Image sensor with dummy pixel or dummy pixel array

[0002] This application is related to the following copending U.S. patent applications and patents; U.S. Pat. No. 5,898,168, entitled IMAGE SENSOR PIXEL CIRCUIT; Ser. No. 08/876,694, entitled IMAGE SENSOR WITH DIRECT DIGITAL CORRELATED DOUBLE SAMPLING; and U.S. Pat. No. 5,877,715, entitled CORRELATED DOUBLE SAMPLING WITH UP/DOWN COUNTER, all of which are by the same inventors as this application, filed on the same day as this application, assigned to the assignee herein and expressly incorporated herein by reference.

[0003] This invention relates generally to image sensors. More specifically, the present invention relates to a complementary metal oxide semiconductor (CMOS) image sensor with a dummy pixel or dummy pixel array to enhance image data accuracy.

[0004] CMOS imager sensors are becoming more competitive to charge coupled device (CCD) imagers in terms of low cost and easy fabrication with current CMOS processes. These image sensors may be used as the input device in a wide variety of camera systems. One advantage of CMOS image sensors (or imagers) is that signal processing circuits can be readily integrated on the same chip as the imager, thus enabling the design of smart, single-chip camera systems. CMOS imagers are contemplated for use with computers and communication devices to realize low cost videoconferencing. Other applications include robotics, machine vision and automotive electronics. CMOS imager cells (or pixels) store light information as a photocharge on a photodiode or photogate, and then an electrical signal corresponding to the photocharge is read out through a bit line, typically a column data line. Unlike DRAM cells carrying digital data, the imager cell should carry a high resolution analog signal for subsequent conversion to a digital code as high as an 8-bit or 10-bit code with today's technology. The bit lines have a large capacitance as compared to that of the storage device in the cell. Consequently, the storage cells are usually active cells, which means that the output signal of the cells are amplified or buffered to the bit-line in order to be unaffected by the bit-line capacitance. However, non-linearities in the amplification or buffering leads to errors in the image data.

[0005] FIG. 1 schematically illustrates one example of prior art active-pixel CMOS imager circuitry. Imager 10 is an image sensor as disclosed in an article by Mendis et al., entitled “A 128×128 CMOS Active Pixel Image Sensor for Highly Integrated Imaging Systems”, IEEE Electron Devices Meeting, p. 583, 1993. A MOS photogate 6 is employed as the light sensitive element in each pixel (imager cell) 18. The transistors within each pixel 18 are typically NMOSFETS. The overall imager 10 is considered a CMOS imager since CMOS electronics are used in conjunction with the imager cells. For example, the transistors 3 within readout circuit 31 are typically PMOSFETS. Imager cells 18 are arranged in rows and columns to form an array, with typically only one row being activated for image data readout at any time. Within each cell 18 photo-charge “q” collected under photogate transistor 6 is transferred through a dc-biased transfer gate transistor 8 to a floating diffusion diode 7 formed beneath substrate surface 9. This floating diode 7 is periodically dc-restored by the application of a logic high RESET pulse on the RES bus to the gate of reset FET 11, thus resetting the potential of diode 7 (i.e., at circuit node 17) to the power supply voltage (VDD). Following each diode reset cycle, the photo-charge is transferred to floating diode 7. The voltage on diode 7 then corresponds to the intensity of light incident upon the associated imager cell 18. This voltage sets the potential of the gate of source follower FET 13, which amplifies or buffers the voltage appearing at its gate terminal for subsequent readout. When row select transistor 12 is turned ON by a VROW pulse on row select line RSL, the voltage at circuit node 17 is detected by readout circuit 31 detecting corresponding voltage on column bus 15.

[0006] To remove reset noise of reset switch 11, a correlated double sampling (CDS) operation is performed. This involves subtracting the reset level measured on column bus 15 from the signal level on the column bus, where the reset level is obtained just prior to the transfer of photo-charge to diode 7. As such, any offset errors due to transistor mismatches are canceled since both levels are measured at the same circuit point. In the embodiment of Mendis et al., the reset level and the signal level are stored on two separate sample and hold capacitors CR and CS, via two separate switches S1 and S2, respectively. Two identical readout circuit portions are required, one for the reset level, and one for the signal level. Further differential amplification at the multiplexed column output is required to complete the reset noise removal operation.

[0007] As mentioned above, the source follower FET 13 amplifies or buffers the voltage level on circuit node 17 such that the image data will be unperturbed by the column data bus capacitance. A problem with this approach is that the source follower is non-ideal, such that the transfer characteristic between the pixel voltage at node 17 and the output voltage on column bus 15 is non-linear. This non-linearity stems from the relationship between threshold voltage and gate voltage. If the source follower is an NMOS fabricated by conventional CMOS processes, the backgate is typically not tied to the source, resulting in a threshold voltage that changes with input (gate) voltage. The larger the source voltage is relative to the substrate voltage, the larger is the threshold voltage, leading to the nonlinear characteristic. Thus, a higher gate voltage produces a higher threshold voltage, resulting in an output signal that is attenuated relative to the ideal case. A similar nonlinearity occurs if the source follower is a PMOS. In either case, the nonlinear transfer characteristic produces a conversion error from the light source to the output image data.

[0008] Electronic shuttering capability can be incorporated within CMOS image sensors by the addition of a shutter transistor within each pixel. When this shutter transistor is pulsed on, the photocharge stored by the photosensitive element is transferred to a storage node within the pixel. The storage charge is then read out sequentially row by row. A problem with this approach, however, is that leakage current at the storage node can severely degrade the signal integrity thereat. As the number of pixels grows to realize high resolution pictures, the time required to scan all the pixels also grows. Hence, the total amount of charge lost by leakage becomes significant, thereby degrading the overall picture quality. In addition, since the charge is read sequentially row by row, the lost charge in the pixels are all different according to their row order. One common solution to this problem is to reduce the leakage current; however, leakage current reduction is difficult to implement. It usually involves expensive manufacturing equipment to minimize leakage current at the p-n junction (source or drain to substrate) of the shutter transistor.

[0009] Accordingly, there is a need to overcome the deficiencies of image quality degradation due to threshold voltage variation in CMOS image sensors, and due to leakage current in image sensors with electronic shuttering.

[0010] The present invention relates to an image sensor including one or more dummy pixels that produce a reference signal which is used to compensate for errors within the devices of the main pixel cells. In a first embodiment, at least one dummy pixel is used in conjunction with other circuitry to correct for nonlinearities in the transfer characteristic of a source follower transistor within each pixel. The dummy pixel may be used in conjunction with A/D converters connected directly to column data lines of the main pixel array. A plurality of comparators and up/down counters may be used to perform directed correlated double sampling with threshold voltage compensation. In a second embodiment, an array of dummy pixels is used to correct for leakage current within the pixels which is present during an electronic shutter mode of operation. The dummy pixel array may be used in conjunction with a dummy pixel compensating for threshold voltage variation, whereby errors due to both leakage current and threshold voltage mismatch are corrected.

[0011] Preferred embodiments of the present invention are described herein with reference to the drawings, in which like reference numerals identify similar or identical components throughout the several figures, wherein:

[0012] FIG. 1 schematically illustrates a prior art image sensor and pixel circuit;

[0013] FIG. 2 is a schematic block diagram of a first embodiment of an image sensor in accordance with the present invention;

[0014] FIG. 3 illustrates an imager cell that may be used in the image sensor of FIG. 2;

[0015] FIG. 4 is a graph showing voltage transfer characteristics of a source follower within a dummy pixel or main pixel;

[0016] FIGS. 5 and 6 are timing diagrams of waveforms applied to the image sensor cells of FIG. 3;

[0017] FIG. 7 schematically illustrates an alternative embodiment of the present invention employing up/down counters for correlated double sampling;

[0018] FIG. 8 schematically illustrates another embodiment of the present invention with electronic shuttering capability;

[0019] FIG. 9 illustrates pixel circuitry that may be used within the pixels of the image sensor of FIG. 8;

[0020] FIGS. 10A-10C are timing diagrams for the image sensor of FIG. 8;

[0021] FIG. 11 illustrates another embodiment of the present invention which compensates for leakage current as well as threshold voltage nonlinearities; and

[0022] FIG. 12 shows a comparison or reference voltage waveforms with an without leakage current compensation.

[0023] FIG. 2 schematically illustrates an exemplary embodiment of an image sensing device in accordance with the present invention, designated as 20. This embodiment employs a single dummy pixel DP in conjunction with other circuitry to compensate for a nonlinear transfer characteristic of a source follower transistor within each imager cell, thus improving image data quality. Device 20 includes a plurality of imager cells (pixels) 30 arranged in M rows R₁to R_Mby N columns C₁to C_N. Several differences are apparent between image sensor 20 and the prior art image sensor discussed above. The prior art analog readout circuits 31 at the bottom of each column have been eliminated and replaced by a set of comparators 40₁-40_Nand registers 42₁-42_N. These comparators and registers operate to provide a direct analog to digital conversion function in conjunction with a digital code generator/controller 50, a digital to analog converter (DAC) 52 and dummy pixel DP. Dummy pixel DP is included to substantially reduce image data errors due to the nonlinearity of the source followers within each imager cell 30, as will be explained below. The overall circuit arrangement of sensor 20 advantageously eliminates the prior art analog readout circuits, which are susceptible to noise and gain errors that reduce the accuracy of the image data. In addition, the prior art capacitor CR used to store the reset sample must have a large area to minimize noise. Image sensor 20 excludes these large capacitors CR, as well as the signal capacitors CS, providing room for other electronics and/or reducing the overall chip size.

[0024] In operation, timing and control circuit 14 activates one row at a time for image data readout. Typically, the image data from all imager cells 30 in the activated row Ri are read out simultaneously on the column data lines 15₁-15_N. The rows are selectively activated by proper application of ROW SELECT signals on lines RSL₁-RSL_N, signals on lines RES₁-RES_M, and supply voltage pulses V_Ron lines 34₁-34_M. During each row activation time interval, digital code generator/controller 50 (hereafter, controller 50) generates a ramped digital output code D1 representing an ideal reference voltage V_REF,IDEAL. The digital output code, which may be an 8-bit or 10-bit code, for example, is applied to DAC 52 and to all registers 42₁-42_N. DAC 52 draws current I_DACat a current level in accordance with digital code D1. The output current line of DAC 52 connects directly to a reference circuit node within dummy pixel DP, such that the potential of that reference node is set to the ideal reference voltage V_REF,IDEAL. This voltage is applied to the input of a source follower FET within the dummy pixel whose size and characteristics are the same as in each pixel 30. As a result, dummy pixel DP produces a nonlinear output voltage V_REFon line 53, which is applied to the inverting terminals of each comparator 40₁-40_N. Column data lines 15₁-15_Nconnect to the noninverting terminals of comparators 40₁-40_N, respectively, whereby each comparator such as 40icompares the corresponding image data output signal V_OUTiwith V_REF. When the output logic of comparator 40iflips state, indicating that V_OUTiis about equal to V_REFat that instant in time, the comparator output logic enables the associated register 42_i, whereby the digital value of D1 at that instant of time is latched in register 42_i. The register then transfers the stored digital code representing the image data of the activated pixel 30 to a register within N:1 column select/scan logic block 44 for subsequent processing.

[0025] With reference now to FIG. 3, there is shown exemplary pixel circuitry for any imager cell 30 in the “ith” row and “jth” column of imager 20. Exemplary circuitry for the dummy pixel DP is also presented. The circuitry within dummy pixel DP is chosen in correspondence with that of cells 30 to enable the voltage transfer characteristics to track that of each cell 30 and thus, to correct for the nonlinearities of the source followers within the cells. Imager cell 30 of FIG. 3 is also disclosed in our commonly assigned, copending U.S. patent application Ser. No. 08/876,694, entitled IMAGE SENSOR PIXEL CIRCUIT, mentioned above (hereafter, the “H. Wong et al. I” application). Imager cell 30 is an improvement over the prior art cell of FIG. 1, in that the large row select MOSFET 12 is eliminated, thereby allowing more space for the photosensitive element, in this case, photodiode 26. In addition, less bus lines are required, which likewise creates more space for the photosensitive element. Only three bus lines—RESET line RES_i, VR line 34iand Row Select line RSL_i—are connected to the imager cells 30 of a common row R_i. Hence, imager 20 can exhibit improved resolution over the prior art by utilizing the extra space created in each pixel to provide a larger area for the photosensitive element and/or reduce the overall pixel size.

[0026] The pixel circuit of cell 30 eliminates the separate row selection transistor by employing FET 22 to perform both a charge transfer function and a pixel selection function. The row select line RSL_iconnects directly to the gate of FET 22. In addition, the pixel selection function is carried out in conjunction with reset transistor 21. As such, reset transistor 21 is functionally different from reset transistor 11 of FIG. 1, which, in the prior art, was used strictly for resetting the photosensitive element. In the prior art, the bus line RES connecting the gate of FET 11 is typically tied to multiple rows, whereby the cells of multiple rows are reset together. In the embodiments disclosed herein, each reset bus RES₁-RES_Mis dedicated for a single row and is not tied to multiple rows, thereby enabling the reset device 21 to perform a partial pixel selection function as well as a reset function.

[0027] FET 21 operates to reset the reference circuit node 25 to nearly the voltage level VR on bus line 34_iwhen the RESET signal on the RES_ibus is pulsed. Specifically, the node 25 voltage is reset to VR−V_DS21, where V_DS21, is typically about 0.4-0.5 volts. As will be explained more fully below, following the application of the RESET pulse, the voltage at reference node 25 is indicative of reset noise during a first sampling time interval in which charge transfer device 22 is OFF. During a second sampling interval in which device 22 is ON, the reference node 25 voltage is indicative of photocharge collected by photodiode 26, and hence, the intensity of light incident upon cell 30. The gate of FET 22 is tied to the RSL_iline to receive the ROW SELECT signal. In the prior art, the RSL line is tied to the separate MOSFET 12. With imager cell 30, the relatively small charge transfer FET 22 performs a partial pixel selection function formerly performed by the large MOSFET 11. Source follower FET 23 is used to amplify and buffer the voltage at reference node 25 to transfer an amplified version of the charge to column bus line 15j. Column load transistor or resistor L_jsinks output current on the column data line. At any given time, only one pixel 30 per column is active, as controlled by both the row select signal and the RES signal. As such, the voltage V_OUTjprovided on column bus 15jto comparator 40jis directly related to the voltage on reference node 25 of the active pixel.

[0028] Dummy pixel DP includes a source follower FET 23_DPof the same size and characteristics as the source follower FET 23 of the imager cells 30. As discussed above, D/A converter 52 draws current I_DACin accordance with the instantaneous digital word D1 outputted by controller 50. A resistor 19 is coupled between circuit node 25_DPand FET 21_DPof generally the same characteristics as FET 21 within each cell 30. The drains of FETS 21_DPand 23_DPare tied to the VR supply line 34_DP. A load resistor or transistor L_DPof the same resistance as any of column bus loads L₁-L_Nconnects to the source of FET 23_DPvia line 53 to sink current. The output current I_DACforces a voltage drop across resistor 19 to produce the ideal voltage V_REF,IDEALat circuit node 25_DP. Resistor 19 is selected such that, when codeword D1 is ramped between a minimum and maximum value during each row activation time interval, the range of voltage levels of V_REF,IDEALat node 25 will correspond to at least the maximum expected range within each imager cell 30. As a result, the output voltage V_REFwill be essentially the same as voltage V_OUTjon column bus 15jwhen active pixel 30jhas the same circuit node 25 voltage. It is noted that FET 21_DPis used to make the condition of the dummy pixel similar to that of the pixel 30. A RESET, DP voltage on line RES_DPand a VR, DP voltage on line 35_DP, both originating from timing and control logic 14, are both set high during each readout interval. As such, FET 21_DPis always ON and the voltage at node 25_DPcan be properly varied by varying I_DAC. FIG. 4 illustrates voltage transfer characteristics between the absolute gate voltage V_G(voltage at circuit node 25 or 25_DP) and the source voltage of source follower FET 23, i.e., V_OUTjor V_REF, for several conditions. The range of V_Gis from V_MINto V_MAXwhich is the selected range of V_REF,IDEALduring image sensor operation. Curve 63 would be the output if V_OUTperfectly tracked V_IN, i.e., if the threshold voltage V_Tof FET 23 were always zero. Curve 62 represents an ideal linear output which would result if V_Tremained constant with varying V_G. Curve 61 illustrates the nonlinear transfer characteristic of a typical source follower FET with a back gate effect. Without dummy pixel DP to compensate for the nonlinearity, the image data readout for darker light conditions (higher V_G) would be an attenuated representation of the actual condition.

[0029] Referring again to FIG. 3, during each row activation interval, D1 ramps from a maximum to a minimum, causing V_REF,IDEALto ramp from V_MAXto V_MIN. During this time, when V_OUTjis less than V_REF, comparator 40joutputs a logic low to register 42j,which does not latch the register. As soon as V_OUTjexceeds V_REF, comparator 40joutputs a logic high, causing register 42jto latch, whereby the last digital code D1 inputted to register 42jwill be stored. Each register as 42jreceives clock and control inputs from logic block 44 to control data storage and transfer. Once latched, the contents of register 42jare transferred to a register within logic block 44 upon a subsequent command from logic block 44, whereupon register 42jis reset for the next data readout.

[0030] It is noted that image sensor 20 of FIG. 2 may alternatively utilize other pixel circuit configurations for the imager cells 30 used in conjunction with dummy pixel DP and the other image sensor electronics. The present invention is not limited to an image sensor using the imager cell of FIG. 3. In essence, for image sensors using correlated double sampling to reduce noise, any pixel circuit which allows the separate sampling of the reset and signal samples can be used within the image sensor 20 of the present invention. For example, any of the other pixel circuit configurations disclosed in the H. Wong et al. I patent application cited above may be used. That patent discloses various modifications to the pixel circuit of FIG. 3 herein, such as: employing a phototransistor in place of photodiode 26; including an anti-blooming transistor within each cell; and including a shutter transistor within each cell to perform electronic shuttering. Each of these pixel circuits avoids the use of a separate row select transistor 12 as in FIG. 1—as such, the dummy pixel circuit shown in FIG. 4 can be used for any of these cases. If a separate row transistor is employed, as in the Mendis et al. pixel circuit described above, the dummy pixel can still be used, but should be modified to include an additional transistor following FET 23 which has the same characteristics of row selection transistor 12 in order to more closely simulate the voltage transfer characteristics of the main pixels.

[0031] Operation of image sensor 20 of FIG. 2 (with imager cells 30 of FIG. 3) will now be described in conjunction with the timing diagram of FIG. 5. The solid lines of the VR, RESET and ROW SELECT waveforms in the diagram represent the waveforms applied to the active pixel 30 of column C_j. The dotted lines 35-37 represent the respective waveforms applied to inactive pixels 30 in that column (i.e., inactive during the time interval from time t₀to t₉). The ensuing discussion pertains to the active pixel. The time prior to time to represents a photocharge collection period for all the pixels in a row (e.g., for the “ith” row, R_i). In this period, VR is low, RESET is high, and ROW SELECT is low. “High” potential for the various waveforms discussed herein is typically 1.2, 1.8, 2.5, 3.3 or 5 volts; “low” is typically zero volts. Photocharge is collected in the photodiode during the photocharge collection period and the potential of photodiode 26 (i.e., the potential of the cathode well relative to the substrate potential) is lowered corresponding to the amount of photocharge collected. Since ROW SELECT is low, FET 22 is OFF and the photocharge does not get transferred out during this time. Also, RESET is high, turning FET 21 ON, whereby reference node 25 is set at nearly VR, which is low. As such, the gate of FET 23 is low, turning FET 23 OFF, such that no voltage from that pixel is put on column bus 15_j. Meanwhile, VR is high for the pixels 30 in the preceding row R_i−1as indicated by waveform portion 35_i−l, to permit data readout from the pixels in that row.

[0032] At time t₀, VR applied to the pixels 30 of row R_iis raised to begin a base reset/read period. When VR is high, reference node 25 is set to a high potential since FET 21 is still ON. The RESET signal is then brought low at time t₁, e.g., five microseconds after time t₀, isolating reference node 25 from the VR bus 34_i. Due to parasitic gate to source capacitance of the reset FET 21, noise on the RES (clock) line feeds through to reference node 25. As such, this “clock feedthrough” due to RESET going low is also stored in circuit node 25. When RESET is brought low at time t₁, the voltage V_OUTjon column bus 15_jis directly related to the reference node 25 voltage due to the gate to source voltage drop of the source follower FET 23. After an appropriate “column settle” time between t₁and t₂, V_REF,IDEALis caused to linearly ramp from a maximum value V_MAXto a first minimum value V_MIN1by code generator 50 sequentially changing the digital code, i.e., sequentially decreasing the code count. (If DAC 52 operates in an “inverse” mode, an increasing output count from code generator 50 will produce the negatively sloping ramp voltage V_REF,IDEAL). The voltage range of V_MIN1to V_MAXis set based on the expected extremes of the voltage range at reference node 25 within the pixels 30 during the reset period. This range is preferably established based on experimental observation of the voltage for typical pixels. It is noted that, in the time prior to t₂, the value of V_REF,IDEALis immaterial since no sampling is being performed. As V_REF,IDEALis ramped down between t₂and t₃, V_OUTjis sampled by means of comparator 40_jcomparing V_REFproduced by the dummy pixel to V_OUTjand changing its logic output as soon as V_OUTjexceeds V_REF. The latest count of D1 is then latched within register 42_jas explained above, and the register contents are transferred to logic block 44 prior to the next sampling interval (i.e., prior to time t₆). The transferred codeword corresponds to the reset level value for that pixel for a subsequent correlated double sampling computation to remove the reset noise from the image data by subtracting the reset level from the signal level. The reset level data within all registers 42₁-42_Nmay be transferred simultaneously to logic block 44 upon appropriate command. The reset level data accurately corresponds to the reset levels at the reference nodes 25 of the active pixels, due to the compensation for threshold voltage mismatch.

[0033] It is noted that sampling is avoided during the column settle interval (t₁to t₂) to avoid sampling transient noise associated with the reset FET 21 being rapidly turned off. The column settle interval t₁-t₂may be about one microsecond in duration.

[0034] During the time interval immediately following time t₁, it is important that no voltage is put on the column bus by the pixels in the other rows, which would otherwise diminish accuracy of the reset noise and signal data readout from the active pixels. As such, the VR, RESET and ROW SELECT signals applied to the inactive pixels are low, high and low, respectively, as indicated by respective waveform portions 35, 36 and 37. The source follower FET 23 of the inactive pixels will then be off during this time. As such, the output voltage provided on each column bus 15₁-15_Ncorresponds only to the reference node 25 voltage of the active pixel 30 in that column.

[0035] A charge transfer interval for the active pixels commences at time t₄by raising the ROW SELECT signal applied to row R_i. This turns on the charge transfer FET 22 of the pixels 30 in row R_i, thereby allowing the charge collected in photodiode 26 to be shared among the photodiode capacitance and the capacitance at reference node 25 (or to be completely transferred to the reference node capacitance, depending on the photodiode design). The potential of reference node 25 will fall corresponding to the amount of electrons collected in photodiode 26. V_OUTjwill fall corresponding to the reference node 25 potential. The charge transfer interval is typically on the order of two microseconds. After a second column settle interval between time t₅and t₆(e.g., one microsecond) a signal sampling interval begins during which V_REF,IDEALis ramped down from V_MAXat time t₆to V_MIN2at time t_6a. (V_REF,IDEALwas previously raised from V_MIN1to V_MAXsometime between time t₃and t₆via appropriate command from logic block 14 or 44 to code generator 50). The value of V_OUTjduring this interval is the “signal” value (including noise) corresponding to the node 25 voltage, which corresponds to the amount of light incident upon the active pixel 30. As such, V_MIN2is selected to be below the minimum potential of reference node 25 expected during this interval. The minimum potential expected essentially corresponds to the brightest light condition, since the higher the photocharge collected, the lower the voltage at reference node 25. V_MIN2is preferably selected based upon experimental observations for the pixel 30 design. As soon as V_REFproduced by the dummy pixel is below V_OUTjduring the signal sampling interval, the output of comparator 40_jflips logic state, causing register 42_jto be latched to the latest count D1 (second codeword). After time t_6a, the register contents of all registers 42₁-42_Nare transferred to registers within logic block 44. The reset value corresponding to the first codeword is then subtracted from the signal value corresponding to the second codeword to obtain an image datum with the reset noise and threshold voltage mismatch removed. This completes the correlated double sampling operation. The subtraction can be performed for all columns in parallel, in either a bit serial or bit parallel fashion. Alternatively, the subtraction can be performed sequentially. Column select/scan logic 44 outputs the final image data corresponding to each pixel 30 in the array to image storage and processing electronics.

[0036] After the signal read interval at time t₇, photodiode 26 needs to be reset to a high potential before the next cycle of photocharge collection can begin. This is accomplished by raising the RESET signal at time t₇high, while keeping the ROW SELECT and VR bus signals high. As such, reference node 25 will be high, and the photodiode is reset to a high potential since FET 22 is ON. Following this operation, ROW SELECT is brought low again at time t₈to turn FET 22 OFF and isolate reference node 25 from photodiode 26, which begins another cycle of photocharge collection. RESET remains high and VR goes low at time t₉, thus inactivating source follower FET 23 and allowing FET 23 of other pixel rows to put signals on column bus 15_j. This is illustrated by waveform portions 35_i+1and 36_i+1representing the respective waveforms applied to the next row R_i+1, and waveform portion 38 of the V_REF,IDEALsignal.

[0037] If a pinned photodiode is used for the photosensitive element, the photocharge collected in the pinned photodiode can be completely transferred to reference node 25 without charge sharing. In the pinned photodiode case, the photodiode reset period is not necessary because the signal read results in complete charge transfer and the photodiode is reset as it is read.

[0038] FIG. 6 illustrates a timing diagram of alternative waveforms for the RESET and ROW SELECT signals. Image sensor 20 can operate with either one or both of these waveforms applied as alternatives to those presented in FIG. 5. Referring first to the alternate RESET waveform, prior to onset of the reset period at time t₀, the RESET signal is brought low at time t_abefore raising VR. This approach reduces feedthrough from the RES line to the reference node 25 due to the gate to source capacitance of reset FET 21. In addition, after the photodiode reset period, the RESET signal is brought low at time t₈before VR goes low at time t₉. After VR goes low, the RESET signal has to be brought high again to hold reference node 25 to the low potential at VR so as to isolate the readout FET 23 from the column bus 15 and allowing the readout FETs of other pixel rows to put signals on the column bus.

[0039] Referring now to the alternate ROW SELECT waveform of FIG. 6, after ROW SELECT is brought high at time t₄to transfer the photocharge to reference node 25, ROW SELECT is brought low at time t5 before the second column settle interval (t₅-t₆). ROW SELECT then remains low during the second sampling interval. This technique allows the clock coupling due to the row select FET 22 to be canceled before signal read. Such clock coupling would otherwise be present due to the parasitic gate to source capacitance of FET 22. The ROW SELECT signal is then brought high again during the photodiode reset period to allow the photodiode to be reset through FET 22 and FET 21.

[0040] With reference now to FIG. 7, another embodiment of the present invention, image sensor 200, is schematically illustrated. Image sensor 200 is essentially the same as image sensor 20 of FIG. 2 discussed above, except that the registers 42₁-42_Nhave been replaced by up/down counters 201₁-201_N, respectively. Unidirectional counter 210 is basically the same or similar to digital code generator/controller 50 of FIG. 2. The technique of using an up/down counter in a correlated double sampling operation, and in particular, with a CMOS image sensor, is described in detail in co-pending U.S. patent application entitled CORRELATED DOUBLE SAMPLING WITH UP/DOWN COUNTER, mentioned above and incorporated by reference. Employing an up/down counter in a CDS operation is advantageous in that the subtraction of the reset sample from the signal sample is performed automatically without the need to separately store both samples in a register and transfer them out separately. Also, the processing circuitry is relieved from the task of performing the subtraction.

[0041] Image sensor 200 combines the advantage of threshold voltage compensation with that of automatic CDS subtraction by the up/down counters. In operation, the waveforms of FIG. 5 or 6 are essentially the same as described above. Briefly, referring to FIG. 5, after the base reset and column settle intervals for the active row (times t₀-t₂) the reset level is sampled by means of unidirectional counter 210 starting to count at time t₂. Counting is initiated by means of timing and logic block 44′ applying a logic high, COUNT/{overscore (RESET)} (C/{overscore (R)}) signal to counter 210 at time t₂. Also at time t₂, each up/down counter 201₁-201_Nstarts counting down from a predetermined reference count (e.g., all zeroes) under the command of a COUNT DOWN signal provided by logic block 44′. As unidirectional counter 210 counts, DAC 52 converts the output count to I_DAC, thus producing V_REF,IDEALwhereby dummy pixel DP produces VREF including the effect of threshold voltage, in the same manner as described above. Thus, V_REF,IDEALis ramped down from V_MAXto V_MIN1as in the case above. During the reset sampling interval t₂-t₃, as soon as V_OUTjof column C_jexceeds V_REF, the logic level output of the associated comparator 40_jflips state, thereby commanding up/down counter 40_jto stop counting. As such, the stopped count corresponds to the reset level for the active pixel in that column with the source follower threshold voltage mismatch compensated for.

[0042] The unidirectional counter is reset to zero at time t₃or shortly thereafter (e.g., at time t₄in FIG. 5) whereas each up/down counter 201₁-201_Nis not reset and remains latched at the last count. Following the charge transfer and column settle intervals (times t₄-t₆) the signal sampling interval commences at time t₆by causing counter 210 to begin counting again via application of a C/{overscore (R)} pulse by logic block 44′. Also, at time t₆, logic block 44′ flips the logic state of a COUNT DOWN signal applied to each up/down counter 201₁-201_Nto command each counter to commence counting upward from the previous count obtained in the reset sampling interval. Then, during the signal sampling interval, as soon as V_OUTjexceeds V_REF, the logic level output of comparator 40_jflips state again, commanding the associated up/down counter 201_jto stop counting. At this point, the count of up/down counter 201_jcorresponds to the signal level minus the reset level for the active pixel in column C_jand the correlated double sampling operation is complete, with the threshold voltage nonlinearity of the source follower FET compensated for. The output counts for all the up/down counters are then transferred to logic block 44′ following the signal sampling interval, whereupon all up/down counters and unidirectional counter 210 are reset to permit image data readout from the subsequent row.

[0043] FIG. 8 schematically illustrates another embodiment of the present invention, image sensor 70, which includes electronic shuttering capability. Image sensor 70 employs an array of dummy pixels DP1′-DPM′ to compensate for leakage current within pixels 80 of the main array. Such leakage current is common when an electronic shutter transistor is used within the pixels. Since image data is read out sequentially row by row, e.g., top to bottom, the pixels in the bottom rows exhibit more leakage current than those in the top rows. The dummy pixel array compensates for this disparity, thus improving the overall image quality. The dummy pixel array 71amay be physically located adjacent the first column C₁of the main array 71bas shown (or alternatively, adjacent the last column C_N) to form a M×(N+1) pixel array 71. This configuration facilitates connection to the same bus lines such as RES₁-RES_Nthat connect to pixels 80. In the embodiment of FIG. 8, A/D converters ¹¹⁰DP and 110₁-110_Nconnect to respective column data lines 15_DPand 15_i-15_Nto directly digitize the reset and signal data outputs on the column data lines. Registers 112_DPand 112_i-112_Ntemporarily store the digitized samples for subsequent transfer to registers in logic block 44 for subsequent processing.

[0044] Referring to FIG. 9 in conjunction with FIG. 8, an exemplary pixel 80_iand dummy pixel DPi in a common row Ri are shown. Pixel 80_idiffers from pixel 30 of FIG. 3 in that a shutter transistor 99 has been added between the charge transfer device 22 and the photodiode 26 to perform an electronic shuttering function. The shutter function enables a “still” image to be taken whenever an external switch is activated, similar to a mechanical shutter in a camera. For motion picture applications, e.g., camcorders, the shutter function allows for versatility in the shutter speed whereby higher quality video is achievable for fast moving subjects. The circuitry of dummy pixel DP_iis identical to that of pixel 80_i, except that the photodiode 26_DPis shielded against light such that no photocharge is developed due to light. As such, dummy pixel DP_ifunctions solely to generate an output signal VH_REF,iwhose value is related to the amount of leakage current within the dummy pixel. This output signal is used as a reference to compensate for leakage current in pixels 80_iof the same row. A SHUTTER signal is applied on a common bus line SH connected to the gates of all devices 99 in the entire M×N array 71b.The SHUTTER signal is also simultaneously applied to each device 99_DPof all the dummy pixels. After a manual or electronic shutter switch (not shown) is activated, the SHUTTER signal is pulsed high a first time to reset the photodiode. Then, after a predetermined time interval, the SHUTTER signal is pulsed again to transfer the photocharge collected to a storage node 101 at the drain of the shutter transistor 99. Storage node 101 stores voltage on capacitance C101 corresponding to the light intensity on the pixel 80i.This voltage is stored until the row select transistor 22 is pulsed to read out the image data. While the charge is being stored at storage node 101, leakage currents I1 and I2 flow to ground, thus reducing the potential at the storage node. Without the dummy pixel array to compensate for this leakage current, the total charge lost by leakage current is significant, whereby the overall picture quality would be degraded.

[0045] Operation of image sensor 70 will now be explained further with reference to the timing diagrams of FIGS. 10A-10C. When a manual or electronic shutter switch (not shown) is depressed, all the photodiodes are reset between time t₀′ and t₁′. This is accomplished by raising, at t₀′, the SHUTTER signal as well as all the ROW SELECT signals, ROW SELECT₁-ROW SELECT_M, all VR signals VR1-VRM and all RESET signals RESET₁-RESET_Mall rows R₁-R_M, respectively. Photocharge is then collected between times t₁′ to t₂′ to capture an image of the subject (still image or image frame of a moving picture). At time t₂′, SHUTTER is pulsed again, thereby transferring the collected photocharge to storage node 101 of each pixel 80. This transferred photocharge lowers the potential of storage node 101, which was previously at some reference voltage level due to capacitance C₁₀₁. In the dummy pixels, the storage node 101 essentially remains at that reference level since the photodiodes 26_DPare shielded from light. At time t₃′, SHUTTER is brought low again, isolating the photodiode 26 from storage node 101. Sequential row by row readout of the image data then begins. At time t_3a′, the VR pulse applied to row R1 (VR1) is raised, which is followed by RESET₁going low at time t_3b′ and ROW SELECT₁going low at time t₄′. As shown more clearly in FIG. 10C, a base reset occurs between times t_3a′-t_3b′. Following a column settle interval between times t_3b′-t_3c′, the reset level is sampled between times t_3c′-t_3d′ by the A/D converters 110_DPand 110₁-110_N. Each A/D converter such as 110_jprovides a digital output code representing the voltage V_OUTj(hereafter, V_OUTj,RES). This output code is used as the reset sample for a correlated double sampling operation. The digital output is stored in register 112_jfor subsequent transfer to logic block 44. A/D converter 110_DPdigitizes the dummy pixel reset level VH_REF,RESifor the active row R_iand subsequently transfers the digital code to register 112_DPfor ensuing transfer to logic block 44.

[0046] Next, ROW SELECT₁is brought high at time t₄′ to transfer the potential of storage node 101 to reference node 25 in the pixels 80 of row R₁. As such, the reference node 25 potential is lowered in correspondence with both the amount of light incident upon the pixel and the voltage drop at storage node 101 due to leakage current that occurred since time t₃′. The potential of the dummy pixel DP₁′ storage node 101_DPis likewise transferred to reference node 25_DP. In the dummy pixel, the potential at reference node 25_DPis lowered in accordance with the voltage drop due to the current leakage since time t₃′. Following a column settle interval between t_4a′-t_4b′, the output signal V_OUTon each column bus is sampled again, and each A/D converter such as 110_jprovides a second digital output code corresponding to the signal level for the active pixel in row R_i, i.e., V_OUTj,SIG. Likewise, A/D converter 110_DPprovides a digital output code representing V_HREF,SIGi, during the signal sample interval. After the digital codes are transferred to logic block 44, the CDS operation is completed via simple arithmetic operation with leakage current compensated for. The image datum IM_ijfor the pixel in row R_iand column C_iis computed as:

[0047] IM_ij=(V_OUTj,SIG−V_OUTj,RES)−(VH_REF,SIGi−VH_REF,RESi)  (1)

[0048] where (VH_REF,SIGi−VH_REF,RESi) represents the leakage current component of each image datum in row R_i. Accordingly, the correlated double sampled image datum for each active pixel has the leakage current component extracted out.

[0049] Following signal sample, RESET₁is brought high at time t₁′, ROW SELECT₁is brought low at time t₆′, and VR1 is brought low at time t₇′. Referring again to FIG. 10A, the image data of the next row R₂is then read out beginning at time t₈′ in the same manner as described above for row R₁. Sequential row by row readout continues to the last row R_M. Since the image data for each row is read out at a different time, the voltage drop in the image data due to leakage current is different for each row. This is illustrated in FIG. 10B, where is it seen that the dummy pixel output voltage VH_REF(i.e., VH_REF,SIG) on column line 15_DPdiminishes with row scan time. VH_REF,1is approximately the ideal value, VH_REF,IDEALat row scan time t₀₀=t₄′-t₃′. This drops to VH_REF,2at row scan time t₁₀=t₈′-t₃′ and eventually to VH_REF,Mat row scan time t₂₀=t_M′t₃′. However, the drop in VH_REFon column bus 15_DPat any given time is essentially the same voltage drop that occurs within the pixels of the activated row due to leakage current. Therefore, the change in image data due to leakage current is canceled by the subtraction operation in accordance with eqn. (1), which occurs in logic block 44 or the image processing circuitry.

[0050] Turning now to FIG. 11, yet another embodiment of the present invention, image sensor 300, is schematically illustrated. This embodiment includes the above-described dummy pixel array 71ato compensate for leakage current, in conjunction with the dummy pixel DP to compensate for threshold voltage nonlinearity. In addition, comparators 40₁-40_Nand up/down counters 201₁-201_Nare used to facilitate correlated double sampling as discussed above.

[0051] In operation, unidirectional counter 210 counts up or down in both the reset and signal sample intervals, as described above, whereby the current I_DACis produced in accordance with the count. This current produces the voltage V_REF,IDEALat the reference node 25_DPof dummy pixel DP. Without leakage current compensation, V_REF,IDEALwould be generated as the same waveform for every row R₁-R_M. The leakage current compensation circuitry modifies the voltage V_REF,IDEALby means of current I_LEAK,EQwhich is caused to flow in parallel with I_DAC. The current I_LEAK,EQresults in V_REF,IDEALbeing lowered by an amount ΔVH_REFi, as shown in FIG. 12, where ΔVH_REFiis proportional to the amount of leakage current in the pixels of row R_i. Thus, ΔVH_REFiis minimal for row R₁and the largest for row R_M. In FIG. 12, waveform 310 represents V_REF,IDEALwithout leakage current compensation, and waveform 320 is V_REF,IDEALwith exemplary leakage current compensation. The two waveforms are substantially the same during the reset sample interval since the reference (readout) nodes 25 and 25_DPin the main and dummy pixel arrays, respectively, were just reset. The leakage current only affects the reference node potential after the row select transistor transfers the charge towards it (following the charge transfer interval).

[0052] With continuing reference to FIG. 11, I_LEAK,EQmay be generated using the voltage to current conversion circuitry consisting of operational amplifier OP1, transistors M1 and M2, resistors R1 and R2 and dummy pixel DP1 which may be the same as dummy pixel DP. Only a single dummy pixel DP and a single dummy pixel DP1 are needed for the entire image sensor. Dummy pixel DP operates as was described above in reference to FIG. 2. Op amp OP1 has its negative input terminal connected to line 15_DPto receive VH_REFi. The positive input is connected in a negative feedback loop so as to buffer the voltage VH_REFi, whereby an output voltage V_SENSEis maintained at the same level as VH_REFi. The output of OP1 is connected to the gate of current source transistor M1. The load resistor of M1 is R_LEAK, which has the same resistance as resistor 19 in dummy pixel DP. Since V_SENSEis the same as VH_REFi, the current through R_LEAKis an amount which can produce the same voltage drop as occurs at the reference (readout) node in the leakage dummy pixel DPi′. The current in M1 is replicated in M2 and the current I_LEAK,EQflows in parallel with I_DACto generate the shifted reference voltage as shown in FIG. 12.

[0053] For applications in which it is only necessary or desirable to compensate for leakage current without compensating for threshold voltage nonlinearity, the image sensor of FIG. 11 can be easily modified for use in conjunction with the up/down counters and comparators without the dummy pixels DP and DP1. This can be implemented by replacing the current type DAC 52 with a voltage type DAC, connecting the analog output of the voltage type DAC directly to line 53 and applying the output of op amp OP1 to the voltage type DAC. The op amp output would be used as a reference voltage which modifies the DAC analog output in correspondence with the leakage current.

[0054] While the above description contains many specifics, these specifics should not be construed as limitations on the scope of the invention, but merely as exemplifications of preferred embodiments thereof. For example, other pixel configurations can be utilized as alternatives to those disclosed herein. Further, the timing and logic circuitry may employ bipolar devices in BiCMOS technology as an alternative to using strictly FETS in CMOS technology. Those skilled in the art will envision many other possible variations that are within the scope and spirit of the invention as defined by the claims appended hereto.