 shows an image taken in this last focus mode.

Zoom In Zoom Out Reset image size

The camera was mounted at the prime focus, and due to the limited weight supported by the telescope in this position, it was necessary to mount the whole readout electronics power supply at the Nasmyth level. This limitation required more than 30 m of cables to connect the front-end electronics and demanded a dedicated filter stage to stabilize the power supplies and guarantee the required low noise in the power sources.

Besides that, it was necessary to avoid any uncontrolled grounding loop around the readout electronics system, since the camera is equipped with several servomotors, pumps, sensors, and cryocoolers that are all controlled by a programmable logic controller (PLC). Therefore, from the very beginning, the camera design was conceptually built around a star-ground grounding configuration with enough flexibility to be adapted to the telescope conditions.

PREAMP and MIX boards. The PREAMP board (Figure 12) is located inside the cryostat (vacuum), and it is devoted to amplifying the video output and connecting the clocks and bias signals coming from the CCD through the D-SUB50 pins feed-through connector installed in the vessel wall toward the MIX board located outside the cryostat (atmosphere).

Zoom In Zoom Out Reset image size

The final PREAMP design is based on a fully differential preamplification stage with 12 layers in order to separate bias, clocks, and video signals with shielding layers between them to avoid electrical coupling (cross talk). A gold-coating finishing was used in order to reduce the thermal radiation and board outgassing. The PREAMP was designed at IFAE and CIEMAT in order to provide low readout noise (<25 μV rms), moderated cross talk (<1/10.000), reduced power consumption and power dissipation (∼200 mW), and low-to-moderated outgassing and thermal radiation. A picture of the PREAMP boards as installed inside the camera is shown in Figure 23.

The MIX board (Figure 13) was designed at CIEMAT and resides outside the cryostat (atmosphere) just behind the D-SUB50 pins feed-through connector. Its main aim is to fan out the CCD signals coming from the PREAMP board and route them toward the MONSOON front ends. The MIX board collects bias, temperature, and video signals and is able to drive up to five CCDs on the MIX5 version or up to three CCDs using the MIX3.

Zoom In Zoom Out Reset image size

Both the PREAMP and MIX boards were manually assembled at IFAE facilities. Once ready, the MIX boards were redirected to CIEMAT for testing, whereas the PREAMP boards were tested at IFAE. The PREAMP boards were tested electrically for shortcuts, open circuits, cold solders, and any other defects produced during the PCB's manufacturing and assembly processes. Once they were electrically functional, different performance tests were done: readout noise, linearity, readout speed noise, and signal stability. These tests were performed using the MONSOON's own front end as a pixel generator by using a clock signal with the same level as expected in the CCD (i.e., ∼100 mV for bias level) and synchronized with the correlated double sample of the video board. PREAMP boards with low performance were discarded.

4.3.1. Electronics System Performance

Gain and bias. The gain for each amplifier was measured at the laboratory just a few days before delivery of the camera to the WHT, using the same electronics but with different cable and ground distribution. The mean value for the individual gains, bearing in mind that each board has similar but different amplifiers and the response of each CCD/amplifier can be different, is 0.71 e⁻ ADU^–1 with an rms of 0.02 e⁻ ADU^–1. Using this averaged gain, the mean averaged bias is 22 ke⁻, which is, in fact, quite high for an 18 bit analog-to-digital converter (ADC) but good enough to allocate the 150 ke⁻ of the CCD full well. Figures 14 and 15 show the distribution plots of the gain and bias levels, respectively, for the whole set of PAUCam CCDs.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Readout noise. The readout noise is a complex issue to deal with in a CCD mosaic, since several noise sources interact with each other. Even a very well-planned grounding scheme can show drawbacks once the instrument is finally installed on the telescope; i.e., the electromagnetic interference from telescope motors and drives can induce strong noise patterns over the readout electronics despite the mean noise still being at the lowest levels. Readout noise changes also during the night for different telescope positions.

To foresee these issues from the very first camera design, a star-ground scheme (a single ground point) was selected. Therefore, every component and electrical device was configured following this guideline and, preferably, having more than one path to reach it. Thereby, during the camera tests at IFAE's laboratory, a readout noise level below 8 e⁻ rms was measured.

Once installed at the telescope, and after a few attempts to improve the camera readout noise during the commissioning run (see Section 9), the mean noise for the whole set of CCDs stabilized at around 9.3 e⁻ rms, although CCD 2 and CCD 3 (see Figure 3 for the CCD numbering scheme of the FPA) showed unexpectedly high noise levels (<11–12 e⁻ rms) related to the readout electronics themselves (see Figure 16).

Zoom In Zoom Out Reset image size

The history of the readout noise in the different observation periods shows a change in the behavior depending on the camera installation and other instrumentation present in the telescope. Figure 17 shows the changes during the years.

Zoom In Zoom Out Reset image size

Cross talk. The readout electronics cross talk (that is, the charge induced in a readout channel from the charge collected from a source in a different channel) was measured in the laboratory using a couple of CCDs installed on a test-bench cryostat. At that time, the worst case of cross talk was reported to be smaller than 3/10,000 in electron charge units. At the WHT, operational conditions are different, since 18 CCDs are connected and four MONSOON crates are synchronized to read out the whole FPA. Different sources of cross talk are being investigated in the installed system. In fact, we have been able to detect cross talk only in the presence of very large signals somewhere in the focal plane, when a substantial number of pixels reach saturation in a CCD. These effects and their correction are described in S. Serrano et al. (2018, in preparation).

As stated in Section 3, the scientific goals and operational constraints of PAUCam required the use of some challenging and nonstandard mechanical technologies, notably, the need to use carbon fiber whenever possible to minimize the weight, have a filter exchange system inside the vacuum vessel and in cryogenic conditions, and design a dual cooling system to minimize the waiting time to have the camera ready for observations after its installation. The next sections describe these mechanical aspects. An overall sketch of the camera is given in Figure 18.

Zoom In Zoom Out Reset image size

4.4.1. Camera Vessel

The camera vessel has to be able to maintain a vacuum pressure below 10⁻⁵ mbar during operation. The instrument should work in an environment between 5°C and 40°C and with a humidity between 10% and 90%. In addition, the vessel design should include all possible ports for electronic connections, sensors, tray movement motors, vacuum pumps, and cryogenic systems; ensure a low vibration transmission to all subsystems; control all electrical connections to provide a star-grounding scheme; and minimize the total weight to account for the limitation for instruments working at the WHT prime focus.

In order to minimize the weight of the camera, the material chosen for the vacuum vessel was carbon fiber. A similar vessel made with aluminum would have more than doubled the total camera weight. We made a detailed study to validate that the material works under vacuum conditions. The carbon fiber is a composite of fibers of carbon and non-outgassing epoxy. For our application, the air tightness was made by the last layer of epoxy, with an approximate thickness of 0.1 mm. To validate the material, we used a residual gas analyzer (RGA) in two equivalent cubic chambers of 20 cm³ made of aluminum and carbon fiber. The results of the RGA showed minimal differences between the two.

In order to further minimize the vessel-wall thickness, we decided to build it in an egg shape. After thorough simulations, we chose a wall-thickness value of 6 mm, which was sufficient to hold the vacuum pressure forces. We used Viton O-rings for all of the service ports. Special molds were made to ensure a smooth surface in the joints in contact with the Viton O-rings. This was especially important for the nonflat (to follow the egg shape) superior cover joint. Figure 19 shows pictures of the camera mold and finished empty vessel. All carbon-fiber parts for PAUCam were made at Magma Composites.³⁴

Zoom In Zoom Out Reset image size

The empty vessel tightness was verified with a leak detector and reached a vacuum pressure of 5 × 10⁻⁶ mbar. After the complete assembly of the camera, it was baked out inside a tent at 60°C for 3 days. The low baking temperature was limited because of the CCD specification requirements. Nowadays, during operation and in cold conditions, the PAUCam pressure is stable at (2–3) × 10⁻⁷ mbar.

The vacuum in the camera vessel is obtained by means of an Agilent³⁵ TV-301 turbo-molecular pump installed in the cryostat lid. An additional Agilent IPD-15 scroll pump is placed in the WHT ring. Both pumps are connected through a corrugated tube. For safety reasons, the camera is also equipped with a Saes Getters³⁶ GP 500 MK5 NEG getter pump, which is regenerated at the start of every operation period.

The camera is equipped with a hybrid cooling system. During regular operations, a set of two Polycold³⁷ PCC PT-30 cryo-tigers with PT-30 gas maintain the temperature of the cold plate absorbing the power output of the CCDs and preamplifiers mounted inside the cryostat. The system also maintains the NB filter trays (all installed in the same jukebox) close to the equilibrium temperature when extracted and positioned next to the focal plane. This avoids systematic effects arising from the variation of the filter spectral transmission with temperature for these NB filters. After the installation of the camera during the instrument change period at the WHT, the cryo-tigers would take too long (around 12 hr) to cool down the complete system to the operational temperature. Therefore, an additional liquid nitrogen system is installed to allow this process to be performed in less than 4 hr and ensure that the camera is operational during the first night after installation. Figure 20 shows the scheme of the cooling system inside PAUCam, with the two cryo-tiger heads and two nitrogen cooling tubes connected to the evaporator inside.

Zoom In Zoom Out Reset image size

The heat transfer to the CCD focal plane and the active NB filter-tray cooling are made by thermal conduction, connected to the cooling system through flexible copper straps in order to absorb the thermal contractions.

The FPA support is made of polished aluminum in order to avoid thermal losses by radiation with the rest of the camera body, which is black with a radiation emissivity coefficient of 0.6. The CCDs are fixed to the focal plane with stainless-steel screws, and their aluminum nitrate bodies are thermally connected to the FPA support. In order to avoid stresses in the detectors due to differential contractions, the FPA alignment pins have been thoroughly designed. The FPA is mounted in the vessel structure with flexible brackets to reduce deformations and maintain flatness, and it is electrically isolated from the carbon-fiber vessel in order to ensure a controlled grounding scheme.

Since the filters cannot be touched, their heat transfer works by radiation. We use an aluminum chiller with several plates face to face to the filters trying to keep the view factor as close as possible to 1 and with a black matte anodized surface of 10 μm thickness. The temperature control of the whole system is done through thermal resistance on the top of the chiller. The aluminum plates are connected between them with metal plates of different thickness to ensure a homogeneous temperature among all of the filter trays. A special filter tray with temperature sensors connected to dummy filters was installed. It allows calibration of the system by analyzing the changes in temperature when the filter tray is extracted from its jukebox and checking the temperature homogeneity in all filters in a tray.

The FPA and the NB cooling system have active heaters to stabilize the temperature. This operation is done by the slow control system (see Section 5). During operation, the temperature of the CCDs is set at 173 K and of the filters at around 246 K, with fluctuations of the order of ±0.2 K.

4.5.1. Filter Exchange System

As specified in Section 3, we decided to place the filter exchange system inside the camera vessel. It consists of two jukeboxes installed on the sides of the camera, each containing seven tray slots. One contains the six BB filter trays, and the other (which is actively cooled) contains the five NB filter trays. Two other slots are occupied with the dummy filter tray for the temperature monitoring described in Section 4.5 and with a shielding.

The majority of the jukebox and filter-tray materials are made of carbon fiber. The BB filters cover the whole focal plane and are attached to an aluminum frame interface mounted in the filter-tray structure. The NB filter trays have a filter per CCD, and all of the structure is made of carbon fiber. Figure 21 shows a picture of both kinds of filter trays.

Zoom In Zoom Out Reset image size

The up and down (vertical) movements of the trays are guided by a lead screw coupled with a servomotor located outside the vessel. The servomotors are controlled by the slow control system (Section 5). With these lead screws, one selects the tray to be inserted close to the FPA. The range of movement is 120 mm. Once the tray is positioned vertically, a magnetic coupling takes it and a horizontal motor moves it 280 mm to place it in front of the FPA. The positioning tolerance was verified to be below 0.1 mm.

Figure 22 shows one of these jukeboxes already mounted with all of the BB filter trays. Figure 23 shows the jukebox inside the camera with the focal plane and preamplifier boards on top of the FPA. During observations, the filters are placed 4 mm above the surface of the detectors, sandwiched between the cryostat window and the detectors.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

4.5.2. Shutter

The shutter of the camera is installed outside the cryostat. It contains two blades sliding in and out of the optical beam. The hardware and control software used are the same as for the filter exchange system (Section 5.1.2), with a lighter transmission system based on pulleys and traction chains (Figure 24). It has an opening aperture of 220 × 220 mm and a minimum opening time of 0.5 s.

Zoom In Zoom Out Reset image size

The shutter is made of aluminum covered with a matte black zinc laminar coating made at Asacoat.³⁸ This process was chosen instead of anodized aluminum because of its better antireflective properties. In order to avoid water condensation on the lens, a dry air blowing system was integrated into the shutter. It was thoroughly designed to distribute the air in a laminar flow in order to avoid turbulence in the light path.

Extensive mechanical stress and light tightness tests were done prior to the shutter installation with satisfactory results. Figure 25 shows a picture of the finished shutter during the light tightness tests prior to being mounted in the camera.

Zoom In Zoom Out Reset image size

4.5.3. Camera Services and Swan Neck

Several services and electronics are mounted on top of the camera vessel: the two MONSOON electronics racks with all of the readout boards, the vacuum pumps, several vacuum pressure sensors, the vacuum electro valves, and two boxes with the needed actuators and power supplies for the control system. The first scroll vacuum pump is at the telescope ring and connected to the turbo-molecular pump with a 10 m corrugated tube. The rest of the services are outside the ring: the electronics power supplies, cooling compressors, and slow control system electronics are mounted in one of the Nasmyth WHT rooms, and the computing systems are in the WHT control room and connected to the camera through optical fibers.

Fixed service cables running from a patch panel at the ring to the Nasmyth room are permanently installed in the telescope structure. They include motor control, cooling pipes, power supply cables, optical fibers, and the Profibus³⁹ link for the slow control system. From the ring patch panel to the camera, all of these connections are routed through the telescope spider thanks to a dedicated swan neck with all cables preinstalled. The swan neck is made of black-painted aluminum and helps to speed up the installation of the camera. Figure 26 shows PAUCam mounted in its testing station at the WHT prior to its first installation at the prime focus of the WHT. All of the services and cables described above can be seen in this picture.

Zoom In Zoom Out Reset image size

The final verification of the system was done at IFAE. A simulator (see Figure 27) of the telescope was built, reproducing the interfaces of the final assembly. It can move the complete camera in every position (elevation and rotation) to simulate real conditions and verify that the tray movement system works properly in real observing conditions.

Zoom In Zoom Out Reset image size

In order to align the focal plane with the WHT interface, the Stil OP10000 confocal sensor used to check the FPA planarity, together with the x–y table, was mounted in the simulator and modification of the camera–telescope mounting interfaces was done to minimize the tip-tilt corrections needed after installation at the WHT.

The main requirements of the PAUCam SCS are related to motion control, as the filter trays are inside the vessel and inserted only a few millimeters away from the CCDs. Therefore, extreme care has to be taken in managing the acceleration and speed of the trays and the torque to be applied to the motors. The monitored torque is used as a main indicator of the correct behavior of the moving parts inside the vessel. The shutter motion control also needs to ensure that the exposure time of the camera is consistent.

5.1.1. Slow Control of the Filter-tray Jukeboxes

Each of the two tray jukeboxes inside the vessel uses two motors. One motor is used to select the filter tray (vertical motor), and the other is used to slide the filter tray into position in front of the focal plane (see Figure 29).

Zoom In Zoom Out Reset image size

All movements inside the vessel have their torque limited by hardware in order to minimize the creation of particles in case of friction or collision.

Vertical movements are allowed only to discrete positions. Only at these positions can the filter tray slide out of the jukebox. They are calibrated for each of the filter trays, and their values are stored in nonvolatile memory. As there are no hardware limits to the movements inside the vessel, the calibration of the system is a critical step of the development.

5.1.2. Shutter Control

The PAUCam SCS is in charge of opening and closing the shutter in order to take exposures. The PAUCam data acquisition system (DAQ; see Section 6.2) provides a TTL signal to open and close the shutter for a given time.

All shutter movements follow a configured profile that tries to minimize jerking movements and transit time. Two motors control each of the shutter blades. With the usual exposure times of PAUCam, first a blade is opened, and at the end of the exposure time, the other blade is closed. For shorter times, the first blade slightly opens, and the second one starts closing shortly after, without waiting until the first one is completely closed. In all cases, special care must be taken to equalize the profiles of both blades and ensure that the open and close operations are equal. Figure 30 shows this profile, calibrated during the PAUCam assembly process. A video camera at 50 frames s^–1 was used to estimate the exposure time error at 500 ms. It resulted in an error below 1%.

Zoom In Zoom Out Reset image size

PAUCam has several sensors to monitor its status. Based on sensor values and the camera state, the SCS can take safety measures to protect the CCDs, starting the relevant protocols to put the camera in a safe mode. Temperature sensors are placed in the cold heads, focal plane, filter-tray jukeboxes, etc. They allow the SCS to monitor the current temperature and its change rate and decide any needed action. There are two redundant vacuum sensors in the camera vessel and additional ones in between the turbo-molecular and scroll pumps.

To allow for several responses of the system, alarms and security routines depend on the camera state (parking, cooling, operation), which is set by the PAUCam camera control system (CCS; see Section 6). For example, at any given time, the SC might have to heat the CCD's cold plate if the vacuum level surpasses a configured limit.

The SC has limited access to nonmotor actuators. The basic actuators on the camera system are vacuum pumps, cooling compressors, and heaters.

The turbo-molecular vacuum pump is permanently enabled, and the SC is used for its activation only.

The cooling compressors ensure that the focal plane and NB filter trays are at the required temperature. The compressors are monitored, activated, and disabled by the SC. To correct the temperature on the focal plane and allow the warm-up, with the CCDs following a controlled profile, the heaters inside the camera are used. The heaters are the only active control over the temperature of the system.

The OCS is the core of the PAU CCS. It is in charge of coordinating all other subsystems to acquire exposures. It has two types of interfaces: one to connect to the satellite nodes and another for the user interfaces (UIs). It is in charge of building the system when a new configuration is loaded, launching each of the satellites defined in the configuration at the specified IP address and executing and coordinating every observation during the night. In order to perform this task, different levels of executions are defined (see Figure 32).

Zoom In Zoom Out Reset image size

Macros. The OCS loads macros defined at specific locations. Macros are the base commands made available to the user, who has no direct access to satellite functionality. Macros can call other macros. The OCS, when loading macros, checks that there is no infinite recursive call.

Observation blocks (OBs). The OCS can define several types of exposure (bias, flats, science, and others), which are also loaded dynamically at boot. Each exposure is defined in terms of high-level macros and can itself be considered a macro. When a macro defines an exposure, it is called an OB.

Night observation definitions (NODs). Observers can define a night observation plan in advance, setting which calibration exposures and targets they want to acquire. An NOD is composed of a set of parameter templates and a list of OBs. Once loaded, PAUCam can take all exposures automatically.

The system used to read out the PAUCam CCDs is based on MONSOON+PanVIEW (see Section 4.3 for details). The function of the DAQi is to provide the PAU CCS with an interface to several PanVIEW instances to be able to use it from the OCS as a block.

The PanVIEW system is in charge of getting pixels and header information out of a detector system through a digital link to the MONSOON crates installed in the camera.

The DAQi provides functionality to configure the readout electronics for several readout modes. It implements the control for regular and guiding exposures, depending on the exposition mode used. As several PanVIEW instances are used, each one generating a FITS file, the DAQi is in charge of creating a single FITS file once all parts are received. When a new FITS file is ready, it publishes its location, so any other satellites requesting it can use it.

PAUCam is a visitor instrument of the WHT. The TCSi is in charge of transmitting commands to the WHT telescope control system (TCS) and accessing the WHT meteorological database. The TCSi has several interfaces with the WHT.

As can be seen in Figure 33, the WHT offers a CORBA⁴⁴ -based software interface to operate with its TCS. The WHT TCS enables the PAU CCS to point the telescope toward a target, set the focus, enable or disable auto-guiding, and send the corrections computed by the guider process (see details in Section 7.2).

Zoom In Zoom Out Reset image size

The WHT has a local weather station that stores its data to a PostgresSQL⁴⁵ database. The TCSi accesses this database periodically and publishes its latest results to the PAU CCS system, so its data can be integrated on resulting FITS files.

The QA is the satellite in charge of calculating the quality of the images acquired and giving the possibility of generating a visual check of the acquired FITS for the observer. The QA starts as many draw, analyze, and check observations (DRACO) instances as required either in the local host or in remote machines, and it is in charge of distributing jobs to them. Each DRACO then writes its results into a database. The different analyses performed with DRACO are detailed in Section 7.3.

The PAUCam SCi is the satellite in charge of sending commands to the SCS (Section 5) and publishing the values of all sensors (many sensors and not only one).

The PAU CCS has a main state machine and several sensors that determine the status of the camera. On the main server of the system, there is a service that stores most of the sensor data continuously and organizes it to be able to retrieve time series of it.

A website has been created to follow the evolution of the camera and its status throughout its complete operation history. There is also a database that stores the OBs executed by the OCS and its parameters, as well as any exception that has occurred during the night execution. Finally, under certain circumstances, a service sends emails to a list of trained personnel when the camera is in an emergency state.

In addition to the interface used for all satellites described above, the OCS has a socket interface to enable UIs to send commands and receive its return values. The PAU CCS has two types of UIs: a terminal command and a graphical UI (GUI).

The GUI is the main interface between the operator and the system. It contains all of the system information and keeps a history of the system's use. The GUI is organized hierarchically to allow the user to have the control needed for each operation (execute different exposure types, control the telescope, NODs, etc.). The GUI also integrates displays that show the alarms, detailed logging, OBs already executed, etc.

Integrated into the execution system of PAUCam is an event database. These events are stored on a PostgreSQL database during execution time and can come from several sources, the most important being the macro execution system. The most typical events registered are alarms, telescope pointing status, and observation executions, but any new event can be defined dynamically.

The second part of the system is the PAUCam logbook, which is the front end of the event database. The logbook displays events versus time with optional filters by event type or date. It also lets the operator introduce log messages into the database. Another feature of the logbook is that it can make online plots of the evolution of the live QA results from exposures taken during the night, together with the environmental conditions at the observatory. An example of this information for a complete night can be seen in Figure 34. The PAUCam logbook is programmed in Python and uses the well-known pyramid framework⁴⁶ as a back end, whereas the front end uses the javascript framework Angular.⁴⁷ Plots are produced using the plotly library.⁴⁸

Zoom In Zoom Out Reset image size

In order to set the camera in focus, we have designed a specific readout and analysis system that provides accurate measurements of the focus position at each detector. As described in Section 4.3, to speed up the acquisition of the image at different focus positions, we produce a single exposure image, where charges are shifted between the different steps, producing an image where the same source (star) is exposed several times, each at a different focus position. Figure 11 shows an example of such an image, where the same star is seen in different places on the CCD separated by a fixed number of pixels (a jump). The first focus is identified because the distance to the second one is doubled (a double jump).

Once the image is read and built, the control system automatically triggers the analysis of this "stacked" focus exposure. In the first place, the raw image is quickly reduced by applying a gain and overscan correction. Over the reduced focus image, we run SExtractor (Bertin & Arnouts 1996) to obtain a list of detected sources. As we need to identify the pattern of one double readout jump, plus N jumps, we create an R-Tree (Guttman et al. 1984) of all detected stars, which allows making a very fast spatial match, keeping the sequences that match all of the jumps. The measured FWHMs of each star in all of the focus sequences identified for each CCD detector are median combined to produce a master curve. This curve is fitted to a sixth-degree polynomial that allows one to automatically detect the focus position that minimizes the FWHM of the point-spread function (PSF). Figure 35 shows an example of a focus sequence analysis automatically performed by the PAUCam control system. The top plot shows the FWHM as a function of the focus position for all analyzed stars. Some of the stars have low signal, or they are in a more distorted area of the focal plane. The red curve is the mean of all measurements. The middle plot shows the average per CCD. One can see small variations because of the tolerances in the CCD fabrication and installation process. The bottom plot shows the results with the fitted polynomial and the preferred focus value.

Zoom In Zoom Out Reset image size

The algorithm runs automatically in a special DRACO instance (Section 6.4), and the results are presented in the GUI application to allow the observer to perform any needed focus correction.

PAUCam has a built-in auto-guiding system to allow precise tracking of long exposures with minimal PSF size and optimal image quality.

When the guiding system is enabled, two of the outermost corner detectors of the focal plane (G12 and G17) are devoted to guiding, leaving the science exposures to be done with the remaining 16 CCDs, including the most illuminated detectors in the focal plane.

When the shutter opens at the beginning of the exposure, the guiding detectors continuously take images with exposure times between 1 and 5 s, depending on the sky conditions and stellar density. The guider software then analyzes the images from these two guiding detectors and, using the stars available in the field, identifies the telescope tracking error and sends the correction signal to the TCSi.

To minimize the latency of the correction signal sent to the TCS, only a subset of the detector is read out. This subsection is called the ROI, which is typically adjusted to a 1000 × 2000 pixel subset in the most illuminated side of the detector. In this area, the image is quite distorted and partly vignetted.

Depending on the stellar density of the field, a number of guide stars from three to 15 are selected to measure the centroid correction. In the first place, the guiding algorithm will quickly reduce the image, subtracting the bias signal and correcting the gain. The image is then analyzed by SExtractor, providing a list of source detections. These detections are filtered to ensure that stars are neither blended nor very close to each other, in the border, unaffected by cosmic rays, etc. The remaining list of candidate guide stars is selected and will be searched in the next iterations when the new guiding images arrive. Stars that are not found in new iterations will be removed from the list, assuming that they might be moving objects, badly selected cosmic rays in the first place, or too-low signal-to-noise ratio (S/N) stars that went below the detection threshold. Having multiple sources as guide stars provides more accurate guiding, even under this area with more distortion of the focal plane. The individual tracking errors found in each star are combined using a weighted average by each star's S/N. After integration, readout, and analysis, the error signal is sent to the TCSi every 6–8 s.

In any case, the WHT's tracking performance is very good, even in an open loop (without any auto-guiding), with an error of less than ∼01 drift minute^–1. Therefore, for exposures shorter than 3 minutes, auto-guiding might not be necessary, and the two guiding detectors will act as the rest of the scientific CCDs, allowing PAUCam to observe with the full 18 detector set.

Figure 36 demonstrates how the guider corrects the telescope drift when guiding in a closed loop. The figure shows the guiding correction computed in X and Y for the case where the signal is sent to the TCSi (closed loop) or when the correction is not applied. One can see that the tracking of the telescope can track the stars even without auto-guiding up to around 4 minutes, but the stars are completely lost when longer exposures are made.

Zoom In Zoom Out Reset image size

During the observations and before they are transferred to the PAU data center (Tonello et al. 2019), the FITS files produced by PAUCam have to pass a series of checks to monitor the quality of the images. DRACO (see Section 6.4) is the satellite responsible for this task and consists of three layers: a manager that orchestrates the commands received by the OCS, defines which checks should be performed, and formats the response of the satellite; a series of algorithms to compute the analysis and quick reduction; and a PostgresSQL to organize the FITS files metadata and quick-reduction results.

The manager receives from the OCS the parameters of each FITS file that has been created: its location, file name, and kind (calibration, scientific, test, focus, etc.). For each kind of file, a specific series of checks to be performed is defined. The DRACO manager is responsible for sending the execution commands to check the files according to their kind.

The algorithms for the online analysis are written in Python and organized into two groups: basic checks and quick reduction.

Basic tests are performed over all of the raw FITS files and are aimed at checking the file integrity and overscan values (electronic noise). The file integrity test checks for consistency of the information given in the header in terms of number of extensions and data type, as well as for the readability of the content. The overscan test computes the mean, median absolute deviation, and standard deviation values of the overscan region of each amplifier and returns the values to the manager.

Quick reduction is performed only on scientific images and consists of more advanced tests that give the observer a better understanding of the quality of the data taken. The reduction procedure consists of running the Astromatic packages SExtractor (Bertin & Arnouts 1996) and PSFEx (Bertin 2011) over each CCD, resulting from combining the data of the four amplifiers after overscan and gain correction. The values returned are the mean background value, the number of detections, the mean elipticity of the detected objects, and the mean PSF for each CCD of a mosaic.

The comparison to the reference values for every CCD (each of them with a different NB filter) in each single image is shown graphically. The manager receives the results from the overscan check and quick reduction and formats them to be displayed in the GUI monitor. When a value is outside the valid range, the graph turns red, and the operator is in charge of taking the proper action. Figure 37 shows the result of DRACO for a normal science exposure. It shows the results of the overscan per amplifier, PSF, background, ellipticity, and number of detections per CCD.

Zoom In Zoom Out Reset image size

DRACO registers all of the information related to the raw and reduced files as soon as the FITS files are created and the check is performed, organizing them in a relational database (PostgresSQL). The DRACO database consists of five main tables: raw mosaic and raw image contain the metadata regarding the raw FITS files and the results of the basic checks, reduced mosaic and reduced image collect the results of the quick reduction, and observation set collects the information and eventually notes regarding the group of mosaics taken the same night, by the same operator, and under the same camera conditions. The DRACO database is queried to monitor the evolution of the image parameters during the night (see Section 6.8).

The quick reduction is the most time-consuming process, but it has been designed to deliver results in real time. The hardware dedicated to the online analysis is a 12 core, 32 GB computer, allowing DRACO quick-reduction algorithms to work with multiple threads. The DRACO response is displayed to the operator just 1–2 minutes after the image has been taken.

DRACO has been designed such that it is easy to integrate new algorithms of analysis to run over PAUCam FITS files. As an example, the algorithm for the stacked focus analysis (Section 7.1) has been integrated in the satellite, and the results are formatted by the manager and displayed in the GUI in real time.

Since its installation at the prime focus of the WHT in 2015 June, we have been taking a photon transfer curve (PTC) in each semester to check the behavior of the camera, determining the current gain, charge transfer efficiency, linearity, and other parameters of the camera. Monitoring the camera regularly, we can determine its stability and update the map of the camera defects due to change or problems with the electronics and detectors.

Our PTC standard calibration set is composed of 32 dome-flat images. Images are grouped in pairs of the same exposure time, which varies logarithmically between 1 and 40 s. Each amplifier of each CCD is analyzed independently, obtaining 72 PTCs. Figure 48 shows the PTC for the first amplifier of CCD 1.

Zoom In Zoom Out Reset image size

With each PTC, we can determine the gain for each amplifier of the camera. Figure 49 presents the mean gain obtained from two PTC tests carried out in 2018 January. Differences between amplifiers in the same CCD are expected because the electronics are different for each one.

Zoom In Zoom Out Reset image size

The linearity curve for the first amplifier of CCD 1 is shown in Figure 50, presenting the signal versus the exposure time. After about 30 s, the ADC is saturated. The dispersion in each point in the graph is due to vignetting. The photoresponse nonuniformity (PRNU) of all CCDs is below 1%.

Zoom In Zoom Out Reset image size

The image quality delivered by PAUCam has the contributions of the optical system and the camera itself. The WHT prime focus corrector was not optimized to cover the wide FoV of the PAUCam focal plane. It delivers a nonflat focal plane introducing distortion in the PAUCam images. By now, we have collected enough images to be able to map this distortion accurately. Figure 51 shows the average pixel scale across the eight central CCDs. This map has been calculated using reference stars during the astrometric solutions calculated in the main data reduction pipeline and provides an accurate representation of the distortion in the focal plane. The optics provide this smooth pixel scale pattern, which is dominated by the prime focus corrector contribution.

Zoom In Zoom Out Reset image size

The best images we have measured with PAUCam have PSFs of around 055–060 FWHM. The external seeing monitors measure FWHMs of around 030 when this happens. This would imply that the system introduces a contribution of approximately 050 to the observed PSF.

We calibrate our astrometry against external stellar catalogs. Currently, we use the Gaia DR2 (Prusti et al. 2016; Brown 2018). We also check our internal consistency with repeated observations of the same fields. Figure 52 shows the internal astrometric residuals between overlapping exposures. We find an average rms value below 50 mas, significantly smaller than the average pixel scale (shown with a dashed line). The accuracy achieved allows precise photometry even in the more distorted regions of the focal plane.

Zoom In Zoom Out Reset image size

PAUCam was designed to obtain accurate photometric redshifts for galaxies using photometry in NB filters. The relative photometric calibration is therefore very important to obtain robust SEDs of the galaxies. The photometric calibration is extensively discussed in F. J. Castander et al. (2019, in preparation). Here we outline the most salient points. The most contributing factors to the error on the photometry are the derivation of the zero-points and the way the photometry is performed. The zero-points (the conversion factor from electrons detected in the detector to fluxes or magnitudes) reflect the sensitivity of the whole system. We compute the zero-point for each image from stellar photometry comparison to SDSS. We fit stellar templates to the SDSS photometry to compute stellar SEDs. We then synthesize the PAUCam NB photometry from the SED fit and compare the fluxes to the ones measured. We obtain the image zero-point as the factor that we need to apply to convert the observed fluxes to the synthetic ones for each single image. We also compute global zero-points from spectrophotometric standard photometry. These need to include the observing conditions to be applied to each image. These zero-points are the basis for our exposure-time calculator. Figure 53 shows the computed zero-points for the NB filters. The zero-points for the BB filters are similar to those of the DECam at the CTIO 4 m Blanco Telescope. This is expected, as both cameras have similar detectors and are mounted on similar-sized telescopes.

Zoom In Zoom Out Reset image size

The relative calibration of the image zero-points of all filters is necessary to build reliable SEDs. Figures 54 and 55 show examples of quasar, galaxy, and stellar SEDs. The PAUCam measurements (circles) are compared to the corresponding SDSS spectra binned at the same resolution (solid lines). The agreement between both measurements demonstrates the capabilities of NB imaging to obtain accurate SEDs.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The photometric calibration has allowed us to measure the photo-z accuracy in the COSMOS field (Eriksen et al. 2019) whose value is σ₆₈/(1 + z) ≃ 0.0035 for i_AB < 22.5, which agrees with the design specifications of PAUCam.