Bionic sampling pixel circuit and pulse coding readout method thereof

Abstract

The application discloses a kind of bionic sampling pixel circuit and its pulse coding reading method, it is related to neuromorphic vision sensor, present technical problems are proposed in this scheme.For existing technology. Including photoreceptor, integral discharge circuit and pulse coding module;Photoreceptor is used to obtain light intensity information and is converted into net photocurrent Iout after output to integral discharge circuit;Integral discharge circuit is based on the feedback signal Vf of pulse coding module to reset membrane capacitance, or discharge net photocurrent Iout to integrate membrane capacitance;Pulse coding module is based on membrane capacitance node potential and threshold voltage Vth is compared, and the comparison result of opposite phase is fed back as feedback signal Vf to integral discharge circuit, and comparison result is buffered and exported as pulse signal Vout outside.Occupational advantage is in, discharge period is determined by integral time and reset pulse width, strong light end discharge frequency tends to be upper limit determined by reset pulse width, pulse discharge frequency is high and adjustable, naturally avoid frequency infinite increase.

Description

Technical Field

[0001] This invention relates to the field of neuromorphic vision sensor technology, namely, a novel image sensor that mimics the information processing method of the biological retina at the chip level, rather than using the traditional camera's "frame-by-frame exposure-readout at a fixed frame rate" method. In particular, it relates to a biomimetic sampling pixel circuit and its pulse-code readout method. Background Technology

[0002] 1. Broad technical background The biological retina is roughly composed of three layers of cells from the outside in: photoreceptors (sensing light and converting it into electrical signals), bipolar cells (integrating the photoreceptor signals and characterizing light intensity with continuously changing membrane potential amplitudes), and ganglion cells (acting as output terminals, converting continuous analog signals into a series of discrete electrical pulses). The retina has a clear functional division: the fovea at the center of the visual field is responsible for perceiving fine textures and brightness details in static scenes; the periphery is responsible for detecting motion and changes in brightness. Inspired by this, neuromorphic vision sensors have developed two complementary pathways: an integral sampling pathway that simulates the perception of fine textures in the fovea, and a differential sampling pathway that simulates the perception of motion in the periphery (represented by dynamic vision sensors such as DVS). This invention only relates to the integral sampling pathway; the differential pathway is not within the scope of this invention.

[0003] 2. Minimal technical background (1) "Points-based" mechanism When a photoreceptor receives light, it generates a photocurrent, which continuously charges a membrane capacitor Cm (corresponding to the gradient potential of bipolar cells that gradually increases with light intensity). When the potential Vm of the membrane capacitor rises to a preset threshold Vth, the subsequent comparator flips, and after being shaped by a buffer stage, a pulse is sent out from the output terminal (corresponding to a single firing of a ganglion cell). At the same time, a feedback reset signal is triggered, driving a reset transistor to discharge the membrane capacitor, causing Vm to drop back, and the circuit then begins the next round of integration. The stronger the light, the larger the photocurrent, the faster the charging, and the shorter the time to reach the threshold. Therefore, the stronger the illumination, the higher the pulse frequency, and the weaker the illumination, the lower the pulse frequency. Continuous light intensity is encoded into the frequency of discrete pulses. This cycle of "photocurrent → membrane capacitor integration → over-threshold firing → feedback reset → re-integration" is the essence of integral visual sampling.

[0004] (2) Problems caused by high "dark current" Even in complete darkness, any photosensitive device will experience a tiny, light-independent current called dark current due to thermal excitation or other reasons; this current is temperature-dependent. In traditional frame-by-frame cameras, dark current only raises the background level, which can be subtracted later. However, in an integral firing mechanism, dark current continuously charges the film capacitor along with the photocurrent—meaning that even in the absence of light, the film capacitor will gradually be charged by the dark current, triggering "firing." The consequences are: unwanted pulses appear in the absence of light / weak light (false firing in dark / weak light), the background pulse noise is raised, and the low-frequency band that should characterize weak light is occupied, reducing the effective light intensity range (dynamic range) that a pixel can distinguish. Therefore, the dark current of photosensitive devices is much more severe in integral pixels than in ordinary pixels.

[0005] (3) Advantages and disadvantages of PD-MOS PD-MOS (Photobiased MOSFET) is a novel photosensitive device based on an n-type MOS field-effect transistor using a lightly doped p-type substrate. Upon illumination, a "photobiased substrate effect" is excited within the device—the substrate potential is raised by the light, and this, combined with the amplification effect of the internal parasitic transistor, causes the drain output current to increase significantly with increasing light intensity while the gate and drain voltages remain constant. This mechanism gives PD-MOS two outstanding advantages over ordinary photodiodes: high photoelectric gain (approximately 10⁻⁶ Ω·cm). 4 It boasts a large current output even under weak light and a wide dynamic range (approximately 140dB). However, PD-MOS also has a significant drawback: its dark current in the subthreshold region is relatively large—which clashes with the previous point, preventing it from being directly used for integrating and firing pixels without processing.

[0006] (4) How does the PID structure suppress dark current? To suppress the excessively large dark current of the PD-MOS transistor, the applicant has previously disclosed a PID structure, such as... Figure 1 As shown: Two PD-MOS transistors are paired, one receiving light normally (light pixel, Light PD-MOS) and the other blocking light (dark pixel, Dark PD-MOS). Both have identical structures and temperature environments, resulting in approximately equal dark current components, while the dark pixel does not generate photocurrent. Therefore, by applying an appropriate gate bias to the light and dark pixels and subtracting their outputs, the similarly sized dark current components can be canceled out, retaining only the net current generated by light in the light pixel, thus improving the signal-to-noise ratio. It should be emphasized that PID control is essentially still a "photosensor-level" structure; it addresses "how to output a clean photocurrent."

[0007] (5) Floating gate type variant Based on PD-MOS, if a conductive layer (floating gate) wrapped in an insulating layer and electrically suspended is added to its control terminal structure, it constitutes a floating gate variant of PD-MOS. By injecting or removing charge into the floating gate, its threshold voltage can be changed and maintained, thereby adding storage and programmable modulation capabilities in addition to photosensitive properties, making it suitable for applications that require programmable synaptic weights.

[0008] Among the existing technologies, the following technical solutions are most similar to the present invention: Existing technical solution 1: Pulse Frequency Modulation (PFM) type photoelectric sensor / pixel.

[0009] This is the most classic circuit implementation for converting light intensity into pulse frequency, and it is most similar to the circuit topology of this invention. Each pixel consists of a photodiode, a reset transistor, and a series of inverters (and a feedback branch): the photocurrent of the photodiode discharges through the integrating capacitor; the capacitor potential is shaped by the inverter chain; when it exceeds the threshold, a pulse is emitted and a reset is triggered; this cycle repeats, making the output pulse frequency proportional to the incident light intensity. Therefore, it is widely used in applications such as artificial retinal prostheses to directly stimulate retinal cells with pulses. This type of solution has been fully and publicly implemented: Kagawa et al. provided a circuit model of the PFM photoelectric sensor (analyzing the relationship between the output pulse frequency and light intensity using the device parameters of the photodiode, reset transistor, and inverter chain), and fabricated a 128×128 pixel PFM image sensor with a photosensitivity of 0.15 Hz / lux using standard 0.35μm CMOS technology (Source: Kagawa K., Yoshida N., Tokuda T., et al., "Building a Simple Model of a Pulse-Frequency-Modulation Photosensor and Demonstration of a 128×128-pixel PFM Image Sensor Fabricated in a Standard 0.35-μm CMOS Technology", OpticalReview, 2004, 11(3): 176–181); Ohta, Ng, et al. used PFM photoelectric sensor arrays for subretinal implantation of retinal prostheses, operating under standard CMOS and a low voltage of 1.2V, with a measured dynamic range of approximately 120dB for input light (Source: Ohta J., Ng DC, Tokuda T., et al., a series of IEEE / Electronics Letters publications on PFM retinal prosthesis visual chips). The technical limitations of this scheme are as follows: First, using ordinary photodiodes as photosensors results in limited photoelectric gain, narrow dynamic response, extremely low photocurrent in low light, slow integration, low pulse frequency, and insufficient frequency difference (frequency resolution) between adjacent light intensity levels, making it difficult to effectively encode the brightness and fine texture of low-illuminance areas. Second, these pixels lack differential cancellation mechanisms for dark current in the device, which directly enters the integration process, causing false discharges in low light / no light conditions. Third, although a reset transistor and inverter chain can be placed within the pixel, there is a lack of means to utilize the physical mechanism of the photosensor itself to achieve light intensity self-adaptation and further expand the dynamic range of the frequency domain.

[0010] Existing technical solution 2: Wide dynamic range PFM digital pixel sensor (with intra-pixel variable reference voltage / intra-pixel magnification).

[0011] To address the issues of insufficient low-light resolution and limited dynamic range in Solution 1, the industry has improved PFM pixels by introducing a variable reference voltage generator within the pixel to shorten the integration time and improve low-light resolution in low illumination; or by introducing in-pixel amplification to enhance low-light response. These improved PFM digital pixels have been simulated and verified on a TSMC 0.18μm, 1.8V CMOS sensor. When the integration time is set to 60μs, the dynamic range reaches 150dB, an improvement of approximately 70dB compared to conventional PFM digital pixels (Source: relevant IEEE open conference paper, "A new wide dynamic range CMOS pulse-frequency-modulation digital image sensor with in-pixel variable reference voltage"). The technical limitations of this approach are as follows: This type of approach improves dynamic range and low-light resolution by adding "external knobs" such as variable reference voltage or in-pixel amplification at the circuit level. However, the photosensor remains a standard photodiode, failing to achieve light intensity adaptation capability at the physical mechanism level of the photosensor. It also lacks node-level differential cancellation for dark current; furthermore, its variable reference voltage and in-pixel amplification require additional circuit overhead. In other words, this approach verifies the circuit approach of "adjusting the reference voltage and adding in-pixel amplification to widen the dynamic range," but its principle differs from the mechanism of this invention, which uses the PD-MOS photo-induced substrate polarization effect to make the photosensor substrate potential adapt to light.

[0012] Existing technical solution 3: Vidar-like pulse camera / pulse sampling model with a central fovea.

[0013] This scheme directly simulates the integral sampling of the fovea of ​​the retina at the biomimetic level: each pixel continuously accumulates the incident light intensity, and when the light intensity accumulated by the photosensitive device exceeds a given threshold, a pulse is generated, thereby continuously recording the scene brightness in a pulse stream for high-speed reconstruction of static textures. The Vidar camera, developed by Huang Tiejun et al., is a prime example. It is about 1,000 times faster than traditional cameras by using only consumer-grade CMOS sensors and integrated circuits. The photoelectric conversion speed of the photosensitive device is about 10 ns, and the output can reach a binary pulse current of about 40,000 Hz (Source: Huang T., Zheng Y., Yu Z., et al., "1000× Faster Camera and Machine Vision with Ordinary Devices", Engineering, 2023, 25(6): 110–119). Its biomimetic sampling model was clearly established on the integral firing pathway of "photoreceptor-bipolar cell-ganglione cell" in the fovea of ​​the retina in earlier work (Source: Zhu L., Dong S., Huang T., Tian Y., "A Retina-Inspired Sampling Method for VisualTexture Reconstruction", IEEE ICME, 2019). Technical limitations of this scheme: This scheme is a work at the biomimetic sampling model and system level. Its pixel photosensitive still relies on ordinary consumer-grade CMOS optoelectronic devices. The photoelectric gain and low light sensitivity are limited by ordinary devices. Similarly, it does not involve node-level cancellation of device dark current, nor does it utilize novel photosensitive devices with inherent gain and photo-induced polarization mechanisms to simultaneously improve sensitivity and frequency domain dynamic range.

[0014] Existing technical solution four: PD-MOS / PID novel photosensitive device (device level).

[0015] To address the issues of low gain, insufficient dark current suppression, and inadequate dynamic range in the aforementioned solutions using conventional photodiodes, the applicant has previously proposed a PD-MOS photosensitive device and a PID dark current suppression structure based on it: the PD-MOS utilizes the photo-induced polarization effect and parasitic transistor effect to achieve high photoelectric gain and wide dynamic response; the PID uses the difference between the bright and dark PD-MOS pairs to suppress dark current and improve the signal-to-noise ratio. The relevant device structure and characteristics have been disclosed in prior published patents.

[0016] The technical limitations of this solution are as follows: the previously disclosed results are all at the level of "photosensing devices", and the disclosed content is the device structure, gain characteristics and dark current suppression characteristics, which solves the problem of "how to obtain a high-gain, low-dark-current photocurrent output"; it does not involve or disclose how to use this type of PD-MOS / PID device in an integral-emission type visual sampling circuit, nor does it disclose specific circuit-level solutions such as real-time dark current cancellation at the film capacitor node and light intensity adaptation using photo-induced polarization effect. Summary of the Invention

[0017] The purpose of this invention is to provide a biomimetic sampling pixel circuit and its pulse code readout method to solve the problems existing in the prior art.

[0018] The biomimetic sampling pixel circuit described in this invention includes a photoreceptor, an integral firing circuit, and a pulse coding module; The photoreceptor is used to acquire light intensity information and convert it into a net photocurrent Iout, which is then output to the integral distribution circuit. The integral distribution circuit resets the membrane capacitor based on the feedback signal Vf from the pulse code module, or integrates the membrane capacitor by distributing the net photocurrent Iout. The pulse coding module compares the membrane capacitor node potential with the threshold voltage Vth, feeds back the inverted comparison result as the feedback signal Vf to the integrator circuit, and buffers the comparison result as the pulse signal Vout for external output.

[0019] The pulse code readout method described in this invention is applied to the biomimetic sampling pixel circuit.

[0020] The biomimetic sampling pixel circuit and its pulse-coded readout method described in this invention have the following technical advantages: the firing period is jointly determined by the integration time and the reset pulse width; the firing frequency at the strong light end tends to the upper limit determined by the reset pulse width; the pulse firing frequency is high and adjustable, naturally avoiding the infinite increase of frequency. A reference bias transistor with a gate-connected reference voltage Vref is provided to set the resting potential after reset and the conduction intensity of the discharge branch, thereby determining the reset pulse width and the refractory period after firing. Since a complete firing cycle is composed of the integration time and the reset pulse width, when the illumination is very strong and the integration time tends to be small, the reset pulse width gradually becomes the lower limit of the firing cycle. This makes the firing frequency of the pixel at the strong light end tend to an upper limit determined by the reset pulse width, thus naturally avoiding the infinite increase of the firing frequency with the light intensity without adding a dedicated frequency limiting circuit, which is beneficial for the readout end to collect pulses with limited time resolution; at the same time, the consistent resting potential limited by the reference bias transistor ensures that each round of integration starts from the same starting point, guaranteeing the consistency of adjacent firing cycles. Attached Figure Description

[0021] Figure 1 This is a schematic diagram of the structure of a PID photoreceptor in the prior art.

[0022] Figure 2 This is a schematic diagram of the structure of a biomimetic sampling pixel circuit described in this invention.

[0023] Figure 3 This is a timing diagram of a biomimetic sampling pixel circuit described in this invention.

[0024] Figure 4 This is a biomimetic sampling pixel circuit described in this invention, and a comparative net photocurrent-incident light intensity curve.

[0025] Figure 5 This is a substrate potential-incident light intensity curve of a biomimetic sampling pixel circuit described in this invention.

[0026] Figure label: VDD - Operating voltage, VPD - Bias voltage of the shared photosensitive junction for bright and dark pixels, VSS - Operating ground; Vt - Photoreceptor output, Vf - Feedback signal, Vm - Potential of membrane capacitor node, Vref - Reference potential, Vth - Threshold voltage, Vout - Pulse signal; Iout - Net photocurrent; M1 - Charging switch transistor, M2 - Discharging switch transistor, M3 - Reference bias transistor; U1 - First inverter, U2 - Second inverter, U3 - Third inverter; Comparator; Cm - Membrane capacitance. Detailed Implementation

[0027] The biomimetic sampling pixel circuit structure described in this invention is as follows: Figure 2 As shown, it includes a photosensor, an integral firing circuit, and a pulse code module; The photoreceptor is used to acquire light intensity information and convert it into a net photocurrent Iout, which is then output to the integral distribution circuit. The integral distribution circuit resets the membrane capacitor Cm based on the feedback signal Vf from the pulse code module, or integrates the membrane capacitor Cm by distributing the net photocurrent Iout. The pulse coding module compares the node potential of the membrane capacitor Cm with the threshold voltage Vth, feeds back the inverted comparison result as the feedback signal Vf to the integrator circuit, and outputs the buffered comparison result as the pulse signal Vout.

[0028] Specifically, the integral issuing circuit has the following structure: The first channel terminal of the charging switch M1 is connected to the output terminal Vt of the photosensor, the second channel terminal of the charging switch M1 is connected to the film capacitor node, and the control terminal of the charging switch M1 is connected to the output terminal of the first inverter U1. The first channel terminal of the discharge switch transistor M2 is connected to ground, the second channel terminal of the discharge switch transistor M2 is connected to the film capacitor node, and the control terminal of the discharge switch transistor M2 is connected to the output terminal of the first inverter U1. The first channel terminal of the reference bias transistor M3 is connected to ground, the second channel terminal of the reference bias transistor M3 is connected to the output terminal of the first inverter U1, and the control terminal of the reference bias transistor M3 is connected to the reference potential Vref. The input terminal of the first inverter U1 is connected to the feedback signal Vf; The upper plate of the membrane capacitor Cm is connected to the membrane capacitor node, and the lower plate of the membrane capacitor Cm is connected to ground.

[0029] The structure of the pulse coding module is as follows: The non-inverting input of comparator Comp is connected to the film capacitor node, the inverting input of comparator Comp is connected to the threshold voltage Vth, and the output of comparator Comp is connected in series with the second inverter U2 and the third inverter U3. Among them, the output terminal of the second inverter U2 outputs a feedback signal Vf, and the output terminal of the third inverter U3 outputs a pulse signal Vout.

[0030] Preferably, the photosensor is a PID photosensor.

[0031] The output Vt of the photoreceptor provides a net photocurrent Iout with intrinsic gain, after dark-state cancellation, to the potential Vm of the film capacitor node via the charging switch M1. The film capacitor Cm serves as the integrating element for time integration of this net photocurrent. The discharge switch M2 and the reference bias switch M3 form a controlled discharge and reset branch. The comparator output is buffered by two stages of inverters and outputs a pulse signal Vout. The feedback signal Vf from the intermediate node between the two stages of inverters in the buffer is driven and fed back to the control terminals of the charging switch M1 and the discharge switch M2, forming a "emission → reset → re-integration" feedback loop. Thus, the entire pixel constitutes a self-excited current-to-frequency conversion unit: as long as there is a net photocurrent, this unit sustainably and periodically emits pulses without requiring an external clock to drive each emission; the emission rhythm of each pixel is entirely determined by the light intensity it receives.

[0032] The charging switch M1 is turned on when the feedback signal is at a non-discharge level and turned off when the feedback signal is at a discharge level, thus disconnecting the charging path to ensure that the film capacitor can be cleanly discharged during the reset phase without being countered by continuous charging.

[0033] The potential at the membrane capacitor node rises with integration, and its rate of rise is dVm / dt=Iout / Cm, which is directly proportional to the net photocurrent Iout and inversely proportional to the capacitance value of the membrane capacitor. The capacitance value of the membrane capacitor Cm determines the potential change caused by a unit charge in the design, and together with the net photocurrent, it determines the integration slope and the scale of the emission frequency.

[0034] The discharge switch M2 is turned on when the feedback signal is at the discharge level, providing a discharge path to ground for the membrane capacitor, so that the charge on the membrane capacitor is quickly discharged and the potential Vm of the membrane capacitor node drops rapidly; when it is not at the discharge level, it is turned off, disconnecting the discharge path and preventing the membrane capacitor from being discharged during the integration phase.

[0035] The reference bias transistor M3 is used to set the resting potential to which the membrane capacitor node falls back after discharge reset and the bias current of the discharge branch: on the one hand, it limits the termination potential of the discharge, i.e., the resting potential, so that each round of integration starts from the same starting point and ensures the consistency of adjacent discharge cycles; on the other hand, it adjusts the conduction strength of the discharge branch through the reference potential Vref, thereby adjusting the reset speed and reset pulse width.

[0036] The comparator Comp continuously compares the potential Vm of the membrane capacitor node with the threshold voltage Vth. When the potential Vm of the membrane capacitor node is lower than the threshold voltage Vth, the comparator output remains at a non-discharge level; when the potential Vm of the membrane capacitor node rises above the threshold voltage Vth, the comparator output switches to a discharge level. The comparator here implements the threshold-crossing decision function.

[0037] The buffer consists of two inverters connected in series to shape the comparator output into a pulse with steep edges and sufficient driving capability. A feedback signal Vf is led out from the intermediate node between the two inverters, and after being driven, it is fed back to the charging switch M1 and the discharging switch M2. The charging switch M1 and the discharging switch M2 employ a complementary conduction method, meaning that only one of them is active at the same feedback control level: when not at the firing level, the charging switch M1 is on and the discharging switch M2 is off, at which point the pixel is in the integration state; when at the firing level, the charging switch M1 is off and the discharging switch M2 is on, at which point the pixel is in the reset state. This complementary relationship ensures that at any given time only one of the charging or discharging paths is active, preventing the integration and reset processes from overlapping and creating a shoot-through current.

[0038] For the biomimetic sampling pixel circuit described in this invention, a corresponding pulse code readout method is also provided. The working timing and operation process of integration-firing-reset are as follows: Figure 3 As shown, the pixel operates self-sustainingly under continuous illumination by cycling through "integration → firing → resetting → re-integration" without requiring an external clock to trigger it sequentially.

[0039] During the integration phase, incident light illuminates the PID bright pixel, generating a channel current I that increases with the light intensity.DS1 Dark pixels generate dark current I. DS2 Subtracting the two values ​​from the output Vt yields the net photocurrent Iout = I DS1 I DS2 During this stage, the feedback signal Vf is at a non-discharge level. Charging switch M1 is on, and discharging switch M2 is off. The net photocurrent Iout flows through charging switch M1 into the potential Vm of the membrane capacitor node, charging the membrane capacitor Cm. The potential Vm of the membrane capacitor node rises approximately linearly from its resting potential at a rate of dVm / dt = Iout / Cm; where the resting potential is set by the reference potential Vref. Because discharging switch M2 is off during this stage and the membrane capacitor is not discharged, the net photocurrent is fully integrated onto the membrane capacitor Cm. The light intensity information is reflected in the rate of rise of the potential Vm of the membrane capacitor node in the form of the integration slope.

[0040] During the discharging phase, when the potential Vm of the membrane capacitor node rises above the threshold voltage Vth, the output state of the comparator Comp flips, and after being shaped by the buffer, a pulse signal Vout is output; simultaneously, the feedback signal Vf flips to the discharging level. The previous pulse of the pulse signal Vout represents the completion of one full integration of the membrane potential from the resting potential to the threshold.

[0041] During the reset phase, the feedback signal Vf, which flips to the firing level, turns off the charging switch M1 and turns on the discharging switch M2. The charge on the film capacitor Cm is discharged to ground through the discharging switch M2 and the reference bias switch M3, causing the potential Vm of the film capacitor node to drop rapidly. The reference bias switch M3, due to its gate connection to the reference potential Vref, limits the resting potential at which this drop terminates and determines the conduction strength of the discharge branch, thus determining the reset speed and reset pulse width. The reset pulse width determines the length of the refractory period after firing; during this period, if the film potential is below the threshold, the pixel stops firing.

[0042] During the re-integration phase, when the potential Vm of the membrane capacitor node falls below the threshold voltage Vth, the output of comparator Comp flips to the non-discharge level, the pulse signal Vout ends, the feedback signal Vf is reset to the non-discharge level, the discharge switch M2 is turned off, the charging switch M1 is turned on again, and the net photocurrent Iout charges the membrane capacitor Cm again, thus returning to the integration phase.

[0043] This process repeats continuously, with each pixel outputting a series of pulses. Within a short timeframe during which the net photocurrent Iout remains approximately constant, the duration of an integration phase is t. int Approximately t int ≈ Cm·(Vth Vrest) / Iout; where Vrest is the resting potential. Add the reset pulse width t. rst A complete distribution cycle T≈tint +t rst The corresponding pulse firing frequency f≈1 / (t) int +t rst When the light intensity is high, the net photocurrent Iout is larger, and t int The emission frequency is relatively short, mainly determined by the integration time and increases with increasing light intensity; when the light intensity is extremely strong, t int The pulse width t tends to decrease and reset rst This gradually becomes the lower limit of the period, causing the emission frequency to tend towards the upper limit determined by the reset pulse width, thus preventing the frequency from increasing indefinitely. Therefore, the incident light intensity is monotonically encoded as the pulse emission frequency of the pulse signal Vout over a wide range. For example... Figure 3 Comparison of the density of sawtooth waves of the potential Vm of the membrane capacitor node and the density of the pulse sequence of the pulse signal Vout under medium-intensity light and weak light conditions.

[0044] The effect of high gain and dark current cancellation on integral output: Due to the photopolarization effect and parasitic transistor amplification, the PID has relatively high photoelectric gain. Under the same low light conditions, its output net photocurrent Iout is larger than that of a regular photodiode. From the integral slope dVm / dt=Iout / Cm, it can be seen that the larger the net photocurrent, the faster the film potential rises and the longer the integral phase t... int The shorter the time window, the more pulses can be obtained within a given observation time window. The frequency resolution corresponding to different light intensity levels in the low-light area increases accordingly, improving the ability to encode the brightness levels and fine textures in the low-light area. At the same time, the observation time window required to achieve the target frequency resolution is relatively widened, providing a more flexible trade-off between low-light resolution and response speed.

[0045] Adaptive behavior to light intensity is achieved by utilizing the photosensitive polarization effect of bright pixels, such as... Figure 4 Figure 5 As shown. Bright pixel substrate potential V subIn a bias state that can fluctuate with light intensity, the photosensitive polarization effect causes the substrate potential to rise with the incident light intensity as the incident light intensifies. This change in substrate potential alters the device threshold and operating point through the bulk effect, causing the photocurrent conversion slope of the bright pixel to adaptively slow down and exhibit compression characteristics in the strong light range, while maintaining a relatively high conversion gain in the weak light range. Its role in the integral firing process is as follows: in the weak light range, the higher conversion gain ensures a sufficiently large net photocurrent and a sufficiently fast film capacitance integration to maintain frequency resolution at the weak light end; in the strong light range, the conversion characteristics are adaptively compressed, preventing the net photocurrent from increasing linearly with unlimited light intensity, thus avoiding premature saturation of the pulse firing frequency at the strong light end. The combination of these two factors allows a single pixel to obtain a relatively wide effective encoding range in the pulse frequency domain without relying on an additional variable reference voltage generator or switching external gain. This adaptation originates from the photosensitive polarization physical mechanism of the photoreceptor itself, fundamentally different from the purely circuit-based approach of adding variable reference voltage generators inside or outside the pixel or amplifying within the pixel in existing technologies. Figure 4 In the diagram, the dashed line represents a contrasting example using an existing integrating sampling pixel with a common photodiode.

[0046] about Figure 4 Explanation of the division into three lighting ranges: low to medium light, medium light, and strong light. There are no absolutely unified physical boundaries for dividing lighting intensity ranges; in this field, the scene's illuminance is usually used as a rough guideline. Low light / low illumination generally refers to approximately 1 lux or less, even up to 10 lux. -3 Moderate illuminance refers to light levels in the lux range, such as at night, under moonlight, or under starlight. Medium illuminance generally refers to light levels from about 1 lux to several hundred lux, such as indoor lighting or on cloudy days. High illuminance generally refers to light levels above several hundred lux, even up to 10 lux. 4 ~10 5 Lux level, such as sunlight on a sunny day or in a direct, strong light environment. (Appendix) Figure 4 The three zones of weak light, medium light and strong light along the horizontal axis from left to right are roughly divided according to this magnitude to illustrate the response characteristics of the same pixel when it spans an extremely wide range of illumination. The three vertical dashed lines are only schematic boundaries and not precise thresholds.

[0047] For those skilled in the art, various other corresponding changes and modifications can be made based on the technical solutions and concepts described above, and all such changes and modifications should fall within the protection scope of the claims of this invention.

Claims

1. A biomimetic sampling pixel circuit, characterized in that, Includes a photosensor, an integral firing circuit, and a pulse code module; The photoreceptor is used to acquire light intensity information and convert it into a net photocurrent Iout, which is then output to the integral distribution circuit. The integral distribution circuit resets the film capacitor (Cm) based on the feedback signal Vf from the pulse code module, or integrates the film capacitor (Cm) by distributing the net photocurrent Iout. The pulse coding module compares the potential Vm of the membrane capacitor node with the threshold voltage Vth, feeds back the inverted comparison result as the feedback signal Vf to the integrator circuit, and outputs the buffered comparison result as the pulse signal Vout.

2. The biomimetic sampling pixel circuit according to claim 1, characterized in that, The structure of the integral issuing circuit is as follows: The first channel terminal of the charging switch (M1) is connected to the output terminal Vt of the photosensor, the second channel terminal of the charging switch (M1) is connected to the film capacitor node, and the control terminal of the charging switch (M1) is connected to the output terminal of the first inverter (U1); wherein the charging switch (M1) and the discharging switch (M2) are transistors with complementary channel types. The first channel terminal of the discharge switch (M2) is connected to ground, the second channel terminal of the discharge switch (M2) is connected to the film capacitor node, and the control terminal of the discharge switch (M2) is connected to the output terminal of the first inverter (U1). The first channel of the reference bias transistor (M3) is connected to ground, the second channel of the reference bias transistor (M3) is connected to the output of the first inverter (U1), and the control terminal of the reference bias transistor (M3) is connected to the reference potential Vref. The input of the first inverter (U1) is connected to the feedback signal Vf; The upper plate of the membrane capacitor (Cm) is connected to the membrane capacitor node, and the lower plate of the membrane capacitor (Cm) is connected to ground.

3. The biomimetic sampling pixel circuit according to claim 2, characterized in that, The structure of the pulse coding module is as follows: The non-inverting input of the comparator (Comp) is connected to the film capacitor node, the inverting input of the comparator (Comp) is connected to the threshold voltage Vth, and the output of the comparator (Comp) is connected in series with the second inverter (U2) and the third inverter (U3). The second inverter (U2) outputs a feedback signal Vf, and the third inverter (U3) outputs a pulse signal Vout.

4. The biomimetic sampling pixel circuit according to claim 3, characterized in that, The photoreceptor is a PID photoreceptor.

5. A pulse code readout method, characterized in that, Applied to the biomimetic sampling pixel circuit as described in any one of claims 1-4.

6. The pulse code readout method according to claim 5, characterized in that, It operates in a self-sustaining cycle of accumulating, distributing, resetting, and accumulating again under continuous illumination.

7. The pulse code readout method according to claim 6, characterized in that, During the integration phase, the feedback signal Vf is at the non-discharge level, the charging switch (M1) is turned on and the discharging switch (M2) is turned off, and the net photocurrent Iout flows into the membrane capacitor node through the charging switch (M1) to charge the membrane capacitor (Cm). The potential Vm of the membrane capacitor node rises from the resting potential; where the resting potential is set by the reference potential Vref.

8. The pulse code readout method according to claim 7, characterized in that, During the discharging phase, when the potential Vm of the membrane capacitor node rises above the threshold voltage Vth, the state of the comparator (Comp) output flips, and after being shaped by the buffer, it outputs a pulse signal Vout; the feedback signal Vf flips to the discharging level.

9. The pulse code readout method according to claim 8, characterized in that, During the reset phase, the feedback signal Vf, which is flipped to the output level, turns off the charging switch (M1) and turns on the discharging switch (M2). The charge on the membrane capacitor (Cm) is discharged to ground through the discharging switch (M2) and the reference bias transistor (M3), and the potential Vm of the membrane capacitor node drops back.

10. The pulse code readout method according to claim 9, characterized in that, During the re-integration phase, when the potential Vm of the membrane capacitor node falls below the threshold voltage Vth, the output of the comparator (Comp) flips to the non-discharge level, the pulse signal Vout ends, the feedback signal Vf is reset to the non-discharge level, the discharge switch (M2) is turned off, the charging switch (M1) is turned on again, and the net photocurrent Iout charges the membrane capacitor (Cm) again, cycling back to the integration phase.

Images