Source (Simulated Star Field and Transit Generator)
Camera (Optics, CCD and Controller)
Structure (Important for photometric Stability)
Thermal Control (Very important for photometric Stability)

To perform the end-to-end system demonstration and to incorporate all the confounding factors that represent the flight system operations, a carefully designed laboratory facility was built at NASA Ames Research Center. The photometry facility includes: a simulated star field with an approximate solar spectrum, fast optics to simulate the space borne telescope, a thinned back-illuminated CCD similar to those to be used on the spacecraft operating at 1 Mpix/sec/amp read rate and shutterless operation. The testbed facility is thermally and mechanically isolated. Each source of noise can be introduced in a controlled fashion and evaluated. Pointing noise or changing thermal conditions in the spacecraft can cause star-image motion at the millipixel level. These motions are imposed by piezoelectric devices that move the photometer. The facility is described below:

Schematic of the source

The Source simulates all of the important features of the real sky for this experiment. The Source provides a simulated star field that represents:

• the same flux as will be expected in the Kepler Mission (after accounting for the aperture of the optics) of 10⁹ counts/sec for m_v=12 star
• the same spectral color as solar-like stars
• the same range in brightnesses (9<m_v<14) for target stars
• the same density of stars for m_v<19
• a field covering a large fraction of the CCD
• several bright stars (m_v=4) as is to be expected in the Kepler Mission

Finally, the source also has:

• the ability to generate Earth-size transits in selected stars.

Lamp

The light source is made up of two Labsphere integrating spheres with a quartz-tungsten-halogen (QTH) lamp in the 15-cm sphere, connected to the 50-cm integrating sphere with an iris, diffuser and spectral filters in between. The spectral filters BG34, KG4 and an OCLI "hot mirror" (with a 750nm cutoff) selectively attenuate the light, especially in the red to provide a fairly close match to the solar spectrum (see the figure below). There is a slight deficiency in the blue, since the lamp is not nearly as hot as the Sun.

The reason for using two spheres is to thermally isolate the heat of the lamp from the star plate. Separate cold strapping and thermal electric coolers are attached to the small sphere to remove the heat from the experiment.

A closed-loop control to 0.1% of the lamp brightness is provided by the Oriel controller and power supply. The amplitude of these fluctuations is similar to what is expected from the Sun and similar stars. However, unlike the real sky where the fluctuations of the stars are uncorrelated, in this simulation, all the fluctuations are all the same due to the common lamp. Note though that this brightness variation is ten times greater than an Earth-size transit. The use of ensemble relative photometry reduces the effect of the lamp variations by about a factor of one hundred. Maintaining a constant lamp current is also important for maintaining the same color of the light source.

Spectral Response Curves

Star Plate

The star plate has a large number of holes of various sizes (used to perform time-variant relative photometry) and they are placed in many locations across the field-of-view to support the suite of tests described earlier. The plate is made of 50-micron thick stainless steel and opaque (transparency of less than one part in a million). The hole pattern was drilled with a laser beam by Lenox Laser, with some holes as small as 3 microns diameter (for the m_v=19 stars). There are 84 holes for the 9<m_v<14 target stars in the uncrowded region of the plate. These are used to isolate the effects of faint background stars, bright stars, smearing, etc. Some of these have very nearby stars as faint as m_v=19 to demonstrate that stars five magnitudes fainter than the target star are not a problem even when spacecraft jitter is simulated. There is a crowded portion of the plate with 1540 stars having the same star field density to m_v=19 as the actual Cygnus region to be viewed by the Kepler Mission. This region was used to demonstrate the ability to perform the high-precision relative photometry even in crowded fields.

The bright (m_v=4) stars were used to demonstrate the maintenance of the relative photometric precision in the presence of an occasional bright foreground star. These bright stars are produced using fiber optic bundles driven by external bright LEDs.

Star plate
Wiring for the 42 transit stars is in blue. The bright-star fiber optics are black.

An Image of Star-Plate Star Field
The image taken with the CCD shows the set of 84 isolated stars used to test various parameters as well as the vertical region of dense stars equivalent to the star density in the galactic plane. The regions labeled "Bias strip" and "Smear calculation strip" are used to make corrections to the image.

Transit Generation

Earth-size transits cause a very small change in brightness on the order of 8x10^-5 for several hours on individual stars. An innovative concept to simulate a transit was developed to produce this small change. A fine fixed wire was mounted across the star hole (see the figure below). To cause the required small change, a current is passed through the wire. The small resistivity of the wire causes it to heat to a few degrees above ambient. Due to the heating and the small coefficient of thermal expansion (CTE) of the wire, the wire temperature increases by about 5°C and the wire expands by about 10 nm to reduce the amount of light by the required small fraction. Transit wires were mounted across 42 of the star holes, but only about a dozen were used during any one test. The switching on and off of the transits and monitoring were controlled with a Macintosh computer running LabView software with National Instrument's digital I/O boards.

Microscope photograph of a transit wire
A transit wire across an m_v=9 star hole with an m_v=14 nearby background star in the upper left.

The Camera simulates all of the functions performed by the photometer on the Kepler Mission spacecraft.

Schematic of the Camera

The camera consists of the optics, CCD, CCD controller electronics, camera operation computer and software, and a CCD cooling system shown schematically above and in cross section below.

Cross sectional view of the Camera

The Optics
The optics are used to form an image of the sky. The light from the star plate is first collimated to make it parallel; similar to star light from a great distance. A Cooke triplet is used to focus the light onto the CCD. All of the optical surfaces, lenses and windows have been anti-reflection coated. As in the Kepler Mission the f-number is small, equal to 1.5. A central obscuration was added inside the triplet to simulate the central obscuration of the Schmidt system for the Kepler Mission. For the Testbed Facility a thermal compensation sleeve was incorporated so that the image remains at the same focus even if the optical components change temperature. As in the space mission, the image was defocused to spread the light over many pixels. The optimum defocused image has about 80% of the light in a five pixel diameter aperture.

The CCD
The key component in the demonstration was the CCD (charged coupled device) and its operation. For the purposes of the test a commercially available CCD was selected that meets the criteria for the Kepler Mission. The device chosen is manufactured by Marconi (formerly EEV) and is shown below.

The Marconi 42-80 CCD

The CCD has 2048x4096 pixels with the pixels being 13.5 microns square. The overall size is 27 by 54 mm (about 1 x 2 inches). In actual operation the pixels are binned on the CCD to 27 micron square. In effect it was used as a 1024x2048 device. The binning improves both the readout speed and the photometric precision.

The most significant feature of the device is that it is a "back-illuminated" device. Usually CCDs are illuminated through the polysilicone wires on the front of the CCD. The wires provide the electrical voltages to control and readout the device. But these wires absorb light and modulate the image, much like looking through a picket fence. To overcome this problem several additional manufacturing steps are utilized. A glass plate is bonded to the front; the device is turned over and the backside is thinned; then it is doped and anti-reflection coated. This provides a device with a much higher quantum efficiency and much less photometrically sensitive to image motion.

The CCD can also be readout very rapidly. Since there is not a shutter in the system, all images have some amount of smearing effectively adding to the system noise. We are currently reading out the CCD at 1 mega pixel per second per amplifier.

Test results discussed on the following page confirm that the end-to-end system performance meets the Kepler Mission requirement.

How do CCDs (charge coupled devices) work?

CCDs are at the heart of each "HandyCam" TV camera and special purpose CCDs are what we use for Kepler.

When light strikes a piece of silicon, it produces electrons that are free to move about the silicon material. These electrons form a charge or a current which is measured to determine the amount of light that has fallen on to the silicon.

In a CCD, the silicon region is divided electrically into small individual picture elements or pixels with about four hundred elements per cm in each direction, like a very finely divided sheet of graph paper. The free electrons are kept from moving around by permanent channel stops (the vertical lines in the figure) and externally applied voltages (the horizontal lines in the figure). Each pixel can then be thought of as an individual bucket or well that collects electrons.

As shown in the animation, first the CCD is exposed to light from a telescope or camera lens. Overtime this produces an image made up of electrons in the CCD.

To readout an image that has been captured with the CCD requires shifting the information out of the pixels. First, the columns of pixels are all shifted down one row. The bottom row of pixels is shifted into a readout register. Each pixel in the readout register is shifted out to an amplifier and the number of electrons in each pixel is recorded. This produces a series of 1's and 0's that represent the image. This is repeated over and over until all the pixels have been read. The stream of 1's and 0's is then digitally processed to reproduce the image that is later displayed.

In the Kepler Mission the 1's and 0's are recorded onboard the spacecraft and sent to the ground, where the data are processed to look for changes in the brightness of each star that may be caused by a planetary transit.

Controller
The CCD requires control and data acquisition hardware and software to operate. The readout is performed by applying a sequence of clock signals to the CCD. On the CCD chip itself are preamplifiers to convert the charge read out from each pixel into a voltage. The sequence of voltages is then feed into the CCD controller.

The CCD controller was built by Dr. Robert Leach's CCD group at San Diego State University. The CCD controller is programmed to generate the clock signals for the readout of the CCD. It takes the analog voltages readout and digitizes them. These are then transferred to a Sparc 5 computer where the data are accumulated and written to disk. The software used to perform these operations is the Lowell Observatory Instrument System (LOIS) software (Taylor, et al., 2000).

The CCD is split down the middle (along the long dimension) and the controller has two channels to control and process each half of the CCD independently thereby doubling the overall readout speed.

CCD Operations
To improve readout speed and mitigate pixel saturation, the Kepler Mission uses relatively large pixels, on the order of 25 microns square. To achieve this with the 42-80 CCD which has 13.5 micron pixels, the CCD was binned 2x2 on the chip, that is, as the CCD was clocked, the charge from each 2x2 bin group was summed. Since only about half of the CCD was illuminated by the star field, only one half of the CCD was read out each time, providing an approximately square image. The binning and use of only half the CCD and use of two readout channels resulted in each channel handling one-half million pixels per 0.5 sec read. The clock speed is 1 megapixel/sec, so one image takes just one-half second to readout.

To prevent saturation of stars as bright as m_v=9, the CCD must be read out every 3 seconds. However, to achieve the required shot noise level to detect an Earth-size transit the signal must be integrated for 6.5 hours. This is accomplished in two steps; the 3 second CCD read outs are integrated for some length of time. (In the lab this is selectable. We have chosen values of 3 or 15 minutes. For the Kepler Mission this is 15 minutes.) In the ground processing software these integrations are co-added for longer times to match the transit duration being searched (2 to 16 hours).

To simulate the effects of cosmic ray hits on the CCD on the overall system noise, this additional noise factor was added to the data in software. Cosmic rays were optionally added into the 3 sec readout data and then an independent cosmic ray despiker algorithm was used to identify them and remove them before the data from each 3-second integration are co-added.

CCD Cooler
To reduce the dark current in the CCD, it was cooled to well be -40 °C. In previous lab setups we have used liquid nitrogen. Although this is a fairly simple method, it has the disadvantage of producing a varying mechanical load on the structure and thereby causing small structural deflections. For the Testbed Facility we chose a Cryotiger closed-cycle cooling system. By tying the cold strap from the Cryotiger cold finger to a rigid insulating isolator before connecting to the CCD we prevented any mechanical vibration from the refrigerator from causing the CCD to vibrate. The temperature of the CCD was controlled by a proportionally controlled heater attached to the CCD. The CCD can be operated below -60°C for many weeks at a time. The dewar has to be pumped out about once per month to maintain these low temperatures.

Mechanical Structure
The Testbed without the exterior walls is shown in a CAD drawing. The Labsphere (red sphere) and star plate (green disk) are shown in the bottom. The CCD is housed in the dewar (purple cylinder) at the top.

The mechanical structure for the Testbed Facility had to meet several important requirements. These are listed below and discussed.

• Mechanical stability between the "Source" and the "Camera":
  To provide the mechanical stability, the metering truss was made of Super-Invar, a material that has an extremely low coefficient of thermal expansion. In addition, the entire structure was surrounded by a thermal enclosure made out of 6mm thick aluminum to maintain a stable temperature (see Thermal Control below). The entire structure was supported by pneumatic vibration isolators.
  
• Simulated spacecraft motion:
  All inertially-pointed spacecraft experience some residual motion at some level, which is called jitter. For photometry, this can have an adverse effect if it is not carefully treated. Since the light from each star is measured with a fixed set of pixels, the amount of light falling on the pixels varies with the star image motion caused by the spacecraft motion. To simulate this effect in the Testbed Facility, piezoelectric transducers (PZT) were incorporated in the mounting of the Camera to the support structure. The PZTs are driven electronically to reproduce the random jitter of the spacecraft or also driven in a fixed pattern to map out the photometric effects of motion. The latter (known as dithering) could then be used to test methods for correcting for the motion during data analysis processing. The PZTs have a full range of motion of one-half of a pixel in each direction.
  
• Focusing capability:
  The Testbed Facility has the capability to adjust the focus. The adjustment is unlike many optical systems, such as a camera lens, where the optical elements are moved by some cam-type of mechanism. This would be a problem, since the mechanisms do not hold the optical elements sufficiently rigid for our purposes and the focus would tend to creep. Focus is changed in the Testbed Facility using commercially available "feeler" gauges used as shims with thickness changes of 25 microns (0.001 inches or 1 mil). The adjustable components are then clamped in place to hold the focus rigidly.
  
• Image rotation/translation capability:
  During flight operations the spacecraft/photometer must be rotated 90° to compensate for the annual apparent motion of the Sun around the spacecraft. This results in the star field rotating on the focal plane. The CCD layout was designed to be 90° rotationally symmetric so that the same set of stars can always be monitored. However, this does result in the star brightnesses being read from a different set of pixels. To demonstrate that the photometric measurements do not depend on the use of any one specific set of pixels, the Testbed Facility was built to allow for both translation of the CCD (by moving the dewar to another set of mounting holes) relative to the star field image and rotation of the star field (by turning the star plate).
  Rotation by a small amount (up to 7°) is used to demonstrate that the photometry is not affected by either having or not having multiple stars in the same CCD readout columns.
  
• Light tightness and dust free environment:
  To prevent stray light affecting the measurements, the outer enclosure (6mm thick aluminum) is opaque to room light. In addition there is a light tight baffle between the Source and the Camera to both block any outside stray light and to prevent light from the star plate scattering back into the imaging optics.
  The baffle also has a window (AR coated) to prevent dust from settling onto the star pin holes in the plate and causing unwanted variability in an individual star. In one case during assembly of the transit wires, dust did get onto one star hole with a wire. It is barely visible under a microscope, but is large enough to cause major variations in the brightness of this single star. Hence, this star has been eliminated from all test results.
  
• RF (radio frequency) shielding
  The truss structure is grounded to the room's instrument ground by a heavy copper strap. The camera dewar is grounded through the Cryotiger lines and isolated from the truss to avoid ground loops. The truss and camera are surrounded by the thermal enclosure (6mm aluminum) which is electrically isolated except for a single copper strap to instrument ground.

To achieve the required mechanical stability and mitigate unwanted motion of the star field across the CCD, the entire Testbed was surrounded by an actively controlled thermal enclosure. The entire enclosure is covered with 10cm of foam insulation. Additionally, the CCD, the photometer dewar and the CCD electronics controller have separate proportional temperature controllers with much tighter control than that of the overall enclosure.

• CCD and dewar thermal control:
  The CCD temperature is monitored by a platinum resistance thermometer which is read by a proportional control device that drives a heater that is in close thermal contact with the CCD. Temperature variations are limited to ±0.06°C. A similar system controls the temperature of the base of the dewar (enclosing the photometer) to ±0.06°C.
  
• CCD electronics thermal control:
  The aluminum case enclosing the CCD drive electronics has six thermoelectric units (TEUs) attached to it. Four are run continuously in the cooling mode. Two more are operated with a proportional control system similar to that used on the CCD except that they are used only to cool. The case temperature is controlled to within 0.2°C.
  
• Source thermal control:
  Heat straps from the 15cm Labsphere connect to a separate pair of TEUs to remove heat from the lamp. The 15cm Labsphere is thermally isolated from the 50cm Labsphere.
  
• Enclosure thermal control:
  Thirteen thermoelectric units (TEUs) attached to equal areas of the enclosure are operated in an on/off mode under control of the LabView software. They receive temperature inputs from resistance temperature detectors (RTDs) thermally attached to the enclosure plates. Each RTD is placed some distance away from the TEU whose cooling/heating is to be controlled. The aluminum plates temperature's are controlled to ±0.5°C.