MLX75411-75412 MLX75412VTF-M MLX75412VTF-G MLX75412VTF-I MLX75412VTF-R - Datasheet Archive

 

Image Sensor Datasheet Avocet Image Sensor Data Sheet (summary) Feature List · · · · ·

 

MLX75411-75412 MLX75411-75412 (Avocet) Image Sensor Datasheet Avocet Image Sensor Data Sheet (summary) Feature List · · · · · · · · · · · · Melexis's low noise, low power rolling shutter CMOS imaging technology 1024 x 512 Image Array, with programmable subwindow ~ optical format for 1024x512 Autobrite® Wide Dynamic Range (>150db), with fully automatic, semiautomatic and manual control Monochrome, standard Bayer and custom color filter arrays available Integrated image optimization and noise reduction tuned for automotive applications TM Autoview Histogram Remapping 2-Wire serial interface for control Parallel data (8/10/12 data bits + CLK/HSYNC/VSYNC) provides direct connect to common video components and video enabled DSP's Start of production Q1 2010 Operating Temperature Range: -40C to +85C full performance (>+85C to +115C degraded performance) Storage Temperature Range: -40C to +125C Specification Avocet Active Pixel Resolution 1024 x 512 Optical format ~"(6.45mm) Pixel size 5.6 square Pixel type 3T Maximum frame rate 60fps (full resolution) Input clock range 20 ­ 54 MHz Exposure time range at full resolution and speed 1.0s ­ 16.7 ms Control interface 2 Wire Serial Slave Video interface 12-bit LVTTL (data, hsync, vsync, and clock) Signal processing Defect pixel interpolation FPN correction Histogram Optimization Dark Current correction Sharpening Progressive Subsample Subwindow Scanning modes Applications · · · · · · · Night vision Forward looking ADAS applications; collision warning, pedestrian detection, lane departure warning, traffic sign/light recognition Blind spot monitoring and detection Color rear-view cameras Driver monitoring for drowsiness Environments that require high sensitivity, wide dynamic range (WDR). Machine Vision applications that require high quality monochrome or color with high sensitivity CONFIDENTIAL This document is being provided on a confidential basis and is intended strictly for use by a limited number of interested parties for the sole purpose of determining potential interest in pursuing a transaction with the Company. By accepting this document, each recipient agrees to treat the information contained herein in a manner consistent with its own information of a confidential nature MLX75411-75412 MLX75411-75412 (Avocet) Image Sensor Datasheet 1. Ordering information Glass-BGA packaged imagers: Part Number Description MLX75412VTF-M MLX75412VTF-M Avocet mono WDR imager in ARC coated glass-BGA package, including Autobrite and Autoview MLX75412VTF-G MLX75412VTF-G Avocet RGBG color WDR imager in ARC coated GBGA package, including AB and AV MLX75412VTF-I MLX75412VTF-I Avocet RGBi color WDR imager in ARC coated GBGA package, including AB and AV MLX75412VTF-R MLX75412VTF-R Avocet RCCC WDR imager in ARC coated GBGA package, including AB and AV Bare die imagers: Part Number Description MLX75412VUC-M MLX75412VUC-M Avocet monochrome WDR imager- Die on wafer (unsawn), including AB and AV MLX75412VUC-G MLX75412VUC-G Avocet RGBG color WDR imager- Die on wafer (unsawn), including AB and AV MLX75412VUC-I MLX75412VUC-I Avocet RGBi color WDR imager - Die on wafer (unsawn), including AB and AV MLX75412VUC-R MLX75412VUC-R Avocet RCCC color WDR imager - Die on wafer (unsawn), including AB and AV See 3.6 for detailed information which algorithms are included in MLX75411 MLX75411 or MLX75412 MLX75412. 2. Document Version Version Date Notes 1.1 05/9/2009 Initial version 1.12 06/01/10 Updated stereo operation flow 2.0 06/03/10 75412 added 2.1 07/30/10 . Updated order info Updated default registers for 75412 status Updated reset polarity 2.2 08/10/10 Updated register map Page 2 of 26 MLX75411-75412 MLX75411-75412 (Avocet) Image Sensor Datasheet Table of Contents Feature List. 1 Applications. 1 1. Ordering information. 2 2. Document Version . 2 3. Device Overview . 5 3.1. Avocet Image Sensor Overview . 5 3.2. Sensor Architecture . 6 3.3. Package Options. 7 3.4. Pixel Array Description & Operation. 7 3.5. Interfaces . 8 3.5.1. Pin description . 8 3.5.2. Pixel data output interface . 9 1.1.1.1. 12/10/8 bit digital video mode (default) . 9 1.1.1.2. Packetized video mode . 10 3.5.3. Two Wire Serial Slave Interface . 11 3.5.4. Two Wire Boot Loader Interface. 11 3.5.5. GPIO Interface. 11 3.6. On-chip Algorithms . 12 3.6.1. Autobrite® Wide Dynamic Range . 13 1.1.1.3. Autobrite® Semi automatic mode . 13 1.1.1.4. Autobrite® automatic mode Feedback Loop. 13 1.1.1.5. Autobrite® Advantages . 14 3.6.2. Column FPN Correction . 15 3.6.3. Dark Current Correction . 16 3.6.4. Spatial Filtering: Defective Pixel Correction . 17 3.6.5. Spatial Filtering: Sharpening . 18 3.6.6. AutoviewTM Histogram Optimization. 19 3.6.7. Context switching . 19 3.6.8. Stereo support . 19 3.7. Device power up behavior and initialization . 20 4. Device Electrical Parameters . 22 4.1. Maximum Electrical Ratings. 22 4.2. DC Electrical Characteristics. 23 4.3. AC Electrical Characteristics (Non-inverted PIXCLK, default) . 24 4.4. AC Electrical Characteristics (Inverted PIXCLK, with 18ns clock period) . 24 4.5. Optical Characteristics . 25 5. Disclaimer. 26 Page 3 of 26 MLX75411-75412 MLX75411-75412 (Avocet) Image Sensor Datasheet Table of Figures Figure 1 ­ Avocet Architecture Block Diagram . 6 Figure 2 ­ pin GBGA package and bare die. 7 Figure 3 ­ Digital Video Interface Line Timing . 10 Figure 4 ­ Digital Video Interface Frame Timing . 10 Figure 5 ­ Autobrite® feedback and control mechanism . 14 Figure 6 ­ Images captured without (left) and with (right) Autobrite . 15 Figure 7 ­ Sharpening Filter (Unsharpened image). 18 Figure 8 ­ Sharpening Filter (Unsharpened image). 18 Figure 9 ­ Avocet Sharpening: Matrix multiplication . 18 Figure 10 ­ Avocet Sharpening Matrices . 18 Table of Tables Table 1 - Avocet image sensor highlights . 5 Table 2 - Avocet pinout. 8 Table 3 ­ I2C address . 9 Table 4 - On chip algorithms . 12 Table 5 - Electrical Specifications: Maximum Ratings . 22 Table 6 - Electrical Specifications: DC Electrical Characteristics. 23 Table 7 - Electrical Specifications: AC Electrical Characteristics. 24 Table 8 - Electrical Specifications: AC Electrical Characteristics. 24 Table 9 - Specifications: Optical Characteristics . 25 Page 4 of 26 MLX75411-75412 MLX75411-75412 (Avocet) Image Sensor Datasheet 3. Device Overview 3.1. Avocet Image Sensor Overview The Avocet image sensor is manufactured using advanced 0.18m CMOS imaging process, integrates a high-sensitivity array, a fully-featured digital imaging processing pipeline and camera control functions into a single module. Avocet is a 1024 x 512 pixel, ~" optical format CMOS image sensor. For VGA applications, the center 640 x 480 pixels can be used in a ¼" format which can reduce the size and/or cost of the matching optical components. The device captures either Monochrome or Bayer-pattern color still pictures. Table 1 - Avocet image sensor highlights Specification Avocet Comments Active Resolution 1024 x 512 Wider horizontal resolution to meet the next generation ADAS requirements Optical format ~"(6.45mm) Center ¼ " can be used for VGA resolution Pixel size 5.6 square Optimized for sensitivity at 1024x512 resolution Pixel type 3T Optimized for WDR and Sensitivity Max frame rate 60fps At full resolution Input clock range 20 ­ 54 MHz Options for clocking: Crystal input, Oscillator Exposure time range 12.5s ­ 16.7 ms Control interface 2 Wire Boot Loader Interface At 54MHz and 60fps, at full resolution and speed. Minimum barrier time 1.22us. Used to load register settings on recovery from a reset. Must be accessible (for programming of serial PROM) via other control interfaces. 2-Wire, low speed, serial control interface used for short distances. Does support broadcast writes for writing multiple imagers. 12-bit Monochrome or Raw Color. Pixel clock, vsync and hsync compatible with the DSP's. (i.e. TI DaVinci or ADI Blackfin) Sensor provides on-chip processing required for vision applications or monochrome display applications. (Color processing is not included in on-chip functions.) 2 Wire Serial Slave Video interface LVTTL Signal processing Defect pixel interpolation FPN correction Histogram Optimization Dark Current correction Sharpening Progressive Scanning modes Required to support machine vision applications Subsample 2x and 4x vertical subsampling Subwindow Single rectangular region. The starting point of the x- and y-address is programmable, as well as the window size. Sync to external event Event Synchronization Note: For some of the features, an external PROM must be connected to the boot loader interface. Table 4 on page 12 summarizes for which algorithms an external PROM is needed. Page 5 of 26 MLX75411-75412 MLX75411-75412 (Avocet) Image Sensor Datasheet 3.2. Sensor Architecture The Avocet CMOS image sensor is a digital image sensor targeted at video capture in automotive, industrial and transportation safety applications. The Avocet sensor is available as either a color or monochrome imaging device. In a monochrome system, image processing is performed on the Avocet as a single chip solution and the Color Filter Array (CFA) is eliminated to maximize device sensitivity. In a color system Avocet acts as a slave in a system controlled by a separate Image Signal Processor chip through one of the external interfaces delivering raw single images or video-like streams of color Bayer-patterned images. The sensor captures the images at the required speed, exposure, and gain, and the external device is responsible for all further image processing such as color demosaicing, white balance, and automatic exposure/gain control. Figure 1 ­ Avocet Architecture Block Diagram Page 6 of 26 MLX75411-75412 MLX75411-75412 (Avocet) Image Sensor Datasheet 3.3. Package Options Avocet is offered in two package solutions: a 55 pin Glass-BGA and a bare die (tested wafers) for use in chip-on-board. Figure 2 ­ pin GBGA package and bare die 3.4. Pixel Array Description & Operation The pixel array is composed from 5.6um square sensing elements based on Melexis's 3T pixel architecture. The Sensor array can either be monochrome or have a set of color filters applied depending on the application requirements. The image sensor operates using an electronic rolling shutter. This maximizes the amount of integration time available by overlapping the readout with the "reset" of a pixel to begin integration. Avocet operates in an adaptable, programmable Wide Dynamic Range (WDR) mode maximizing sensitivity while providing uncompromising dynamic range. Page 7 of 26 MLX75411-75412 MLX75411-75412 (Avocet) Image Sensor Datasheet 3.5. Interfaces Avocet has four main interfaces: pixel data output, two wire serial slave, two wire serial master, and GPIO. Each of these interfaces is described below. 3.5.1. Pin description Table 2 - Avocet pinout 55 GBGA Ball 6, 7, 8, 12, 11, 10, 9, 2, 3, 4, 47, 46 29 15 52 26 25 28 27 48 49 34 42 1, 18, 20, 36 5, 32, 38, 54 17, 37, 55 16, 39, 53 13, 40, 51 14, 41, 50 30 31 19 24 23 22 45 44 43 21 33 35 56 CBGA Ball [H3, E2, F2, E1, F1, G1, H1, C2, B4, A4, C4, B5] H8 F3 C3 E8 H7 G8 F8 D1 D2 C6 C5 B1, B8, G4, H6 A1, A8, G5, H5 B2, B7, G3 A2, A7, H4 B3, B6, H2 A3, A6, G2 F6 F5 F4 G7 F7 E7 A5 D8 C8 G6 D7 C7 C1 Symbol PIXD[11:0] Type OUTPUT Description Parallel pixel data output bit 11 (MSB) to 0 (LSB) HSYNC VSYNC PIXCLK SDAT SCLK BLIDAT BLICLK ADDR0 ADDR1 XTALIN RESET_N VDDA GNDA VDDD GNDD VDDIO GNDIO VREFM VREFP VCM GPIO GPIO2 GPIO3 RSVD_1 RSVD_2 RSVD_3 RSVD_4 RSVD_5 RSVD_6 NC OUTPUT OUTPUT OUTPUT I/O INPUT I/O OUTPUT INPUT INPUT INPUT INPUT SUPPLY SUPPLY SUPPLY SUPPLY SUPPLY SUPPLY REF REF REF INPUT INPUT INPUT INPUT INPUT INPUT OUTPUT INPUT OUTPUT N/C Line valid. Asserted high when PIXDAT data is valid Frame valid. Asserted high when PIXDAT data is valid Pixel clock out. 2-wire slave serial data interface. 2-wire slave serial clock interface. 2-wire serial boot loader interface data. 2-wire serial boot loader interface clock. 2-wire slave serial interface address bit select [0] 2-wire slave serial interface address bit select [1] System clock input. Active low image sensor reset. Analog power supply 3.3V Analog Power supply ground Digital power supply 1.8V Digital Power supply ground I/O pad power supply 3.3V I/O Power supply ground Analog reference voltage Analog reference voltage Analog reference voltage Connect to GND Connect to GND Connect to GND Connect to GND Connect to GND Connect to GND Connect to GND Connect to VDD Leave unconnected No connection NOTES: 1. PIXDAT data is valid on rising edge of pixel clock by default. Pixel clock can be inverted via register write to bit [6] of the output_option_enables register (0x8205). 2. The value of Rsi resistor will vary depending on the capacitive load on the systems serial bus. Since all systems will vary, an absolute value cannot be specified. For most systems, a 2k resistor is recommended unless system requirements demand a stronger/weaker pullup. Page 8 of 26 MLX75411-75412 MLX75411-75412 (Avocet) Image Sensor Datasheet 3. A 2k pullup resistor is required for the boot load interface wires. A minimum128k EEPROM such as Microchips 24LC128 24LC128 is also required. 4. The following truth table applies: Rsetx_x Table 3 ­ I2C address ADDR1 ADDR0 Two-Wire Interface Slave Address 0 (Rset1_0) 0 (Rset1_0) 1 (Rset1_1) 1 (Rset1_1) 0 (Rset0_0) 1 (Rset0_1) 0 (Rset0_0) 1 (Rset0_1) 0101_100[R/W] 0101_101[R/W] 0110_110[R/W] RSVD 5. With default power-up frame dimensions, 27MHz provides 30fps full resolution 1024x512 operation, and 54MHz provides 60fps full resolution 1024x512 operation. Frequency range is < or = to 54MHz. 6. Avocet has an internal POR (Power On Reset) circuit. Either a reset controller device or an RC filter can be used to adjust the RC ramp time of the reset_n signal as desired. 3.5.2. Pixel data output interface The image data from Avocet is sent via a pixel interface. The interface is designed to be compatible with standard interfaces to DSP's (TI DaVinci and ADI Blackfin) for image data transfer. Data is transferred as frames (images), one line at a time from the top of the image to the bottom of the image. Each line is transferred in contiguous data bursts from the left of each line to the right. The interface consists of the following pins: · · · · PIXD[11:0] ­ 12/10/8 bit pixel data interface. In the 10 and 8 bit modes, the lower order bits are not used and are held to zeros. HSYNC ­ This signal indicates horizontal sync or data valid. It can be configured as active high or active low by changing on-chip configuration registers. VSYNC ­ This signal indicates vertical sync. It can be configured as active high or active low by changing on-chip configuration registers. PIXCLK ­ This is the output clock that should be used to clock in HSYNC, VSYNC, and PIXD[11:0] on the receiving end. By default, the output registers are clocked on the rising edge of this clock. This can be changed so that the negative edge corresponds to the data changing by an on-chip configuration register. 1.1.1.1. 12/10/8 bit digital video mode (default) When configured in digital video mode, the chip outputs pixel data on the PIXD bus and signals start of line and frame with HSYNC and VSYNC. Page 9 of 26 MLX75411-75412 MLX75411-75412 (Avocet) Image Sensor Datasheet The chip will output one pixel per clock at the input clock rate; with data changing on the rising edge of PIXCLK by default (can be changed to negative edge if desired). HSYNC will be asserted for every valid pixel of a line (1024 by default), and de-asserted during hblank and v-blank intervals. By default, h-blank will be for 298 clocks between valid lines. VSYNC will be asserted for every valid line, and will rise coincident with the rise of the first HSYNC of the frame. It will drop coincident with the fall of the last HSYNC of the frame. By default, Vsync is low for 168 rowtimes. The digital video timing is as follows: Figure 3 ­ Digital Video Interface Line Timing Figure 4 ­ Digital Video Interface Frame Timing 1.1.1.2. Packetized video mode The chip can also be configured to transmit all relevant information on the PIXD bus (including start of line and start of frame indicators). The video data is sent in one line per "packet", with packet start, stop, and CRC. Additional status information is also transmitted in the video stream. Page 10 of 26 MLX75411-75412 MLX75411-75412 (Avocet) Image Sensor Datasheet In this mode, the start of line and start of frame are sent framed in separate "packets" on the PIXD bus. As a result, the HSYNC and VSYNC lines are not used. For further information on the packetized video mode and its format, please contact Melexis. 3.5.3. Two Wire Serial Slave Interface Avocet provides an interface to all configuration and control of the imager sensor and image processing functions. This two wire interface is comprised of a single clock line and a single data line. A serial protocol is used to address the chip and to read and write data. 3.5.4. Two Wire Boot Loader Interface Avocet is capable of using an external PROM to control the behavior of the device after power up and to augment the internal processing. It does this through a 2 wire interface that uses the same protocol as the two wire serial slave interface. The only supported use of this interface is for the connection of the external PROM to control the power up behavior. Note: These pins should not be physically connected to the slave interface pins. 3.5.5. GPIO Interface Avocet has 3 GPIO pins for future expansion of device capabilities. As these are currently unused, recommendation is to connect these to GND. Page 11 of 26 MLX75411-75412 MLX75411-75412 (Avocet) Image Sensor Datasheet 3.6. On-chip Algorithms The image processing algorithms for Avocet are described in this section. Each algorithm is individually controlled with an enable/disable. All of the algorithms also have individual control bits that allow tuning and control of the behavior of the individual algorithms. The Spatial Filtering algorithms are not color aware, and should be disabled if color filters are present. Some algorithms are only available on MLX75412 MLX75412 and not on MLX75411 MLX75411. The algorithm availability for both devices can be found in the table below. Table 4 - On chip algorithms Algorithm 75411 75412 External PROM needed Description Autobrite® semi automatic Yes Yes No User only has to set two registers to fully control the sensitivity and response curve Autobrite® full automatic No Yes Yes An Automatic Exposure (AE) controller that regulates the dynamic range compression of high-dynamic-range pixels. Column FPN removal Yes Yes No Uses electrical black from DAC's to compensate for ADC and column circuitry offset mismatches. Dark Current subtraction Yes Yes No Performed using optical black (dark row) averages. Spatial Filtering Yes Yes No Done within a single 3x3 kernel Image Statistics Yes Yes No Feeds image information to apply to the next frame for Autobrite, spatial filtering, and Melexis's histogram optimization AutoviewTM No Yes Yes Performs automatic histogram equalization, or gamma correction, or a user-programmable transfer function Context switching Yes Yes No The imager will toggle each frame between two configurations. For cases where the output stream is used by several applications. Stereo support Yes Yes No To synchronize the output stream of two imagers Page 12 of 26 MLX75411-75412 MLX75411-75412 (Avocet) Image Sensor Datasheet 3.6.1. Autobrite® Wide Dynamic Range Autobrite uses the variable height/multiple reset method and a feedback loop to meet the criteria for wide dynamic range cameras. Autobrite uses combined technologies to expand dynamic range and a straightforward standard, three transistor CMOS pixel to provide reliability and cost-effectiveness. No additional frame buffers or post processing is required. Based on research conducted at the Massachusetts Institute of Technology, Autobrite controls the pixel through Melexis proprietary variable height/multiple reset method. In automatic mode, a complete feedback loop simultaneously controls the integration time and dynamic range expansion for total adaptability and programmability. 1.1.1.3. Autobrite® Semi automatic mode In semi automatic mode, the user keeps full control of the integration time and amount of compression. The algorithm will set the low level registers to implement these settings with optimized iSNR. The algorithm will not update the settings dynamically, it will only change the settings when the user changes either the exposure time or the compression level. In this mode, the user has only two inputs: Exposure time and level of compression. The exposure time can be set as a two-byte register which controls the exposure time in number of rows. The amount of compression can be set by setting the time between the last compression barrier and the pixel readout. Placing this barrier closer to readout will increase the compression ratio. Based on the position of this last barrier, the firmware will place the other five barriers, while keeping iSNR maximal. By using this mode, the amount of I2C communication can be reduced significantly: In full manual mode, it would take 32 bytes to set the exposure time and compression curve. In semi automatic mode, this is reduced to 6 bytes (one 16b and one 32b register). 1.1.1.4. Autobrite® automatic mode Feedback Loop Achieving wide dynamic range through an image sensor with linear response at low illumination and non-linear response at high illumination solves only part of the problem. To complete the solution, the automotive camera must be able to decide which response curve to use. Furthermore, system designers must be able to override the automated decision to customize the response for specific applications. Autobrite uniquely meets these criteria with its key features: adaptability and programmability. Autobrite includes a mechanism to dynamically adjust both the response curve and the total integration time based on the scene being observed. Essentially, the dynamic range is expanded in real time by changing the timing and height of the reset signal. The control mechanism can be configured to automatically adapt to each environment or programmed for a specific application, thereby providing performance that is unmatched by other approaches for achieving wide dynamic range. "Figure 5 ­ Autobrite® feedback and control mechanism" illustrates the Autobrite control mechanism. Page 13 of 26 MLX75411-75412 MLX75411-75412 (Avocet) Image Sensor Datasheet The control loop starts with the image sensor capturing an image. Registers acquire statistics of the scene such as number of pixels that exceed a threshold. A proprietary control mechanism, which can be tailored to a specific application, uses the statistics to select the optimum response curve and integration time. A multiplexing device inputs the signals and allows either the calculated values or user-supplied values to be fed to the voltage and timing control, which generates the barrier voltage(s) for the image sensor. Simultaneously calculating both the integration time and the required dynamic range expansion allows the image sensor to settle on optimal settings very quickly. This fast response time is critical in applications where dramatic changes in the lighting conditions occur quickly, such as the appearance of headlights from other vehicles. Another advantage of this approach is that the collection of the statistics and the control algorithms can be tailored to a specific application. Users can program Autobrite to meet specific requirements of their application. Figure 5 ­ Autobrite® feedback and control mechanism For example, system engineers can program Autobrite to: · Select a specific region of interest within the image frame that Autobrite will use to optimize the integration time and response curve. · Select a specific integration time and response curve, overriding the automated adjustment. · Select a maximum integration time or response curve that the automated adjustment is not to exceed. · Adjust the speed of adaptability to respond more quickly or slowly to lighting changes. · Manually adjust the height and timing of the barrier voltages. 1.1.1.5. Autobrite® Advantages Figure 6 provides a side-by-side comparison of images captured with and without Autobrite. In the image on the left, the intra-scene dynamic range clearly exceeds the dynamic range of the camera, resulting in lost details in both the light and dark regions. In the image on the right, Page 14 of 26 MLX75411-75412 MLX75411-75412 (Avocet) Image Sensor Datasheet Autobrite enables the same scene to be captured with complete visual details even in the extremes of brightness and darkness. Figure 6 ­ Images captured without (left) and with (right) Autobrite 3.6.2. Column FPN Correction Fixed Pattern Noise or FPN correction is meant to correct for the differences in the ADCs that are used to convert the pixel voltages to pixel values and the column circuitry that connects the pixels to the ADCs. Because ADCs are not perfect, they do not always convert a given voltage to the same value. The differences in the values for any single ADC tend to be small, but the differences between two ADCs converting the same voltage can be noticeable. The transistors and wires that carry the pixel voltages to the ADCs can experience similar differences. This causes pixels in one ADC or column to appear brighter or darker than another. Differences between two ADCs lead to banding in the image with the bands being equal to the width of the multiplexers that feed the ADCs. Differences in the column circuitry lead to differences between adjacent columns or striping. This kind of noise is very visible in the output image because it is not random noise. It also creates false edges or differences in contrast between columns. These differences will be exaggerated by pixel processing algorithms like sharpening. They can also cause machine vision applications problems with object detection because of the extra false edges in the image. To correct for this, the differences between ADCs and column circuits are measured by applying known voltages to each column and using the ADC to measure the voltage. A near black signal is applied to each column to allow measurement of the offset. Black is not used because some of the offsets might be negative. These measurements are saved for each column in a RAM structure and used to calculate the correction factors for each column. Variables that affect the FPN values: · Process differences - Because ADCs, transistors, and wires are not manufactured identically, these differences can lead to differences in the ADC output when converting a given voltage. · Temperature - Changes in temperature will change the analog behavior of the ADCs and transistors. Higher temperatures typically lead to higher differences between ADCs, transistors, and wires and therefore higher FPN. · Voltage - Changes in voltage will change the analog behavior of the ADCs and transistors. Lower voltages typically lead to higher FPN. Page 15 of 26 MLX75411-75412 MLX75411-75412 (Avocet) Image Sensor Datasheet Because both temperature and voltage can vary over time, FPN correction is designed as a dynamic correction process that constantly measures the differences and adjusts the correction factors. 3.6.3. Dark Current Correction Dark current correction is meant to correct for the average leakage of all pixels. Because pixels are not perfect, they leak current over time even if there is no light shining on them. This causes pixels to appear unnecessarily bright. To correct for this, some pixels on the array are covered with a metal shield that is meant to block all light from reaching them. These pixels can then be read and the results averaged to obtain the "average" leakage value for all the pixels in the array. This "average" leakage is used to correct the active area. Variables that affect the dark current values: · Process differences - Because not all pixels manufactured identically, some leak more than others. · Temperature - The higher the temperature, the more current a pixel will leak. Temperature is a relatively slow change on the imager in comparison to frame times. Quick changes in the dark current values are probably due to other non-leakage causes. To help reduce the effects of these other fast changing causes, a low pass filter is included on the dark current average. Without this low pass filter, minor changes in the dark current value could lead to a fast frame to frame change of the correction value. This would cause frame to frame flickering. The hardware uses a low pass filter and caps the maximum dark current value to counteract any short term variation in the dark current measurements. The device also offers a manual mode of dark current correction. When in manual mode, the dark current measurements and the correction value can be written via the 2-wire serial bus. Page 16 of 26 MLX75411-75412 MLX75411-75412 (Avocet) Image Sensor Datasheet 3.6.4. Spatial Filtering: Defective Pixel Correction When the image sensor is manufactured, not all pixels are created identically. There are often pixels in the image array that respond differently than the others. Sometimes these pixels are stuck off or stuck on. This results in white or black pixels in the output image. But more often, there are pixels that respond to the light more quickly or more slowly than average. These pixels do not appear white or black all the time, but appear lighter or darker than the surrounding pixels when exposed to the same amount of light. Both types of pixels can be alone or "clustered" with other defective pixels in groups. Avocet implements "neighbor comparisons" which takes advantage of the fact that two pixels next to each other in the array are unlikely to be significantly different than all of their neighbors. Even when the image has a black or white spot that would be one pixel in size, the use of imperfect optics leads to light being scattered to adjacent pixels. This tends to smooth out the transitions from one pixel with its neighbors. The advantage of this algorithm is that the comparisons are done in real time, without the need for a PROM to record the pixel map. The algorithm is also capable of handling pixels that become damaged. Note: The on-chip defective pixel correction algorithm is not color aware, and should be disabled if color filters are present. Page 17 of 26 MLX75411-75412 MLX75411-75412 (Avocet) Image Sensor Datasheet 3.6.5. Spatial Filtering: Sharpening Sharpening is meant to enhance the contrast differences between adjacent pixels. This has the visual effect of "sharpening" the focus of an image. Below is a section of an image. Figure 7 ­ Sharpening Filter (Unsharpened image) Figure 8 ­ Sharpening Filter (Unsharpened image) The filter operation is done with a matrix linear convolution operation. A matrix of source pixels is convolved with a matrix of constants to generate a single value: Figure 9 ­ Avocet Sharpening: Matrix multiplication The possible selections for sharpening matrices are (from weakest to strongest): Figure 10 ­ Avocet Sharpening Matrices Page 18 of 26 MLX75411-75412 MLX75411-75412 (Avocet) Image Sensor Datasheet 3.6.6. AutoviewTM Histogram Optimization Avocet includes a histogram remapping function that maps from 12 bit pixels to 8 or 10 bit pixels to facilitate data processing and transfer in systems that do not implement the full 12 bit pixels. The algorithm emphasizes areas in the histogram that contain most of the information while compressing areas with limited information. This is particularly useful in scenes that naturally have a somewhat sparse histogram. An automotive night scene is an example where most of the information is lumped into three luminance bands: headlights and taillights which are very bright; traffic signs that are of medium intensity; and pavement which is relatively dark. By emphasizing the regions containing the most information, 12 bits can be reduced to 8 with a minimum loss of information. The on-chip hardware is capable of remapping 8 individual segments of the histogram to new areas. Both the offset and the gain of the remapping are controllable through register settings. The automatic algorithm uses the image statistics to calculate remapping constants to provide a good mapping when reducing the number of bits of the pixels. 3.6.7. Context switching Context switching is a feature which allows parameters on the image sensor to be switched on alternating frames. This enables an output stream which switches between two sets of parameters on alternating frames. One example would be to alternate between frames with a long and a short exposure time. This feature allows the output stream to be used by two different applications at the same time, each having their own configuration of the imager. The effective framerate for each application will be half the output framerate. E.g. by running the imager at 60FPS 60FPS, each application could still use 30FPS 30FPS of data. The parameters which can be alternated are: · Integration time (in semi-automatic Autobrite mode) · Amount of compression (last barrier time in semi-automatic Autobrite mode) · Hardware feature enables (FPN, DC correction, Sharpening, Histogram remapping) · Sharpening filter strength 3.6.8. Stereo support For applications where the output of two imagers has to be combined, it is often needed to synchronize the output streams. Avocet has the possibility to do this synchronization and compensate for vertical misalignment. By doing this vertical compensation, the two imagers output the same horizontal line of the scene at the same time, even in case they are misaligned. To do this synchronization, both imagers should have a common clock source. The external I2C master will follow a prescribed routine to align both imagers. Page 19 of 26 MLX75411-75412 MLX75411-75412 (Avocet) Image Sensor Datasheet 3.7. Device power up behavior and initialization Avocet has an internal power-on reset circuit that will reset the chip after power has reached acceptable levels. After power-on, the user may optionally download a set of application specific register values. These values configure the internal circuitry (analog bias levels, register settings .) for optimal application specific performance. This initialization may be accomplished using a standard serial PROM or over the 2 wire interface from a host controller. From a high level, the power up sequence is as follows: 1. Power is applied to the chip. 2. A clock is applied to the chip. The default programming expects a 27MHz clock, which will give 30FPS 30FPS video on the output interface. 3. The on-chip power-on-reset logic holds the chip in reset until the power is stable. 4. All on-chip registers are reset to the default states. 5. If the reset pin level is low (0V), the chip will be held in reset and the boot sequence is held at this step. If the reset pin level is high (3.3V), the chip comes out of reset with all the power on values and proceeds to the next step in the sequence. 6. The firmware boot sequence takes over. The boot sequence comprises: a. Query the 2 wire serial master interface to see if a valid PROM is located at device address 0xA0. b. If a valid PROM is present, load the contents of the PROM into the chip. If no valid PROM is present, continue with the boot sequence. The contents of the PROM can be used to change the default behavior of the device. See section 3.5.4 for more information about the use of the boot loader interface and the PROM. c. Enable the image capture and enable the output to begin transmission of video. d. Complete the firmware boot by setting register 0x8500 to 0x00, which indicates the boot process is complete. (Until the boot process is complete, 0x8500 will read as non-zero.) 7. At this point, the user can change the device operation by using the 2 wire serial slave interface to set the chip control registers. Page 20 of 26 MLX75411-75412 MLX75411-75412 (Avocet) Image Sensor Datasheet Page 21 of 26 MLX75411-75412 MLX75411-75412 (Avocet) Image Sensor Datasheet 4. Device Electrical Parameters 4.1. Maximum Electrical Ratings Table 5 - Electrical Specifications: Maximum Ratings Symbol Description Vdda Analog supply voltage Idda Minimum Nominal Maximum Unit -5% 3.3 +5% V Analog supply current mA Vddio I/O supply voltage -5% 3.3 +5% V Iddio I/O supply current Vddd Digital supply voltage Iddd Digital supply current Tstg Storage Temperature -40 +125 C Topl Operational Temperature -40 +115 C mA -5% 1.8 +5% V mA Page 22 of 26 MLX75411-75412 MLX75411-75412 (Avocet) Image Sensor Datasheet 4.2. DC Electrical Characteristics VDDIO = 3.3V (+/- 5%); Tamb=Ambient=25C Table 6 - Electrical Specifications: DC Electrical Characteristics Symbol Definition VIH Condition Minimum Nominal Maximum Unit Input high voltage 2.4 - VDDIO + 0.3 V VIL Input low voltage -0.3 - 0.8 V IIN Input leakage current No pull-up resistor; Vin = VPWR or VGND -2 - 2 uA VOH Output high voltage IOH = -4.0mA VDDIO ­ 0.4 - - V VOL Output low voltage IOL = 4.0mA - - 0.4 V IOH Output high current VOH = VDDIO -0.7 -7 - - mA IOL Output low current VOL = 0.7 - - 7 mA VDDA Analog supply voltage Default settings 3.135 3.3 3.465 V IDDA Analog supply current Default settings; XTALIN=27MHz - 40 62 mA VDDIO I/O supply voltage Default settings 3.135 3.3 3.465 V IDDIO I/O supply current Default settings; XTALIN=27MHz - 20 20 mA VDDD Digital supply voltage Default settings 1.71 1.8 1.89 V IDDD Digital supply current Default settings; XTALIN=27MHz - 30 32 mA Page 23 of 26 MLX75411-75412 MLX75411-75412 (Avocet) Image Sensor Datasheet 4.3. AC Electrical Characteristics (Non-inverted PIXCLK, default) VDDIO = 3.3V (+/- 5%); Tamb=Ambient=25C Table 7 - Electrical Specifications: AC Electrical Characteristics Symbol Definition Condition XTALIN Input system clock frequency Clock duty cycle tRCLK Input clock rise time tFCLK Input clock fall time tPDXP XTALIN to PIXCLK CLOAD = 10pF propagation delay tPDPD PIXCLK to valid DOUT[11:0] CLOAD = 10pF propagation delay tSUD Data setup time tHD Data hold time tPDPH PIXCLK to HSYNC CLOAD = 10pF propagation delay tPDPV PIXCLK to VSYNC CLOAD = 10pF propagation delay 4.4. Minimum 20 45 0.5 0.5 1.6 Nominal 27 50 2 2 2.6 Maximum 54 55 3 3 3.5 Unit MHz % ns ns ns 2.5 3.8 5 ns -1 0 1.2 -0.5 0.25 2.4 -0.1 0.5 3.5 ns ns ns 1.2 2.4 3.5 ns AC Electrical Characteristics (Inverted PIXCLK, with 18ns clock period) VDDIO = 3.3V (+/- 5%); Tamb=Ambient=25C Table 8 - Electrical Specifications: AC Electrical Characteristics Symbol Definition Condition XTALIN Input system clock frequency Clock duty cycle tRCLK Input clock rise time tFCLK Input clock fall time tPDXP XTALIN to PIXCLK CLOAD = 10pF propagation delay tPDPD PIXCLK to valid DOUT[11:0] CLOAD = 10pF propagation delay tSUD Data setup time tHD Data hold time tPDPH PIXCLK to HSYNC CLOAD = 10pF propagation delay tPDPV PIXCLK to VSYNC CLOAD = 10pF propagation delay Page 24 of 26 Minimum 20 45 0.5 0.5 10.8 Nominal 27 50 2 2 11.8 Maximum 54 55 3 3 12.8 Unit MHz % ns ns ns -4.5 -6 -6.5 ns 4.5 8 -6 6 8.5 -6.5 6.5 9 -7 ns ns ns -6 -6.5 -7 ns MLX75411-75412 MLX75411-75412 (Avocet) Image Sensor Datasheet 4.5. Optical Characteristics Table 9 - Specifications: Optical Characteristics Parameter Definition Condition Chief Ray Chief ray angle the sensor Angle has been optimized for. Page 25 of 26 Minimum Nominal 10 Maximum Unit degrees MLX75411 MLX75411 (Avocet) Image Sensor Datasheet (Advanced Information) 5. Disclaimer Devices sold by Melexis are covered by the warranty and patent indemnification provisions appearing in its Term of Sale. Melexis makes no warranty, express, statutory, implied, or by description regarding the information set forth herein or regarding the freedom of the described devices from patent infringement. Melexis reserves the right to change specifications and prices at any time and without notice. Therefore, prior to designing this product into a system, it is necessary to check with Melexis for current information. This product is intended for use in normal commercial applications. Applications requiring extended temperature range, unusual environmental requirements, or high reliability applications, such as military, medical life-support or life-sustaining equipment are specifically not recommended without additional processing by Melexis for each application. The information furnished by Melexis is believed to be correct and accurate. However, Melexis shall not be liable to recipient or any third party for any damages, including but not limited to personal injury, property damage, loss of profits, loss of use, interrupt of business or indirect, special incidental or consequential damages, of any kind, in connection with or arising out of the furnishing, performance or use of the technical data herein. No obligation or liability to recipient or any third party shall arise or flow out of Melexis' rendering of technical or other services. © 2010 Melexis NV. All rights reserved Page 26 of 26