A CCD ANTI-BLOOMING TECHNIQUE for USE in PHOTOMETRYA. William Neely NF/ Observatory Ltd. Rt. 15 Box 760 San Lorenzo, NM 88041 Electronic mail: email@example.com James R. Janesick Jet Propulsion Laboratory California Institute of Technology 4800 Oak Grove, Pasadena, CA 91109 FAX (818)393-0045
ABSTRACT.A new anti-blooming technique was used to collect images for a BL-lac monitoring project. The anti-blooming was accomplished by clocking two of the collection phases (in a three phase chip) back and forth during the integration of an image. Calibrations were done using VRI standard stars to insure that the technique did not affect normal photon collection and photometric data. New sources of noise were identified which must be subtracted from the processed image. In contrast, the technique reduced dark current noise, which is of benefit with a thermo- electric-cooled camera.
The technique eliminated blooming across the stars of interest, while preserving photon-collection efficiency. It also reduced the total noise of the exposures.
CCD cameras offer several advantages in photometry; including linearity, sensitivity, and a large dynamic range. However, bright sources may overwhelm the full-well capacity of the chip and bloom charge across the surface. This may make photometry in crowded fields difficult, if there are relatively bright stars near the object of interest.
Anti-blooming structures have been incorporated into CCD's. However, they are complicated and degrade quantum efficiency. Scientific CCD's rarely are equipped with these structures. A new technique of clocking two of the collecting phases was recently demonstrated (Janesick & Elliot 1992, Janesick 1992). The technique uses existing multi-phase CCD's. The software is changed to allow clocking of two phases (in our case, phase one and two) during the integration of an image. One of our observing projects involves monitoring PKS 0215+015, a BL-lac object with a resting 18.3 mag in V. The object is being monitored for a flare which will facilitate a spectral analysis of interposed hydrogen clouds with the Hubble Space Telescope. The object and field comparison stars are very close to SAO 110456, which shines at 5.58 in V. The SAO star causes severe blooming across the field during our normal 12 minute exposures (Figure 1). We decided to reconfigure the camera, and test the ability of the new anti-blooming technique to provide good photometric information and eliminate the interference of the SAO star.
NF/ Observatory uses an automated remotely operated telescope with a digital control and data link to Silver City, NM, 30 Km away (Neely 1991). The mirror is a 44.5cm F4.5 newtonian with the camera at prime focus. The camera is unguided (other than normal siderial tracking during exposures).
The CCD used in this study is the JPL/NASA CRAF/Cassini CCD. It is a 1024x1024 pixel, front-side illuminated device employing 12 micron pixels. It was originally developed for the CRAF/Cassini project (which has been downgraded to the Cassini Mission), for a visit to Saturn. A more recently developed sensor will now make the trip.
The device is an n-buried channel, three-phase CCD. The pixels are arranged in a square with a top and bottom read-out. Only the top readout is employed in our application. The center of the CCD serial register has an addition doped channel leading to improved charge transfer efficiency (Bredthauer 1991).
The CCD is driven by a IBM-486 which allowed easy manipulation of clocking patterns. The camera routines are written in assembly language. The system clock and memory refresh are disabled during sampling to insure accurate timing. The video signal is sent through a double-correlated sampler, then converted to 16-bit digital data. The full well of the CCD chip is 100,000 electrons. The floor noise of the on-chip amplifier is 5 electrons. The total noise for the signal chain is 27 electrons at -50 C. Dark current adds additional noise. During normal exposures, phase one is biased to full surface well (slightly above optimum full surface well). Phases two and three are inverted to act as a barrier. In an anti-blooming exposure, phases one and two are alternately used as the collection phase (Figure 2). Phase three remains inverted during the entire exposure. The clocking frequency most often used was 100 Hz. The serial register was not clocked during either type of exposure.
During the initial calibration run, it was noted that the 'dark current' was decreased compared to a normal exposure. Various clocking schemes were then employed to investigate the character of the change in noise. Clocking frequencies from .1 to 30,000 Hz were employed at various chip temperatures. All cooling was done thermo-electrically. All investigations of noise sources were done with 5 minute 'dark' exposures.
It was essential to be sure that clocking did not affect collection efficiency. The photometry runs were done on PG2213-006 (Landholt 1992) at a chip temperature of -50 C. Exposures were used which kept pixels in a non-saturated range. Differential magnitudes were used across the field of the CCD to eliminate any effects of extinction. All photometric reductions were done using PCVISTA (Treffers 1989).
Estimation of anti-blooming efficacy was done on several bright sources including The Orion Nebula, NGC 1976. Long exposures were done at different clocking frequencies.
The usual noise sources for the chip were affected by anti-blooming clocking. At 100 Hz clocking the 'dark current' appeared to be significantly reduced. This effect was then explored over several chip temperatures (Figure 3). This apparent 'beneficial' effect of clocking seemed to plateau at lower temperatures, suggesting another noise source.
To dissect out the 'other noise source', different clocking rates were tried. The data shows a significant increase in noise as the clocking frequency increases (Figure 4). This noise is known as spurious noise and is generated electrons being liberated in the clocking process (Janesick 1989). Holes are trapped in the Si-SiO2 interface. As the gates invert, the holes are released and electrons are liberated by impact ionization. The slope of the clocking pulse is important in the magnitude of this effect. Faster rise times means more acceleration and more noise. Subtracting the spurious noise from the total noise during 'dark' exposures separated out the true dark current. The dark current is significantly reduced at clocking frequencies above 10 Hz (Figure 5).
Different deep sky objects, exposures, and clocking frequencies were used to grade the efficacy of the anti-blooming. The chip resisted blooming at V 4.0 at a clocking frequency of 100 Hz (Figure 6). This is equal to a suppression of 3 X 105 electrons/pixel/second. At JPL, 5 X 106 electrons/pixel/second suppressions was achieved at a clocking frequency of 2000/sec. Simulation frames of a cluster, with and without anti-blooming are seen in Figure 7.
The results of the photometry of PG2213-006 are shown (Table I). The differential magnitudes were in good agreement with Landholt's values and were independent of whether the chip was clocked during exposures. The number of photons collected was also independent of the anti-blooming clocking (Table II). It is important to note that the images must remain unsaturated to maintain linearity. In trials of our BL-lac object, PKS 0215+015, the anti-blooming scheme worked well. Figure 1 shows good suppression of the blooming from the SAO star.
The noise contribution of the clocking scheme is a complex one. The dark current noise is pixel specific, and a template of noise must be subtracted from the image before flat-fielding. The spurious noise is more uniform and acts like shot-noise. Therefore, these two noise sources must be separated out and subtracted individually from the image.
The contribution of the two noise sources will vary, depending of the camera system. With our camera, the noise was lowest in the 10 to 100 Hz range (Figure 8). Since the anti-blooming effectiveness started to plateau at 100 Hz, this frequency is optimum for our system. Other cameras will optimize at different clocking frequencies. If the camera is cooled by liquid nitrogen, dark current will be insignificant. In MMP chips the dark current will also be lower overall, but some reduction in dark current will still be seen (Janesick 1989). In this case, the spurious noise, if significant, may greatly increase the overall noise of the camera. Spurious noise can be reduced by carefully adjusting the parallel clock waveform. In a thermo-electrically cooled device, the clocking decreases the actual dark current because it takes time for the 'thermal holes' to form. The 'holes' are then filled with the next clocking inversion. It is essential to determine the relative contributions of the noise sources for later reduction of images. The simplest way is to take dark images at different clocking frequencies. At very high frequencies, the spurious noise will predominate, which then can be extrapolated back to the working frequency.
It is also important to visually inspect the 'dark frame' and see if the field is uniform after the dark current is subtracted. Another potential noise source comes from luminescence. This occurs during clocking when long-wavelength photons can be produced. The problem is once-again reduced with a decrease in rise time of the clock voltage. Luminescence is non-uniform and will appear as 'Jim Dots'(Janesick & Elliot 1992). This effect was not seen in our camera system.
A new anti-blooming technique was effectively used in a photometry project. The photon collection and photometric data were not affected. The overall noise of the system was reduced. However, another noise source was introduced and had to be separately subtracted from the image to keep the standard deviation of the flat-field to a minimum. The anti-blooming effect was sufficient to eliminate blooming completely for stars V 4.0 and dimmer using a clocking frequency of 100/sec. More rapid clocking would allow anti-blooming of even brighter sources.
The CCD chip was supplied by JPL for field evaluation. This work was supported by a grant from NASA administered by American Astronomical Society. Thanks to Lori Neely for her photographic reproductions of the illustrations. Additional thanks to Larry Miller for his improvements on PCVISTA.
References:Bredthauer, R.A., Janesick, J.R., and Robinson, L.B. 1991 , in SPIE/SPSE's Electronic Imaging Science and Technology Conference, CCD & Solid State Optical Sensors II, San Jose, CA February 14, 1991, "Notch and Large Area CCD Imagers".
Bredthauer, R.A. 1991, SPIE Proc., Vol. 1447, p310.
Janesick, J.R., and Elliot, S.T. 1992, in Astronomical Society of the Pacific Conference Series, Vol 23, Astronomical CCD Imaging and Reduction, Ed. Steve B. Howell, D. Harold McNamara, Managing Ed. of Conference Series, Bookcrafters, Inc. p.1.
Janesick, J.R., 1992, Nasa Tech Briefs, Vol.16, No. 7, Item #71, pp 20-22.
Janesick, J.R., Collins, S., Blouke, M.M., and Freeman, J. 1989, Op. Eng., 26(8), 692-714.
Landholt, A.U. 1992, A. J., 104, pp.340-371 and pp.436-491.
Neely, A.W., and Treasure, F.A. 1989, Remote Access Automatic Telescopes, Fairborn Press, Mesa, Ariz., p141-150.
Treffers, R.R., and Richmond, M.W. 1989, Pub. A. S. P, 101, p725.
TABLE I Landolt Cluster Photometry PG 2213 UD 11/27/92 No-Bloom Clocking at 100/sec Predicted Magnitudes Measured Differential Magnitudes Normal No-Bloom V R V R V R A 14.178 13.772 B 12.706 12.279 C 15.109 14.683 A-B 1.472 1.493 1.47ñ.02 1.46ñ.02 1.45ñ.02 1.45ñ.02 C-B 2.403 2.404 2.33ñ.04 2.39ñ.04 2.37ñ.04 2.44ñ.04 TABLE II PHOTON COLLECTION for STAR A over 333 pixels IN DN (one DN = 3.2 electrons) Star Sky A - Sky V normal 160484 - 89628 = 70856 V no-bloom 164604 - 94044 = 70560 R normal 320148 - 198238= 121916 R no-bloom 338964 - 217832= 121132 TEXT FOR FIGURES (a) (b) Fig. 1 - 480 second exposures of BL-lac PKS 0215-015 (bl). a) Normal Exposure b) Exposure with anti-blooming at 100 Hz clocking rate The comparison stars for the project have now appeared from behind the blooming. The bright star is SAO 110456. Fig. 2 - Clocking Diagram for 3 phase CCD. The collected electrons are shifted between phase 1 and 2. Phase 3 remains low as a barrier. Fig. 3 - Average "dark current"(dark current plus other noise sources) vs temperature with and without anti-blooming clocking. Fig. 4 - Spurious noise production as a function of clocking rate at -50 C. Fig. 5 - Real dark current production vs clocking rate. After 10 Hz the electron production is sharply reduced. Fig. 6 - Minimum V Magnitude at which blooming can be suppressed with our optical and camera system as a function of clocking rate. (a) (b) Fig. 7 - JPL test exposures of a simulated crowded star field. a) Normal exposure. b) Exposure with anti-blooming clocking at 2000 Hz. Fig. 8 - Total noise production as a function of clocking rate. This include both dark current and spurious noise.