`
`LU et al., “On-Chip Automatic Exposure Control Technique,”
`
`Proceedings – Seventeenth European Solid-State Circuits Conference 1991,
`
`pp. 281-284 (1991)
`
`TRW Automotive U.S. LLC: EXHIBIT 1050
`PETITION FOR INTER PARTES REVIEW
`OF U.S. PATENT NUMBER 8,599,001
`IPR2015-00436
`
`
`
`On-chip Automatic Exposure Control Technique
`M. Lu, G. Wang, D. Renshaw, and P. D. Denyer
`University of Edinburgh
`Department of Electrical Engineering
`Mayfield Road
`Edinburgh, EH9 3JL, UK.
`Tel. 31 650 5661 Fax. 31 662 4358
`
`ABSTRACT
`
`This paper introduces techniques for the electronic variation & automati¬
`cal control of exposure in solid-state image sensors. A novel electronic aperture
`and a simple exposure controller are described, which have been integrated on
`chip with a CMOS image sensor array. The technique can achieve a wide range
`exposure control of 40,000:1. This is equivalent to 15 stops of a mechanical iris
`system.
`
`1. INTRODUCTION
`Automatic exposure control is of significance to avoid mechanical iris control in electronic
`vision systems. This paper discusses two related issues: 1) how to electronically settle exposures
`(electronic aperture), and 2) how to automatically control the exposure (electronic controller).
`This paper emphasizes on-chip design because recent development in CMOS sensor-processor
`technology has provided a new method for implementing control logic with the sensor array on
`the same substrate[l,2].
`An on-chip electronic aperture, equivalent to eight stops of a mechanical system has been
`reported before[3]. We present here a novel scheme, which enables exposure adjustment over a
`much wider range of 40,000:1, equivalent to 15 stops of a mechanical system. The electronic
`aperture is achieved by varying the integration period, thus varying the sensitivity of the photo
`array. The integration time is defined to be the sum of a variable number m of line intervals
`plus a variable number n of pixel clock intervals. Therefore, the maximum exposure setting is
`a field time, and the minimum setting is a few pixel clock periods.
`If we now alter the exposure setting in response to the monitored image, we can implement
`fully automatic electronic exposure control. A simple control mode is described in this paper,
`which costs approximately 1,000 gates to implement. The control is achieved by monitoring the
`image pixel stream and estimating the fractions of each picture which are very white and very
`black. On the basis of this information, the device decides whether the picture contrast is
`acceptable, or too bright, or too dark. If necessary, the exposure time is then changed in the
`appropriate direction.
`The techniques have been used in several designs, covering applications from single-chip
`CMOS video cameras[4], to single-chip "smart" vision systems such as burglar-alarm verification
`
`cameras.
`
`281
`
`1050-001
`
`
`
`(n xtclk<tline ,ra xtline<tfield). We
`
`2. ELECTRONIC APERTURE
`Algorithm The architecture of the MOS image sensor is shown in Figure 1. The light sensing
`area consists of a photodiode array. The photodiodes are pre-charged to a fixed bias voltage
`during a reset cycle and then isolated for a suitable exposure time. During this time incident
`light partially discharges the junction capacitances, through the generation of photo-current in
`the diodes. For each pixel, the exposure is determined by the pixel integration time i.e. the
`time between resetting and sampling. When using the normal scheme of scan registers this
`integration time has to be a fixed number of clock cycles (usually a field time).
`The novel scheme is to define the sample and reset signals in such a way that the time between
`them can be varied. The integration time tint is then defined to be the sum of a variable
`number m of line intervals plus a variable number n of clock intervals:
`tim = w x tline + n x tclk
`where tclk is the pixel clock period and tline is the line period,
`refer m *tline as coarse settlement, and n xtcUc define settlement.
`At a particular line time, row i
`is being sampled and then reset, rows
`-coarse settlement:
`i+l through i+m are integrating and all other rows are being reset during the line period, as
`shown in Figure 2.
`Fine settlement is achieved by varying the reset period. The reset time can
`-fine settlement:
`range between a few clock intervals and nearly one line time, resulting in extra integration time
`n x tclk added to the coarse settlement. The fine settlement becomes more important when
`m xthne is smaller. In fact, the exposure is dominated by the fine settlement when m equals 0.
`Circuitry The problem is then to generate and decode suitable signals in such a way as to
`enable the correct rows and columns of the array in sequence. The vertical scan register has
`been replaced by a scan register with decoding, as shown in Figure 3 and Figure 4. The
`single-bit data-stream has been augmented with other signals, such as scan, sample, and reset.
`3. ELECTRONIC EXPOSURE CONTROLLER
`Algorithm The image pixel stream is compared with two DC references to pick up "very white
`pixels" and "very black pixels". A "very black pixel" means its value is below a black reference
`and a "very white pixel" is above a white reference. These occurrences are counted, and a
`threshold number is set to judge the present exposure. If "very black pixel" number is greater
`than the threshold number and "very white pixel" number is less, the picture is thought to be
`too dark, and the exposure should be increased. On the other hand, if the "very white pixel"
`number is greater than the threshold number and the "very black pixel" number is less, the pic¬
`ture is thought too bright, and the exposure should be decreased. When the numbers are both
`greater or both less than the threshold, the exposure is thought to be acceptable. No action will
`needed in this case.
`The new integration time is calculated according to the following formula:
`Tnew = Tpre(l±step)
`, if exposure is increased/decreased;
`= Tpre
`, if no action.
`where, Tnew is new integration time for the next frame, Tpre is the present integration time.
`Circuitry
`Figure 5 shows the block diagram of a simple exposure controller, which costs
`approximately 1,000 gates.
`-comparator: The video stream is fed into this block. Two DC voltage references are set to
`identify "very white pixels" and "very black pixels".
`
`282
`
`1050-002
`
`
`
`-judge: This block judges the present exposure by counting numbers of "very black" and "very
`white" pixels according to the above algorithm.
`-calculator: The new integration time is calculated here according to the above formula.
`-driving block: This block produces the driving signals, according to Tnew, needed by elec¬
`tronic aperture, such as scan, reset and sample.
`
`4. APPLICATIONS
`We report here two working cameras, as examples of on-chip automatic exposure control tech¬
`niques.
`The first example is a single-chip CMOS video camera[4]. It has a 312x287 pixel sensor array,
`together with the necessary sensing, addressing, and amplifying circuitry. The chip has a 2,000
`gate logic processor. Half of these gates generate synchronization timing to format a standard
`composite video output. The other half of the gates are the exposure controller. The chip
`measures 7.58x7.56mm, using 1.5 jim, 2 level metal CMOS technology. The exposure con¬
`troller and decode circuitry occupies 10% of the area.
`Our second example is a low-resolution camera for use in security applications. The chip has a
`smaller sensor array (156x100) but more control functions, measures 5.57x4.00mm in the
`same technology as the first example. The exposure controller and decode circuitry in this case
`occupies 20% of the area.
`Both camera modules include the camera chip with an attached miniature lens, a 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. Satisfac¬
`tory exposure control performance has been achieved for both. The exposure range is 40,000:1
`and quality of the pictures are good across the entire range. The automatic adjustment is fairly
`smooth and fast.
`
`5. CONCLUSIONS
`We have developed a novel electronic aperture and a simple exposure controller, which can be
`integrated with image sensor to form single-chip vision systems. Comparing with today's solid-
`state cameras, the control range is much wider, and the cost, power consumption and size are
`dramatically reduced.
`
`6. ACKNOWLEDGEMENTS
`We acknowledge support received from the Science and Engineering Research Council (Grant
`GR/F 36538 IED2/1/1159).
`
`7. REFERENCES
`[1]
`D. Renshaw, et. al., "ASIC Vision", Proc. IEEE Custom Integrated Circuits Conference,
`1990, pp 3038-3041.
`D. Renshaw, et. al., "ASIC Image Sensors", Proc. IEEE International Symposium on Cir¬
`cuits and Systems, 1990, pp 7.3.1-7.3.4.
`A. Asano, "Solid Sensors Continue to Improve Their Image", Journal of Electronic
`Engineering, 25 (1988) Nov., Tokyo, pp 64-67.
`G. Wang, et. al., "CMOS Video Cameras", Euro ASIC 91, Paris.
`
`[2]
`
`[3]
`
`[4]
`
`283
`
`1050-003
`
`
`
`1
`
`i-1
`i
`i+1.
`
`i+ro-
`i+m+1-
`
`N -
`
`reset
`
`sample
`and reset
`
`integrate
`
`reset
`
`c
`*<0
`
`a-.
`
`1Ot
`
`r >
`
`o/p
`amp&
`buffer
`
`horizontal addressing
`
`column sense amps
`
`*F
`#:# light
`
`sensing
`area
`
`Figure 1. Architecture of the image sensor
`
`Figure 2. Function at each row
`
`sample
`
`Xi-i
`
`reset
`
`Figure 4 . Decode circuitry
`
`sample reset
`decoder cell
`tL ~l
`ti
`
`V(l)
`
`V(¡).
`
`Ù.
`
`V(N>»H
`
`^ U
`
`^VN
`
`vertical shift
`scan register cell
`
`Vik.
`
`Figure 3 . Vertical shift register with docoding
`ii references
`
`VWP
`
`inc/dec
`
`comparator
`
`analogue
`output
`
`photoarray
`
`VBP
`
`word
`lines
`
`judge
`
`electronic
`aperture
`
`±_±
`
`calculator
`
`integration
`time
`
`4_t
`
`driving block
`
`enable
`
`sample
`scan
`
`284
`
`reset
`Figure 5. Block diagram of electronic exposure controller
`
`1050-004