Optical sensor

Abstract

The present application concerns an optical sensor that includes one or more charge transfer pixels (10) each including a buried photodiode (11) generating a photoelectric charge when illuminated, a conversion element (12) receiving at least a portion of the photoelectric charge and tending to impose, on the photodiode, a potential satisfying a non-linear relationship with the intensity of generation of the photoelectric charge, and a charge transfer element (14) for reading the charge stored by the photodiode (11) such that the residual charge in same is zero after the reading by transfer.

Claims

1. An optical sensor that includes one or more charge-transfer pixels each including a buried photodiode generating a photoelectric charge under illumination, a conversion element having a non-linear conductivity dependent on at least one control signal that is applied thereto, receiving at least some of this photoelectric charge and impressing on the photodiode a potential respecting a non-linear relationship with the intensity generating the photoelectric charge so that the non-linear relationship corresponds to a linear progression at low light levels then a logarithmic progression at higher light levels, and a charge-transfer element for reading the charge stored in the photodiode such that the residual charge in the photodiode is zero after the read-out by transfer, the conversion element being a MOS transistor, the one or more control signals being applied to the gate and/or to the drain of the transistor, the sensor including a doped confinement well under a drain/collector region of the transistor.

2. The sensor as claimed in claim 1, the conversion element being a bipolar transistor, the control signal being applied to the collector of the transistor.

3. The sensor as claimed in claim 1, said at least one control signal being dependent on temperature.

4. The sensor as claimed in claim 1, including at least one reference pixel having the same structure as the one or more illuminated pixels, this reference pixel being protected from an incident light and receiving an injection of charge simulating an illumination condition.

5. The sensor as claimed in claim 1, the charge-transfer element being a MOS transistor.

6. The sensor as claimed in claim 1, the charge-transfer element being a bipolar transistor.

7. The sensor as claimed in claim 1, the control signal being the collector or drain voltage.

8. A method for compensating for the temperature of a sensor as claimed in claim 1, comprising the step of acting on said at least one control signal in order to compensate for the influence of the temperature on the leakage of photoelectric charge from the photodiode.

9. An optical sensor that includes one or more charge-transfer pixels each including a buried photodiode generating a photoelectric charge under illumination, a conversion element having a non-linear conductivity dependent on at least one control signal that is applied thereto, receiving at least some of this photoelectric charge and impressing on the photodiode a potential respecting a non-linear relationship with the intensity generating the photoelectric charge so that the non-linear relationship corresponds to a linear progression at low light levels then a logarithmic progression at higher light levels, and a charge-transfer element for reading the charge stored in the photodiode such that the residual charge in the photodiode is zero after the read-out by transfer, the conversion element being a MOS transistor, the one or more control signals being applied to the gate and/or to the drain of the transistor, the drain/collector of the transistor being formed by a doped well having a depth larger than or equal to that of read transistors of the pixel.

10. An optical sensor that includes one or more charge-transfer pixels each including a buried photodiode generating a photoelectric charge under illumination, a conversion element having a non-linear conductivity dependent on at least one control signal that is applied thereto, receiving at least some of this photoelectric charge and impressing on the photodiode a potential respecting a non-linear relationship with the intensity generating the photoelectric charge so that the non-linear relationship corresponds to a linear progression at low light levels then a logarithmic progression at higher light levels, and a charge-transfer element for reading the charge stored in the photodiode such that the residual charge in the photodiode is zero after the read-out by transfer, the conversion element being a MOS transistor, the one or more control signals being applied to the gate and/or to the drain of the transistor, the conversion element including a buried collector or drain, common to a plurality of pixels of the sensor.

11. An optical sensor that includes one or more charge-transfer pixels each including a buried photodiode generating a photoelectric charge under illumination, a conversion element having a non-linear conductivity dependent on at least one control signal that is applied thereto, receiving at least some of this photoelectric charge and impressing on the photodiode a potential respecting a non-linear relationship with the intensity generating the photoelectric charge so that the non-linear relationship corresponds to a linear progression at low light levels then a logarithmic progression at higher light levels, and a charge-transfer element for reading the charge stored in the photodiode such that the residual charge in the photodiode is zero after the read-out by transfer, the conversion element being a MOS transistor, the one or more control signals being applied to the gate and/or to the drain of the transistor, a doped region corresponding to the cathode of the photodiode extending below a doped region corresponding to the drain/collector of the transistor.

12. An optical sensor that includes one or more charge-transfer pixels each including a buried photodiode generating a photoelectric charge under illumination, a conversion element having a non-linear conductivity dependent on at least one control signal that is applied thereto, receiving at least some of this photoelectric charge and impressing on the photodiode a potential respecting a non-linear relationship with the intensity generating the photoelectric charge so that the non-linear relationship corresponds to a linear progression at low light levels then a logarithmic progression at higher light levels, and a charge-transfer element for reading the charge stored in the photodiode such that the residual charge in the photodiode is zero after the read-out by transfer, the conversion element being a MOS transistor, the one or more control signals being applied to the gate and/or to the drain of the transistor, a doped region corresponding to the cathode of the photodiode extending below a doped region corresponding to the drain/collector of the transistor, the photodiode being illuminated via a back side of a substrate.

13. An optical sensor that includes one or more charge-transfer pixels each including a buried photodiode generating a photoelectric charge under illumination, a conversion element having a non-linear conductivity dependent on at least one control signal that is applied thereto, receiving at least some of this photoelectric charge and impressing on the photodiode a potential respecting a non-linear relationship with the intensity generating the photoelectric charge so that the non-linear relationship corresponds to a linear progression at low light levels then a logarithmic progression at higher light levels, and a charge-transfer element for reading the charge stored in the photodiode such that the residual charge in the photodiode is zero after the read-out by transfer, the sensor including at least one reference pixel having the same structure as the one or more illuminated pixels, this reference pixel being protected from an incident light and receiving an injection of charge simulating an illumination condition, said at least one control signal being controlled automatically depending on the signal output from the reference pixel or a group of reference pixels, in order to maintain the output signal of the reference pixel or of this group of reference pixels at a constant value.

14. An optical sensor that includes one or more charge-transfer pixels each including a buried photodiode generating a photoelectric charge under illumination, a conversion element having a non-linear conductivity dependent on at least one control signal that is applied thereto, receiving at least some of this photoelectric charge and impressing on the photodiode a potential respecting a non-linear relationship with the intensity generating the photoelectric charge so that the non-linear relationship corresponds to a linear progression at low light levels then a logarithmic progression at higher light levels, and a charge-transfer element for reading the charge stored in the photodiode such that the residual charge in the photodiode is zero after the read-out by transfer, the charge-transfer element including drain-induced-barrier-lowering means.

15. An optical sensor that includes one or more charge-transfer pixels each including a buried photodiode generating a photoelectric charge under illumination, a conversion element having a non-linear conductivity dependent on at least one control signal that is applied thereto, receiving at least some of this photoelectric charge and impressing on the photodiode a potential respecting a non-linear relationship with the intensity generating the photoelectric charge so that the non-linear relationship corresponds to a linear progression at low light levels then a logarithmic progression at higher light levels, and a charge-transfer element for reading the charge stored in the photodiode such that the residual charge in the photodiode is zero after the read-out by transfer, the charge-transfer element including collector-induced-barrier-lowering means.

16. An optical sensor that includes one or more charge-transfer pixels each including a buried photodiode generating a photoelectric charge under illumination, a conversion element having a non-linear conductivity dependent on at least one control signal that is applied thereto, receiving at least some of this photoelectric charge and impressing on the photodiode a potential respecting a non-linear relationship with the intensity generating the photoelectric charge so that the non-linear relationship corresponds to a linear progression at low light levels then a logarithmic progression at higher light levels, and a charge-transfer element for reading the charge stored in the photodiode such that the residual charge in the photodiode is zero after the read-out by transfer, the conversion element being a virtual base or gate transistor the potential of which is set to that of a substrate.

17. An optical sensor that includes one or more charge-transfer pixels each including a buried photodiode generating a photoelectric charge under illumination, a conversion element having a non-linear conductivity dependent on at least one control signal that is applied thereto, receiving at least some of this photoelectric charge and impressing on the photodiode a potential respecting a non-linear relationship with the intensity generating the photoelectric charge so that the non-linear relationship corresponds to a linear progression at low light levels then a logarithmic progression at higher light levels, and a charge-transfer element for reading the charge stored in the photodiode such that the residual charge in the photodiode is zero after the read-out by transfer, the conversion element being a bipolar transistor, the control signal being applied to the collector of the transistor, an emitter of the bipolar transistor formed by a doped region corresponding to the cathode of the photodiode and a base of the bipolar transistor corresponding to the anode substrate of the photodiode.

Description

(1) Other features and advantages of the present invention will become apparent on reading the following detailed description of nonlimiting examples of implementation thereof, and on examining the appended drawing, in which:

(2) FIGS. 1 to 4, which were described above, illustrate the prior art;

(3) FIG. 5 is a circuit diagram of a pixel according to the invention;

(4) FIG. 6 illustrates the use of a CMOS transistor to produce the non-linear conversion element;

(5) FIG. 7 shows the cathode voltage of the photodiode as a function of photoelectric current;

(6) FIGS. 8 and 9 illustrate the charge stored in the photodiode corresponding to two different voltages Vpin;

(7) FIG. 10 illustrates the variation in the response of the charge Q.sub.PD of a pixel as a function of the variation in the gate voltage V.sub.G or the voltage Vpin;

(8) FIGS. 11 and 12 are analogous views to FIGS. 8 and 9, but corresponding to two different operating temperatures of the pixel;

(9) FIG. 13 illustrates the variation in the response of a pixel as a function of temperature;

(10) FIG. 14 illustrates the drop in the amount of charge stored as a function of gate voltage and as a function of illumination level, due to the effect of temperature;

(11) FIG. 15 illustrates the variation in the response of the sensor as a function of the gate voltage V.sub.G;

(12) FIG. 16 illustrates the injection of a current into a reference pixel;

(13) FIG. 17 shows an example of processing of signals generated by reference pixels of the sensor using a feedback loop;

(14) FIG. 18 illustrates the relationship between I.sub.D and V.sub.GS in the presence of the DIBL effect;

(15) FIG. 19 illustrates modulation of the current I.sub.D and of the voltage V.sub.S via the DIBL effect;

(16) FIG. 20 illustrates the use of a bipolar transistor to produce the conversion element;

(17) FIGS. 21 to 27 show examples of embodiments of pixel structures;

(18) FIG. 28 is a pixel read-out timing diagram;

(19) FIG. 29 illustrates that it is possible to associate a plurality of photodiodes with the same capacitive node;

(20) FIG. 30 illustrates that it is possible to produce a buffer structure in a pixel; and

(21) FIG. 31 shows an example optical sensor according to the invention.

(22) FIG. 31 shows an example sensor 100 according to the invention. This sensor 100 includes, as is conventional, a matrix-array (N rows and M columns) of active pixels 10 and one or more reference pixels 10 protected from incident light. The one or more pixels 10 are identical to the active pixels 10, except for the fact that they are protected from incident light, and that a non-zero illumination level is simulated by injection of a current, as explained below.

(23) The sensor 100 may include means allowing the pixels to be addressed, such as a row controller 101, and a read circuit 102 and a column-scan circuit 103, as is conventional, and an interface circuit 105. The row controller 101 generates control signals for each selected row. For each selected row, the timing diagram of FIG. 2 may be applied. The entire output of this selected row enters into the read circuit 102 in which the double readouts are processed. This circuit is controlled by the column-scan circuit allowing the pixels to be read. The read circuit may either be analogue, or digital and comprise an analogue-digital converter.

(24) According to one advantageous aspect of the invention, the sensor may include a temperature-compensation circuit 106. The reference pixel or group of pixels 10 receives the same control signals and generates an output signal that is processed in the compensation circuit by comparison with a predefined reference level. This circuit generates a drive signal that is sent to all the pixels, including the one or more reference pixels, in order to compensate for the variation in the response of the pixel as a function of temperature, as will be explained in detail below.

(25) For applications in which the variation in the pixel response as a function of temperature is tolerated, it is possible to remove this compensation circuit and the one or more reference pixels. In this case, a set electric signal is applied to the pixels in the matrix-array. One example of such a case is a sensor mounted on a temperature-regulated thermal electric cooler (TEC). This set control signal for example aims to preserve a linear progression in the response of the sensor at low light levels, depending on the targeted applications.

(26) The invention is not limited to a particular number of pixels nor to pixels arranged in a matrix-array; the pixels may be reduced to one row; at the least, the sensor may include only a single active pixel.

(27) FIG. 5 schematically shows an example pixel 10 of an optical sensor according to the invention.

(28) This pixel 10 includes a photodiode 11, which is preferably buried, connected to a non-linear current-voltage conversion element 12 that preferably has a current-voltage relationship that is close to a logarithmic function, in particular under high light levels in order to generate a non-linear voltage on the cathode 13 of the photodiode. A charge-transfer device 14 for transferring charge to a reception device 15 for receiving the charge allows access to be gained to the charge stored in the photodiode, which is representative of the level of exposure to light of the photodiode.

(29) The non-linear conversion element 12 may be controlled by an electrical signal originating from a control circuit (not shown in FIG. 5) such as produced by the aforementioned circuit 106, this electrical signal preferably being a function of temperature so as to compensate for its influence but possibly also as a variant resulting from intentional programming. This control circuit is preferably integrated into the same chip as the pixels.

(30) In one example implementation of the invention, which example is illustrated in FIG. 6, a MOS transistor forms the non-linear conversion element 12 and the charge-transfer device 14 for its part is also a MOS transistor.

(31) In a pixel of common size ranging from a few microns to a few tens of microns, the photocurrent generated by the photodiode is very small. For example, a photodiode of 10 m by 10 m generates only about 10 nA with a direct illumination of 100 kLux.

(32) At this level of current, the non-linear conversion transistor 12 operates in subthreshold mode. The drain current I.sub.D may then be expressed by the following equation:

(33) $\begin{array}{cc}{I}_D-I_O{e}^{\frac{V_G}{V_T}}\left(e^{-\frac{V_S}{V_T}}-e^{-\frac{V_D}{V_T}}\right)& \left(1\right)\end{array}$

(34) In this equation, V.sub.G is the gate voltage of the conversion element, V.sub.S the source voltage, V.sub.D the drain voltage and V.sub.T=kT/q, k being Boltzmann's constant, T absolute temperature and a constant slightly higher than 1.

(35) When a voltage Vd that is sufficiently high with respect to V.sub.T is applied, since 11 is close to 1, formula (1) may be simplified to:

(36) $\begin{array}{cc}{I}_D=I_O{e}^{\frac{V_G-V_S}{V_T}}& \left(1^{}\right)\end{array}$

(37) Given that V.sub.S=V.sub.PD and that I.sub.D=I.sub.ph+I.sub.dark, where I.sub.D is the drain current of the conversion element, I.sub.ph the photocurrent and I.sub.dark the dark current, the voltage on the cathode of the photodiode is given by the relationship:
V.sub.PD=V.sub.S=V.sub.GV.sub.T log(I.sub.ph+I.sub.dark)+V.sub.T log I.sub.O(2)

(38) The photoelectric-charge-forming intensity corresponds to the photocurrent Iph. FIG. 7 is a graphical representation of the photodiode cathode voltage as a function of the photoelectric current Iph, which voltage is logarithmic, except at very low light levels. This decreasing voltage indicates that the number of freed electrons stored in the cathode of the photodiode increases as a function of light intensity, essentially logarithmically if the variation in the capacitance of the photodiode junction is ignored. This approximation is valid because the variation in the voltage is small because of the logarithmic compression obtained via the conversion transistor.

(39) When the photodiode is a buried photodiode, the cathode is completely depleted of electrons when the cathode voltage reaches the value Vpin. FIG. 8 illustrates the case where the voltage V.sub.pin is lower than the maximum possible voltage if the photodiode is not a buried photodiode. In this case, when the photocurrent is below a certain level and the voltage on the cathode reaches the value Vpin, there is no longer any charge stored in the photodiode. All the photo-electric charge is removed via the non-linear conversion transistor.

(40) When the voltage V.sub.pin is higher than the maximum possible voltage on the source of the non-linear conversion transistor, as illustrated in FIG. 9, the cathode of the photodiode is charged to the voltage Vpin by the transfer transistor after a complete charge transfer. In this case, the source voltage of the transistor, which is also the voltage of the cathode of the photodiode, is also biased to a value V.sub.pin. According to relationship (1) above, the current in the conversion transistor is almost 0. The excess current corresponding to I.sub.ph+I.sub.dark creates an accumulation of charge in the photodiode at the start of its exposure to the light, after the complete transfer of its charge by the transfer transistor TX.

(41) This accumulation is essentially linear until the photodiode voltage V.sub.PD touches the non-linear response curve set by the conversion transistor at which point practically all the photoelectric charge is removed by this conversion transistor. Afterwards, the variation in the amount of charge stored follows the logarithmic triangle as in the case of FIG. 8.

(42) If the amount of charge stored in the photodiode is plotted as a function of the variation in the voltage V.sub.pin and the voltage V.sub.G, a family of curves such as illustrated in FIG. 10 is obtained. The maximum possible value of the cathode voltage V.sub.PD of the photodiode is dependent on the voltage V.sub.G on the gate of the conversion transistor according to relationship (3) above.

(43) Since the voltage V.sub.pin is set by construction, the variation in the gate voltage V.sub.G of the conversion transistor causes a variation of the same nature as the variation in the voltage V.sub.pin but in an opposite direction. A higher V.sub.pin value or a lower V.sub.G value create an essentially linear initial response followed by a logarithmic response. It is preferable to create a non-linear response having a linear initial section because it allows all the photo-electric charge to be preserved at low light levels.

(44) The charge-reception device may be a floating-diffusion node, as is known for a 4T pixel.

(45) Before the charge of the photodiode is read out, the capacitive node FD is pre-charged to an initial voltage that is higher than the voltage Vpin of the photodiode. The transfer transistor TX is then actuated.

(46) Some of the charge stored in the junction of the photodiode is transferred to the capacitive node FD via this transfer transistor TX. The voltage of the photodiode increases and that of the capacitive node FD decreases until equilibrium is reached. The variation in the voltage of the capacitive node FD provides information on the amount of charge transferred from the photodiode to the capacitive node FD. Preferably, the initial voltage of the capacitive node FD is high enough for the voltage on the photodiode to be able to reach the voltage V.sub.pin at the end of the transfer. In this case, the transfer is complete and the photodiode is completely emptied of mobile charge.

(47) The voltage on the gate of the MOS conversion transistor may be chosen so that when the voltage V.sub.pin is reached, the current in the MOS conversion transistor is very low, and ideally generates less than one electron with the exposure time used. After a complete charge transfer has taken place, a new acquisition cycle may commence.

(48) In one variant, as illustrated in FIG. 30, the charge-reception device is a buffer structure PD2 having the same structure as the buried photodiode PD1 and playing a role of charge memory, being protected from incident light by a mask 210. This buffer structure PD2 has a voltage V.sub.pin higher than that of the photodiode PD1, so as to allow charge to flow from the photodiode PD1 to the buffer structure PD2. It may be of smaller size than the photodiode from which it receives charge, because it does not have to generate photoelectric current. This configuration is useful for producing what is called a snapshot sensor, in which the respective charges of all the photodiodes are simultaneously transferred to the associated buffer structures, which are then read sequentially, thereby avoiding deformation of the image in the case of moving objects. Charge may be transferred from a buried photodiode PD1 to the associated buffer structure PD2 via a transfer element such as a MOS transistor for example, controlled by a signal TX1. The transfer of the charge from the buffer structure PD2 to a capacitive node FD with a view to readout thereof may be achieved using another transfer element such as another MOS transistor for example, controlled by a signal TX2. The signal TX1 is applied to all the transfer transistors controlling transfer to the buffer structures PD2, whereas the signal TX2 is applied sequentially.

(49) The thermal dependence of the response of the pixel of the present invention, according to equation (2) above, is due to V.sub.T, Io and I.sub.dark.

(50) I.sub.dark may be neglected because the dark current in a buried photodiode may be decreased to a very low value using a modern fabrication process. The dependence of V.sub.T on temperature is proportional to absolute temperature. It may be corrected quite easily with a gain correction such as conventionally encountered within a video camera.

(51) The current Io of the conversion transistor is highly dependent on temperature because typically it doubles every 6-7 C. This variation deforms the response curve by changing the maximum possible voltage on the cathode of the photodiode, as FIGS. 11 and 12 show. The amount of charge stored in the photodiode is also impacted by temperature, as illustrated in FIG. 13 in which it may be seen that an increase in temperature decreases the amount of charge stored in the cathode of the photodiode. Lastly, although the voltage Vpin is set by construction by the physical structure of the photodiode, it also changes with the temperature T.

(52) FIG. 10 shows that by modulating the gate voltage V.sub.G of the conversion transistor, it is possible to produce the same effect as the variation in the voltage Vpin, and the variation in the response with temperature, illustrated in FIG. 13, may thus be compensated for via a suitable modulation of the voltage V.sub.G as a function of temperature.

(53) FIG. 14 shows that by lowering the gate voltage V.sub.G to V.sub.GdV.sub.G when the temperature increases, it is possible to increase the initial charge storage capacity of the photodiode and thus obtain the same response curve.

(54) From formula (3), it may be seen that a modulation of the gate voltage V.sub.G allows this variation to be compensated for if it is possible to keep the sum V.sub.G+V.sub.T*Log Io constant.

(55) FIG. 15 illustrates one possible way of implementing a temperature compensation, using one or more reference pixels masked by a layer that is opaque to light, such as a layer of metal for example, like the one or more aforementioned pixels 10. An electrical current is injected, as illustrated in FIG. 16, into the one or more photodiodes of these reference pixels, thereby simulating the photoelectric current. This current is sufficiently large that this or these reference pixels work in the logarithmic regime. This current may be injected into the photodiode of a reference pixel through an electrical resistor connected to a voltage source that generates a current in the same direction as the photoelectric current generated by the photodiode under the effect of its illumination. As a variant, this current may be injected into the photodiode of a reference pixel through a capacitor connected to a voltage ramp generator that generates a current in the same direction as the photoelectric current generated by the photodiode.

(56) Thus, instead of creating a reference level corresponding to the dark level, a reference level corresponding to a certain illumination level is created by electrically simulating this illumination of the one or more reference pixels. Thus drawbacks related to the effect on the reference signal of parasitic light, which is relatively small at a high illumination level, is avoided. The injected current for example simulates an illumination level of at least 10000 lux.

(57) Any temperature variation shifts the linear-logarithmic transition and has an impact on the response of this or these reference pixels. It is possible to adjust the gate voltage V.sub.G so that the response of this or these reference pixels is always at a predefined level, by virtue of a feedback loop such as illustrated in FIG. 17. This operation allows the dependence on temperature of the response of the pixels to be successfully eliminated. It may be seen in FIG. 17 that the pixels 10 and 10 receive the same control signal, on the conversion element, in the present case the same gate voltage V.sub.G in the illustrated example. This control signal is generated so as to maintain the signal generated by the reference pixels at a setpoint value V.sub.setpoint. In the case of a plurality of reference pixels 10, the average value generated by these pixels may be used as output signal.

(58) The one or more reference pixels 10 are preferably placed on the substrate of the sensor in order to obtain a temperature that is as close as possible to that of the active pixels. A plurality of reference pixels or groups of reference pixels may be placed in various locations on the sensor in order to correctly measure the temperature of the substrate. The average value of the reference pixels may be used to set the compensation.

(59) The regulation may be achieved by analog means. As a variant, it is possible to use digital means including an ADC and a DAC associated with a processor.

(60) In one variant, the compensation is achieved using a temperature probe that delivers the temperature to a circuit including a controller equipped with a memory in which is stored a look-up table that directly generates, for each temperature value, a corresponding value of the control signal, for example the gate voltage V.sub.G of the conversion transistor in the example just described with reference to FIG. 6. The look-up table may be obtained beforehand in a calibration phase.

(61) In the case where temperature does not vary greatly during the use of the sensor or in the case where the temperature dependence is tolerated, it is possible to apply a set control voltage to the gate V.sub.G depending on the wanted non-linear response.

(62) The one or more reference pixels are advantageously used to obtain a temperature compensation.

(63) The conversion element may take multiple forms, of MOS or bipolar structure.

(64) Advantageously, in particular when it is a question of a MOS transistor, the design of the conversion element takes advantage of the effect referred to as drain induced barrier lowering (DIBL), which may be quite pronounced in a short-channel MOS transistor. The same effect is also observed in a BJT transistor when the base is weakly doped, for example to lower than 10.sup.17 atoms per cm.sup.3. This effect is an electrostatic influence of the drain on the source. A higher drain voltage makes it easier for electrons to exit from the source. As FIG. 18 shows, the increase in the drain voltage V.sub.D is equivalent to an increase in the gate voltage V.sub.G with the same drain voltage. In other words, the DIBL effect modulates the current Io in a sub-threshold MOS transistor. This modulation of the current Io may cancel out the temperature-related contribution and thus achieve a compensation of the response of a pixel according to the present invention.

(65) In a standard CMOS fabrication process, the DIBL effect is minimized by special doped wells in the transistor structure. The chapter on the MOSFET structure of the book Physics of Semiconductor Devices by S. M. Sze published by John Wiley & Sons Inc. 1981 gives a quite detailed description on these optimizations. In the context of the present invention, it is preferable to preserve this effect sufficiently in order to be able to create a compensation of the current in the range of variation in temperature. One effective method for achieving this is to decrease the doping of the substrate under the gate. For example, it is possible to decrease the doping of the substrate under the gate to a level comprised between 10.sup.14 and 10.sup.17 atoms per cm.sup.3 and to increase the depth of doping for the drain, for example to the same value (within 30%) as the length of the channel, without too greatly decreasing the channel length, the depth thus for example being comprised between 0.25 microns and 2 microns.

(66) The advantage of this approach exploiting the DIBL effect is that it makes it possible to place a set voltage on the gate and to create the compensation by modulating the drain voltage of the conversion transistor. In this case, it is possible to choose to use a voltage V.sub.G equal to 0 for the gate of the conversion transistor. A gate biased to 0 V is advantageously replaced by a thin layer 110 a surface region of which is highly P-doped, this layer making contact with the P substrate such as illustrated in FIG. 19, for a P-type substrate in which in the cathode of the photodiode consists of an N-doped well.

(67) The virtual gate thus formed does not suffer from a dispersion in the threshold voltage induced by charge trapped in the oxide of the gate as in a conventional transistor. A sufficiently high dopant concentration is preferred, for example higher than 10.sup.17 atoms per cm.sup.3 in order to make the surface Fermi level stable and also to eliminate surface leakage current. It is also possible to use a substrate without additional doping for this conversion transistor, i.e. what is called a native transistor. The doping is insured simply by uniform doping during the fabrication of the silicon wafer, the most common dopant concentration being 10.sup.15 atoms per cm.sup.3.

(68) The drain modulation voltage of the conversion transistor may be generated with the same methods as those proposed above for the compensation achieved by acting on the gate voltage V.sub.G, here replaced by a modulation of the drain voltage V.sub.D.

(69) The non-linear conversion element may also be a bipolar transistor as illustrated in FIG. 20. In this case, the collector is equivalent to the drain, and the base to the gate and the emitter to the source of the MOS transistor described above. Formula (3) may be rewritten in the form:
V.sub.PD=V.sub.E=V.sub.BV.sub.T log(I.sub.ph+I.sub.dark)+V.sub.T log I.sub.O(4)

(70) where V.sub.E is the emitter voltage and V.sub.B the base voltage.

(71) The same compensation mechanism also applies thereto, and it is possible to act on the base voltage by way of control signal, or better still to achieve control via the collector voltage, as described below.

(72) The structure illustrated in FIG. 21 a) is incompatible with a buried photodiode because it is necessary to create an explicit interconnection between the emitter and the cathode of the buried photodiode. To form a contact on the cathode requires a portion of the cathode to touch the surface of the silicon.

(73) The preferred configuration with a buried photodiode is to connect the base to the substrate, which forms the anode of the buried photodiode, as illustrated in FIG. 21 b). In this case, modulation of the voltage of the collector alone allows a temperature compensation to be achieved.

(74) As with a virtual gate, the doping concentration in the base region is advantageously low for two reasons. The first is that it is advantageous to have a relatively pronounced DIBL effect, for example of more than 50 mV/V (1V on the drain inducing a change in the threshold of 50 mV) for the temperature compensation. The second is that a quite high maximum photodiode cathode voltage is needed in order to be able to reach the voltage V.sub.pin of the buried photodiode. If this dopant concentration is not low, it is necessary to work with a voltage V.sub.pin that is very close to zero, or even negative, in order to ensure the initial response is a linear progression, allowing a good sensitivity at low light levels to be ensured.

(75) It is advantageous to use a weakly doped substrate grown epitaxially without additional doping for the base of this transistor. The use of native doping in a substrate grown epitaxially also ensures a better dopant uniformity in a sensor making use of this type of pixel.

(76) It will also be noted that the drain or the collector of the non-linear conversion transistor is biased to a high voltage, typically close to the supply voltage of the sensor. Therefore the collector or the drain diffusion also sucks up electrons created by the incident light. This competition decreases the quantum efficiency of the photodiode. A P-doped well may form an electrostatic screen repulsing some of these electrons toward the buried photodiode.

(77) FIG. 22 shows an embodiment according to the invention with a conventional MOS conversion transistor. It is possible to use, as illustrated, a floating diffusion FD to receive and to measure the amount of charge stored in the buried photodiode, under illumination. The voltage-follower portion for the read-out of the voltage on the floating diffusion FD is shown schematically because it may easily be made with available conventional transistors in a conventional CMOS fabrication process.

(78) FIG. 23 illustrates a variant with a doped confinement well 140 under the doped drain region of the transfer transistor and under that of the conversion transistor. This doped confinement well is, in the described example, of P.sup.+-type, and the substrate is of P-type. The doped confinement well lies at a depth larger than the doped region forming the cathode of the photodiode.

(79) In the embodiment illustrated in FIG. 24, the control signal consists of the drain voltage of the conversion transistor. The gate is at the same potential as the substrate, and it is qualified a virtual gate. The doped surface region of the photodiode is extended by 115 beyond the doped region of the photodiode forming the cathode. This embodiment takes advantage of the DIBL effect in a conversion transistor the virtual gate of which is biased to zero. In order to accentuate the DIBL effect, the drain of the conversion transistor is doped to a depth d larger than that of the read transistors of the pixel, which are conventional transistors; in particular the selection transistor SEL is a conventional transistor. This depth may be at least equal to that of the region NPD of the buried photodiode, as illustrated. The P.sup.+-doped virtual gate connects with the protective layer 110 of the buried photodiode.

(80) The N-doped region NPD of the buried photodiode plays the role of the source of the conversion transistor. It is optionally possible to leave a gap 150 between the virtual gate and the drain in order to limit the leakage current due to the tunneling effect caused by their very high dopant dose.

(81) A doped confinement zone is advantageously produced under the drain of the conversion transistor, and likewise under the transfer transistor.

(82) In the embodiment illustrated in FIG. 25, which is relatively compact, a bipolar transistor, the base of which is connected to the substrate 169, which forms the anode of the buried photodiode, plays the role of the non-linear conversion element. The collector 170 of the accentuated-DIBL-effect conversion transistor is placed under the buried photodiode. Since the collectors of the transistors of the various pixels share the same voltage, they may be merged into one. The buried collector 170 may be biased for all the pixels via a surface contact and a sufficiently deep N-doped well (not shown in FIG. 25) passing through the P-doped first substrate 169.

(83) The embodiment illustrated in FIG. 26 has a configuration that is suitable for a backside-illuminated pixel 171. The conversion transistor is a bipolar transistor, the base of which is connected to the substrate, which forms the anode of the photodiode, and the temperature compensation is achieved via the DIBL effect between the collector and the buried photodiode (the de facto emitter of this transistor).

(84) In the embodiment illustrated in FIG. 27, the transfer transistor is a BJT or virtual-gate MOS transistor, the gate voltage of which is that of the substrate. The DIBL effect is exploited to extract the charge stored in the photodiode. A capacitor CX is connected between the signal TX controlling transfer and the capacitive node FD. After the capacitive node FD has been reset, the signal TX passes to a higher level. This overvoltage is transmitted to the capacitive node FD via the capacitor CX and causes a current to pass via the extreme DIBL effect referred to as punch through (or PT). This current allows the charge to be transferred from the buried photodiode to the capacitor CX. In this figure, the photodiode is illuminated from the back side, and the doped region forming the cathode of the photodiode is extended by a region 130 under the drain 120 of the conversion transistor. The doped surface region, of P.sup.+-type in the case of a P-type substrate, is also extended by 115 above the region 130 toward the drain 120, in order to eliminate surface leakage current.

(85) The advantage is that the transfer by punch through occurs in the bulk of the silicon and therefore the P.sup.+ layer may better cover the cathode of the buried photodiode. This further decreases dark current. The fabrication process may possibly be simpler.

(86) In all these embodiments, the read out is carried out in the same way as for a conventional 4T pixel, the timing diagram of the signals of which is illustrated in FIG. 2. FIG. 28 gives an operating timing diagram applicable to all the embodiments.

(87) The compactness of the embodiments described above may be increased by merging the capacitive node FD and the transistors involved in the read-out of the voltage of the capacitive node FD. In this case, a plurality of buried photodiodes associated with their non-linear conversion element may be connected to the same capacitive node FD via respective transfer transistors, as illustrated in FIG. 29. These various photodiodes may be read sequentially, one after the other, with the control timing diagram of FIG. 2. This compacting technique is commonly used in small-pitch image sensors.

(88) The invention is not limited to the examples just described. In particular, the N and P carrier types may be inverted, a P-type substrate becoming an N-type substrate, an N-doped well becoming a P-doped well and vice versa.

(89) Various modifications may be made to the structures described. For example, the embodiment of FIG. 24 may have no doped confinement well under the non-linear conversion transistor and the transfer transistor.