Biasing scheme for large format CMOS active pixel sensors

[0001] 1. Field of the Invention

[0002] The invention relates to active pixel sensor arrays and, more particularly, to a biasing arrangement for complementary metal oxide semiconductor (CMOS) active pixel sensor arrays.

[0003] 2. Description of the Related Art

[0004] Active pixel sensor (APS) imaging devices are described in U.S. Pat. No. 5,471,515, for example. APS imaging devices include an array of pixel cells, generally arranged in rows and columns. Each pixel cell includes a photodetector (e.g., photodiode) that converts light energy into electric signals. The pixel cell also includes one or more transistors. The transistors typically provide amplification, read-out control, and reset of the photodetector signal, and operate to provide reset signal and a photodetector pixel signal as output signals from the pixel cell.

[0005] The architecture of a conventional imaging device with a CMOS APS 2 is illustrated in FIGS. 1 and 2. APS 2 includes a row and column array of pixel cells 4. As illustrated in FIG. 1, array rows are oriented horizontally, and array columns are oriented vertically. The 4×4 array pictured in FIG. 1 is merely illustrative. Typical APS arrays are fabricated as much larger arrays.

[0006] Referring more specifically to FIG. 2, a representative three transistor (3T) pixel cell 4 is shown as including a photodiode 6 (PD) and a readout portion which includes a source follower transistor 8 (M_D) and a row select transistor 10 (M_SEL). Pixel cell 4 has a drain of source follower transistor 8 (M_D) connected to voltage line 12 (V_{AA—PIX—COL.j}). The source of the row select transistor 10 (M_SEL) is connected to a column line 14 of the cell array to which a load transistor 16 (M_LD) also is connected. The load transistor 16 (M_LD) acts as a current sink for a bias current I_BIAS—PIX. Thus, each pixel 4 includes a source follower arrangement of transistors 8, 10, with the pixel and a common current sink transistor 16 for each array column.

[0007] The operation of APS 2 will be described with reference to the timing diagram of FIG. 3. The photodiode 6 is reset by pulsing a reset transistor 18 (M_RS) with a reset pulse Φ_RS. During the ensuing integration (exposure), the photodiode 6 voltage is decreased by the photo current. After the pre-determined integration period, a row select pulse Φ_SEL turns ON and an output signal appears at the node denoted by V_OUT (FIG. 2).

[0008] The output signal voltage is given by Equation 1, as follows:
V_OUT,sig =A_V·(V_RS,pix−ΔV_sig−(V_th−δV_th))   (1)
where A_V, V_th, V_RS,pix and ΔV_sig are the voltage gain (A_V) and the threshold voltage (V_th) of the source follower, the initial photodiode voltage (V_RS,pix) just after the reset, and the voltage swing (drop) (ΔV_sig) caused by the accumulation of the signal electrons on the photodiode, respectively.

[0009] V_th represents a deviation of the effective threshold voltage, which gives rise to fixed pattern noise (FPN) in the array. FPN (also called nonuniformity) is spatial variation in pixel output values under uniform illumination due to device and interconnect parameter variations (mismatches) across the sensor. It is fixed for a given sensor, but varies from sensor to sensor. FPN increases with illumination, but causes more degradation in image quality at low illumination. CMOS (APS) sensors have higher FPN than charge-coupled devices (CCDs) and suffer from column FPN, which may appear as shadow stripes in the image and can result in image quality degradation.

[0010] The output signal voltage V_OUT,sig for a column pixel is sampled and held on a first capacitor in a sample and hold (S/H) circuit 20 (FIG. 1). Then the photodiode is reset again, and the output reset voltage is sampled and held on a second capacitor in sample and hold (S/H) circuit 20.

[0011] The output reset voltage is given by Equation 2, as follows:
V_OUT,rs=A_V·(V_RS,pix−(V_th−δV_th))   (2)
The signal voltage swing, representing the light-induced signal on the photodiode, can be extracted by subtracting (1) from (2), which yields:
V_SIG=A_V·ΔV_sig   (3)
Thus, variations in V_RS,pix and V_th are eliminated, in principle.

[0012] Referring again to FIG. 1, a row select circuit 22 selects a row to be reset or readout. All rows are selected for read out in sequential fashion. The S/H circuit 20 performs sample and hold operations for the reset and pixel signals for each of the array columns, which are represented by equations (1) and (2). The outputs of the S/H block 20 are fed to a differential amplifier which subtracts the two signals to produce V_SIG in accordance with equation (3). This signal is then digitalized and sent to an image processor with other signals from the pixels of array 2.

[0013] The FIG. 3 pulse timing diagram illustrates the operation of the image sensor shown in FIG. 1 having photodiode active pixels. Conventionally, the row select and sample-and-hold operations are performed in a row-by-row fashion, represented in FIG. 3 by these successive rows i−1, i, and i+1. A period at the start of each frame between the end of the last row select and the beginning of the first row select is called the vertical blanking period (V_BL).

[0014] In a large format array based on the representative architecture shown in FIG. 1, a problem occurs due to the voltage (I-R) drop along the V_SS line, which couples ground to all the column circuit sink transistors 16 as illustrated in FIG. 4. During the readout period, voltage drops along the V_SS line, which is typically grounded at terminal AV_ss, cannot maintain the established (ground) level for each column line due to parasitic resistances R_SS denoted in FIG. 4. The graph of FIG. 5 illustrates the voltages on column lines 1 to j_max as shown, those column lines farthest from terminal AV_ss have a voltage above ground while those closest to AV_ss have voltages at or near ground along the line V_SS.

[0015] The voltage drops along the line V_SS impact the bias currents I_BIAS—PIXproduced by column line transistors 16. The bias current is given by Equation (4), as follows:
I_{BIAS—PIXS,i}=β/2.(V_GS,i−V_TH)²=β/2.(V_LN−V_{SS—PIX—COL,i}−V_TH)²  (4)
where β, V _GS,i, and V_{SS−COL, i}are a process and size dependent parameter (β), the gate-source voltage for the j-th load transistor (V _GS,i), and the V_SS voltage at the column J (V_{SS—PIX—COL, i}), respectively. Thus, as V_{SS—PIX—COL, i} becomes higher than V_SS, bias current decreases. The change in the output voltage due to the higher V_SS is given approximately by Equation (5) as follows:

[0016] $\begin{array}{cc}\Delta V_{\mathrm{OUT}}=\sqrt{\frac{{\beta }_{\mathrm{LD}}}{{\beta }_D}}·\Delta V_{\mathrm{SS}}& \left(5\right)\end{array}$
where β_LD,β_{D, VSS} are β for M_LD and M_D, and the voltage change on the V_SS line, respectively. If a fixed pattern noise (FPN) suppression operation, where the signal voltage swing is extracted by subtracting Eq. 1 from Eq. 2, is not performed, shading appears in a reproduced image. By applying such an FPN suppression operation, shading caused by the voltage change on V_SS line, expressed by Eq. 5, may be suppressed. However, other shading will still be present.

[0017] Shading may be caused, for example, by decreased bias currents due to the reduced effective gate-source voltage V_GS of the load transistor 16 (M_LD), which may result in variations in time constants for charging and discharging the sample and hold capacitors C_SHS or C_SHR within the sample and hold circuit 20. The capacitor C_SHS and C_SHR respectively sample and hold the pixel output signal and reset output signal. The time constants are given approximately by Equation 6, as follows:

[0018] $\begin{array}{cc}\tau =\frac{C_{\mathrm{SHX}}}{g_m}=\frac{C_{\mathrm{SHX}}}{\sqrt{2·\beta ·I_{\mathrm{BIAS\_PIX}}}}& \left(6\right)\end{array}$
Thus, the resulting voltages on the sample and hold capacitors may change when the pulse width of the sample-and-hold pulse is comparable to the time constant. The pulse width of the sample-and-hold pulse may be comparable to the time constant since higher resolution image sensors require shorter pulse widths to obtain a given frame rate. Another related concern is the source follower gain variation due to the decreased bias current.

[0019] Similar voltage drops occur along a voltage line 24 which supplies the pixel voltage V_AA—PIX, as a result of parasitic resistances R_AA along that line, as shown in FIG. 4. The voltage changes here affect the pixel output voltage little, however, since the driver transistor 8 (M_D) of each pixel cell (FIG. 2) operates in the saturation region and thus the drain voltage does not affect its drain current. A problem arises, however, during the reset operation.

[0020] There are two modes of pixel reset operation, “hard” and “soft.” A “hard reset” refers to a reset where the reset transistor 18 (M_RS) operates in its linear region. The initial voltage of the photodiode V_RS,pix is given by Equation 7, as follows:
V_RS,pix=V_{AA—PIX—COL}   (7)

[0021] To make reset transistor 18 (M_RS) operate in its linear region, a pulse height of the reset pulse Φ_RS should be higher than V_{AA—PIX—COL} by V_th, where V_th is the effective threshold voltage of the reset transistor 18. Noise associated with the ‘hard’ reset is given approximately by Equation 8, as follows:
V_hard²=kT/C_PD   (8)
where k is the Boltzmann's constant, T the absolute temperature in K and C_PD is the capacitance of photodiode 6, respectively.

[0022] A “soft reset” refers to a reset where the reset transistor 18 (M_RS) operates first in its saturation region, and then in the sub-threshold region. An initial voltage V_RS,pix of photodiode 6 is given by Equation 9, assuming the pulse height of the reset pulse Φ_RS is equal to V_AA—PIX, as follows:
V_RS,pix=V_AA—PIX—−V_th   (9)

[0023] When the pulse height of the reset pulse Φ_RS is equal to V_AA—PIX, reset transistor 18 (M_RS) operates in its saturation region. The final stage at the reset operation is dominated by the sub-threshold operation. Reset noise associated with the ‘soft’ reset is given approximately by Equation 10, as follows:
v_soft²=kT/2C_PD   (10)

[0024] Although the noise associated with the soft reset is less than that associated with the hard reset, there are other concerns with the soft reset. For example, the soft reset yields non-linearity in low light level photo-conversion characteristics of the pixel cell. The soft reset also yields image lag due to the fact that the final stage at the reset operation is dominated by the sub-threshold operation of the reset transistor 18. Therefore, the hard reset is preferable to the soft reset, especially for motion picture applications, for example.

[0025] In order to perform the hard reset, the pulse height of the reset pulse Φ_RS should be higher than V_{AA—PIX—COL} by V_th, which requires an additional voltage pump circuit in a row select pulse generation circuit (assuming that V_AA—PIX is the highest voltage supplied to the image sensor.) There may be a case, however, in which the soft reset is applicable when noise performance and dynamic range are the most important performance parameters.

[0026] If a large format array is built based on the architecture shown in FIG. 1, a problem associated with the hard reset occurs due to the voltage (I·R) drops along the V_AA—PIX line. When a row is selected for readout, the pixel bias currents I_bias,pix flow from V_AA—PIX to AVSS as shown in FIG. 4. During this row select period, voltages along the V_AA—PIX line cannot maintain an appropriate level, as shown in FIG. 6, due to parasitic resistances R_AA along the V_AA—PIX line.

[0027] The pixel reset Φ_RS is applied after the signal voltage sampling (Φ_SHS=‘H’) while the row select pulse Φ_SEL is kept on, as shown in the timing diagram of FIG. 3. Since the pixel reset voltage, which is a drain voltage of M_RS, is V_AA—PIX, as expressed by Eq. 7, the initialized voltage of each photodiode V_RS,pix on a row is not identical, as shown in FIG. 6. In addition, if the V_AA—PIX line is contaminated by noise, that noise is sampled on a photodiode.

[0028] The present invention provides readout circuitry for each column of an APS device array in which substantially equal sink currents exist on each column line or in which sink currents are not employed.

[0029] In a first exemplary embodiment of the invention, a capacitor loaded source follower transistor is provided for each pixel so that no bias currents are needed. In a second exemplary embodiment of the invention, a level shift circuit is associated with each column line load transistor to apply a constant gate-source voltage to the load transistor. In a third exemplary embodiment a clamp circuit is associated with each column line. In a fourth exemplary embodiment, a differential amplifier is associated with each column line to sink a constant current.

[0030] In another exemplary embodiment, a transistor is placed between the line V_AA—PIX and the column line V_{AA—PIX—COL}. By making the transistor, preferably an n-MOS transistor, operate in its saturation region, noise on the V_AA—PIX line can be filtered out. Further, by adding a p-MOS transistor in parallel with the n-MOS transistor, a more reliable hard or soft reset operation of the pixel cell can be obtained.

[0049] Referring initially to FIG. 7, an APS 30 according to an exemplary embodiment of the present invention is shown. APS 30 includes a number of pixels 34 arranged in an array 32. Pixels 34 may employ many different architectures, including the 3 transistor (3T) architecture shown in FIG. 2 or 4 transistor (4T) or higher order transistor architectures known in the art. APS 30 does not use the FIG. 2 current sink transistor 16, but rather employs a capacitor loaded source follower formed by connecting the column lines 44, 46, 48 of the array to sample and hold circuit 36 in the manner shown in FIG. 7. Consequently, the load transistor 16 of FIG. 2 is eliminated from the circuit. For clarity, certain features of APS 30 not critical to the invention have been eliminated in the drawing of FIG. 7.

[0050] Column read-out circuitry of this embodiment will be described for column 48 as being representative of other columns in the array. Connected to column line 48 is a read-out circuit 51 provided in S/H circuit 36. The read-out circuit 51 includes two capacitors 52 (C_SHS) and (C_SHR) connected in series with column line 48. These two capacitors effectively replace the load transistor 16 (M_LD) shown above in the conventional arrangement of FIG. 2,forming two capacitor-loaded source follower circuits with transistor 8 (M_D), connected to capacitor 52 (C_SHS) or 54 (C_SHR) through column select transistor 10 (_MSEL).

[0051] Signal acquisition and reset of the read-out circuit are controlled by a sample and hold signal Φ_SHS, supplied to switch 56, a sample and hold reset signal Φ_SHR, supplied to switch 58, and a capacitor reset pulse Φ_RSC supplied to switches 60 and 62.

[0052] Operation of the embodiment of FIG. 7 is described below. Although the description is for the conventional PD APS pixel cell 4, shown in FIG. 2, any other pixel cell structure can be used.

[0053] Just before the sample and hold operation for the pixels on any one of the rows i, the hold capacitors 52 (C_SHS) and 54 (C_SHR) are reset to ground (V_SS) by turning Φ_RSC on. Then the row select pulse Φ_SEL,i is turned on (see FIG. 8 timing diagram). When the sample pulse for signal Φ_SHS is turned on, switch 56 closes and the pixel output V_OUT,sig is sampled onto capacitor 52 (C_SHS). After Φ_SHS is turned off, the pixels on the i-th row are reset by the reset pulse Φ_RS,i. Then, the sample pulse for reset signal Φ_SHR is turned on and switch 58 closes, and the reset signal V_OUT,rs is sampled on capacitor 54 (C_SHR). Lines 53 and 55 represent output lines to other circuitry which processes the signals held on capacitors 52 and 54.

[0054] The signal voltage V_OUT,sig and V_OUT,rs are given by Equations (1) and (2), above. In this capacitive loaded source follower, the hold capacitors 52, 54 are charged up to a voltage level expressed by Equations (1) or (2). With this arrangement, the current flow through the driver transistor 8 (M_D) of a pixel does not cause current flow and thus voltage drops on the V_SS line, which affect the sampled and held signals.

[0055] FIG. 9 shows a modification of the embodiment of FIG. 7 in which a line 64 connects column line 48 to ground (V_SS) by way of a switch 66 controlled by a signal Φ_RS—COL. A timing diagram is shown in FIG. 10. The arrangement of FIG. 9 allows the use of a reset signal Φ_RS—COL to reset column line 48 to ground (V_SS), before the sample-and-hold pulses are turned on and the row select pulse is applied.

[0056] Referring to FIG. 11, another exemplary embodiment of the present invention is shown in which a level shift circuit 52 is used to maintain a constant gate-source voltage as applied to a load transistor 50 (M_LD). FIG. 11 shows a block diagram of a column readout bias portion of an APS, such as the conventional APS described above in connection with FIGS. 1 and 2. Load transistors 50 (M_LD), corresponding to load transistor 16 in FIG. 2, are arranged with a level shift circuit 52 inserted between the V_SS line and the gate of each load transistor 50 (M_LD). The bulk (p-well) of each load transistor is connected to its source.

[0057] The input of the level shift circuit is V_SS—COL—i and the output of is V_SS—COL—i+ΔV, where ΔV is a pre-determined voltage. With this circuit configuration, the gate-source voltage and the gate-bulk voltage are kept constant even if the V_SS line voltage raises due to the I-R drop mentioned above. Thus, the bias current for source follower circuits (which are formed by transistor 8 (M_D) and load transistor 50 (M_LD) through transistor 10 (M_SEL)) can be kept constant.

[0058] Referring to FIGS. 12 and 13, another exemplary embodiment of the present invention is illustrated in which a clamp circuit 63 is employed. The circuit diagram of FIG. 12 shows a column readout bias portion of an APS, and is similar to the embodiment described above in connection with FIG. 11, the level shift circuit 52 having been supplanted by a clamping circuit 63. FIG. 13 is a pulse timing diagram for the circuit of FIG. 12.

[0059] Each clamping circuit 63 includes a clamping capacitor 64 (C_CL) connected between the gate of a load transistor 66 (M_LD) and the V_ss line. The gate side of clamping capacitor 64 also is connected to a voltage line V_CL by way of a switch 68 controlled by a signal Φ_CL.

[0060] Referring to the timing diagram of FIG. 13, during the vertical blanking period when no currents flow from V_AA—PIX to V_SS (FIG. 2), the clamp pulse Φ_CL is turned on, thereby charging the clamp capacitor 64 at the clamp voltage V_CL. This configuration makes the gate-source voltage and the gate-bulk voltage constant even if tie V_SS voltage rises due to the I-R drop when a row select pulse turns on.

[0061] Although the clamp pulse Φ_CL is applied during the vertical blanking period in FIG. 13, it is apparent that the clamp pulse can be applied any other time when the row select pulses are turned off.

[0062] Referring to FIG. 14, a simplified circuit diagram of another exemplary embodiment of the invention is shown in which a differential amplifier 70 is placed on each column readout line. The inputs for the differential amplifier 70 are bias voltages, V_REF1 and V_REF2, and the output of each differential amplifier 70 is connected to the column line. The differential amplifier is powered by V_AA, which preferably is different from V_AA—PIX.

[0063] The differential amplifier is designed so that it draws current, which is expressed by Equation 11 as follows:
I=I₂(V_REF2)−I(V_REF1)   (11)
and voltage change or noise on V_SS line does not affect the value of the current I.

[0064] The proposed embodiments above are especially useful for large format image sensors since identical bias currents for columns can be obtained. Shading in a reproduced image, which is caused by I-R drop on V_ss line, can be eliminated.

[0065] Further modifications to the above-described exemplary embodiments of the invention provide for filtering of a pixel reset signal and provide for the ability to operate a ‘hard’ or ‘soft’ reset. An embodiment is shown in FIG. 15, in which an n-MOS transistor 80 (M₁) is connected between the V_AA—PIX line and the drain of M_D and M_RS of a pixel. The gate of transistor 80 (M₁) connects to a bias line V_{AA—PIX—CTL} which is usually biased at the same value as V_AA—PIX so that transistor 80 (M₁) operates in its saturation region. Since no currents flow on the V_{AA—PIX—CTL} line, the gate voltage of transistor 80 (M₁) is maintained constant. On the other hand, voltages along the V_AA—PIX line change due to the IR drop during the period when Φ_SEL is turned on as mentioned above in connection with FIG. 6.

[0066] The source voltage of transistor 80 (M₁), which is V_{AA—PIX—COL—j} is described by Equation 12, as follows:
V_{AA—PIX—COL—j}=V_{AA—PIX—CTL}−V_th   (12)
where V_th is the threshold voltage of transistor 80 (M₁). Thus, the source voltage is constant regardless of the voltage variation or possible noise on the V_AA—PIX line.

[0067] In this configuration where V_AA—PIX=V_{AA—PIX—CTL}=V_AA, the pixel reset can be considered to be in between the hard reset and the soft reset modes of operation described above, since the drain voltage and the gate voltage of the reset transistor in a pixel M_RS is V_AA-V_th and V_AA, respectively, thus M_RS operates around a boundary between the linear and saturation regions. If the hard reset mode is required, V_{AA—PIX—CTL} should be slightly lower than V_AA—PIX, so that the reset transistor inside a pixel M_RS operates in the linear region.

[0068] Another embodiment is shown in FIG. 16, where both the hard reset and soft reset operation modes can be obtained. A p-MOS transistor 90 (M₂) is added in parallel to n-MOS transistor 80 (M₁). When V_{AA—PIX—CTL} is high (V_AA), transistor 90 (M₂) is off, and operation of the circuit is the same as in the previously described embodiment. When V_{AA—PIX—CTL} is low (V_SS), transistor 80 (M₁) is off and transistor 90 (M₂) is on. In this case, V_{AA—PIX—COL} becomes V_AA—PIX and the pixel reset is performed in the soft reset mode, assuming the pulse height of Φ_RS is V_AA.

[0069] FIG. 17 shows system 300, a typical processor based system modified to include an image sensor IC 308 including an APS array having compensating circuitry as described above in connection with FIGS. 7-16. Processor based systems exemplify systems of digital circuits that could include an image sensor. Examples of processor based systems include, without limitation, computer systems, camera systems, scanners, machine vision systems, vehicle navigation systems, video telephones, surveillance systems, auto focus systems, star tracker systems, motion detection systems, image stabilization systems, and data compression systems for high-definition television, any of which could utilize the invention.

[0070] System 300 includes central processing unit (CPU) 302 that communicates with various devices over bus 304. Some of the devices connected to bus 304 provide communication into and out of system 300, illustratively including input/Output (I/O) device 306 and image sensor IC 308. Other devices connected to bus 304 provide memory, illustratively including random access memory (RAM) 310, hard drive 312, and one or more peripheral memory devices such as floppy disk drive 314 and compact disk (CD) drive 316.

[0071] Image sensor IC 308 can be implemented as an integrated image sensor circuit on a chip with circuitry to compensate for current gradients, as described above. Image sensor IC 308 may be combined with an image processor, which receives digitized pixel signals from a pixel array and provides digital image output signals. The process can be a CPU, digital signal processor, or microprocessor. The image sensor IC 308 and processor can be combined in a single integrated circuit.

[0072] The present invention provides methods and applications for active pixel sensors (APS) having photoreceptors arranged in an array to form a plurality of columns of image sensors. Columns of image sensors have circuitry that compensates for voltage drops in at least one of a voltage supply line and a ground line connecting the plurality of columns of image sensors.

[0073] While preferred embodiments of the invention have been described and illustrated above, it should be understood that these are exemplary of the invention and are not to be considered as limiting. Additions, deletions, substitutions, and other modifications can be made without departing from the spirit or scope of the present invention. For example, circuitry can provide no reset capability, or either a hard reset alone, a soft reset alone, or both alone, or in combination with each of the exemplary embodiments for addressing readout circuitry voltage drops. Accordingly, the invention is not to be considered as limited by the foregoing description but is only limited by the scope of the appended claims.