`
`WANG et al., “CMOS video camera,”
`
`Euro ASIC ’91, pp. 100-103 (1991)
`
`TRW Automotive U.S. LLC: EXHIBIT 1048
`PETITION FOR INTER PARTES REVIEW
`OF U.S. PATENT NUMBER 8,599,001
`IPR2015-00436
`
`
`
`CMOS Video Cameras
`
`G. Wang, D. Renshaw, P. B. Denyer and M. Lu
`
`University of Edinburgh
`Department of Electrical Engineering
`Mayfield Road
`Edinburgh, EH9 3JL, UK.
`
`Abstract
`A single chip CMOS video camera is presented,
`along 1vith design technique and characterization
`resulrs. The chip comprises a 312X 287 pixel
`phorodiode array together with all the necessary
`sellsi,zg, addressing and amplifying circuitry, as
`well as a 1,000 gate logic processor, which
`implements synchronization timing to deliver a
`fully-formatted composite video signal and a
`further 1,000 gate logic processor, which imple-
`111e111s allfomatic exposure control over a wide
`rallge. There are also simple solutions for 'Y
`correction and test.
`
`1. Introduction
`
`We introduce a new capability that extends the
`CM OS ASIC marketplace in a sector of high
`growth rates. This market sector is that of image
`sensing and processing, covering applications
`from electronic cameras to 'smart' vision sys(cid:173)
`tems.
`
`Camera and vision systems addressed by today's
`CCD technology appear cumbersome, power(cid:173)
`hungry and expensive. The experimental work
`reported here demonstrates
`that high-quality
`image sensors can be implemented entirely in
`commodity ASIC CMOS technology, operating
`from single 5v supplies.
`
`The reported chip is a highly-integrated CMOS
`VLSI camera, shown in Figure 1. Most of the
`core area is a 312 x 287 pixel image sensor array,
`together with the necessary sensing, addressing
`and amplifying circuitry. The output signal can
`be either linear or 'Y corrected.
`'Y correction is
`achieved by a simple solution which uses the
`nonlinear ID-V GS characteristic of an MOS
`
`TH0367-3/91/0000/0100$01.00 © 1991 IEEE
`
`100
`
`transistor. The layout of the sensor is custom
`designed to make it as compact as possible.
`
`At the top (Figure 1) is the 2,000 gate logic pro(cid:173)
`cessor, laid out using a semi-custom standard(cid:173)
`cell compiler. Half of these gates generate syn(cid:173)
`chronization
`timing,
`including
`line-sync and
`frame-sync signals to format a 625line/50Hz
`standard composite video output. The other
`half of the gates are included to electronically
`control exposure over a wide range (40,000:1),
`enabling the use of a single fixed-aperture lens.
`The chip measures 7 .58mm x 7 .56111111 , using 1.5
`µm, 2 level metal CMOS technology.
`
`Figure 1. Photo-micrograph of single chip
`video camera
`
`1048-001
`
`
`
`A video camera has been built using this chip
`along with a 6 MHz clock source, a 5 volt
`power supply, plus one bipolar transistor and a
`small number of
`resistors
`and capacitors
`required to match the line impedance to the
`monitor and decouple the power supply. The
`picture quality
`is subjectively excellent, and
`compares well with commercially available cam(cid:173)
`eras.
`
`2. Image Sensor Block
`
`The architecture of the image sensor is shown in
`Figure 2. The light sensing area consists of a
`312 x 287 diode array matrix, schematically indi(cid:173)
`cated by the columns and rows of individual
`photodiodes. The pixel size is 19.6µm x 16µm,
`giving a light sensing area of 6.12mm x 4.59mm .
`This corresponds to the standard 1/2" format.
`
`cvo
`
`supplies
`
`clock
`
`gain
`setting
`
`auto-exposure
`control
`logic
`
`video format
`timing & control
`circuits
`
`o/p
`mux
`
`o/p
`amp&
`buffer
`
`horizontal addressing
`
`column sense amps
`
`• CVO -- composite video output
`
`Figure 2. Architecture of the image sensor
`
`The photodiodes are accessed on the basis of
`sequential selection of each row through a verti-
`
`JOI
`
`cal shift register. At the top of each column is a
`sense amplifier. The sensed information is read
`out sequentially along the x-direction under con(cid:173)
`trol of a horizontal shift register. At the end of
`the path there is an output amplifier [1,2].
`
`The sense amplifier is a single-ended differential
`charge integrator. Its performance demands an
`accurate capacitor, formed by metall/metal2
`and metall/poly. However, commodity ASIC
`CMOS technology sometimes can not guarantee
`the resulting capacitance values. We designed a
`gain-controllable integrator, shown in Figure. 3,
`which allows wide range of programmable varia(cid:173)
`tion of the capacitance value.
`
`Vref
`
`self
`compensation
`
`Figure 3.
`
`Integrator with programmable gain
`and self compensation
`
`The main concern in the output stage design is
`the read-out speed required to achieve high
`resolution. A 6 MHz clock was chosen for this
`design; this gives a horizontal resolution of 312
`pixels. The resultant picture quality is assured
`by a two stage output buffer with sample and
`hold function.
`
`3. Automatic Exposure Control
`
`The device automatically controls its exposure
`over a range of 40,000:1. Control is achieved
`by varying the integration time prior to reading
`each row of pixels. The integration time can be
`as long as one field, or as short as three cycles
`of the pixel clock(about 500ns).
`
`1048-002
`
`
`
`The exposure is set by monitoring the video
`stream and estimating the fractions of each pic(cid:173)
`ture which are very white and very black. On
`the basis of this information, the device decides
`whether the picture contrast is acceptable, or too
`white, or too dark. If necessary, the exposure
`time is then changed, in the appropriate direc(cid:173)
`tion.
`
`4. Generation of the Video Format Signal
`
`Figure 4 shows a block diagram for the genera(cid:173)
`tion of the video formatted signal. The -y
`corrected image data is multiplexed with the
`sync-level and blanking-level, controlled by tim(cid:173)
`ing control signals, which are provided from the
`bipolar
`transistor
`video
`t1mmg block.
`A
`(emitter follower) is needed to provide a low
`impedance output.
`
`Video
`
`timing
`
`Image signal
`
`Image
`
`sensor
`
`Vdd
`
`Sync
`level
`
`Blanking
`level
`
`Figure 4. Generation of the video output
`
`5. Simple Solution for -y Correction
`
`to be
`image data needs
`The analogue
`-y
`corrected, to compensate for the nonlinearity of
`monitor tubes [3]. This is usually implemented
`using discrete components e.g. a ladder-network
`reference voltages.
`resistors and
`of diodes,
`Unfortunately, this is not suitable for integra(cid:173)
`tion. In this design -y correction is achieved by
`a simple solution which uses the nonlinear ID(cid:173)
`V GS characteristic of an MOS FET, as shown
`
`102
`
`in Figure 5.
`
`Vdd
`
`Corrected
`output
`
`Linear
`input r-f
`
`Figure 5. Gamma corrector
`
`SPICE simulation was carried out and a simula(cid:173)
`tion result is shown in Figure 6. A theoretical
`curve of ideal -y correction ( -y =0.45) is also
`shown in Figure 6.
`
`ideal curve
`
`1.
`
`0.8
`
`0.6
`
`0.4
`
`0.2
`
`0.2
`
`0.4
`
`0.6
`
`0.8
`
`1.
`
`Figure 6. Gamma correction curves
`
`6. Simple Solution for Test
`
`Special consideration has been given to make it
`possible to carry out digital wafer test which is
`as complete as possible. The analogue parts arc
`also tested by making them produce digital out(cid:173)
`puts, so avoiding a requirement for full analo(cid:173)
`gue test. The test includes bit-line tests and
`word-line tests. Only a 0.78% increase in chip
`area was required to implement the on chip
`
`1048-003
`
`
`
`hardware necessary for this form of testing (Fig(cid:173)
`ure 2). The individual photo pixels may be
`tested if a sufficiently long vector set is allow(cid:173)
`able.
`
`The chip can also self-generate a checkerboard
`pattern which may be displayed on a monitor
`screen, or captured by a frame grabber. This
`pattern can be used not only to find defective
`pixels, but also to check analogue performance
`parameters, such as read out speed and unifor(cid:173)
`mity.
`
`7. Eliminating Noise
`
`Complete guard rings are put around all analo(cid:173)
`gue parts to minimize interference from the digi(cid:173)
`tal parts. Routing is arranged with priority to
`analogue output and analogue power supplies.
`Analogue power supplies and digital supplies are
`separated, and supplies to different analogue
`parts are divided where necessary.
`
`There are two sources of fixed pattern noise:
`threshold variation in
`the MOS pixel access
`transistors causing speckles, and mismatches
`between
`the column sense amplifiers causing
`vertical stripes. The solution to the pixel thres(cid:173)
`hold variation is to reduce the pixel reset voltage
`below (Vdd-Vt) so that the reset voltage is
`insensitive to the variation of the threshold Vt.
`
`Column fixed pattern noise arises mostly from
`offset mismatches in the column sense amplif(cid:173)
`iers. We have successfully eliminated this prob(cid:173)
`lem by automatically compensating each amplif(cid:173)
`ier to give zero offset during each line synchron(cid:173)
`ization interval.
`
`8. Characterization
`
`An optical test measurement set-up was used to
`characterize the camera. The following table
`summarize the measured results of the perfor(cid:173)
`mance characterization experiments. The param(cid:173)
`eters of typical monochrome CCD cameras are
`also given for comparison.
`
`·-
`
`parameter
`operating voltage
`for camera
`power dissipation
`for chip
`power dissipation
`for camera
`s.n.r.
`exposure ran_ge
`saturation level
`antibloomin_g factor
`dark current*
`
`------·-~·--
`CMOS
`CCD
`12v
`5v
`
`50mW
`
`200mW
`
`lW
`
`51dB
`40,000:1
`20lux
`lOOx
`0.0004
`
`52dB
`300:1
`20lux
`lOOx
`0.005 ·-
`
`* as fraction of saturation at room temperature,
`20msec integration time
`
`9. Conclusions
`
`We have developed several design techniques to
`achieve a single chip camera. in unmodified
`CMOS
`technology. which matches
`the
`performance of CCD cameras. The design has
`proven that three technical barriers which most
`greatly
`influence new product development;
`cost, power consumption and size, are all
`dramatically reduced over
`today's solid-state
`camera technologies.
`
`10. Acknowledgements
`
`the
`We acknowledge support received from
`Science and Engineering Research Council
`(Grant GR/F 36538 IED2/1/1159).
`
`11. References
`
`[1] D. Renshaw, et. al., "ASIC Vision", Proc.
`IEEE
`Custom
`Integrated
`Circuits
`Conference, 1990, pp 3038-3041.
`
`Image
`"ASIC
`al.,
`et.
`[2] D. Renshaw,
`Sensors",
`Proc.
`International
`IEEE
`Symposium on Circuits and Systems, 1990,
`pp 7.3.1-7.3.4.
`
`[3] Eugene Trundle, Television and Video
`Engineers Pocket Book, Heinemann, 1987.
`
`103
`
`1048-004