EBAPS®: Next Generation, Low Power, Digital Night Vision1
Transcription
EBAPS®: Next Generation, Low Power, Digital Night Vision1
Presented at the OPTRO 2005 International Symposium May 10, 2005, Paris, France. EBAPS®: Next Generation, Low Power, Digital Night Vision1 Verle W. Aebi, Kenneth A. Costello, Philip W. Arcuni, Patrick Genis, and Stephen J. Gustafson Intevac Corporation Santa Clara, CA USA 1. Abstract Intevac has developed a new Low Light Level Camera sensor technology for application to a variety of low light level imaging applications. The new sensor is an Electron Bombarded Active Pixel Sensor (EBAPS). EBAPS technology is based on use of a GaAs photocathode derived from Generation-III image intensifier technology in proximity focus with a high resolution, backside thinned, CMOS Active Pixel Sensor (APS) imager anode. The electrons emitted by the photocathode are directly injected in the electron bombarded mode into the CMOS APS anode. In this approach low noise gain is achieved in the CMOS anode via conversion of the high energy photoelectron (1 – 2 KeV resulting from the high voltage bias applied between the photocathode and CMOS anode) to electron-hole pairs in the anode via the Electron Bombarded Semiconductor (EBS) gain process. The electrons are collected in the APS pixel and subsequently read out. The EBS gain process is inherently low noise with an excess noise factor (Kf) of less than 1.1. This is substantially less than a microchannel plate based Generation-III image intensifier (MCP, Kf of 1.8) or the avalanche gain process in an Electron Multiplying CCD (EMCCD, Kf of 1.4). The low noise EBS gain process eliminates the need for an MCP and enables higher SNR at the lowest light levels. This offers the possibility of higher performance for an EBAPS based camera relative to a standard Image Intensified camera based on Gen-III tube technology using an MCP for gain or EMCCD based cameras. This low noise gain advantage has been combined with modern semiconductor packaging and manufacturing approaches to enable a small integrated EBAPS module which can be mass produced at low cost in an automated ultra high vacuum production packaging system. This new sensor manufacturing approach allows high volume, cost sensitive, markets to be addressed. It also enables a variety of sensor formats to be easily addressed as it allows combination of standard CMOS APS imaging chips with a GaAs photocathode in the EBAPS configuration. This allows customization of the EBAPS for a given camera application. EBAPS technology will be described with its application in a first generation EBAPS sensor and low light level camera (NightVista) developed for commercial security camera applications. The NightVista camera has a 1/2 inch optical format and a VGA (640 x 480) array with a 12µm pixel. The camera incorporates a gated high voltage power supply for automatic gain control. It also incorporates 2 point non-uniformity correction (NUC), bad pixel replacement, and histogram equalization image processing functions. The EBAPS sensor, high voltage power supply and camera electronics combined weight is 45 grams (not including camera housing). This is approximately 60% of the weight of a Generation-III image intensifier module as used in a standard night vision goggle. The EBAPS sensor and electronics are also ideally suited to head mounted system packaging and enable system designs with minimum forward projection relative to currently fielded night vision goggles. Results will also be presented for a next generation EBAPS camera based on the ISIE6 (Intevac Silicon Imaging Engine), SXGA (1280 x 1024 array, 6.7 µm pixel), EBAPS sensor with a 2/3 inch optical format. The ISIE6 EBAPS has lower readout noise than the NightVista EBAPS sensor for improved low light level performance and supports a 27.5 fps readout rate. Finally performance modeling will presented on a larger 1 inch optical format, SXGA, ISIE10 EBAPS sensor under development for an EBAPS camera targeted for future high performance head mounted night vision applications. 1 It should be noted that the U.S. Government makes no official commitment nor obligation to provide any additional information or an agreement of sale on any of the systems/capabilities described in this paper. 1 Presented at the OPTRO 2005 International Symposium May 10, 2005, Paris, France. 1. Introduction Low light level cameras have a number of significant, dual use applications. These include traditional military head mounted night vision and commercial applications including surveillance, medical, and scientific applications. Modern night vision systems are rapidly transforming from the presently used direct view systems to camera based systems as evidenced in a number of US military programs such as Future Force Warrior or Digital Enhanced Night Vision Goggle (DENVG). These systems are driven by advances in video display and processing. Video based systems allow image processing including fusion with other imagery such as from a FLIR sensor in addition to image transmission for remote display and image recording or local display of imagery from a weapons mounted sensor or imagery from a remote sensor. Surveillance applications are predominately video based where camera cost, size and performance are often critical. Scientific applications require cameras with good photon sensitivity over a large spectral range and high frame rates. These applications, and others, are driving the need for improved low light level sensors with direct digital video output. Two technology approaches are used today for high performance low light level video cameras. The first is based on a Generation-III (GaAs photocathode) or Generation-II image intensifier fiber optically coupled t o silicon imager (either CCD or CMOS) to form an Image Intensified (I2) camera. The second uses an Electron Multiplying CCD (EM-CCD) as the low light level sensor. Both technologies have different technology advantages and disadvantages relative to a camera based on the EBAPS sensor. The traditional I2 camera utilizes an image intensifier originally optimized for direct view night vision military system applications. In this approach the scene to be imaged is focused by the input lens onto the photocathode faceplate assembly. The light energy liberates photoelectrons from the photocathode to form an electron image. The electron image is proximity focused onto the input of the microchannel plate (MCP) electron multiplier, which intensifies the electron image by secondary multiplication while maintaining the geometric integrity of the image. The intensified electron image is proximity focused onto a phosphor screen, which converts the electron image back to a visible image. A fiber optic then transfers this visual image to a standard CCD or CMOS image sensor, which converts the light image into electrons t o form a video signal. In these existing I2 cameras, there are four interfaces at which the image is sampled, and each interface degrades the performance of the camera. In addition the camera is composed of a high cost image intensifier which is composed of a number of custom, high cost, parts (MCPs, fiber optic, and vacuum assembly, in addition to the photocathode). Finally size and weight are not optimized relative to a modern focal plane based camera. Recently EM-CCD devices have been developed and commercialized for low light level imaging applications.1, 2 In these devices a low noise avalanche gain process is employed in the charge domain in a modified readout register of the CCD. This gain mechanism increases the signal sufficiently to effectively mitigate the noise effects of the on-chip output amplifier. This enables high sensitivity, low light level, operation at video readout rates. The EM-CCD requires high clock voltages (on the order of 20 to 35 volts) on some of the register gates to achieve charge multiplication gain. Dark current reduction in the CCD by cooling is also required to achieve good low light level sensitivity. The cost, size, weight, and performance disadvantages of the I2 and EM-CCD low light level cameras has been addressed by development of Electron Bombarded (EB) silicon imager technology.3, 4 In this technology photoelectrons from a photocathode are accelerated to and imaged in a silicon imager anode (CCD or CMOS imager) directly (Figure 1.1). Gain is achieved by electron multiplication resulting when the high velocity electron beam dissipates its energy in the silicon of the imager chip to produce electron-hole pairs by the electron-bombarded semiconductor gain process. The EB gain is high enough to mitigate the noise effects of the on-chip amplifier and other camera electronics as is the case with the EM-CCD. Noise is generated in all the elements in the multiplier chain of the conventional I2 camera, particularly in the MCP where the electron multiplication statistics result in an excess noise factor on the order of 1.8 for a 2 Presented at the OPTRO 2005 International Symposium May 10, 2005, Paris, France. modern Generation-III image intensifier and 1.4 for a Generation-II image intensifier. EM-CCD cameras have an excess noise factor determined by the electron multiplication statistics of the avalanche gain process in silicon. This is on the order of 1.4 for a well designed electron multiplication gain register.1 In contrast, very little noise is generated by the Electron Bombarded gain process in an EB silicon imager, where the electron multiplication occurs by electron-hole pair formation as the accelerated electron beam travels through the silicon. The EB gain process is essentially deterministic with a resulting excess noise factor near 1.035 , substantially less than for EM-CCD or I2 cameras. The EB silicon imager eliminates the MCP, phosphor screen, and fiber optics, and does not require focal plane cooling to reduce dark current due to the low emitted dark current from the photocathode at ambient temperature and as a result both improved image quality and increased sensitivity can be obtained in a smaller sized camera relative to an I2 or EM-CCD camera.6 The initial approach selected for the EB silicon imager was based on a backside thinned CCD mounted behind the photocathode in a standard 18mm diameter image intensifier tube package, but with the phosphor output screen replaced with the packaged CCD imager. The EBCCD effort demonstrated the viability of the integrated EB imager concept and reduced camera size and weight relative to CCD based I2 cameras through the elimination of the fiberoptic output and its associated optical bonding issues, but the costs were not sufficiently reduced to meet high volume military and commercial market requirements. Further size, weight, and power reductions were also required to meet the needs of head mounted night vision applications. Large format CCD camera electronics consume several watts of power due to the CCD clocking requirements making them impractical for battery operated applications and require external electronics for a complete camera. The size of the external camera electronics presents an obstacle to applications that would benefit from miniaturization of the camera. EM-CCD cameras consume significant additional power due to the need to cool the focal plane, required to reduce dark current noise. Finally CCDs require specialized semiconductor processing lines that are not compatible with mainstream CMOS semiconductor fabrication technology. This substantially complicates application of the silicon foundry model so successfully used in the CMOS industry to the fabrication of specialized CCDs for Low Light Level cameras further increasing the cost of EM-CCD and EBCCD cameras. These disadvantages have been addressed by moving to new CMOS imagers or Active Pixel Sensors (APS) as a replacement for the CCD and re-designing the vacuum package to take full advantage of packaging advances in the semiconductor industry, resulting in the Electron Bombarded Active Pixel Sensor (EBAPS®) concept. This approach has addressed the performance, size, power, and cost disadvantages of present Low Light Level cameras. FIGURE 1.1 EB Silicon Imager SILICON IMAGER PHOTOELECTRONS CONTROL SIGNALS VIDEO OUT VACUUM ENVELOPE PHOTOCATHODE IMAGE PHOTONS 2. EBAPS® Design The recent development of high performance CMOS imagers enables the EBAPS sensor and allows it t o address some of the key deficiencies in previous low light level cameras. This includes substantial reduction in electronics size and weight due to the ability to integrate much of the camera electronics on-chip with 3 Presented at the OPTRO 2005 International Symposium May 10, 2005, Paris, France. CMOS technology. Power is substantially reduced in a CMOS imager based camera with an order of magnitude reduction possible relative to EM-CCD and EBCCD cameras. Overall camera size is reduced relative to an I2 camera by application of industry standard semiconductor sensor packaging approaches that reduce sensor size significantly relative to a military image intensifier tube. Ultimate performance of an EBAPS sensor will be determined to a large extent by the CMOS imager architecture and design. First, it is essential that the CMOS imaging area have 100% fill factor (no dead area). Any reduction in active area will result in lost photoelectrons. This is equivalent to a reduction in photocathode quantum efficiency or sensitivity. At the lowest light levels (starlight or overcast starlight), low light level camera performance is dictated by photon statistics. It is essential for the imager to detect the maximum number of photons for adequate low light level resolution and performance. Second, the CMOS imager architecture must maximize integration of the image photons with close to 100% duty cycle. This requirement when combined with high fill factor enables the collected signal to be maximized for good low light level performance. Essentially 100% fill factor can be achieved in a CMOS imager with a properly designed backside illuminated design. A backside thinned format enables 100% fill factor for an arbitrary pixel size for back illuminated operation. A standard CMOS imager cannot be used in a frontside illuminated electron bombarded mode since the metal and dielectric stack (typically 4 to 5 microns thick for a modern sub-micron CMOS process) will block the electrons from reaching the silicon at moderate acceleration voltages (2 kV typical). In addition fill factor would be restricted to the optical fill factor of the CMOS imager which is typically less than 50%. In a backside thinned CMOS imager the chip is flip-chip bonded onto a carrier substrate and the silicon substrate is removed by mechanical and chemical thinning. The free silicon surface is then passivated to reduce carrier recombination at the surface. A properly designed pixel will allow a majority of the generated charge to be collected by the photodiode in the pixel regardless of photoelectron impact position in the pixel. This enables high single photoelectron signal-to-noise (SNR) to be obtained with 2 kV electron energy EB gains on the order of 200. SNRs above one are achieved for a single photoelectron if the on-chip electronics noise (pixel referenced) is less than the EB gain. A rolling shutter approach is implemented in the CMOS image chip to enable integration of the image signal at close to a 100% duty cycle. The above approach of a properly designed backside thinned CMOS imager with a rolling shutter architecture enables essentially full utilization of the available signal from the photocathode. This is essential for high performance low light level imaging where the ultimate performance is determined by the photocathode quantum efficiency and the signal limited shot noise. The other critical requirement for a low light level camera is high dynamic range to accommodate the intrascene dynamic range of a nighttime scene with lighting (on the order of 105 or 106 ) as is often the case in urban environments. CMOS imagers with extended dynamic range capabilities are common today.7, 8 In addition the anti-blooming structures used in CMOS imagers are effective and do not impact fill factor. The programmable extended dynamic range capability of CMOS imagers is not available in CCD sensors which are inherently linear devices. The dynamic range of I2 cameras is set by the MCP and is not programmable. This level of dynamic range will result in better intra-scene performance than that obtained with either an EM-CCD, EBCCD or an I2 camera. This will result operationally in the capability to better observe scene detail in dark areas of scenes which contain light sources. Low voltage operation is also important to ease power supply requirements for a gated power supply. Gating can be used to reduce duty cycle for exposure control in high light level conditions. Gating is controlled by the camera automatic gain control (AGC) algorithm. Minimization of the total voltage swing enables a smaller, more power efficient, power supply. This is important for low power, miniature camera design. Other design requirements for the EBAPS are determined by requirements for high performance, reliability, minimum size, and low cost. High performance is achieved through use of a high quantum efficiency GaAs photocathode. GaAs has good sensitivity in the near IR (600nm – 900nm) region where a higher photon flux is available at night than in the visible region of the spectrum. Reliability considerations are driven by 4 Presented at the OPTRO 2005 International Symposium May 10, 2005, Paris, France. the GaAs photocathode requirement for an ultra-high vacuum (UHV) environment for photocathode stability. A critical aspect of this program has been selection of UHV compatible materials and the development and demonstration of cleaning procedures and processing techniques that allow good photocathode life to be achieved while maintaining acceptable CMOS imager performance. Low cost has been demonstrated through a sensor development process that has considered manufacturing cost throughout the product development cycle. Part count has been minimized in the EBAPS with the device consisting essentially of a packaged CMOS imager with an input window incorporating a photocathode as shown in Figure 1.1. This has been achieved by adopting a proximity focused sensor design using standard semiconductor packaging approaches for the sensor. This has enabled the overall package size to be minimized and has also enabled adoption of automated manufacturing approaches. Video based head mounted night vision system requirements also drive sensor design choices for an optimum CMOS chip for this application. Overall constraints are low light level performance optimization, system size, weight and power consumption. System angular resolution versus light level is a key system performance metric that should be maximized with the goal of equal or better performance relative t o presently deployed direct view night vision goggles. Ultimately angular resolution is determined by sensor pixel format and system field of view. Today the accepted field of view for a head mounted system is 40° horizontal. Pixel format is limited by presently available microdisplays suitable for head mounted applications. Today SVGA (800 x 600 pixel) format displays are available with SXGA (1280 x 1024) displays now reaching the market. There is no near term prospect for larger format microdisplays in the near term (next 2-3 years). The field of view requirement when combined with the SXGA format fundamentally limits ultimate system resolution, regardless of light level. This limit is 0.92 cycles per milliradian. Today typical direct view goggles with a 40° circular field of view have limiting high light level resolution of >1 cycle per milliradian. At low light level (1/4 moon illumination and below) system resolution as measured by standard tasks such as maximum range for man recognition begins to be limited by the SNR of the system, not pixel count. System SNR is a function of the lens f/#, lens transmission, and effective focal length.9 For a fixed system field of view and lens f/#, system SNR for a given scene object subtense is proportional to the product of the lens clear aperture diameter and the square root of its transmission.9 Typically the lowest manufacturable f/# lens for night vision applications is on the order of f/1.2. Lens aperture diameter will thus be directly proportional to focal plane size with the above assumptions and low light level performance will increase with focal plane size. The optimum CMOS format for best low light level performance is the largest size determined allowed by the application and other system constraints. Size constraints have limited head mounted image intensifier tube format to 18mm diameter tubes. Larger format 25mm image intensifiers have not been used for head mounted applications, but have been used for drivers viewers on vehicles or weapon sights. An 18 mm diagonal format has been chosen as the optimum CMOS focal plane size for head mounted applications as this is consistent with presently fielded head mounted night vision systems. This requirement combined with the SXGA pixel array format determines an optimum pixel size of 10 – 11µm for head mounted night vision applications. 3. EBAPS Sensor and Camera Family The requirements stated above have been used as design architecture guidelines for the EBAPS sensors and cameras developed at Intevac. Two generations of EBAPS sensors have been developed with the third generation EBAPS in development. The first generation EBAPS, NightVista, is based on a 1/2 inch image format, VGA (640 x 480, 12µm pixel), CMOS imager. The NightVista CMOS chip has an integrated high performance analog signal processor comprised of a programmable gain amplifier (PGA), a high speed 10 bit A/D converter, and fixed pattern noise elimination circuits. The second generation ISIE6 camera has three key improvements over the NightVista: first the optical format has been increased to 2/3 inch; the focal plane is based on a SXGA (1280 x 1024, 6.7µm pixel) CMOS imager; and the read noise has been substantially reduced. The combination of these improvements results in substantially improved performance at all light levels. The ISIE10 camera further optimizes low light level performance relative t o 5 Presented at the OPTRO 2005 International Symposium May 10, 2005, Paris, France. the ISIE6 for head mounted night vision with an increased sensor size (1 inch optical format) obtained by enlarging the pixel size to 10.8µm with further incremental reduction in read noise and increased frame rate. The key CMOS imager chip specifications of the NightVista, ISIE6, and ISIE10 EBAPS are summarized in Table 3.1. Common characteristics of the EBAPS sensors which are determined by the photocathode are given in Table 3.2. TABLE 3.1 NightVista, ISIE6, and ISIE10 key specifications Format Pixel Size Optical Format Frame Rate Video Output NightVista VGA 640 x 480 12.0µm x 12.0µm 1/2" (9.8mm diagonal) 30 frames per second RS-170 or interlaced digital video ISIE6 SXGA 1280 x 1024 6.7µm x 6.7µm 2/3” (11mm diagonal) 27.5 frames per second 10 bit Digital Output, progressive scan ISIE10 SXGA 1280 x 1024 10.8µm x 10.8µm 1” (17.7mm diagonal) Up to 37 frames per second 10 bit Digital Output, progressive scan TABLE 3.2 Common EBAPS Sensor Characteristics Photocathode High Voltage Power Supply 24 Hour Capability GaAs (500nm – 900nm Band) Gated for Dynamic Range Control Daytime imaging with High Voltage off The proximity focused sensor design combined with semiconductor style packaging results in small form factor for the sensor. A photograph of the three EBAPS sensors developed by Intevac is shown in Figure 3.2. NightVista ISIE6 ISIE10 FIGURE 3.2 GaAs EBAPS®: NightVista, ISIE6, and ISIE10 4. EBAPS Performance An EBAPS based camera has some significant performance differences relative to a standard I2 camera. In particular since the sensor does not utilize a microchannel plate it can be operated in a day only mode with no high voltage applied to the sensor. This is a result of the longer cutoff wavelength of silicon relative t o GaAs. The GaAs photocathode acts as a long pass filter in front of the backside thinned CMOS image chip. The GaAs photocathode begins transmitting light at 750nm with close to 100% transmission for wavelengths longer than 900nm. Silicon has some sensitivity out to 1100nm wavelength. Thus the CMOS image sensor directly detects photons in the 750nm to 1100nm wavelength band in an un-intensified mode of operation. Typical spectral response curves of the intensified night mode of operation with the GaAs 6 Presented at the OPTRO 2005 International Symposium May 10, 2005, Paris, France. photocathode and high voltage applied is shown in Figure 4.1 (blue curve). The un-intensified day mode of operation with high voltage off with direct photon detection by the Silicon CMOS imager is also shown (red curve). This mode of operation enables high resolution near IR imagery to be obtained in the day with no impact on sensor operational life. When operating in day mode the degradation in image quality resulting from the proximity focused electron optics is no longer present and the resolution is much improved with MTF similar to a standard CMOS image sensor. AGC in the day mode sensor is obtained through integration time control on the CMOS chip. Figure 4.2 is an example of typical day mode imagery obtained with the NightVista camera. For comparison purposes Figure 4.3 is an example of night mode imagery obtained by stopping down the lens used for the image obtained in Figure 4.2 at essentially the same time and illumination conditions. The variation in image contrast is a result of the different spectral sensitivity bands for the two images. FIGURE 4.1 GaAs EBAPS® Spectral Response FIGURE 4.2 NightVista Day Mode Imagery FIGURE 4.3 NightVista Night Mode Imagery Limiting resolution versus light level measurements have been performed in the laboratory for both the NightVista and ISIE6 cameras. In these tests an Optoliner with a calibrated 2856°K light source was used with limiting resolution measured using a backlit, 100% contrast, 1951 Air Force resolution target. The resolution target is read by observing a computer monitor adjusted for optimum brightness and contrast. The results are shown in Figure 4.4. As expected the ISIE6 camera has substantially higher resolution at all light levels. This is a result of the smaller pixel size (6.7µm versus 12 µm) and the lower read noise of the ISIE6 chip. 7 Presented at the OPTRO 2005 International Symposium May 10, 2005, Paris, France. FIGURE 4.4 Limiting resolution vs. faceplate illumination FIGURE 4.5 ISIE6 camera low light level image Night time imagery has been captured with the ISIE6 camera. A night time image of two men in Battle Dress Uniform (BDU) against a green grass background captured with the ISIE6 camera is shown in Figure 4.5. The SXGA format and lower read noise result in substantially higher resolution and image quality than can be obtained with the VGA format NightVista camera. Range performance modeling has been conducted on the EBAPS camera family under development at Intevac. Modeling was performed using the NVESD II2CCD camera model using input parameters given in Table 3.1. The system parameters were held constant with a 40° horizontal FOV using a f/1.2 lens. The task modeled was for recognition of a man dressed in Battledress Uniform (BDU) against a green grass or foliage background. The model results are shown in Figure 4.6. A substantial increase in recognition range is obtained for each camera generation. FIGURE 4.6 EBAPS camera family relative recognition range performance The NightVista and ISIE6 laboratory and field test results support the model predictions. The increase in recognition range from NightVista to ISIE6 to ISIE10 is a result of a decrease in CMOS read noise with EBAPS generation, an increase in focal plane size allowing increased light gathering capability, and an increase in format from VGA for NightVista to SXGA for ISIE6 and ISIE10. The relatively slow drop off in 8 Presented at the OPTRO 2005 International Symposium May 10, 2005, Paris, France. range performance versus light level for all three cameras is one of the advantages of a camera based, indirect view, system versus a standard direct view night vision goggle. This is a result of separating display brightness from the low light level sensor function. In the camera based system using a microdisplay the display brightness can be separately optimized for best eye performance, removing the falloff in eye performance with reduced display brightness as occurs in a direct view goggle at starlight and below illumination levels. 5. EBAPS Product Features The EBAPS camera family offers substantially smaller size and weight than presently available low light level cameras. The commercial NightVista and ISIE6 cameras are shown in Figures 5.1 and 5.2. The NightVista camera EBAPS sensor, electronics, and high voltage power supply weigh 45 grams (not including case and other mechanical mounting components). ISIE6 and ISIE10 cameras will be only slightly higher in mass due to the increased sensor size. A standard 18mm format image intensifier tube in contrast weighs in excess of 80 grams. The addition of CMOS image sensor and electronics along with a fiber optic taper for optical coupling of the tube output to the CMOS image sensor would increase overall weight of an I2 camera to the 150 gram range. The other critical advantage of EBAPS for HMD applications is the low sensor profile. This enables a reduction in forward system projection in excess of 3 cm for a helmet mounted EBAPS camera relative to a standard goggle or I2 camera solution. This reduced forward projection improves ergonomics of a head mounted system by improving the center of gravity and reducing risk of entanglement of the system in an operational environment with branches and other obstacles. FIGURE 5.1 NightVista camera FIGURE 5.1 ISIE6 commercial camera Camera power has also been addressed with the FPGA based NightVista camera consuming 1.1W for the RS170 video configuration. The ISIE6 camera consumes 1.8W while processing 4X the number of pixels per second. FPGA based camera designs for the ISIE6 and ISIE10 cameras are targeting 1.2W power consumption upon completion of on-going power reduction activities. An ASIC based ISIE10 camera would consume <1W of power. Less than 1W of power consumption is targeted for battery operated, head mounted, applications. The EBAPS cameras all have a number of common features that improve overall system performance. These include sophisticated AGC algorithms that control camera exposure through the miniature, gated, high voltage power supply. The AGC algorithm allows the user to select either the entire frame or a user selectable window for exposure control. Average brightness in the window along with the percentage of allowed saturated pixels is also selectable. Non-Uniformity Correction (NUC) to remove fixed pattern noise is performed on a pixel-by-pixel basis using a standard two point correction algorithm. In contrast t o thermal imagers the gain and offset parameters required for NUC correction in an EBAPS are stable over time and can be stored in the camera at the factory. The cameras also perform bad pixel correction using 9 Presented at the OPTRO 2005 International Symposium May 10, 2005, Paris, France. standard approaches. An important camera image processing function is the histogram equalization algorithm. Histogram equalization is critical to optimize display of the imagery on a standard 8 bit display for viewing and to obtain the performance benefits of an indirect view, camera based, night vision system relative to a direct view goggle. It does this by optimizing mapping of the image data to the 8 bit display. The image processing functions in the EBAPS cameras are all performed with only a few lines of latency as required for head mounted applications. 6. Conclusion Planned future military night vision equipment will be video based using a head mounted low light level camera coupled with a microdisplay. Commercial products with significant potential are also primarily camera based for 24-hour security monitoring applications, scientific applications, and medical applications. This program has demonstrated a family of EBAPS based cameras which meet the requirements for both commercial applications where cost and performance are critical and next generation video based head mounted night vision applications. Future work at Intevac will include completion of the ISIE10 EBAPS sensor and camera later this year along with camera power reduction activities. The ISIE10 EBAPS will have optimum performance for camera based head mounted night vision applications. Expected performance for the ISIE10 EBAPS is comparable to presently fielded Gen-III night vision goggle systems. Other work underway includes development of prototype EBAPS based HMD systems which will leverage the packaging advantages of the EBAPS camera technology. 7. Acknowledgements This work has been supported by Intevac with partial support by the US Army. I would also like t o acknowledge the continuing support of the US Army RDECOM and CERDEC and Night Vision and Electronic Sensors Directorate and the many useful discussions with them during the course of this work. 8. References 1 2 3 4 5 6 7 8 9 M. S. Robbins and B. J. Hadwen, “The Noise Performance of Electron Multiplying Charge-Coupled Devices,” IEEE Trans. Electron Devices, Vol. 50, pp. 1227-1232, May, 2003. J. Hynecek and T. Nishiwaki, “Excess Noise and Other Important Characteristics of Low Light Level Imaging Using Charge Multiplying CCDs,” IEEE Trans. Electron Devices, Vol. 50, pp. 239-245, Jan., 2003. V. W. Aebi, K. A. Costello, J. P. Edgecumbe, J. J. Boyle, W. L. Robbins, R. Bell, D. Burt, A. Harris, I. Palmer, and P. Pool, “Gallium Arsenide Electron Bombarded CCD Technology,” SPIE Vol. 3434, pp. 37-44, 1998. M. Suyama, A. Kageyama, I. Mizuno, K. Kinoshita, M. Muramatsu, K. Yamamoto, “An electron bombardment CCD tube”, SPIE Vol. 3173, pp. 422-429, 1997. R. A. LaRue, K. A. Costello, G. A. Davis, J. P. Edgecumbe, and V. W. Aebi, “Photon Counting III-V Hybrid Photomultipliers Using Transmission Mode Photocathodes”, IEEE Trans. Electron Devices, Vol. 44, pp. 672 – 678, 1997. G. M. Williams Jr., A. L. Reinheimer, V. W. Aebi, and K. A. Costello, “Electron-bombarded back-illuminated CCD sensors for low light level imaging applications”, SPIE Vol. 2415, pp. 211-235, 1995. E. R. Fossum, “CMOS Image Sensors: Electronic Camera-On-A-Chip,” IEEE Trans. Electron Devices, Vol. 44, pp. 1689-1698, 1997. Pixim Inc.: datasheet D2000 Imaging System (On-line). Available: http://www.pixim.com/products/Pixim_D2000_Product_Brief.pdf Edward J. Bender, “Present Image Intensifier Tube Structures,” Electro-Optical Imaging: System Performance and Modeling, Lucien M. Biberman, Editor, SPIE Press, pp. 5-1 – 5-96, 2000. 10