`
`Analog-to-Digital Conversion Architecture
`
`A. Dickinson, 5. Mendis, D. Inglis, K. Azadet
`
`(AT&T Bell Labs)
`
`E. Fossum
`
`(Jet Propulsion Laboratory)
`
`This paper presents two implementations of a CMOS digital camera, and a low-
`
`power comparator circuit suitable for focal-plane applications.
`
`The CMOS digital camera consists of a CMOS active pixel iinage sensor (APS)
`
`integrated with an array of single-slope analog-to-digital converters (ADCs). Single
`
`slope analog-to-digital (A/ D) conversion was selected as it
`
`is simple and easy to
`
`implement on the focal plane since the requirements on the circuit components are not
`
`severe and the real estate requirement is moderate for low-resolution application. The
`
`digital camera employs a column parallel architecture where each column of pixels
`
`shares a single readout circuit and ADC. The irnage sensor consists of an array of
`
`CMOS APS pixels with row and column decoders and clock generator circuits. Both
`
`row and column decoders were designed to give full random access of the imager array
`
`to easily implement electronic pan/zoom functions. A clamping and sample/ hold
`
`circuit in the readout signal chain reduces fixed pattern noise (FPN) and l<TC noise from
`
`the pixel prior to A/ D conversion. Each ADC consists of a comparator, and a set of
`
`output latches and control logic circuit formed using standard CMOS logic. All the
`
`ADCS share a single off-chip ramp generator and counter circuit that provides the
`
`reference signal.
`
`The digital camera was first demonstrated as a 32 x 31 test array and was
`
`fabricated using a double—poly, double-metal 2 pm n-well CMOS process. The ADC and
`
`the image sensor were designed to achieve 30 Hz frame rate operation of a 128 x 128
`
`element array. The pixel size was 40 um x 40 pm with a 26% fill-factor, and the ADC
`
`channel was 40 p.rn x 2.5 mm. The design was also expanded to a QCIF array (176 x 144
`
`00°‘
`
`Magna 201 l
`
`TRW V. Magna
`IPR20l 5-00436
`
`Magna 2011
`TRW v. Magna
`IPR2015-00436
`
`0001
`
`
`
`pixels) and fabricated using AT&T's 0.9 pm linear CMOS technology which is a double
`
`poly, double metal n-well process. Poly-poly coupling capacitors in the readout circuit
`
`were the only deviation from a standard digital---CMOS process. The pixel size was
`
`reduced to 20 um x 20 um while preserving a fill-factor of 27%. The ADC channel was
`
`also scaled down to 20 um x 1.2 mm.
`
`Both implementations were operated with a single 5V supply voltage and one
`
`additional d.c. voltage. Digital output with 8-bit resolution was achieved at video rate
`
`(30 Hz frame rate). The capability of achieving 10-bit resolution with the same design
`
`was also demonstrated. It was seen that on—chip A/ D conversion actually reduces
`
`power dissipation on the focal plane by eliminating the large transistors needed to drive
`
`an analog output signal off the focal plane. Low-power operation was achieved with a
`
`supply voltage of 3.5V with power dissipation of 2 mW and 35 mW in the test array and
`
`QCIF array respectively. Column-wise FPN (60 mV) due to comparator offsets and
`
`capacitance mismatches was much larger than pixel-to-pixel variations. Temporal noise
`of the image sensor and ADCs was measured to be approximately 1 mV (r.m.s.). The
`
`ADC was characterized using a test circuit on the test array. Both differential and
`
`integral non-linearities were less than 1 / 2 LSB.
`
`The low power, low complexity and low FPN requirements for the comparator
`
`used in the parallel si.ngle~slope ADC motivated a study for improvement of the
`
`comparator structure. The new comparator uses a current-mode approach for the
`storage of one input and reuse of the same transistors for both analog inputs (ramp and
`
`signal), thus avoiding the problem of transistor mismatch encountered in differential
`
`structures. One other major advantage of this new structure is the ability to operate in a
`
`continuous-time mode (no switching) using the fact that one of the inputs (input signal
`
`related to a pixel value) is constant during conversion. The circuit has been fabricated in
`
`a digital 0.9 pm CMOS process and is functional with a power supply down to 1.5V.
`
`The measured offset is 1 mV. The comparator fits into 20 pm for pitch matching with
`
`the pixel array (actual size: 20 mm x 200 pm).
`
`9002
`
`0002
`
`
`
`Digital Camera Architecture
`
`
`
`
`
`
`
`
`ROW IIIIIIIIIIIII
`Decoder
`
`
`
`
`
`
`
`
`
`
`Schematic of Pixel
`
`Schematic of Single-slope ADC
`
`V25
`
`
`
`1”?“
`
`Strohed
`Comparator
`
`0003
`
`0003
`
`
`
`Image from Digital Camera DIGCAM1
`
`Summary of Results
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`echnology
`
`SSADC2
`
`2 pm I1-Well,
`
`
`
`DIGCAM1
`
`0.9 pm n-we-11,
`, double-metal
`
`:-1
`mager array
`32 >< 31
`
`
`
`
`
`
`
`
`“""’”‘°““*
`FPN
`
`60 mV (4% sat.)
`
`
`
`
`
`
`
`
`
`
`
`—_ ”
`
`’ Measured on test ADC in the last coluzxm of the test array.
`
`0004
`
`0004