High dynamic range photodetector and photonic image sensor

Abstract

The application belongs to the technical field of photoelectric detection, and particularly relates to a high-dynamic-range photoelectric detector and a photoelectric image sensor. Mainly comprising: a photodiode (PPD), a diffusion node (FD), a reset switch (RST) and a charge transfer switch (TX). A control device controls these switches to generate initial linear mode (LIN) and logarithmic mode (LOG) response signals of the photoelectric detection element. The control device performs two consecutive exposures on the photodiode with two different working parameters, and a calculation device receiving corresponding signals generates a unified logarithmic response signal by considering the slopes (K1; K2) of the responses and the offsets (OS1; OS2) between the signals.

Description

Technical Field

[0001] This invention belongs to the field of photoelectric detection technology, and particularly relates to a photoelectric detector and a photoelectric image sensor. Background Technology

[0002] Image sensors were originally developed for taking pictures (photography, film photography, etc.), but they are now becoming a key component in intelligent systems such as self-driving cars and mobile robots.

[0003] In such applications, images captured by image sensors are typically analyzed by neural network-based software. The reliability of these analyses depends heavily on the quality and characteristics of the images provided by the sensor. In this case, the required image quality and characteristics are quite different from those understood by the human visual system in photography or cinematography.

[0004] Fundamentally, minimizing variations in image quality and characteristics is crucial for neuromorphic recognition systems. In essence, this recognition process involves comparing the input image with images previously "seen" by the same system.

[0005] It is well known that image sensors with logarithmic response facilitate the miniaturization of such changes, a characteristic present in the visual systems of all living organisms. However, fabricating logarithmic response image sensors with performance similar to that of biological visual systems is challenging.

[0006] The implementation of logarithmic sensors must be based on physical phenomena that are exponential in nature, because a simple mathematical logarithmic transformation of images that are traditionally captured with a linear response cannot cover the operating dynamic range required by logarithmic sensors.

[0007] Figure 1 A logarithmic response photodetector based on a nonlinear conversion element is demonstrated.

[0008] The simplest implementation of this photodetector is based on a nonlinear resistor that converts the photocurrent in a photodiode into a voltage with a logarithmic response. This nonlinear resistor can be a PN junction or a diode-connected transistor. Despite its simple structure, this photodetector suffers from several fatal flaws, particularly: significant differences in response between different photodetectors, strong fixed-space noise, sensitivity loss under low light conditions due to leakage current in the nonlinear resistor, large temperature-dependent response variations, and strong residual noise at low light levels. This is why this basic approach has not achieved industrial success.

[0009] The second method is based on a photodiode operating in solar cell mode, where the open-circuit voltage of the photodiode is used as the image signal. This solution can significantly reduce stationary noise. However, to maintain image resolution, an anti-blooming structure is required. In this case, the photocharge collection efficiency is greatly reduced. Simultaneously, a reset operation is needed to compensate for pixel response dispersion and eliminate residue. The sensitivity of this photodetector is very low compared to conventional photodetectors. Patent EP1354360B1 describes this solution.

[0010] The third approach is to create a photodetector that initially operates in a charge accumulation mode (linear mode) and then its response gradually evolves to a logarithmic mode. Operating the photodetector initially in accumulation mode allows for reading the photodiode in charge mode. Reading in charge mode allows for correlated double sampling to suppress switching noise during the readout. All modern conventional photodetectors use this operating mode. In charge-transfer photodetectors, switching noise can be almost completely suppressed; the only residual noise is the fundamental noise in the readout transistor. For example, the readout noise in modern charge-transfer photodetectors is typically a few electrons (the best performers can be as low as 0.25 electrons), while the readout noise in logarithmic response photodetectors is tens or even hundreds of electrons. Patents US9413991B2, US9843750B2, and US20150281607A1 provide examples of linear / logarithmic hybrid response photodetectors.

[0011] This linear / logarithmic hybrid solution is very meaningful because it can significantly improve sensitivity at low light levels while maintaining logarithmic operation under strong light.

[0012] However, a major drawback is that the transition point between the linear and logarithmic modes is highly sensitive to certain physical parameters of the device; in particular, it changes significantly with temperature. Numerous attempts have been made to correct this inhomogeneity, but none have yielded satisfactory results. Summary of the Invention

[0013] In view of this, the present invention aims to propose a solution that protects a LIN-LOG response photodetector from fixed spatial noise while maintaining its advantages in both low and high light levels, thereby obtaining a photodetector with high dynamic range and low noise level, and more broadly, a sensor with a pixel matrix, which is currently particularly anticipated in intelligent systems based on artificial intelligence vision. This solution can be implemented in various ways, with one or two photodetectors and different control parameters.

[0014] According to the first aspect, a photodetector element is proposed, which is particularly suitable for matrix image sensors, comprising: an embedded photodiode (PPD) exposed to a radiation source and having a certain accumulation capacity; A diffusion node (FD) associated with the photodiode is capable of converting changes in photocharge in the photodiode into voltage; A controlled reset switch (RST) is capable of applying a voltage to the diffusion node; A charge transfer switch (TX) is provided to transfer accumulated charge in a photodiode to a relevant diffusion node, and a photodiode exposure control device is provided to control the reset switch and the charge transfer switch to generate a response signal of the photodetector, the response signal having an initial linear mode (LIN), followed by a logarithmic mode (LOG); with a potentially uncertain transition region (X) between the two modes. The element is characterized in that the control device is capable of performing two consecutive exposures on the photodiode at two different exposure times, and is further characterized in that it includes a computing device that receives two response signals corresponding to the two exposures as inputs and is capable of generating a uniform logarithmic response signal across the entire response range by combining the two response signals. By determining the ratio (R) between two signal values, and based on the value of that ratio: The signal value obtained using the shortest exposure time is converted into logarithmic form using a predetermined slope coefficient (K1) to form the output signal (FS), or The output signal (FS) is formed by applying a predetermined offset (OS1) to the signal value obtained using the longest exposure time.

[0015] According to the second aspect, a photodetector element is proposed, particularly suitable for forming a matrix image sensor, characterized by comprising: The first embedded photodiode (PPD) is exposed to a radiation source and has a certain accumulation capacity. The second embedded photodiode (PPD) is exposed to the same radiation source and has a certain accumulation capacity. The photodetector includes at least one diffusion node (FD) associated with each photodiode, capable of converting changes in photocharge in the photodiode into voltage; at least one controlled reset switch (RST) capable of applying voltage to the diffusion node; two charge transfer switches (TX) capable of transferring accumulated charge in the two photodiodes to the diffusion node; and a photodiode exposure control device capable of controlling the reset switch and the charge transfer switch at different exposure times to generate two distinct response signals from the two photodiodes. These response signals are essentially characterized by an initial linear mode (LIN), followed by a logarithmic mode (LOG), with a potentially uncertain transition region (X) between the two modes. The exposures of the two photodiodes overlap at least partially in time. The photodetector also includes a computing device that receives the two response signals obtained from the two photodiodes respectively as input and is capable of generating a uniform logarithmic response signal across the entire response range by combining the two response signals. By determining the ratio (R) between two signal values, and based on the value of that ratio: The signal value obtained using the shortest exposure time is converted into logarithmic form using a predetermined slope coefficient (K1) to form the output signal (FS), or The output signal (FS) is formed by applying a predetermined offset (OS1) to the signal value obtained using the longest exposure time.

[0016] The two photodetector elements described above may optionally possess the following additional features: The exposure cycle is defined as two exposure times, meaning that each capture ends at approximately the same moment.

[0017] According to the third aspect, a photodetector element is proposed, particularly for forming a matrix image sensor, comprising: An embedded photodiode (PPD) is exposed to a radiation source and has a certain accumulation capacity. A diffusion node (FD) associated with the photodiode, capable of converting changes in photocharge in the photodiode into voltage; a controlled reset switch (RST), capable of applying voltage to the diffusion node; a charge transfer switch (TX), capable of transferring accumulated charge in the photodiode to the associated diffusion node; and a photodiode exposure control device, capable of controlling the reset switch and the charge transfer switch to generate a response signal of the photodetector, the response signal essentially having an initial linear mode (LIN), followed by a logarithmic mode (LOG), with a potentially uncertain transition region (X) between the two modes. The element is characterized in that the control device is capable of performing two consecutive exposures of the photodiode with two different integral capacities, and is further characterized in that it includes a computing device that receives two response signals corresponding to the two exposures as input, and is capable of generating a uniform logarithmic response signal across the entire response range by combining the two response signals, by determining the difference (DS) between the two signal values, and based on the magnitude of the difference: The output signal (FS) is formed by converting the signal value obtained using a larger accumulation capacity to its logarithm and applying a predetermined slope coefficient (K2), or by applying a predetermined offset (OS2) to the signal value obtained using a smaller accumulation capacity.

[0018] The two components mentioned above may optionally possess the following additional features, either individually or in any technically compatible combination thereof: Different integral capacities are fixed values ​​obtained through construction; Different integral capacities are obtained by a variable bias voltage at the charge transfer switch (TX).

[0019] According to another aspect, a pixel matrix photoelectric sensor is proposed, characterized in that each pixel includes a photodetector element as defined above, and characterized in that it includes a common digital processing device forming a computing device associated with each different pixel. Attached Figure Description

[0020] Other aspects, objects, and advantages of the invention will become more apparent from the following description of preferred embodiments of the invention (as non-limiting examples and in conjunction with the accompanying drawings). In the drawings: Figure 1 Three implementation schemes for logarithmic response photodetectors of known types are provided. Figure 2 The left side shows a linear / logarithmic hybrid response photodetector, and the right side shows the signal timing diagrams involved in this type of photodetector. Figure 3AThe upper part shows the physical structure of a traditional charge-transfer photodetector, and the lower part shows the band structure of this type of photodetector. Figure 3B for Figure 3A The signal timing diagram involved in the photodetector, Figure 4A The energy bands of this type of photodetector in two operating modes. Figure 4B for Figure 4A The signal timing diagram involved in the photodetector, Figure 4C for Figure 4A The response of the photodetector under different operating conditions. Figure 5 To investigate the effect of changing the exposure time and the energy barrier height under the gate on the response of the same photodetector, Figure 6 The upper part shows the response of the same photodetector at two different exposure times, and the lower part shows the ratio change between the obtained signals. Figure 7 The upper part shows the response of the same photodetector at two different maximum potential well levels, and the lower part shows the difference in the obtained signals. Figure 8 The upper part shows the response obtained at two different exposure times, and the lower part shows the corrected response obtained by processing the output signal. Figure 9 The upper part shows the responses obtained at two different maximum potential well levels, and the lower part shows the corrected response obtained by processing the output signal. Figure 10 This schematically illustrates two ways to obtain two different signals in a photodetector. Detailed Implementation

[0021] First, let's review the structure and behavior of linear / logarithmic or LIN / LOG photodetectors.

[0022] As previously mentioned, the logarithmic response in a photodetector is generated through controlled photogenerated charge leakage. For example, in a photodetector with a switching diode, charge leakage is achieved by turning on the PN junction of the switching diode. In this case, the voltage of the signal node can be pre-charged to an initial voltage that prevents the PN junction from turning on, thus eliminating charge leakage at the start of exposure. As photogenerated charge gradually accumulates, the voltage of the signal node decreases linearly. The voltage change of the signal node SIG is linearly related to the light intensity until the switching diode begins to conduct, at which point the change in the SIG voltage gradually becomes logarithmic with increasing light intensity. Figure 2A LIN / LOG photodetector is shown, in which the cathode of the photodiode PD is connected to the reference voltage VREF via a reset transistor MRST and to the power supply voltage VDD via a logarithmic NMOS transistor LOG.

[0023] It is understandable that if the initial voltage of the output signal SIG is acquired immediately after the reset transistor MRST is turned off, and then subtracted from the voltage acquired at the end of the exposure, the switching noise introduced by the MRST transistor can be suppressed.

[0024] In this basic version of the LIN-LOG photodetector, it can be seen that: a) Under low light levels, the photodetector operates in the linear region, and the switching noise of the MRST transistor is eliminated by the difference between two readings taken at the beginning and end of the exposure, as described above; thus, good sensitivity is obtained. b) The transition between LIN-LOG regions depends not only on light intensity but also on exposure time: the longer the exposure time, the lower the light intensity at the transition point; c) The transition point between the two regions depends largely on the type of switching element (PN junction, MOSFET junction, etc.) and the disturbance of the reset transistor at the end of the reset.

[0025] In such photodetectors, good image quality is achieved in the linear region due to double sampling and differential processing. However, in the logarithmic region, the image quality is quite poor due to the dispersion and instability of the transition points between the two regions.

[0026] Furthermore, while sensitivity is improved, eliminating reset noise requires readings at the start and end of exposure. This not only necessitates an image memory, but the relatively long time interval between the two readings also leads to degradation of low-frequency noise in the transistor. Finally, the dark current is very large due to the need for contact points on the photodiode. These considerations suggest that a charge-transfer photodetector is preferable.

[0027] Figure 3A The microelectronic structure of a classic charge-transfer photodetector is shown, with energy level representation shown at the bottom. Photocharge accumulates in a buried photodiode PPD, whose N-doped region (denoted as NPD) is protected by a P+ doped layer connected to the substrate SUB. The transfer transistor TX and reset transistor RST are schematically shown in the figure. The PW region represents a P-doped well.

[0028] This configuration ensures very low dark current, thus creating the possibility of complete depletion of the N-region. The transfer transistor TX connects this buried photodiode PPD to the diffusion node FD. The reset transistor RST is connected between the voltage source VDD and node FD. To read the photocharge from the buried photodiode PPD, node FD is first reset via the reset transistor RST. Then, the gate of the transfer transistor TX is biased to a high voltage, thereby forming a conductive channel beneath this transistor TX to transfer charge from the photodiode PPD to node FD.

[0029] The physical parameters of the photodiode PPD and the transfer transistor TX are adjusted so that all the charge of the light is transferred to node FD when the transfer transistor TX ends its operation. It is understood that this read operation also clears the photodiode PPD, making it ready for another acquisition cycle.

[0030] For details, please refer to the following: Figure 3B The timing diagram shows two voltage signals, one after the RST transistor activates and the other after the TX transistor activates. The difference between these two signals is the output signal. This operating mode is called "correlated double sampling," and reset noise can be almost completely eliminated when the sampling interval is very short.

[0031] Now, by introducing a charge leakage path between the photodiode (PPD) and the charge drain during exposure, a hybrid LIN-LOG response can be introduced into this charge-transfer photodetector. It is well known that in charge-transfer photodetectors, charge can naturally leak from the photodiode during exposure. Therefore, special care must be taken to ensure the presence of a charge drain to prevent excess charge from diffusing to neighboring photodetectors in the case of matrix sensors, which would otherwise significantly degrade image quality. For this reason, charge-transfer photodetectors with a LIN-LOG response have a special excess charge drainage arrangement in the PPD diode.

[0032] A simple way to discharge excess photogenerated charge is to lower the energy barrier under the gate of the TX transistor below the energy barrier of the substrate. In this case, the excess charge will flow to the FD node through this path instead of flooding into the substrate and contaminating neighboring pixels. The RST transistor must then remain on.

[0033] like Figure 4A As shown, when the accumulated charge level reaches the energy barrier under the TX gate, the charge overflows to the FD node and is effectively discharged to the voltage source through the RST transistor.

[0034] Figure 4BThe timing diagram shown is very similar to that of a conventional charge-transfer photodetector. The main difference is that, as shown in the timing diagram, the reset signal RST must remain active throughout the entire exposure time of the photodetector.

[0035] Physically, it can be proven that during this overflow period, the amount of residual charge in the embedded photodiode (PPD) is proportional to the logarithm of the light intensity incident upon it. Yang Ni's article "QLog Solar Cell Mode Photodiode Logarithmic CMOS Pixels Using Charge Compression and Readout" in the journal *Sensors*, 2018, Vol. 18, p. 584, explains this operation in detail.

[0036] Figure 4C It shows Figure 4A The photodetector is based on the LIN-LOG response shape of photon intensity.

[0037] The LIN-LOG transition point can be adjusted by either the voltage applied to the transfer transistor TX during exposure or by the exposure time. Figure 5 The left side shows the evolution of the LIN-LOG response with barrier height and exposure time.

[0038] As is well known, the threshold voltage of a MOS transistor depends to a large extent on the interface state between the gate and the substrate channel. This interface state is unstable and highly discrete. Therefore, the switching point of a photodetector with a charge-transfer type LIN-LOG response is highly discrete between individual photodetectors. Consequently, an image sensor composed of such photodetectors (one per pixel) is subject to very strong fixed spatial noise per pixel. It should be noted that much effort has been put into correcting this fixed spatial noise, but a simple and stable solution has not yet been achieved.

[0039] If we analyze the response of a photodetector with a LIN-LOG response, regardless of its specific implementation, the response consists of three regions, such as... Figure 4C As shown (where the horizontal axis is a logarithmic scale): the first linear region (LIN), an intermediate region (X), and then a logarithmic region (LOG).

[0040] As mentioned above, the initial LIN portion of the response is very accurate because the voltage conversion is based on the number of photoelectrons accumulated in the photodiode and is performed on the intrinsic capacitance of the node FD. This process is very stable and relatively accurate. The capacitance variation of the floating diffuse node FD can be easily corrected by gain, and the thermal dark charge can be corrected by a simple offset voltage. Typically, photodetectors with LIN-LOG linear responses implemented using CMOS technology require virtually no correction.

[0041] In the following discussion, it is assumed that the linear response of the LIN-LOG photodetector is accurate, either inherently or after calibration.

[0042] As can be seen, the logarithmic component of the response follows the logarithmic law well because the process is governed by the thermodynamic laws governing electrons. Discrepancies are manifested in the offset and logarithmic slope values. The slope of the logarithmic response is directly related to absolute temperature, and in a specific design, it can also be modulated by reading the gain. It can be measured, and its correction is easy, for example, by measuring the response at two different light levels to determine this slope.

[0043] In the following discussion, we will assume that the slope of the LOG response of a photodetector with a LIN-LOG response is accurate, whether inherent or after correction.

[0044] The problem lies in the intermediate region X between the LIN and LOG regions, as this transitional area is difficult to model and varies with the component's physical parameters and operating temperature. The signal from the LIN-LOG photodetector may lie in one of these regions, with entirely different fixed-space noise: very low fixed-space noise in the LIN region and very strong fixed-space noise in either the X or LOG region. Therefore, this region cannot be identified by the simple response of this photodetector.

[0045] The solution according to the invention will now be described for: Determine the intermediate region X of the LIN-LOG response photodetector. Correct the offset of the LOG response and obtain a uniform logarithmic response across the entire operating range of the LIN-LOG response photodetector.

[0046] It should be noted that these operations are performed by digital processing circuitry associated with the photodetector and receiving its output signal, or, in the case of a pixel matrix image sensor, by circuitry associated with all pixels.

[0047] First, the process of determining the intermediate region X is described.

[0048] We know that the transition point between LIN and LOG responses is affected by different physical or functional parameters. For example, a longer exposure time will advance the transition point, while a shorter exposure time will delay it. A larger integration capacity will delay the transition point, while a smaller integration capacity will advance it.

[0049] Let's recall that the LIN response is accurate and stable in absolute value, the LOG response is accurate and stable in slope, and there is an unstable and non-uniform intermediate response between the LIN and LOG responses, which is difficult to calibrate.

[0050] To correct the discrepancy in the LIN-LOG response, we need to know the exact location of the LIN-LOG transition region, but we cannot know this location by simply looking at the signal value of a photodetector with a LIN-LOG response.

[0051] One method for determining the transition region is to use the same LIN-LOG response photodetector with two different exposure times, T1 and T2. We use R to represent the ratio between the two exposure times: R = T1 / T2, and choose T1 and T2 such that R > 1. We define SIG1 and SIG2 as the signals obtained at exposure times T1 and T2, respectively.

[0052] like Figure 6 As shown: When the ratio: RS = SIG1 / SIG2 equals R, which means that SIG1 and SIG2 are both within the LIN region. When this ratio equals 1, it means that both SIG1 and SIG2 are located in the LOG region. When RS is determined to be between 1 and R, it means that SIG1 or SIG2 is located in the middle region X.

[0053] The second method for determining the transition region is to adjust the maximum integrating capacitance FW (Full Well) of the LIN-LOG response photodetector. The first integrating capacitance FW1 is used to obtain the first response signal SIG1, and the second integrating capacitance FW2 is used to obtain the second response signal SIG2, assuming that the light intensity remains constant during the two exposures.

[0054] When the SIG1 and SiG2 signals are in the LIN region, the response is not affected by the change in the integrating capacitance because the voltage is proportional to the number of photons reaching the photodiode.

[0055] On the other hand, if the light intensity causes the SIG1 and SIG2 signals to be in the LOG region, we observe a relatively constant difference D between SIG1 and SIG2. By analyzing the difference: DS=SIG1-SIG2 can determine the working areas of SIG1 and SIG2, such as Figure 7 As shown.

[0056] First, the working areas of SIG1 and SIG2 are determined using a first method based on the ratio between exposure time and signal.

[0057] Therefore, when the RS ratio is between R-TH1 and 1+TH2 (TH1 and TH2 are experimentally determined tolerance coefficients), it means that SIG1 (corresponding to the longest exposure time) has entered the LOG region, while SIG2 remains in the LIN region. Recall that at this point, SIG2 exhibits a precise and stable response, while SIG1, located in the LOG region, suffers a very scattered offset. Typically, the coefficients of TH1 and TH2 are approximately equal to Rx0.2 and Rx0.8, respectively.

[0058] Now for reference Figure 8 According to the present invention, signal SIG2 is converted into its logarithmic value LSIG2 and weighted by the slope of the logarithmic response of SIG1, that is: LSIG2 = K1 * log(SIG2) When the ratio RS is between R-TH1 and 1+TH2, the value of SIG1 should be equal to LSIG2. In this case, for the same light intensity (assuming it is constant between two exposures), the photodetector has two equal measurements: SIG1 and LSIG2.

[0059] LSIG2 is an accurate and stable response because it is obtained while the photodetector is still operating in the LIN region. Therefore, a correction offset value OS1 can be obtained to align SIG1 with LSIG2: OS1 = LSIG2 – SIG1. This parameter OS1 can be calculated during the calibration phase, or for a given exposure, as long as the RS value falls within the interval between R-TH1 and 1+TH2.

[0060] Therefore, the uniform logarithmic response FS of the photodetector can be generated, and its determination method is as follows: If RS > 1 + TH3, then FS = LSIG2 If RS <= 1 + TH3, then FS = SIG1 + OS1 The threshold TH3 is preferably equal to R / 2. Figure 8 This correction and the acquisition of the unified output signal FS are illustrated graphically.

[0061] Now for reference Figure 9 When using the second method based on the modulation accumulation well depth FW, two exposures are performed on the photodetector using two different TX barrier energy levels.

[0062] Then, the output processing circuit calculates the difference DS between the two obtained response signals SIG1 and SIG2.

[0063] In the case where the DS value is zero or less than the tolerance value TH4 (selected to handle possible measurement noise), this means that the values of signals SIG1 and SIG2 are both in the LIN region. When the difference DS is greater than TH4, then the values of signals SIG1 and SIG2 are in the LOG region.

[0064] Two tolerance thresholds TH5 and TH6 can be set. When it is determined that DS is within this interval, this means that the value of signal SIG1 is in the LIN region while the value of signal SIG2 is in the LOG region. The values of TH5 and TH6 are chosen to be able to definitely judge that SIG1 is indeed in the LIN region and SIG2 is indeed in the LOG region. These values can be determined through simulation and / or actual measurement. Usually, TH5 = Dx0.2 and TH6 = Dx0.8 can be set.

[0065] Then, the digital processing device performs a mathematical conversion of SIG1 to its logarithm by weighting with the logarithmic slope (K2) of SIG2: LSIG1 = K2 * LOG(SIG1) Remember that the LSIG1 value is an accurate and stable response because it is obtained when the photodetector is still operating in the LIN region. It constitutes a reference value. Also remember that when DS is within the interval [TH5, TH6], the value of SIG2 needs to be corrected to be consistent with LSIG1, and the offset value is: OS2 = LSIG1 - SIG2 Therefore, a unified logarithmic response of pixel FS can be generated, and its calculation is as follows: If DS < TH7, then FS = LSIG1 If DS >= TH7, then FS = SIG2 + OS2 The threshold TH7 is preferably substantially equal to D / 2. Figure 9 This correction and the obtaining of the unified signal FS are graphically illustrated.

[0066] Now a feasible method for dynamically updating the correction parameters will be described.

[0067] As described above, the main correction parameters are the logarithmic slopes K1 or K2 and the offset values OS1 or OS2. The parameters K1 or K2 are mainly related to temperature and the gain of the read channel, and can be measured during the calibration phase or extracted from a reference circuit integrated in the sensor.

[0068] The offset values ​​OS1 or OS2 are determined from the responses at two different exposure times or two different integration capacitances, respectively. The initial offset value can be obtained during calibration with stable light intensity. Because its value varies considerably with temperature and operating conditions, it is best to update this value frequently during operation of the LIN-LOG response photodetector.

[0069] It is foreseeable that the processing circuit calculates the value of RS (or DS) during use, and recalculates the value of OS1 (or OS2) whenever the calculated value falls within a given range. This continuous update can track changes in the operating conditions of the photodetector and maintain the effectiveness of the calibration. Alternatively, this update can be performed unconditionally at fixed time intervals.

[0070] It should be noted that in practical use, light intensity will vary, and this variation can introduce errors between two shots. Continuous and adaptive correction can then be used to track these changes in conditions. Furthermore, it is also possible to plan for the digital processing circuitry of the photodetector to employ a more intelligent update procedure. Apply a smoothing function during updates. Excluding abnormal correction values, this situation may occur when there is a large difference in light between two exposures.

[0071] In the context of a pixel matrix image sensor, an example architecture for implementing the present invention includes a substrate on which photoelectric conversion elements and control transistors are fabricated. A dedicated companion processor is connected to the substrate and programmed to apply the aforementioned corrections to the output signal of each pixel, generating an image with a uniform logarithmic response FS. This image can be directly used by conventional downstream processing units.

[0072] We will now describe a method for optimizing the signal-to-noise ratio.

[0073] To recap, when using exposure time modulation methods, the uniform response under low-light conditions uses the SIG2 response obtained with the shortest exposure time. However, this solution is not optimal under low-light conditions. In low-light conditions, both SIG1 and SIG2 signals are in LIN mode, but the signal-to-noise ratio of the SIG2 signal is lower than that of the SIG1 signal generated with a longer exposure time. To optimize performance, a uniform FS response can be generated using the response given by the SIG1 signal (applied by dividing by the value of R) instead of SIG2. This can achieve better image quality at low light levels. The decision threshold between generating the FS response as described above and in this paragraph can be determined experimentally on a real sensor.

[0074] When implementing the integral capacity modulation method, under low-light conditions, the SIG1 and SIG2 signals can be averaged to obtain a better signal-to-noise ratio. In fact, under low-light conditions, the SIG1 and SIG2 signals theoretically have the same value; calculating the average of these two values ​​can reduce random noise. Similar to the previous case, the decision threshold can be determined experimentally.

[0075] We will now describe one embodiment of the invention, which ensures the temporal superposition of two exposure stages, thereby generating signals SIG1 and SIG2.

[0076] As mentioned above, the method for developing a uniform logarithmic response is based on two consecutive readings of the same photodetector with different operating parameters, namely: Changes in exposure time, Or changes in accumulated abilities.

[0077] It is important to note that these two methods can also be combined. This results in an increase in the total duration of a single acquisition, which can sometimes pose a problem for matrix sensor applications equipped with such photodetectors and that must handle fast motion (e.g., in automotive vision).

[0078] In this case, refer to Figure 10 According to the present invention, a dual photodetector is implemented, comprising two photodiodes and two associated control circuits, the imaging parameters of which are either programmed to be different or constructed to be different. Each dual photodetector simultaneously or quasi-simultaneously provides two signals, SIG1 and SIG2, which are then subjected to the same correction method as previously described to obtain a uniform logarithmic response.

[0079] therefore, Figure 10 The left side illustrates two serial methods based on a single photodetector: one by modulating the exposure time (denoted as T1 / T2), and the other by modulating the accumulation capacity (denoted as FW1 / FW2). The right side shows the same method for the same pixel using two photodetectors, which has the advantage of ensuring at least partial simultaneous capture, thereby reducing the overall image acquisition and processing time.

[0080] For example, if a method based on two different exposure times is used, the control signals are programmed so that the two exposure stages end simultaneously or substantially simultaneously. In this case, signals SIG1 and SIG2 are output at the same time, and no buffer memory is needed to store the signal output first in a single photodetector embodiment. If a method based on modulation accumulation capacity is used, this modulation can be implemented in hardware to provide different physical characteristics for the two photodiodes, such as using different doping levels.

[0081] Of course, the present invention is by no means limited to the embodiments described and shown in the accompanying drawings, and those skilled in the art will be able to make many variations or modifications.

Claims

1. A photodetector element, particularly for forming a matrix image sensor, comprising: Buried photodiodes (PPDs) exposed to radiation sources and having a certain accumulation capacity. The diffusion node (FD) associated with the photodiode is capable of converting changes in photocharge within the photodiode into voltage. A controlled reset switch (RST) is capable of applying a voltage to the diffusion node. A charge transfer switch (TX) is capable of transferring the charge accumulated in the photodiode to its associated diffusion node, and The exposure control device of the photodiode can control the reset switch and the charge transfer switch to generate a response signal of the photodetector. The response signal essentially has an initial linear mode (LIN), followed by a logarithmic mode (LOG), with a potentially uncertain transition region (X) between the two modes. The element is characterized in that the control device is capable of exposing the photodiode twice consecutively at two different exposure times, and is further characterized by including a computing device that receives two response signals corresponding to the two exposures as inputs, and is capable of generating a uniform logarithmic response signal across the entire response range by combining the two response signals. The output signal (FS) is formed by determining the ratio (R) between the two signal values ​​and, based on the value of the ratio, converting the signal value obtained using the shortest exposure time into logarithmic form using a predetermined slope coefficient (K1), or by applying a predetermined offset (OS1) to the signal value obtained using the longest exposure time.

2. A photodetector element, particularly used for forming a matrix image sensor, characterized in that, include: A first embedded photodiode (PPD) exposed to a radiation source and having a certain accumulation capacity. A second embedded photodiode (PPD) exposed to the same radiation source and having a certain accumulation capacity. At least one diffusion node (FD) associated with each photodiode is capable of converting changes in photocharge in the photodiode into voltage; At least one controlled reset switch (RST) is capable of applying a voltage to the diffusion node; Two charge transfer switches (TX) are capable of transferring the charge accumulated in the two photodiodes to the diffusion node, and The exposure control device for the photodiodes is capable of controlling the reset switch and the charge transfer switch with different exposure times to generate two different response signals from the two photodiodes. The response signals are essentially in an initial linear mode (LIN), followed by a logarithmic mode (LOG), with a potentially uncertain transition region (X) between the two modes. The exposures of the two photodiodes at least partially overlap in time. The photodetector further includes a computing device that receives two response signals obtained from the two photodiodes as inputs, and is capable of generating a uniform logarithmic response signal across the entire response range by combining the two response signals. The output signal (FS) is formed by determining the ratio (R) between the two signal values ​​and, based on the value of the ratio, converting the signal value obtained using the shortest exposure time into logarithmic form using a predetermined slope coefficient (K1), or by applying a predetermined offset (OS1) to the signal value obtained using the longest exposure time.

3. The element according to claim 2, characterized in that, The exposure cycle is defined as two exposure times, meaning that each capture ends at approximately the same moment.

4. A photodetector element, particularly for forming a matrix image sensor, comprising: Buried photodiodes (PPDs) exposed to radiation sources and having a certain accumulation capacity. The diffusion node (FD) associated with the photodiode is capable of converting changes in photocharge within the photodiode into voltage. A controlled reset switch (RST) is capable of applying a voltage to the diffusion node. A charge transfer switch (TX) is capable of transferring the charge accumulated in the photodiode to its associated diffusion node, and The exposure control device of the photodiode can control the reset switch and the charge transfer switch to generate a response signal of the photodetector, the response signal having an initial linear mode (LIN), followed by a logarithmic mode (LOG), with a potentially uncertain transition region (X) between the two modes; The element is characterized in that the control device is capable of performing two consecutive exposures on the photodiode with two different accumulation capacities, and is further characterized in that it includes a computing device that receives two response signals corresponding to the two exposures as inputs, and is capable of generating a uniform logarithmic response signal across the entire response range by combining the two response signals, and forming an output signal (FS) by determining the difference (DS) between the two signal values, and forming the output signal (FS) based on the value of the difference: either by converting the signal value obtained using the larger accumulation capacity to its logarithm and applying a predetermined slope coefficient (K2), or by applying a predetermined offset (OS2) to the signal value obtained using the smaller accumulation capacity.

5. A photodetector element, particularly used for forming a matrix image sensor, characterized in that... include: A first embedded photodiode (PPD) exposed to a radiation source and having a first accumulation capacity; A second embedded photodiode (PPD) exposed to the same radiation source and having a second accumulation capacity different from the first value. At least one diffusion node (FD) associated with each photodiode is capable of converting changes in photocharge within the photodiode into voltage. At least one controlled reset switch (RST) is capable of applying a voltage to the diffusion node. Two charge transfer switches (TX) are capable of transferring the charge accumulated in the two photodiodes to the diffusion node. The exposure control device for the photodiodes is capable of controlling the reset switch and the charge transfer switch to generate two distinct response signals from the two photodiodes. These response signals essentially have an initial linear mode (LIN), followed by a logarithmic mode (LOG), with a potentially uncertain transition region (X) between the two modes. The exposures of the two photodiodes at least partially overlap in time. The photodetector also includes a computing device that receives two response signals obtained from the two photodiodes as inputs and can generate a uniform logarithmic response signal across the entire response range by combining the two response signals. The output signal (FS) is formed by determining the difference (DS) between the two signal values ​​and, based on the value of the difference, by converting the signal value obtained using the larger accumulation capacity to its logarithm and applying a predetermined slope coefficient (K2), or by applying a predetermined offset (OS2) to the signal value obtained using the smaller accumulation capacity.

6. The element according to claim 5, characterized in that, The different accumulation capacities are fixed values ​​obtained through construction.

7. The element according to claim 4 or 5, characterized in that, The different accumulation capacities are obtained by the variable bias voltage of the charge transfer switch (TX).

8. A pixel matrix photoelectric sensor, characterized in that, Each pixel includes a photodetector element according to any one of claims 1 to 7, and is characterized in that it includes a common digital processing device forming a computing device associated with each different pixel.

Images