CFHT, Instruments, Detectors, IR, Redeye Manual: Chapter 3.


Redeye Design Summary

There are two complete Redeye cameras. The cameras are virtually identical in all respects except one houses 1.7:1.0 reimaging optics, while the other houses 0.7:1.0 reimaging optics. Table 3.1 outlines the plate scales offered by the cameras at the various CFHT Cassegrain foci. Each camera uses a Rockwell NICMOS3 Hg:Cd:Te array with 256 x 256 pixels that are on 40 m centers.

Table 3.1 - Plate scales and fields of view for the two cameras are listed.

The vacuum jacket for each camera is an electron-beam welded structure with 6 flat faces. For convenience, the dewars were designed to permit bottom-feed nitrogen injections. All of the optics are held in tubes that are secured to a vertical cold work surface via detachable pillow blocks. The lenses in each tube are held in spring loaded cells and are surrounded by numerous baffles to minimize stray light. Each lens tube can be removed and split to provide easy access to the pupil stop, which must be changed to accommodate f/8 and f/35 modes. Two filter wheels, each containing 8 filter cells, are driven by a pair of external stepper motors. The array is located near the base of the dewar, supported by rigid fiberglass stand-offs, and cooled with a cold finger attached to the LN2 can. The chip carrier assembly is housed in a separate radiation shield to block stray light.

The same DSP-based controller planned for use with all future CCDs is used to run the Redeye infrared cameras. The "Gen III" controller is coupled to a separate computer (Sparc 1E engine) that serves as an interface between the controller and the summit HP9000 computer. The Sparc engine and its accompanying 16 Mb of RAM and 700 Mb hard disk are linked to a fiber optic transmitter in a single VME Crate. The fiber link is used for high speed data transfer from the controller to the array computer, where images can be buffered, coadded, or manipulated in a variety of ways. Observers are provided with the "Pegasus" window environment to use for camera control and image manipulation.


An overriding goal in the design of the Redeye cryostats was the ease with which they could be reconfigured and maintained. Given the expected ~5 year lifetime of these instruments in the CFHT "arsenal" of array detectors, having cameras that were simple to disassemble yet could meet the science specifications demanded by the Community was of paramount importance in the design phase of the project. Tapping the experience of the IRTF in their construction of Protocam (Toomey et al. 1990) and an assortment of square photometer cryostats, it was decided early in the project to adopt a similar square "theme", with a large vertical cold work surface that would in effect become an optical bench. Various optical components are mounted on this surface, which provides the extremely flat and stable environment needed for the delicate optical alignment in the cameras. Another advantage of this design is that the flat surfaces on the vacuum jacket make attaching the motors and electrical connectors convenient. Electron beam welding was used to join the jacket plates since this technology provides the level of component precision and jacket strength needed by the camera. The nitrogen can is also an e- beam welded structure, composed of simple aluminum plates with an internal reinforcing bar. Most of the machining and complexity of the basic cold structure is in the horizontal work surface, which houses the detector and was milled from a single piece of aluminum. For convenience a bottom nitrogen injection feed was built into the cameras so they could be filled by the CFHT day crew and telescope operators in the same fashion as existing CFHT CCDs.

Figure 3.1 shows the mechanical layout of the cold structure and outer vacuum jacket. The jacket uses a variety of subplates to allow coupling of exterior components to the interior mechanism as well as easy access to key components. For example, the motors are mounted on a separate plate so that, if it is ever decide to relocate the motors or totally reconfigure the interior, this can be done by replacing a single subplate instead of creating a large number of new holes in the jacket. The window is mounted in a subplate as well in order to facilitate different window sizes by simply replacing the cell without making any fundamental changes to the jacket. Access to the array is possible by removing another plate on the bottom of the dewar, then removing a radiation shield on one side of the detector cavity. Finally, a large plate covering an entire side of the dewar permits access to the vertical cold work surface, optics, filters, etc.

Figure 3.1 - Front and side views of the Redeye camera assemblies are shown. The design is based on an uplooking dewar with a bottom nitrogen feed. All optical elements are mounted on a vertical cold work surface. The front view shows the wide field optics while the side view shows the narrow field optics.

Modular components were used throughout Redeye's interior. The array is mounted via fiber glass tabs to the cold structure in a well shielded cold cavity. These tabs provide thermal and electrical isolation of the array from the rest of the camera while holding the array fixed to within 1 pixel of its nominal position, regardless of the camera's orientation. The detector is cooled by a cold finger, with temperature control provided by the Gen III electronics. The array is mounted on a board that can be removed from the entire camera by removing four screws and unplugging a single socket used to transmit all electronic signals to and from the array. The lens barrels are mounted in pillow blocks so that, by simply removing the top of each block and sliding the collar on the end of each tube, it is possible to remove each tube within a few minutes from the cameras. This procedure also preserves critical dimensions of the tubes so they can be replaced or interchanged between cameras while keeping everything in focus. The filters are mounted in cartridge assemblies that can be removed after loosening 4 screws in the vertical work surface. Under normal circumstances, it is not necessary to remove the filter wheels (e.g., to replace filters), but having the ability to remove and interchange wheel cartridges between cameras was seen as a desirable feature during the normal lifetime of the cameras. Access to filters is provided by small plates on the outside of each cartridge. Finally, the entire inner cold structure is mounted within the vacuum jacket on four fiberglass tabs near the bottom of the camera. After unplugging the inner electronics and detaching the cold structure from these tabs it is possible to remove the entire cold structure from the vacuum jacket rapidly and easily so that access to any part of the camera is possible. Achieving this level of modularity, precision, and interchangability between cameras, while keeping a total camera weight under ~20 kg, reflects the central design goals of the Redeye cryostats.

Optical Design Summary

Each Redeye camera houses 5 lenses that serve to reimage the telescope's focal plane onto the infrared array. Unlike many of the infrared cameras built during the late 1980's, Redeye was intended to be used at both a relatively fast f/8 telescope beam and a more conventional f/35 beam. The aberrations created by an f/8 beam (which scale as ~f/#2 to 3) combined with the relatively small 40 m pixels of a NICMOS3 array led to the use of LiF and BaF2 as lens materials. This combination of materials offers much better color correction than the more commonly used ZnX lenses used in many infrared cameras. Also, their low indices lead to very high intrinsic transmission and reduced mechanical tolerances for the lens cells. Unfortunately, LiF is a rather difficult material to use in lens manufacturing and, due to its extremely high thermal expansion coefficient, special precautions were used in the mounting and coating of the Redeye optics. In particular each lens receives radial mechanical support by touching two small bumps 120 apart on the inside walls of the barrels. A small strip of beryllium copper mounted 120 to the bumps is used to provide ~1 lb of radial pressure on the lenses. Most importantly, the bumps are machined so that, when cold, zero net optical decentration is achieved between all lenses to within the ±0.001" machining tolerance used during fabrication. This technique of mounting lenses also eliminates the possibility of crushing the lenses during cool down - a real threat since the lenses have such poor thermal conductivity that they will not contract as fast as the barrels. In a similar manner, the lenses receive axial support through precision spring tab retaining rings which provide ~1 lb of radial force at points separated by 120 around the edge of each lens. These tab supports are held in place by screws so that it is impossible to apply too much pressure on the lenses from the screws, i.e. simply tightening the screws with moderate torque delivers the proper amount of axial force on each lens.

Figure 3.2 shows the design of the Redeye optics, including axial and marginal rays coming from the CFHT Cassegrain f/8 configuration. In each design, the field lens is located ~25 mm behind a CaF2 window in the dewar. The telescope's focal plane is coincident with the front apex of the field lens. This of course makes the system somewhat susceptible to dust being reimaged onto the array, but this problem is expected to be minimal since the field lens is in a sealed environment. Redeye was designed this way in order to allow custom field masks to be installed in the field lens cell, as may be needed for imaging FTS, cold grism slits, or coronagraphic applications. Next in each system is a filter, which is mounted in one of two wheels in the camera. Following the filter is a cold stop, which can be easily removed from the camera and exchanged with stops of various sizes and shapes. Observers who wish to use Redeye with a visitor instrument should specify the f-ratio of the beam they intend to send into the camera so a properly sized cold stop can be placed in the camera prior to their run. In the standard f/8 Cassegrain mode the cold stop has a central occulting disk that blocks thermal flux from the f/8 baffles and hole in the primary mirror from reaching the detector. This stop theoretically reduces the K-band background by ~0.5 mag, while reducing the system throughput ~1%, hence it boosts the system's sensitivity greatly. Behind the cold stop lies 4 reimaging lenses which, in the narrow field optics, reimage the telescope's focal plane onto the array with 0.7:1.0 magnification. In the wide field optics these lenses yield 1.7:1.0 minification of the telescope's focal plane. Not seen in Figure 3.2 are the ~15 baffle stops that surround the beam as it passes through the optics. These baffles, along with most of the interiors of the optical tubes, are painted black to minimize stray light.

Figure 3.1 - The Redeye narrow and wide field optics are shown. They provide 0.7:1.0 and 1.7:1.0 magnification respectively and are composed of fluorite crystalline materials.

Exact optical specifications are provided in Tables 3.2 and 3.3 for observers who would like to insert these optics into models of their own, e.g., to evaluate the performance of visitor instruments in combination with Redeye.

Table 3.2 - Detailed specifications for the Redeye narrow field optical design are listed. All dimensions are in millimeters.

Table 3.3 - Detailed specifications for the Redeye wide field optical design are listed. All dimensions are in millimeters.

The 77 K refractive indices over the 1.0 - 2.5 m wavelength range for LiF and BaF2 are given by the Schott equation:

The constants ax for the lens materials used are listed in Table 3.4.

Table 3.4 - Schott constants for equation 3.1 are shown for the materials used in the Redeye lenses.

Additional details about the Redeye optics are available from the CFHT technical staff, upon request.

Adjusting Redeye's Internal Focus

The design specifications made on the Redeye cold optics make it possible to swap complete tube assemblies between dewars while preserving the focus of the optics with respect to each other and the detector. In order to meet this design goal, the barrels used to mount the lenses have adjustable lengths that can be fixed through the use of a locking collar. Barrel lengths are changed by simply threading them in and out of their subassemblies, which are shown in Figure 3.3 for both the narrow and wide field lens sets. Note that it is crucial that these barrels have the correct warm lengths so that when cooled they contract into a configuration that (1) places the pupil image formed by the field lens in the plane of the cold stop and (2) places the imaging planes of the reimaging quartet on the apex of the field lens and detector respectively. Clearly this is a critical set of adjustments that must be made on the optics since errors in the location of a single component translate into errors across the entire reimaging system. For example, if the pupil plane of the field lens is misplaced then the occulting cold stops used to block light from the f/8 baffles and primary mirror cell will lose their effectiveness, resulting in a slight increase in background thermal flux and a decrease in system throughput. Perhaps the largest problems that can stem from improperly adjusting the reimaging barrel is the resulting shift in the focal plane of the reimaging optics and apex of the field lens. This can be corrected by adjusting the position of the telescope's secondary mirror, but at the expense of the quality of the pupil image at the cold stop. Furthermore, if the reimaging barrel has the wrong length it will be impossible to simultaneously focus objects in the telescope's focal plane and any special field mask used in Redeye, e.g., the masks that will probably be used to support future grism and coronagraphic applications.

Figure 3.3 - The warm lengths (in mm) of the Redeye narrow (bottom) and wide field (top) lens barrels are shown. It is critical that these tubes have the depicted warm dimensions to assure correct internal focus of the reimaging optics.

Referring to Figure 3.3 again, note that dimensions have been listed next to each barrel. These dimensions are the warm barrel lengths that should be used to achieve proper internal focus when the camera is cooled and the entire optical assembly contracts by ~0.4%. Normally these barrels will not change length once the locking collars are secured within the barrel subassemblies, but repeated thermal cycling may over time lead to the collars coming loose and the barrels may inadvertently change length when removed from cameras. If this happens use a precision height gauge (or a set of calipers if that is all that is available) to readjust the barrels back into their correct lengths.

Mounting the Redeye Dewar

Each Redeye dewar has a total of 12 M10 holes tapped into its vacuum jacket that can be used to mount the camera at different CFHT foci. These holes are shown in Figure 3.4 and are split into sets of four on three different flat surfaces of the jacket. Figure 3.4 shows the exact mechanical locations of these holes and can be used as a reference for observers wishing to mount Redeye to visiting instruments. Note that two overall dimensions of the camera are set by the filter wheel stepper motors and a shutter, which are shown as dotted lines in two of the views. The position of the window and optical axis with respect to the jacket is also shown.

Figure 3.4 - The pattern of M10 holes in the vacuum jacket of each Redeye camera is shown. All dimensions are in millimeters and are common to both cryostats except the two in parentheses, which apply only to the wide field camera. At upper right is a perespective view of the vacuum jacket with the filter wheel motors and subplate, array access subplate, window cell, and main cover removed.

There are two different sets of spacers used for mounting Redeye to the CFHT Cassegrain environment, depending on whether or not the camera is at the f/8 or f/35 focus. Figure 3.5 shows the series of spacers needed to mount Redeye at the f/8 focus. In this mode, the standard 100 mm, 280 mm, and Redeye adapter plate are used between the camera and Cassegrain environment. Figure 3.6 shows in detail the Redeye adapter plate, which has a special hole cut in its center to allow the external shutter to lie within the adapter plate. The Redeye adapter plate has 4 holes in it that are on an identical layout with the 4 holes in the top of the Redeye vacuum jacket. Using the M10 bolts that are normally left threaded into each camera yields a fairly accurate radial placement of the camera's optical axis with respect to the telescope's axis. Note that this mode does not require a tip/tilt adjustment and therefore uses only simple spacers to place the camera near the optimal f/8 telescope focus. Also note that two of the holes are 20 mm in diameter, while the other two are 10.5 mm in diameter (and when used with M10 bolts provide accurate registration with respect to the Redeye optics). This difference is due to slight machining errors in the locations of the top mounting holes. This design was adopted In order to use a single plate for all Redeye modes, and preserve optical alignment with the different hole placements in the cryostats.

Figure 3.5 - The standard f/8 and f/35 Redeye Cassegrain configurations are shown.

Figure 3.6 - The mechanical specifications for the Redeye Cassegrain adapter plate are shown. All dimensions are in millimeters.

Figure 3.5 also shows the configuration that should be used at the infrared f/35 CFHT focus. Once again, the same Redeye adapter plate is used to attach the camera to the spacers, except in this mode a special tip/tilt spacer is used that allows the camera's optical axis to become parallel with the telescope's optical axis. This spacer has a thickness of 150 mm so that, when used in combination with the standard CFHT 200 mm spacer and Redeye adapter plate, the Redeye focal plane is nearly coincident with the optimal f/35 focus.

Filling Redeye with LN2

Due to its relatively large thermal mass, cooling Redeye is not as straightforward as the other CCD dewars commonly in use at CFHT. In order to protect the optics from being broken by differential contraction during cool down, all lenses are mounted in spring loaded cells. Unfortunately this creates a poor thermal contact between the lenses and the cells which, in combination with the extremely low conductivity of the crystalline lenses, leads to essentially only radiative cooling for all of the optics. This in turn creates cool down times of roughly 2 - 4 hours for the various lenses. Since the lenses have intrinsically low emissivity in the 1 - 2.5 m range though, excess radiation from the lenses is actually quite small during the cool down period. Lab tests indicate that it takes about 5 hours for everything in the cameras to cool to 77 K. Since the NICMOS array's dark current is a strong function of temperature, typically one sees rapidly changing dark current patterns as the array slowly cools, rather than thermal flux from the inside of the camera.

The dewar should never be cooled with a warm pressure exceeding 10 mtorr so that it can effectively cryo-pump during the cool down phase. In order to reach nominal hold times, the cold pressure in the dewar should be below 1 mtorr, and ideally should drop below 0.1 mtorr so that radiative heating is the dominant thermal load on the cold structure (i.e., not conduction by residual gas molecules). A pressure test port is attached to the outside of each dewar to check the dewar pressure whenever necessary. An even better diagnostic of the dewar's vacuum quality and heat load is the boil-off rate when the camera has reached an equilibrium temperature. This boil-off rate should be around 1-2 liters sec-1 for both dewars, which corresponds to roughly a 20 hour hold time. Boil-off rates exceeding 5 liters sec-1 indicate a serious problem in the camera and a member of the CFHT technical staff should be contacted to investigate the problem.

Referring to Figure 3.7, which shows the Redeye cold structure from the side, note that the dewar uses a LN2 injection feed that is similar to that used by all of the CFHT CCD dewars. In order to maximize the LN2 throughput, dedicated injection tubes should be left in the dewars after they are mounted on the telescope. These tubes are marked and are cut to the exact length needed to splash LN2 into the can once it clears the tube. Redeye can be mounted in either uplooking or side looking configurations, depending on the insert-tube used in the dewar. The insert-tubes thread into the base of the dewar's LN2 can. When an uplooking mode is used, the longest insert tube should be threaded into the can to make the can's capacity to ~2 liters. Do not use this tube if Redeye is going to be used in a side looking mode, since the end of the tube will be immersed in liquid and a geyser of LN2 will be ejected from the can, which can be very dangerous. Only use the short insert-tube for side looking modes, which places the end of the tube in the center of the can so that LN2 is not spilt at any camera attitude. Of course the capacity is reduced to about 1 liter when this shorter tube is in place, leaving a hold time of ~10 hr.

Figure 3.7 - The setup used to fill Redeye with a cold dewar mounted at the CFHT Cassegrain focus is shown. This filling method is essentially identical to that used for CFHT's CCD dewars.

Experience has shown that the minimum injection pressure that will work is ~15 psi, hence be sure the transfer dewar is pressurized to this level before attempting to fill a dewar, otherwise the LN2 will flash-evaporate as soon as it enters the dewar and essentially no liquid will reach the can. Since all of the summit transfer dewars have been outfitted with pressure safety-valves that open at 15 psi, getting sufficient pressure in a transfer dewar should not require any special effort. If for some reason the pressure in a transfer dewar is well below 15 psi yet the dewar is known to contain liquid, carefully place the dewar on its side for a few seconds. This will vaporize some of the LN2 (when it comes in contact with the warm top of the dewar) and restore the pressure level to 15 psi. When cold and mounted on the back of the telescope (i.e. in the normal uplooking mode), it should take about 15-20 minutes to completely fill Redeye. The dewar is full when a steady stream of LN2 exits the spill port on the injection assembly. It is not unusual for a weak steam of liquid (few drops per second) to fall from the spill port only a few minutes into the injection process, hence do not take this to indicate that the dewar is full. This weak stream comes from the slight amount of LN2 that splashes back down the injection tube and the liquid oxygen that naturally condenses (at 90 K) on the extremely cold surface of the brass injection assembly.

Filling Redeye with LN2 when it is warm requires a different technique than when it is already cold. LN2 should be injected into a warm dewar at the highest rate possible to overcome the dewar's initial thermal load and allow the nitrogen can to accept liquid. Since the pressure from transfer dewars cannot exceed 15 psi, the only way to increase the liquid throughput is to turn Redeye upside down, remove the insert-tube, then use that insert-tube (squeeze the rubber hose on the transfer dewar around one end) to inject LN2 into the dewar. This tube has about twice the diameter of the normal injection tube hence the liquid throughput is greatly increased. After roughly 30 minutes, the dewar will top off with liquid using this fill technique. If time permits, allow the dewar to cool for an hour then top it off again before threading the insert-tube into the can and inverting the dewar into its normal uplooking state. Do not let the dewar sit for more than a couple of hours after the initial injection without topping it off since, when the dewar is cooling from room temperature, the boil-off rate is extremely high and all of the LN2 will be lost in only a couple of hours. Unlike filling a cold dewar, filling a warm dewar should only be done by qualified CFHT personnel since the final step will lead to a geyser of LN2 streaming from the dewar before it is inverted.

If for some reason the dewar has boiled off its entire reserve of LN2 while it is mounted on the telescope, two options for cooling the dewar are available. First, the camera can be removed from the Cassegrain environment and cooled as illustrated in Figure 3.8. This can be cumbersome though since removing the camera from the telescope is at least a two-person job. The other option is to leave the camera on the telescope but remove the insert tube and inject LN2 with that into the camera reservoir for 5-10 minutes. This will use several liters of nitrogen while the can is cooled to the point that it will accept liquid. After cooling the can, immediately replace the insert tube and begin injecting LN2 with the injection tube. Under these conditions it could take ~30 minutes to top-off the dewar. Before attempting either of these solutions, first check the dewar's internal pressure. If it exceeds 5 mtorr, the camera must be pumped back down to < 1 mtorr before attempting to cool/fill it. Otherwise, the activated charcoal in the camera will saturate with condensed gas while it cryo-pumps during the cool down period. In any event, the Waimea technical staff should be contacted before a warm camera attached to the Cassegrain environment is dealt with.

Figure 3.8 - In order to fill Redeye when the dewar is starting at room temperature, it is best to remove the insert tube and inject LN2 with that tube with the dewar inverted to maximize the LN2 throughput and overcome the dewar's large thermal mass.

Array Temperature Control

A number of NICMOS3 properties are temperature dependent and therefore demand precise thermal control of the array in order to maintain stable operation over the course of a night of observing. These properties include noise, responsivity, and most of al, dark current. In order to compensate for slight temperature variations in the array that stem from (1) imperfect thermal coupling of the array to the dewar's cold structure and (2) changes in the cold structure's temperature as the dewar boils-off LN2, the array is maintained at 80 K, i.e., a little above the nominal temperature of LN2 (77 K) at the summit. This is accomplished by a 1/4 watt resistor and a temperature sensitive diode that are embedded in the cold finger, just under the array. The array is heated above 77 K by a small current dissipated in the resistor. This current is controlled by the utility card in the Gen II controller, which is monitoring the temperature of the cold finger array with the diode. The entire operation is handled automatically, hence users need not worry about temperature control unless something appears to be wrong with the camera. If the temperature is significantly above 80 K (by just 10 K), a large increase in dark current will occur. This can be detected by gross changes in the appearance of raw frames and/or simple measurements of the mean ADU level in dark frames. Be sure to read-out the array a few times before making a dark current measurement if it has been exposed to a strong source recently. The temperature of the array is most easily checked by looking at the FITS header of raw frames.

Manual Array Temperature Monitoring

If for some reason the temperature controller fails or it is impossible to get reliable temperature measurements through the FITS headers, it is possible to make a direct measurement of the array temperature with the aid of the filter wheel indicator box (see Chapter 4 for instructions). This box has pair of test points that, when used in conjunction with a multimeter in the diode-check mode, yields the temperature of the array. Figure 3.9 shows the relationship between array temperature and sensor voltage. Normally, if the temperature controller fails, Redeye can be operated safely without the controller (at 77 K) until it can be repaired. Until it is repaired, reliable checks of the temperature can be made with this utility box.

Figure 3.9 - The relationship between temperature diode voltage and array temperature is shown.

GEN III Controller - System Overview

Early in the Redeye project it was decided to use the same "Gen III" controller under development for CFHT CCDs with Redeye. The reasons for using this DSP controller with Redeye are numerous and include:

1) By having a single type of controller that operates all of the CFHT array detectors, the long term complexity of the summit data acquisition environment is greatly reduced and hardware redundancy is increased.

2) The cost of the electronics can in large part be absorbed by a system already under development for CFHT's CCDs.

3) The only significant difference between running a NICMOS array and a CCD with the DSP system is the software that is up loaded into the controller and the dewar cabling. As a result, switching between CCD and Redeye control requires minimal effort.

4) Redeye can take immediate advantage of the software already developed for CFHT's CCD program.

Figure 3.10 shows the layout of the entire system used for Redeye. The controller, which is based on that developed by Leach (1991) at San Diego State University, is coupled to a separate computer (Sparc 1E engine) that serves as an interface between the controller and the summit HP9000 computer. The Sparc engine and its accompanying 44 Mb of RAM and 886 Mb hard disk are linked to a fiber optic transmitter in a single VME Crate. The fiber link is used for high speed data transfer from the controller to the array computer. Observers use the CFHT "Pegasus" window environment for camera control and image manipulation.

Figure 3.10 - The CFHT controller system is depicted. This same controller is used for all arrays at CFHT and is based on the San Diego State University design by Leach (1991).

Features and Design

Some of the features of the controller used for Redeye and all array detectors at CFHT include:

The Gen III controller simultaneously controls and reads out up to 20 arrays at rates of up to 105 pixels per second. It can be used either for operating many arrays in a mosaic configuration, or for operating more than one read-out on each array in applications that require total array read-out time to be minimized, or combinations of these two configurations. The controller supplies up to twelve clocked signals and up to seven DC bias voltages per read-out derived from a bank of digital-to-analog converters (DACs). The analog video output of each array read-out is amplified by low noise pre- and post-amplifiers and filtered with a dual-slope integrator of programmable integration time before being processed by a 10 microsecond Analog-to-Digital Converter (ADC). A fully programmable Digital Signal Processor (DSP), the Motorola DSP56001, provides timing and clocking at a 10 MHz rate, and can be programmed to control a wide range of array geometries and read-out requirements. The controller accepts 24-bit commands from a user-supplied computer over a 4 Mbit/second optical fiber data link, and supplies 16-bit pixel data over a 40 Mbit/second optical fiber data link to the computer. A VMEbus interface card (VME INF) connects the fibers from the telescope to the Sparc1E computer and provides an intelligent DMA controller using another Motorola 56001 DSP.

The controller electronics next to the dewar has five separate components including (1) a digital timing board, (2) one or more analog read-out boards, each one handling one read-out, (3) a utility function board that manages detector temperature regulation, shutter operations, and an assortment of other tasks, (4) a backplane communications board, and (5) a power supply. A three board set would suffice for reading out single arrays and a 22 board set would be used for reading out 20 arrays. The digital timing board contains the fiber optic communications circuitry, the DSP, the DSP program, the master clock, and interface circuitry for maintaining the backplane. The backplane is a commercially available VMEbus board whose pins and timing functions have been redefined for this application. The power supply accepts 110 volts AC input and produces DC voltages for each board, with the analog board performing extensive local regulation of noise-critical voltages.

A more detailed breakdown of the array control system is shown in the top half of Figure 3.11. For control of the NICMOS3 array we use a total of 4 analog read-out boards providing one video processor chain per array quadrant. Only one analog read-out board, quadrant one, is used for clocking the array and clocks all four quadrants in parallel. In this way the read-out time is optimized. Thus, for each pixel cycle, each of the four quadrants are clocked in parallel resulting in a burst of four digitized pixel values being shipped down the data fiber to the VME interface board. Upon arrival in the VME INF the four pixels are buffered in RAM until a block of 512 pixels has been collected. At this point a VME bus transfer (DMA write) is executed by the VME INF which places the buffered data into predefined UNIX memory. From there a software driver running on the Sparc1E adds the buffer to a growing image, and when all of the pixel values in the array have been transferred, the image is written to disk.

Figure 3.11 - Layout of the array control system (top) and typical read-out timing sequence (bottom) are shown.

Clocking of the array is controlled by the DSP timing board through an onboard program. This program operates as a waveform generator clocking a bank of DACs. Each DAC output contains diode protection circuitry to prevent accidental excursion out of the NICMOS3 TTL-like operating voltage window. We have found, as noted by the array manufacturer, that direct control of several operating biases can improve device performance by optimizing dynamic range, lowering the noise floor, and reducing noise. These bias voltages are provided by DACs and are actively controlled during various parts of the array read-out sequence.

To achieve optimal system performance we employ a multiple sampling technique that provides low read noise with minimal read-out overhead. The bottom half of Figure 6.2 shows a timing diagram for a typical exposure and read-out cycle. After a rapid reset of the entire array a brief settling period is allowed, then a non-sampled read-out is made, another brief settling delay occurs, then a sampled read-out is executed. This sampled read-out is shipped to and stored in the Sparc1E memory for later use. Next the shutter is opened and an exposure is made. At the end of the exposure time the array is again read-out, but not sampled, the array is allowed to settle, then a final read-out is sampled and transferred to the Sparc1E memory. Simple image processing is then performed whereby the first read out is subtracted from the second yielding a low noise frame (correlated double sample). Finally, this image is descrambled (the pixels are reordered into quadrants) and written to disk with an appropriate FITS header.

Digital Timing Board

The heart of the digital timing board is the Motorola DSP56001, a monolithic, integer digital signal processor with a 24-bit data word. It has a 16-bit address space, a fast ALU, extensive on-chip peripheral support and a Reduced Instruction Set Computer (RISC) architecture that executes most instructions in one clock cycle (100 nanosec in this design). There are separate address spaces for on-chip program and data memory, a synchronous serial interface, boot logic and a simple interface to an external data bus. The DSP56001 functions as a timing generator by writing 24-bit data words from its internal memory to its external bus every 100 nanosec, whereby circuitry on the timing board decodes the data word to support three separate timing functions. A delay function simply halts processor operation for intervals ranging from 50 nanosec to 12.5 microsec in 50 nanosec increments as a convenient way to implement settling delays or to set the integration time constant. The second timing function consists of digital control lines connected directly to the backplane that are updated in several different ways by the DSP. The third function is implemented by writing 16-bit data words to the backplane, from which each analog board decodes the data word for selectively updating its DACs. Four of these 16 bits, plus an additional control signal, select one of the 20 analog boards, while four more bits select which DAC on the board is to be written to with the remaining 8 bits. One of these 32 selections is intercepted by the timing board for updating the digital control lines that are connected directly to the backplane.

On power-up or reset the DSP program is read in from a boot ROM, which is a single byte-wide socketed device for easy re-programming external to the controller. Programming is done on a PC- or SUN-based cross compiler supplied by Motorola in native DSP56001 assembly language. The supplied program consists of initialization code to configure the DSP in the desired mode, a command processor, testing and diagnostic routines, routines to read from and write to internal DSP memory over the fiber optic link, tables containing read-out parameters and timing waveforms, and array read-out code. A description of the commands is supplied in a separate document. Modifications to the code can be done either by re-programming the ROM or by modifying the DSP contents over the fiber optic link after the ROM program is booted. The ROM is an erasable electrically programmable part (EEPROM) that can be reprogrammed from the DSP via the serial link as well, though a special security jumper and password system is implemented to protect against unauthorized intrusion. Furthermore, a backup copy of the DSP code is resident in ROM to ensure that changes to the ROM code are not made that would render it permanently unable to be booted. Rapid, reproducible and non-intrusive changes to the array clocking voltages for optimization of the array device operation can be made while the array is operating by exercising the write DSP memory command.

Support for the backplane is provided by high current drivers and a careful timing design to ensure reliable operation when many read-out boards are installed. Since the DSP also reads the pixel data from the A/D converters on each analog board over the backplane, receivers are also included. The DSP then writes the data to the fast serial transmitter to be received by the instrument control computer. Both the transmitted and received data words consist of a high start bit followed by 24 data bits, high true, with the most significant bits transmitted first.

Analog Read-Out Board

The video signal from the array's output drain is connected to a load resistor to make the array FET acts as a source follower. The signal is DC coupled to a classical JPL-designed preamplifier with a low noise dual FET first stage, a dual transistor for canceling out the Miller effect in the FETs, and an op amp for gain. A clamp circuit follows to keep the input signal close to an average of zero volts so the following amplifiers are well within their optimum operating ranges. Gain selection of 1× or 4× allows operation at high light levels with some compromise in read-out noise due to finite A/D converter resolution or low light levels with no such compromise. A polarity switch implemented with a low resistance JFET analog switch alternates the next amplifier between inverting and non-inverting operation so the resettable integrator will integrate up on the baseline pedestal immediately after the array is reset, and down on the video signal after charge is coupled onto the output node of the array. This is a classical dual-slope integrator and is the optimum signal processing algorithm for array signals that are dominated by white noise in the relevant passband. A buffer stage after the integrator allows the zero signal level of the array to be set close to zero digital counts, and is set with a programmable DAC. This stage drives the 16-bit Crystal Semiconductor A/D converter, which has a 10 microsec combined sample/hold and conversion time. The output from the converter is in serial form, and is converted to parallel form by two shift registers before being placed on the backplane for reading by the DSP and transmission over the fast optical fiber link to the computer.

Clocking signals for the array transfer and reset transistor gate are generated by a bank of twelve 8-bit DACs that were chosen for their speed and low glitch energy. They provide a voltage resolution of about 100 millivolts, and an output over the range of +10 to -10 volts. The DAC output is buffered by a fast op amp that can drive large capacitive loads. A set of 12-bit DACs generates the DC bias voltages - seven of them for the array and one for the offset adjustment of the video processor. Three of these voltages are unipolar high voltages of 0 to +30 volts, while the remaining four are bipolar over the range of +10 to -10 volts. Twelve bit DACs were chosen to provide greater long-term stability than 8-bit DACs, but are only settable to 8 bits, as their four least significant bits are grounded. Long term voltage stability to better than 5 millivolts is achieved. Additional circuitry on each board provides an interface to the backplane and decodes the five board select signals (D12-D15, A0), and the four DAC select signals (D8- D11). One of these 16 codes selects a programming sequence for the 12-bit DC bias supply DACs, which is a two step programming process. Regulators are placed liberally throughout the board to minimize coupled noise and minimize switching glitches, and are located on the power supply input of every DAC, on the supply lines to the A/D converter, and on the ±15 volt supply lines to the video processor.

Note that no potentiometers are used anywhere in the controller, as all adjustable voltages are set digitally by the DSP. The analog board is implemented on a six-layer printed circuit card with careful isolation between digital grounds, the noisy analog and digital ground surrounding the clock drivers and logic circuitry, and the quiet analog grounds in the video processor. Ground planes are placed liberally throughout the circuit, and a careful physical placement of components isolates these circuits as well.

Backplane and Power Supply

The backplane is simply implemented as a VMEbus J1/P1 backplane whose pins and timing have been completely redefined for this application. 96-pin DIN connectors provide a plentiful number of reliable pins, while the multilayer backplane provides good power distribution and noise suppression. DC power is distributed to the boards through the backplane, using the +5 connection for +5 volts, whereas the ±12 volt connection is powered with ±15 volt supply that is then down-regulated to ±12 volts on each analog board. The high voltage supply, nominally +30 volts, that is required for operating the drain of the array on-chip amplifier is brought onto the analog boards separately, to insure low noise operation, and is heavily filtered as well. The boards are all of the standard VME 3U width, that is, roughly four inches wide, but longer than the VME standard by about 50%. The number of VMEbus boards used in a particular installation is determined by the number of analog read-out boards that are required; up to 21 connector VME backplanes are available, allowing 20 read-outs if no other support boards are required. It is expected in typical installations that the backplane and controller boards will be mounted as close as practical to the array, which is typically on the side of the cryogenic dewar containing the arrays. Robust grounding of the analog boards at the array connector end is required for low noise operation in multiple read-out configurations.

The backplane supports full 24-bit data words on both reading and writing. While the analog read-out board only read and writes 16 of these bits, other boards can be built to utilize the full 24-bit capability of the DSP56001. Four address lines, A00-A03, are carried on the backplane in order to address the A/D converter when transferring pixel image data to the DSP. This makes the backplane a D24:A04 system. It is not a bus in the normal sense since only the timing board can be a master, and no bus arbitration circuitry is needed. A complement of 22 timing signals generated by the timing board are also carried over the backplane, while only seven of these are currently used by the analog board. These additional signals could be used to operate such devices as a programmable array temperature controller, a shutter, diagnostic hardware, filter wheels and so on. Two interrupt input lines to the DSP are also available on the backplane, which can be used to implement a hardware timing circuit for overall exposure timing that would be independent of the host computer.

The power supply delivers +5, ±15 and +36 volts to the controller, the first three through the dedicated power distribution pins on the VME backplane. Normally it is expected that the power supply will be mounted remotely from the controller some distance away. Noise pickup along the power supply cables can be minimized by proper bypassing at the controller, while the extensive on-board regulation provided by the noise-critical analog read-out boards should minimize its effect.

Detailed Specifications


Leach, R W. "Interfacing Astronomical Instrumentation to VMEbus Computers," Pub. Astro. Soc. Pacific, 103 (664), pp. 587-596.

Toomey D. W., Shure M., Irwin E. M., Ressler M. E., "ProtoCAM - An Innovative IR Camera for Astronomy," SPIE Proceedings - Instrumentation in Astronomy VII, 1235, ed. D. L. Crawford, pp. 69-81, SPIE, Washington, 1990.