IMAGE SENSOR

Description

Summary

[0001] Imaging 3D time-of-flight sensors (TOF sensors) are particularly well-suited for detecting objects and people in three-dimensional space. This capability stems from their fundamental principle: at each pixel of the sensor, the distance to objects within the sensor's detection range is determined based on the transit time (also known as time of flight or TOF) of light pulses. The sensor therefore provides a sensor output signal that reflects a data set containing a distance value for each pixel, corresponding to the distance between that pixel and the object surface segment depicted on it. The sensor thus delivers a three-dimensional image that can be analyzed using appropriate processing algorithms.For example, it is known that the sensor output signal is analyzed to determine whether objects are located within a predefined spatial sector. In certain applications, one or more imaging 3D time-of-flight sensors can be used to detect whether a person and a machine part are coming dangerously close to each other, e.g., within the machine's operating range.

[0002] In certain applications, particularly in the interaction between humans and machines, machines and machines, or machines and spaces, it is necessary that the detection is carried out in a particularly reliable, i.e., safe, manner.

[0003] The degree of safety is classified in various standards into safety integrity levels (SIL): Automobiles (ISO 26262) QM ASIL-A ASIL-B / C ASIL-D - General (IEC-61508) - SIL-1 SIL-2 SIL-3 SIL-4 Aviation (DO-1781254) DAL-E DAL-D DAL-C DAL-B DAL-A Railway (CENELEC 50126 / 128 / 129) - SIL-1 SIL-2 SIL-3 SIL-4 State of the art

[0004] 3D-TOF sensors are known that comprise at least one optical transmitter for emitting electromagnetic radiation, preferably (but not exclusively) in the near-infrared range, and an image sensor composed of an arrangement of radiation receivers. Each individual radiation receiver corresponds to a pixel (picture element) of the image sensor and thus of the 3D-TOF sensor.

[0005] Various Time-of-Flight sensors are described in US 2017 / 257617 A1, WO 2017 / 050633 A1, EP 3 182 162 A1 and WO 2018 / 172767 A1.

[0006] Typically, the radiation receivers are divided into two groups (radiation receiver A, radiation receiver B) and arranged as a matrix.

[0007] Each radiation receiver has a photosensitive element (e.g. a photodiode or a photogate) and at least one storage capacity.

[0008] A typical operating principle of a time-of-flight (TOF) sensor is as follows: At time t0, the transmitter is activated, emitting a light pulse with a duration corresponding to t1 for a short period t1 (e.g., 30 ns). Simultaneously, the receiver A is also activated for a short activation period tA (e.g., also 30 ns). The light pulse is reflected by an object and registered by the receiver A with a delay of tof. During the activation period tA, the energy of the reflected light pulse causes the release of charges—specifically, photoelectrons—in the photosensitive element. The amount of charge resulting from the registration of the light pulse (corresponding to the number of released photoelectrons) is stored in the storage capacity (A1).Receiver A can only receive that part of the reflected light pulse corresponding to the time difference tA-tof, since the reflected light pulse arrives at the photosensitive element with a delay of the flight time tof compared to the start of the activation period. Therefore, the activation period and the pulse duration of the reflected light pulse only partially overlap for a certain period. The overlap period is inversely proportional to the travel time of the light pulse (see figure). Figure 1 The number of photoelectrons released is therefore at least approximately inversely proportional to the travel or flight time of the light pulse.

[0009] With a time delay Δ = t0 + tA (again 30 ns in the example) – that is, directly following the end of the first activation period tA – the radiation receiver B is activated. The radiation receiver B is active for a second activation period tB (also 30 ns in the example). The receiver also registers only a portion of the reflected light pulse, which corresponds to the time difference Δ = tof + ti - tA (where ti = tA, this is tof), and is therefore proportional to this time difference (the travel time of the light pulse). The corresponding amount of charge is stored on the storage capacity B1.

[0010] Since the reflected signals can be very weak depending on the reflectivity of the reflecting object and the distance, it is necessary to provide an amplification mechanism. This is achieved by repeatedly performing the measurement described above and accumulating the amount of charge recorded each time in the storage capacity until a sufficient signal strength is reached.

[0011] The distance can then be calculated as follows: E1 = P R * kref * T − tof E2 = P R * kref * tof + T − T = P R * kref * tof with: E1; E2 - Energies per sampling for receiver A and BT respectively = tA = tB = ti - max. flight time = 30 ns for a max. distance of 4.5 m (uniqueness range) kref - reflection coefficient of the target, tof - flight time of the pulse, P(R) - reflected pulse power at kref = 1,

[0012] The charge quantities are proportional to the respective energy values ​​for each sampling process.

[0013] In simplified terms, one could write for 200 accumulations: with A = ∑ n = 1 200 pE 1 B = ∑ n = 1 200 pE 2 A; B - amounts of charge accumulated in storage capacity A and B respectively, p - is the proportionality factor, whose specific properties are not relevant for these considerations.

[0014] By normalization (division by the sum of the charge quantities accumulated in the two capacities), normalized charge quantities Q1 and Q2 can be formed that are independent of the degree of reflection. Q1 = A / A + B (elimination of the dependence on the degree of reflection through quotient formation) Q2 = B / A + B From the standardized charge quantity, a value for the flight time can be calculated for each capacity: tof1 = 1 − Q1 · T tof2 = Q2 · T

[0015] In the ideal, error-free case, the two flight times determined in this way, tof1 and tof2, are equal: tof1 = tof2. Any measurement errors can be reduced by averaging: tof = 1 2 tof1 + tof2 Mittelwert

[0016] The measured distance between the sensor and the reflecting object is calculated as follows: S = 1 2 c * tof c - speed of light

[0017] An extended version of the described setup consists of each receiver having two storage capacities (receiver A via Aa and Ab; receiver B via Ba and Bb). This feature results in expanded functional possibilities. First option:

[0018] Receivers A and B are operated identically. The storage capacities Aa and Ab are switched on and off in a staggered manner – Ab takes over the function of B (as described above).

[0019] Instead of two photo receivers, only one photo receiver with two storage capacities is used. Aa = ∑ n = 1 200 pE 1 Ab = ∑ n = 1 200 pE 2 used. The advantage of such an arrangement, with suitable design, can lie in a higher density of receivers (better resolution). Second option:

[0020] The additional storage capacity is used for 4-phase sampling of the signal. While in Figure 1 If a signal sampling with two integration periods is described (2-phase), the signal can now be sampled four times, with the sampling points being offset by 1 / 4 of the period of the signal each time, see Figure 2 . For this implementation variant, periodic pulse sequences with a period T can also be used as the emitted signal.

[0021] The phase shift ϕ of the reflected signal caused by the flight time tof can be calculated from the charge amounts Aa, Ab, Ba and Bb accumulated in the capacitors as follows. φ = tan − 1 Aa − Ab Ba − Bb Strictly speaking, the calculation of the phase angle only applies to sinusoidal signals.

[0022] The distance S between the sensor and the reflecting object can then be calculated from the phase shift ϕ: S = φ ∗ T 2 π

[0023] T is, in this case, the period of the periodic signal.

[0024] The relationships shown for the second possibility also apply if as periodic signal, a sinusoidal one The form is chosen.

[0025] Which method is appropriate depends largely on the specific application and the system design of the TOF sensor.

[0026] While one under Figure 1 The described method is usually used with very powerful pulses at a very low duty cycle (e.g., >25W pulse power; 1:1000). Figure 2 The pulse sequences shown or a duty cycle of 1:1 at pulse powers of 1W to 4W are used. The choice of peak power for sinusoidal modulation of the light source is similar. Limitations on the power used must be observed to ensure eye safety. Use of Active Pixel Sensors (APS)

[0027] It is known to provide an Active Pixel Sensor (APS) for each pixel. An Active Pixel Sensor typically comprises a photodiode and several (field-effect) transistors. Light incident on the photodiode releases charges in the photodiode's depletion region, causing the (reverse) voltage across the diode to decrease as a result of incident photons. An Active Pixel Sensor operates as follows: At the beginning of the activation period, the voltage across the photodiode is set to a defined initial value by means of the reset transistor. During the activation period, the depletion region capacitance of the photodiode is discharged by the photocurrent caused by the incident, reflected light pulse. The voltage across the photodiode decreases proportionally to the intensity of the reflected light pulse and the duration of the portion of the light pulse that falls within the activation period of the Active Pixel Sensor.At the end of the activation period, the voltage drop across the photodiode is read and fed to analog post-processing or directly to an analog-to-digital converter (ADC). For this purpose, each pixel has an amplifier transistor, which is typically switched column by column onto a common readout line for all pixels in a row by means of a selection transistor. Active pixel sensors can be implemented using CMOS (complementary metal-oxide-semiconductor) technology. Since charges in the photodiode's depletion region can also be released, for example, by thermal processes, signal noise typically occurs, which is superimposed on the actual photosignal. Correlated Dual Sampling (CDS)

[0028] To reduce the aforementioned signal noise, it is known to measure the voltage across the photodiode not only once at the end of the activation period, but also again after resetting the photodiode to restore the full reverse voltage (dark voltage), in order to obtain a reference signal with which noise signal components can at least partially be compensated.

[0029] Sensors that implement this type of correlated dual sampling have a corresponding circuit for each pixel, referred to here as a CDS stage. The CDS stage can be implemented, for example, using operational amplifiers. Structure of a pixel

[0030] A pixel of an image sensor can contain the following components. an Active Pixel Sensor (also referred to here as APS cell), a CDS stage assigned to the APS cell, and a Sample and Hold (S&H) stage for reading the pixel. invention

[0031] The aim of the present invention is to design the construction of a TOF sensor as described above in such a way that a dangerous failure is very unlikely and preferably can only occur with a probability of less than 1 in 1100 years (SIL3, or Cat 4, PL e according to ISO 13849-1).

[0032] According to the invention, to achieve this goal, a 3D-TOF with at least one pixel matrix is ​​proposed, which has a pixel structure divided into at least two signal paths (channels A and B), wherein the signal paths are arranged such that the signal propagation times are identical for both channels according to pixel localization (H-tree), and wherein each signal path has its own independent analog signal output (signal A and signal B), and wherein the 3D-TOF sensor also has at least two optical transmitter groups, each transmitter group containing at least one transmitter.

[0033] Preferably, the transmitters are designed to emit infrared light, so that they operate in the infrared spectral range of light.

[0034] Preferably, each signal path of the pixel matrix has its own control logic (logic of signal path A and signal path B). The logic of each signal path is preferably configured to determine amplitudes and distance values ​​by processing the signal values ​​of both channels through cross-exchange of the channel data. According to a first embodiment of the invention, the optically sensitive areas of the pixels of the first channel (pixel A) and the pixels of the second channel (pixel B) are optically connected such that both pixels always receive the same amount of light.

[0035] According to the invention, the optically sensitive surfaces of the pixels of the first channel (pixel A) and the pixels of the second channel (pixel B) are optically connected in pairs such that both pixels receive the same amount of light, for example, by means of a diffuser that distributes the amount of light evenly across the optical surfaces of one pixel from channel A (pixel A) and one pixel from channel B (pixel B). This method is referred to here as optical binning. The signal paths of the pixel groups, each comprising one pixel A and one pixel B, are not electrically connected. In this application, the pixel matrix can be read out by addressing the pixels in channel A and channel B in parallel. Consequently, the signal values ​​of signal A1 and signal B1; of signal A2 and signal B2; ... ; of signal A and signal Bn are always present simultaneously at the output of the pixel matrix.

[0036] Preferably, a comparator stage is provided which is configured to compare the signal values ​​that are applied in parallel to the outputs of the pixel matrix. Here, the digitized output value of the comparator is preferably compared with a preset value in at least one of the logics A or B. If both signal values ​​then show a large difference, and thus, for example, the comparator comparison value for the difference is exceeded, the logic preferably generates a fail signal and the readout process of the pixels of the 3D-TOF sensor is aborted.

[0037] Preferably, an analog-to-digital converter (ADC) is provided, which is assigned to channel A and channel B respectively and configured to digitize the analog signal values ​​A and B in the respective channel and feed them to the logic of channel A and the logic of channel B, respectively. The digitized channel values ​​can then be compared. For this purpose, the channel data is preferably exchanged crosswise. If the calculated difference value exceeds the comparison value, the logic preferably generates a fail signal and the readout process is aborted.

[0038] Preferably, the 3D-TOF sensor is configured to perform an additional, specific signal processing step, such that the functionality of the photoreceptors is tested by comparing signal values ​​A and signal values ​​B in this additional, specific signal processing step. Upon activation of the transmitters (at least one transmitter group), pixels A and B are activated simultaneously; more precisely, the storage capacities of the respective pixels are activated. For two-phase mode, the activation period corresponds to twice the duration of the light pulse, and in four-phase mode, it corresponds exactly to the duration of the light pulse. This signal step ensures that identical signal values ​​are expected at the output of both channels A and B of the optically binned pixel matrix. Identical signal values ​​(i.e., the comparator threshold is undershot) indicate that the pixels are functioning correctly.

[0039] Preferably, the 3D-TOF sensor is configured to perform an additional signal processing step in which the functionality of the transmitter groups is checked. Upon activation of the transmitters (only one transmitter group), either pixel A or pixel B is activated; more precisely, the corresponding memory capacities of the respective pixels are activated. The activation period is twice the duration of the image cycle. In two-phase mode, the duration of the activation period corresponds to twice the light pulse duration, and in four-phase mode, it corresponds exactly to the light pulse duration. As a result of this additional signal step, different signal values ​​are expected at the output of the two channels A and B of the pixel matrix with optically binned pixels.

[0040] The difference in signal values ​​at the output of the analog pixel matrix indicates that the respective transmitter group has sent a light pulse and is functioning.

[0041] Once proof has been provided for one transmitter group, the same procedure can be carried out for the other transmitter group. A preferred method is to alternate the transmitter groups between two image cycles.

[0042] Preferably, the 3D-TOF sensor is configured such that an image cycle begins with the testing of the first transmitter group (claim 9), followed by the testing of the photoreceivers (claim 8), and then the actual image processing begins. Once the image cycle is complete, the next image cycle starts with the next (i.e., at least the second) transmitter group. This switching of transmitter groups is continued as required.

[0043] According to one embodiment, the 3D-TOF sensor can be configured to eliminate the signal step of checking the first transmitter group (claim 9) by having the transmitter groups operate in at least two different wavelength ranges (which may overlap) and by configuring the pixel matrix such that pixel A and pixel B are fixed to one of the two different wavelength ranges by a corresponding optical filter, whereby the respective optical filter allows only one of the two wavelength ranges to pass through. A particularly preferred solution is the use of a wavelength of 850 nm for the first transmitter group and the pixels of channel A, and 940 nm for the second transmitter group and the pixels of channel B. If the signal step according to claim 6 is now performed, the first and second transmitter groups are activated simultaneously in this case.The pixels of channel A receive the reflected light pulse in the first wavelength range (850 nm), and the pixels of channel B receive the reflected light pulse in the second wavelength range (940 nm). If the output of channels A and B then shows equal signal values ​​or signal values ​​in a predefined ratio for the adjacent pixels, this confirms that both the addressed photoreceivers of the two channels and the two transmitter groups are functioning correctly.

[0044] Preferably, optical binning can also be provided to improve the agreement of the signal values ​​of the optical filtering described above with respect to the two wavelengths.

[0045] Preferably, the 3D-TOF sensor can be improved by adjusting the area of ​​the photoreceptors of channel A and channel B according to the selected wavelength so that the different quantum efficiency for different wavelengths is compensated for, i.e., with an intact 3D-TOF sensor, the same signal values ​​are present at the output of channels A and B.

[0046] Preferably, each channel has its own processor connected to the signal path of the channel's logic. Furthermore, the processors are preferably connected to the input of the other channel. Each processor therefore receives the calculated amplitudes and distance values ​​determined by its respective logic.

[0047] The logical operations of each channel logic are preferably verified by each processor calculating at least one test value (random number) completely independently of the other processors. Typically, one test value is calculated for the two-phase mode and two test values ​​for the four-phase mode. Each test value is passed by the processor to the logic of the same channel. Using the passed test values ​​and the test values ​​of the other channel, the logic calculates at least one output value, advantageously two. The same operations are used to calculate the amplitude and distance values ​​for calculating the output value. The output values ​​are passed to the processor of the respective channel. This processor, in turn, calculates the expected output values ​​and compares the output values ​​received from its channel's logic against its own calculated output values. In case of a discrepancy, it outputs a fail signal.

[0048] In a particularly preferred embodiment, one of the two processors is preferably a safety-oriented processor. A safety-oriented processor has at least two independent, non-reactive processing cores and processes two inputs into one output (1oo2 - one out of two). Preferably, this architecture is used to divide the image operations into a safe and a non-safe image processing process. The safe process exclusively checks whether deviations of individual pixels from the expected normal position occur within the defined field of view of the 3D time-of-flight sensor (deformation of the 3D point cloud). If such a deviation occurs, an alarm signal is output.

[0049] The non-safe processor performs complex image processing with the aim of identifying the actual cause of the deformation of the point cloud (e.g., detecting the movement of a person, a hand, a head, etc.).

[0050] According to a second embodiment of the invention, a 3D-TOF sensor is proposed which is preferably characterized by the following features: The 3D-TOF sensor has an image sensor with a pixel matrix, wherein, the pixel matrix has at least two separate, independent blocks and transmits the data into at least two separate signal paths; and / or the pixels of the image sensor's pixel matrix have at least one photosensitive receiver and one memory cell for storing charge carriers; and / or the signal paths consist of at least an analog-to-digital converter (ADC), programmable logic, memory cells, and a connected processor; and / or the image sensor has at least one global shutter that can be triggered by a control signal from the programmable logic and can set a minimum shutter time of less than 100 ns; and / or the signal paths are connected independently and without feedback to each other; and / or the signal paths are connected in such a way that the data of the logics or processors can be exchanged crosswise.and / or each block of the image sensor is assigned its own imaging optic such that the same image of the object side is projected onto the blocks. The distance between the optics is chosen to be so small that the parallax error cannot be resolved, and / or each block is assigned at least one illumination source that is synchronized with the global shutter via the logic.

[0051] The image sensor is preferably characterized by the fact that each block has a separate global shutter which can be controlled by the logic of the associated signal paths.

[0052] The image sensor is preferably characterized by the fact that the global shutters of the blocks depend only on a single logic signal of a single signal path and their dependence is set by adjustable delay times (register values).

[0053] Alternatively or additionally, the image sensor is preferably characterized by the fact that each block has a a narrow optical spectral range is assigned by an optical filter or a system of optical filters, and / or that the system of optical filters is formed by filters directly on the image sensor as well as at least one further filter in the imaging system of the optics, and / or that the optical filters on the blocks of the image sensor, as edge filters, ensure a sharp spectral separation of the blocks, while the filter in the optical imaging system is designed as a band filter and covers the spectral ranges of the blocks, and / or that each spectral range of the blocks is assigned an illumination source that emits light at exactly the wavelength corresponding to the transmission range of the optical filters of the corresponding block.

[0054] This corresponds to a method for reliably acquiring 2D image data with an image sensor of the aforementioned type, comprising the following process steps: In the first processing step, the global shutter is triggered with a predefined exposure time. As a result, a charge image corresponding to the object image is stored in the pixel memory of each of the two blocks. In the second processing step, the image data from the first block is transferred to the processing line of the first block (Readout Block 1), and the image data is transferred to the processing line of the second block (Readout Block 2), and so on. Following this step, the image data is now stored in digital form in the blocks.

[0055] The invention will now be explained in more detail with reference to exemplary embodiments and the figures. The figures show Figure 1: illustrates distance measurement according to the TOF principle, which is based on measuring the time of flight of an emitted light pulse; Figure 2: illustrates a time-of-flight measurement similar to that in Figure 1 with a four-phase scanning of the reflected light pulse; Figure 3: illustrates the basic structure of a TOF sensor according to the invention with two parallel paths and a cross-comparison between them; Figure 4: illustrates how the in Figure 3 The basic structure shown can be implemented at the pixel level of a 3D TOF sensor; Figures 5a-5c illustrate different ways in which the pixels of a pixel matrix can be divided into two channels (Channel A and Channel B); Figures 6a and 6b illustrate in a cross-sectional view how optical binning can be implemented; Figure 7 illustrates, using the example of the pixel matrix from Figure 5a, how the pixels of channels A and B, to be evaluated in pairs, can be optically connected in parallel by optical binning; Figure 8: illustrates the simultaneous sampling of the reflected light pulse; Figure 9: shows a circuit diagram for sampling a pixel matrix according to the Figures 6 and 7 after the in Figure 8 the principle shown; Figure 10: shows that the lighting for emitting the light pulses is also designed in duplicate and can have two transmitter groups with two transmitters each, which are used alternately; Figure 11: illustrates the in place of in Figure 8In the simultaneous scanning of pixels A and B shown, only one of the pixels, either pixel A or pixel B, is scanned in order to obtain a measure of the amount of light from the activated transmitter group; Figures 12a, 12b: illustrate that the respective pixels A and pixel B of the pixel matrix can be sensitive to different wavelengths or different wavelength ranges; Figure 13: shows that in a pixel matrix according to Figure 12a or 12bThe pixels may have different area sizes if the quantum efficiency of the pixels differs for the respective wavelength, in order to obtain the output signals shown with comparable levels by compensating for a lower quantum efficiency with a larger pixel area; Figure 14: shows a logic circuit with an additional comparator to test channels A and B, each operating in two wavelength ranges; Figure 15: shows that instead of two identical processors A and B, two different processors can be used to evaluate the signals of channels A and B, one of which is a safety-related processor; Figure 16: shows the basic structure of an image sensor in which the operational amplifier of the CDS stage also acts as a comparator for comparing the signal values ​​of pixels A and B at the analog pixel level through an extension of the circuit; Figure 17: illustrates the readout cycle;Figure 18: shows the CDS level of the image sensor; Figure 16 in a first switching state; Figure 19: shows the CDS stage of the image sensor Figure 16 in a second switching state; Figure 20: shows the CDS stage of the image sensor. Figure 16 in a third switching state; Figure 21: shows a high-resolution image sensor having several parallel readout channels for block-wise pixel readout; Figure 22: shows an image sensor with two pixel blocks A and B, each sensitive to different wavelength ranges; Figure 23: illustrates that testing the function of the pixel matrix at the block level can be performed in the same way as at the pixel level (compare with Figure 8 Figure 24 shows an image processing circuit for reading a pixel matrix divided into two blocks A and B; Figure 25 shows an image sensor similar to the one shown in Figure 22, in which an additional optical channel is provided; and Figure 26: illustrates the serial scanning of the reflected light pulse by means of an image sensor that has only a single global shutter.

[0056] The invention is based on a general arrangement as described in EN ISO 13849 (machine safety). According to this standard, SIL-3 (Cat. 4; PL e) can be achieved by designing the system with at least two signal paths, so that integrity can be verified by cross-comparing the two parallel signal paths (see figure). Figure 3 .

[0057] According to the Figure 3 The illustrated basic structure of a 3D-TOF sensor 10 has two input units 12A and 12B, which can be formed by pixels; compare Figure 4The output signals generated by the input units, for example, the pixel output signals, are fed to logic units 14A and 14B, which evaluate the output signals of input units 12A and 12B, respectively. Logic units 14A and 14B are connected for cross-comparison purposes. This cross-comparison checks whether logic units 14A and 14B produce the same output signal. If the cross-comparison is successful, the signals acquired by input units 12A and 12B and processed by logic units 14A and 14B are output to their respective output units 16A and 16B.

[0058] The 3D-TOF sensor from the Figures 3 and 4Each system thus has two parallel signal paths, one of which is formed by the input unit 12A, the processing logic 14A, and the output unit 16A, and the second of which is formed by the input unit 12B, the processing logic 14B, and the output unit 16B. The first signal path therefore forms channel A, and the second signal path forms channel B. It is now proposed to apply this parallel system to a 3D time-of-flight (TOF) sensor.

[0059] Therefore, the terms "signal path" and "channel" will be used synonymously in the following. First invention variant

[0060] In the first embodiment of the invention, the principle described above is applied to at least two pixels (pixels A and B) such that continuous parallel processing is achieved by comparing the signal paths A and B. The comparison of the signals is performed at different levels of signal processing, see [reference]. Figure 4 .

[0061] All pixels A and B are arranged as a matrix in equal rows, columns, or as a checkerboard.

[0062] The signal paths are consistently separated into channel A and channel B. This applies to both the analog and digital signal paths, regardless of whether the receiver matrix integrated into silicon is exclusively analog or analog and digital (see [reference]). Figure 5 .

[0063] Therefore, both implementation options exist: integrating only the analog circuit components in silicon and using external logic on an FPGA, or fully integrating all components in silicon or another semiconductor material. Optical binning / Checking the receiver function

[0064] The aggregation of multiple pixels at the circuit level is a technically well-known method. This technique is used to: to reduce the resolution, increase the frame rate, improve sensitivity, and reduce noise.

[0065] Optical binning is introduced as a new feature to ensure that pixels A and B receive identical optical power (see figure).

[0066] Optical binning can be achieved by... homogenizing diffuser The diffuser is positioned over pixels A and B. This arrangement can be such that only the optically sensitive areas (photoreceptors) of pixels A and B are optically connected. Alternatively, the diffuser can be placed over the entire pixel geometry. Another option involves using micro-optics, which are already widely used to increase the fill factor in CMOS image sensors. The micro-optics are then positioned over the diffuser.

[0067] In a further embodiment of the first invention variant, the diffuser and micro-optics can also be realized in one element on a micro-optical basis.

[0068] With respect to the matrix arrangement of pixels A and B, A1 and B1, A2 and B2, ... An and Bn are optically connected to each other, provided that the analog signal path of A1 and B1, A2 and B2, An and Bn, respectively, is identical in length to ensure consistent signal propagation times. This requirement also applies to the digital control signals of the switching transistors of the respective optically connected pixels. bound pixels, see Figure 6 and Figure 7 .

[0069] The functionality of the parallelized photoreceivers is checked by an additional signal step. The purpose of this signal step is to verify that channels A and B deliver an identical signal value. These signal values ​​must be significantly above the noise level (zero signal).

[0070] This is achieved by simultaneously scanning the reflected light pulse. For the two-phase mode, the scanning time is twice as long (see Figure 8 ) the light pulse duration and, in four-phase mode, exactly the length of the light pulse duration. In this case, too, the cycle of sending and sampling the reflected light pulse should preferably be repeated several times until a sufficient amount of charge has accumulated in the memories of both pixel A and B.

[0071] In Figure 9 The circuit diagram is sketched. The optical storage boundPixels (A1 and B2) are selected simultaneously after the integration phase and connected to the output of the analog matrix. The charge levels of pixels A1 and B1 are compared for the first time in the analog signal path (e.g., using a comparator). It is checked whether the voltage difference between A and B is below a predefined threshold. Depending on the number of memory elements in the pixel (two-phase or four-phase), this cycle can be performed once per pixel or twice. Parallelization of the circuitry is also possible. Storage a and b to realize.

[0072] The signal values ​​are then processed in an analog-to-digital converter (ADC A and ADC B; see Figure 9The data is digitized and compared a second time. The operations can be performed by the subsequent logic and / or by the processor A or B following the respective logic A or B. Logic A and logic B differ from processor A and processor B in their implementation. The logics are circuits that implement the logical operations at the circuit level and can, for example, be in the form of an FPGA (Field Programmable Gate Array). The functionality of the processors, on the other hand, is defined by a sequence of functions (e.g., in the form of firmware). The processors can, for example, be digital signal processors (DSPs).

[0073] If the defined difference value is exceeded in the analog or digital domain, then a Fail signal triggered in the logic block and the Image recording stopped. This fail signal is passed directly to the processor interfaces (output) via the signal path.

[0074] The processor output, in conjunction with a safe control system, can then put a machine into the secure hold lead, see Figure 9 . Dual version of the light sources / Testing the functionality of the transmitters

[0075] As soon as the signal path is duplicated, the illumination of a 3D-TOF sensor is also doubled. optically binned Pixels are at least duplicated. At least two transmitter groups are used alternately, see Figure 10 .

[0076] A broadcasting group consists of at least one broadcaster. Figure 10 The diagram shows a double version of the transmitter group, each represented by two transmitters. Transmitters 1 and 3, as well as 2 and 4, are operated in parallel and alternately. LEDs or laser diodes (e.g., VCSEL) can be used as transmitters; see [reference]. Figure 10 . Extended cycle for testing the transmitter function

[0077] Through the optical binningThe additional signal step allows verification that the photoreceiver and the signal processing pipeline deliver identical signal values. Equality of the signal values ​​can be determined via a difference value that should be below a predefined threshold.

[0078] This step does not provide any information about the functionality of the transmitters. A further processing step can provide this information.

[0079] Instead of the in Figure 8 In the described simultaneous sampling of the reflected light pulse from pixel A and pixel B, only one pixel is sampled, i.e., either A or B. Therefore, when determining the difference value in the comparator stage (analog) or in the digital logic section, a signal above a threshold value is generated, which provides a measure of the amount of light from the activated transmitter group; see Figure 11 .

[0080] During the current image cycle, the signal steps are monitored accordingly to check the signal paths. Figure 8 and Figure 11 placed prior to the actual image generation. Check transmitter group 1 Check recipient Image process Check transmitter group 2

[0081] Transmitter group 1 and transmitter group 2 are activated alternately in the cycle. Alternative test mechanism for transmitter function by splitting the transmitter wavelengths

[0082] A particularly advantageous implementation of continuous parallel signal transmission can be achieved by having the transmitter groups transmit at different wavelengths, while pixels A and B can each selectively respond to the respective different wavelengths, see Figures 12a and 12b .

[0083] The advantage of this method is that the one relating to Figure 11The described signal step does not have to be performed iteratively once for the first and once for the second transmitter group, but rather this step can be performed simultaneously, since both transmitter groups can be activated at the same time and selectively detected by the receivers due to the wave selectivity.

[0084] The optical binning This then occurs wavelength-selective. That is, a typical implementation could involve the use of Narrowband filter and homogenization function with a central wavelength of 850nm, as well as for the second receiver group with a central wavelength of 940nm.

[0085] In a special design, the optically sensitive area of ​​the photodetector could be designed to compensate for the different quantum efficiencies of the various wavelengths (in silicon, the quantum efficiency for 940 nm drops by about 30% compared to 850 nm). A larger sensitive area of ​​the receiver with lower quantum sensitivity could compensate for this disadvantage; see [reference]. Figure 13 .

[0086] Figure 12 represents various arrangements of the wave-selective solution. In Figure 12a Both A and B operate at both wavelengths. The functionality of A and B is checked for each wavelength. Consequently, another comparator can be used. Depending on which wavelength is used, either logic A or logic B processes the comparator signal. Figure 14 )

[0087] Is the filter arrangement according to Figure 12bIf the transmitter groups with the two different wavelengths are used and activated simultaneously during the test step, the transmitter and receiver tests can be performed at the same time. If the differential signal at the comparator output between channels A and B is less than a defined threshold, it can be concluded that both transmitter groups and the photoreceivers of both channels are functional. Function of logics A and B

[0088] Logic A provides the control signals for the photoreceivers of channel A, for example, to control the switches in the Figures 18 to 20 The CDS stage shown is for signal path A. Logic B provides the control signals for the photoreceivers of channel B, for example, to control the switches of the [unclear text]. Figures 18 to 20The CDS stage shown is for signal path B. Synchronization of the logic ensures that the correct pixel pairs are addressed during the readout process. The validity of the signal values ​​is determined via the comparator signal.

[0089] In a further processing step, the digital difference signal of the channels is calculated separately in each logic unit. The logic units exchange the digitized channel data for this purpose, provided it is considered valid. If the digitized difference signal of pixels A and B exceeds the threshold value, a fail signal is transmitted to the corresponding processor.

[0090] Both logics independently calculate the signal amplitude and distance values. The calculated values ​​from logic A are passed to processor A, and the values ​​from logic B are passed to processor B. The calculated values ​​are also passed back and forth. Each processor then compares the amplitude and distance values. If the comparison reveals a discrepancy, the corresponding processor outputs a fail signal. Functional test of the logic

[0091] Between image cycles, the processors send a digital input value (random number) to the upstream logic units. These logic units introduce this random number into their processing pipeline and calculate an output value, which is then checked for correctness by a processor routine on each unit. The processing workflow includes the exchange of input values ​​between the logic units. The check process for each logic unit therefore uses the input values ​​from channel A and channel B.

[0092] The calculation functions correspond to the operations for determining the distance and amplitude values. If the comparison results in an error in one processor, a fail signal is output via that processor. Function of processors A and B

[0093] In the first iteration, both processors perform identical operations. Each processor has its own memory and interfaces.

[0094] The outputs OUT A and OUT B are processed at a downstream level (e.g., safety-related control).

[0095] In a second version, which is particularly advantageous for complex image processing operations, e.g., motion analysis of people, a safety-oriented processor is used for one of the processors (e.g., channel A).

[0096] A safety-critical processor internally possesses a parallelized, non-reactive architecture. Such processors are known (e.g., HICore 1 - https: / / www.hima.com / en / products-services / hicore-1 / ) and exhibit, for example, a so-called 1oo2 (one out of two) architecture. In this case, the processing is divided into a safe process and one non-safe process split.

[0097] The division will be explained using the following example.

[0098] Monitoring a work area requires preventing a hand from entering a safety zone or stopping a machine if a hand does enter the area.

[0099] Since hand detection requires certain complex image processing operations (computing time), the process is divided in such a way that a specific area is defined as a three-dimensional safety zone. The safety-related processor then only monitors the position of the point cloud within the defined FOV (field of view) safety zone. A deviation (deformation) of the point cloud would trigger a fail signal. Monitoring the position of the point cloud is a relatively simple operation.

[0100] The other processor analyzes which object has entered the security zone. The corresponding operations are complex and implemented using modern high-level languages. If it is a hand, for example, an instruction is given to the operator, see [reference]. Figure 15 .

[0101] It is also possible to implement the operational amplifier of the CDS stage as a comparator for comparing the signal values ​​of pixel A and pixel B on the analog pixel level by extending the circuit.

[0102] Figure 16 Figure 1 shows the basic structure of a corresponding image sensor. Figures 18 to 20 below illustrate the CDS stages in various switching states in more detail.

[0103] Here, a pixel is built as an example from an APS cell, a CDS and a sample and hold stage (labeled S&H in the figures).

[0104] The CDS stage serves on the one hand as an accumulator for the necessary multiple exposures and at the same time as a stage to suppress the noise components from the reset of the photodiode of the APS cell as well as to separate background light.

[0105] The cross-comparison with the neighboring pixel is now introduced in the CDS stage according to the sequence already explained.

[0106] Figure 17 illustrates the cycle of selection.

[0107] The functionality can be well explained using the function of the CDS stage. Phase 1:

[0108] In pixel A - switches S1a and S3a are closed, switches S2a, S4a and S5a are open.

[0109] In pixel B - switches S1b and S3b are closed, switches S2b, S4b and S5b are open.

[0110] We obtain the following conditions: In pixel A: Q ′ 1 a = C1a Va − Vref ; Q ′ 2 a = C2a Vref − Vouta n In pixel B: Q ′ 1b = C1b Vb − Vref ; Q ′ 2b = C2b Vref − Voutb n

[0111] Figure 18shows pixel A and pixel B in a first switching state of the CDS stages, which they assume in phase 1. Phase 2:

[0112] In pixel A - switches S2a and S4a are closed, switches S1a, S3a and S5a are open.

[0113] In pixel B - switches S2b and S4b are closed, switches S1b, S3b and S5b are open. In pixel A: Q " 1a = C1a Vb − Vref ; Q " 2 a = C2a Vref − Vouta n + 1 In pixel B: Q " 1b = C1b Va − Vref ; Q " 2b = C2b Vref − Voutb n + 1

[0114] Using the node theorem, we can then write: In pixel A: Q ′ 1a + Q ′ 2 a = Q " 1 a + Q " 2a

[0115] Substituting the relationships for the charges into this equation, one obtains the following result after appropriate transformation: Vouta n + 1 = C1a Vb − Va + C2aVout n

[0116] Equivalently, one obtains the ratio at the output of pixel B. Voutb n + 1 = C1b Va − Vb + C2bVout n

[0117] Phase 1 and phase two are repeated several times (N times) to achieve a sufficient output voltage on capacitor C2a and C2b, respectively.

[0118] Figure 19 shows pixel A and pixel B a second switching state of the CDS stages, which they assume in phase 2. Phase 3 - Examination period

[0119] All switches are open except for Sa5 and Sb5, which are closed. During this phase, the operational amplifier acts as a comparator in both pixels.

[0120] The resulting voltage on capacitors Ca2 and Cb2 is then compared with the reference voltage.

[0121] Figure 20 shows pixel A and pixel B a third switching state of the CDS stages, which they assume in phase 3.

[0122] The multiple accumulated differential voltage Vb-Va is present across the capacitance Ca2.

[0123] The differential voltage Va-Vb, accumulated multiple times, is applied across the capacitance Cb2.

[0124] We then compare 3 cases. 1. N x Vouta(n+1) = N x Voutb(n+1) – this would be the ideal case, where pixel A and pixel B deliver the same value. This ideal case is purely theoretical due to the physical properties of real components. 2. N x Vouta(n+1) ≈ N x Voutb(n+1) – The output values ​​of pixel A and pixel B are approximately the same. One of the values ​​is slightly higher or slightly lower than the other. In this case, both pixels can be considered functional.

[0125] A logical 0 appears on both output lines of the comparator.

[0126] 3. N x Vouta(n+1) ≠ N x Voutb(n+1) - The output value of pixel A is either much larger or much smaller than the output value of pixel B. This pixel is considered defective.

[0127] In this case, a logical 1 would be present at pixel output A if Vb>> Va.

[0128] In the case Vb< <Va würde eine logische 1 auf dem Ausgang des Pixel B liegen.

[0129] After the comparison, the CDS level is reset and a new cycle can begin.

[0130] The switches of the respective CDS stage are controlled by the logic A or B assigned to the corresponding signal path. Additional operating mode:

[0131] The introduced difference calculation Vb-Va and Va-Vb can also be used very advantageously in a 4-phase operation. In this case, the difference calculation can be used to calculate the phase angle without reading the individual values, as described above, by calculating the phase shift ϕ as follows: φ = tan − 1 Aa − Ab Ba − Bb

[0132] This requires the use of two additional pixels. Thus, a group of four pixels or a group of two pixels, with a total of four memory cells, results in a complete 4-phase architecture. Second invention variant

[0133] According to a second embodiment of the invention, image sensors (imagers), e.g., in high-resolution CMOS technology, are preferably equipped with several parallel readout channels; see Figure 21 The goal of this architecture is to achieve a high frame rate despite high resolution. The readout channels typically combine either multiple rows or multiple columns of an image sensor matrix. We will subsequently refer to such a combination as a block or imager block; see [reference]. Figure 22 .

[0134] In detail, Figure 22 : 1 - Image sensor, 2 - Optically sensitive area, 4 - Imager Block A with spectral filter A, 5 - Imager Block B with spectral filter B, 7 - Readout channel for Imager Block A, 8 - Readout channel for Imager Block B.

[0135] The second invention variant makes use of this division of the image field into selection blocks.

[0136] The further variant of the invention will be explained in more detail below using a first embodiment as an example.

[0137] The image field, which preferably has a very high resolution (e.g., 1280 x 1024 pixels), is divided into blocks according to the assigned readout channels. Each readout block uses a global shutter that can be triggered separately.

[0138] Each block is assigned its own imaging optic. The distance between the optics is very small, so that parallax is no longer resolved on the image side even at a small object distance. This means that the same image of the object side is produced on each block.

[0139] This arrangement of optics and image sensor is used in a parallelized readout and image processing circuit according to Figure 4 used. Calculation of a depth image:

[0140] This embodiment assumes a two-channel setup. The image sensor (imager) is divided into block A and block B. In this embodiment, each block has 640 x 1024 pixels. The digital readout channels for block A and block B are routed to the corresponding logic blocks A and B. It is irrelevant whether these logic blocks already incorporate parts of an integrated circuit or an external logic circuit, such as an FPGA.

[0141] Logic units A and B control the respective global shutter signals and, accordingly, the illumination sources LA and LB. The illumination sources are operated synchronously. Both sources are triggered simultaneously with the global shutter signal of block A at time t0 for a duration ti = tA. The light pulse is reflected off a target and registered by all receivers A of block A with a delay of the flight time tof. The amount of charge (photoelectrons) resulting from the registration of the light pulse is stored on the storage capacity (A1) belonging to each receiver. The receivers of block A can only receive a portion of the reflected light pulse, which corresponds to the time difference tA - tof.

[0142] With a time delay t0+tA, the global shutter of the receivers in block B is activated. Receiver B is active for time tB (in the example, tA=tB). The receiver also registers only a portion of the reflected light pulse, which is proportional to the time difference tof+ti-tA (if ti=tA, this is tof). The corresponding charge amounts are stored on the storage capacity B1 belonging to each receiver in block B.

[0143] Since the reflected signals can be very weak depending on the reflectivity of the target and the distance, an amplification mechanism is necessary. This is achieved by repeatedly performing the measurement described above and accumulating the recorded charge quantity in the corresponding storage capacities until a sufficient signal strength is reached. This process can be repeated, for example, 100 times.

[0144] As a result, the storage capacities A1 of block A contain an image for the time span t0 to tA, while the storage capacities B1 contain an image for the time span tA to tB.

[0145] These two images are transmitted to the respective logic units via their respective readout channels. The logic units then exchange the image information crosswise and each separately calculates the depth image pixel by pixel according to the aforementioned rules: Q1 = A / A + B (elimination of the dependence on the degree of reflection through quotient formation) Q2 = B / A + B tof1 = 1 − Q1 * T tof2 = Q2 * T tof = 1 2 tof1 + tof2 (Mean) S = 1 2 c * tof

[0146] The calculation rule is applied such that pixels are always calculated according to their position in the receiving matrix. Thus, A0,0 corresponds to B0,0; A0,1 corresponds to B0,1; .... Am,n corresponds to Bm,n according to the chosen arrangement: A 0,0 … .. A 0 , n . … … . . Am , 0 … . Am , n B 0,0 … .. B 0 , n . … … . . Bm , 0 … . Bm , n

[0147] At the end of the calculation, both logics contain a distance image. Both distance images should be identical except for a temperature-dependent calibration factor, which can be determined once. Functional testing by comparing the depth image:

[0148] The identity of the two depth images A and B is verified in a further operation by pixel-by-pixel comparison. This comparison can be performed on both logic units or on the downstream processors.

[0149] A threshold can be defined for comparison, specifying how many pixel pairs may deviate from a defined expected comparison value. This threshold defines whether the pixel pair yields the same values ​​or not, and what constitutes equality or inequality. Furthermore, the number of inequality pairs can be used as a decision parameter.

[0150] If the defined threshold is exceeded on at least one processor, at least one of the processors generates a fail signal at the output.

[0151] Comparing depth images A and B already involves summing several comparison functions. These include pixel sensitivity, gate functions (integration / accumulation), readout, ADC, and logic operations.

[0152] The validity of the complete functionality of the pixel matrix can also be achieved by a pixel-wise comparison of an amplitude image. Testing the function of the pixel matrix in the 2D amplitude image:

[0153] Just like at the pixel level (see above and Figure 8 and 9The function of the pixel matrix in the 2D amplitude image can also be tested at the block level; see Figures 23 and 24. This is achieved by simultaneously sampling the reflected light pulse in both block A and block B. In this case, too, the cycle of sending and sampling the reflected light pulse must be repeated several times until a sufficient amount of charge has accumulated in the memory of the pixels in both block A and block B.

[0154] After the accumulation cycle is complete, the images in blocks A and B are read and compared pixel by pixel on both logic units. The comparison values ​​can be evaluated both on the logic unit and on the downstream processors.

[0155] A threshold can be defined for comparison, determining whether the pixels must be considered equal or unequal. Furthermore, it can be defined how many pixel pairs may differ from each other (be unequal) within a defined range before the fail signal is generated. If the defined threshold is exceeded on at least one channel, at least one of the processors generates a fail signal at its output. Testing the light sources:

[0156] The described test cycle is equally suitable for testing the light sources. In this embodiment, a total of four light sources are used. In this embodiment, two are controlled by the logic of channel A and two by channel B, respectively.

[0157] The lighting sources can now be tested by using either the sources of channel A or the sources of channel B in each test cycle.

[0158] In contrast to comparing receiver signal values ​​by cross-comparison, the comparison of lighting sources is carried out by serial comparison in the respective channel.

[0159] In the present example, the test procedure is as follows: 1. Start a light pulse with twice the length of the depth image acquisition (e.g., 60 ns) of the illumination source in channel A. 2. Simultaneously start the global shutters in both blocks A and B with twice the length of the light pulse (e.g., 120 ns). 3. Store the images of block A (image A0) and block B (image B0) in the logic blocks of their respective channels. 4. Start a light pulse with twice the length of the depth image acquisition (e.g., 60 ns) of the illumination source in channel B. 5. Simultaneously start the global shutters in both blocks A and B with twice the length of the light pulse (e.g., 120 ns). 6. Store the images of block A (image A1) and block B (image B1) in the logic blocks of their respective channels. 7. Each pixel value A(i,j) from the A submatrix and each pixel B(i,j) from the B submatrix is ​​compared with the corresponding values ​​from the exposure with the LB source when exposed with the LA source.A maximum difference value is defined. If this value is exceeded, a defect exists and a fail signal can be triggered. It is also conceivable to trigger the fail signal only when the difference value is exceeded for a predefined number of pixel pairs. The comparison values ​​can be evaluated both in the logic and on the downstream processors.

[0160] The invention is explained using a second embodiment.

[0161] As in the first embodiment, the image field, which preferably has a very high resolution, e.g., 1280 x 1024 pixels, is divided into blocks according to the assigned readout channels. In contrast to the first embodiment, each readout block is further subdivided. That is, each readout block operates with two global shutters that can be triggered separately.

[0162] Each subblock (Aa, Ab, Ba, Bb) is assigned its own imaging optic. The distance between the four optics is very small, so that parallax is no longer resolved on the image side even at a small object distance. This means that the same image of the object side is produced on each subblock.

[0163] This arrangement of optics and imager is integrated into a parallelized readout and image processing circuit according to Figure 24 used.

[0164] Calculation of a depth image: This embodiment still assumes a two-channel setup. However, the image sensor (imager) is now divided into blocks Aa and Ab, and blocks Ba and Bb. In this embodiment, each block has 640 x 512 pixels. The digital readout channels for blocks A and B are routed to the corresponding logic units A and B.

[0165] Logic modules A and B control the respective global shutter signals and, accordingly, the lighting LA and LB. The lighting sources are operated synchronously.

[0166] While in the first embodiment a signal sampling with two integration periods is described (2-phase), the signal is now sampled four times, with the sampling points being offset by ¼ of the period of the signal.

[0167] Referring to the exemplary embodiment with a light pulse of length 30 ns, this would result in a sampling sequence that could be described as follows: Subblock Aa from 0 to 30 ns, for subblock Ba 30 ns to 60 ns, for subblock Ab 60 ns to 90 ns, and for subblock Bb 90 ns to 120 ns; see also Figure 2 .

[0168] The relationships shown also apply when a sinusoidal shape is chosen as the periodic signal.

[0169] Since the reflected signals can be very weak depending on the reflectivity of the target and the distance, it is necessary to implement the amplification mechanism described above. This is achieved by repeatedly performing the measurement and accumulating the recorded charge amount in the corresponding storage capacities of the pixels within the subblocks until a sufficient signal strength is reached. This process can be repeated, for example, 100 times.

[0170] As a result, the storage capacities of the pixels in the subblocks each contain one image for each quarter of the period. From this, the phase and thus the distance S per pixel can be calculated using the following rule. S = φ * T / 2 π T is, in this case, the period of the periodic signal.

[0171] Strictly speaking, the phase angle calculation only applies to sinusoidal signals. For rectangular signals, appropriate corrections must be applied to ensure high linearity of the distance characteristic. One way to improve the linearity of the distance characteristic in 4-phase sampling is by using T / 2 sampling.

[0172] The testing procedures described in the first example are carried out in the same way.

[0173] In the following, we will explain a third embodiment, which is described in Figure 26 is shown.

[0174] Unlike the previous examples, we only have one global shutter available for the entire pixel matrix. Whereas in the previous examples, two global shutters were used for the phases of the depth image (see Figures 1 , 8 , 11 and 23), so that two phase images can be recorded simultaneously (in parallel), is due to the fact that in Figure 26 The illustrated embodiment is based on an image sensor with only a single global shutter. In this case, the two phase images must be acquired sequentially (serially), as shown. Figure 26 This is illustrated. Furthermore, the same optical setup as in the first embodiment can be used. The pixel matrix is ​​divided into two blocks, A and B, along the two readout channels. Each block is assigned its own imaging optic. The distance between the two optics is very small, so that parallax is no longer resolved on the image side, even at a small object distance. This means that the same image of the object side is generated on each block.

[0175] In contrast to the first embodiment, each block is additionally equipped with an optical filter or a system of filters. The filter or filter system limits the transmitted wavelength of each block to a narrow spectral band, which corresponds to the source wavelength of the associated light source; see Figure 25 .

[0176] The exemplary embodiment in Figure 25 differs from the one from Figure 22 through an additional optical channel 3 and an additional readout channel 6. In detail, this shows Figure 25 : 1 - Image sensor, 2 - Optically sensitive area, 3 - Additional optical channel, 4 - Imager Block A with spectral filter A, 5 - Imager Block B with spectral filter B, 6 - Additional readout channel, 7 - Readout channel for Imager Block A, 8 - Readout channel for Imager Block B.

[0177] For all illustrated embodiments, the respective 3D-TOF sensor can be configured to eliminate the signal step of checking the first transmitter group by having the transmitter groups operate in at least two different wavelength ranges (which may overlap) and by configuring the pixel matrix such that pixel A and pixel B (or block A and block B) are fixed to one of the two different wavelength ranges by a corresponding optical filter, whereby the respective optical filter allows only one of the two wavelength ranges to pass through. A particularly preferred solution is the use of a wavelength of 850 nm for the first transmitter group and the pixels of channel A, and 940 nm for the second transmitter group and the pixels of channel B.

Claims

1. 3D TOF sensor with at least one pixel matrix that has a pixel structure divided into at least two signal paths (channels A and B), wherein the signal paths are routed in such a way that the signal propagation times are identical for both signal paths according to the pixel localization and wherein each signal path has its own independent analog signal output (signal A and signal B), wherein the 3D TOF sensor also has at least two optical transmitter groups, of which each transmitter group contains at least one transmitter, wherein the optically sensitive surfaces of the pixels of the first signal path (pixels A) and the pixels of the second signal path (pixels B) are optically connected in pairs by optical binning in such a way that both pixels receive the same amount of light and wherein the signal paths of the pixel groups, each comprising a pixel A and a pixel B, are not electrically connected to one another.

2. 3D TOF sensor according to claim 1, wherein the transmitters operate in the infrared spectral range.

3. 3D TOF sensor according to claim 1 or 2, wherein a separate logic (logic of signal path A and signal path B) for controlling and processing the signal values originating from the pixels of the corresponding signal path is provided for each signal path of the pixel matrix4. 3D TOF sensor according to claim 3, wherein the logic of each signal path is designed to determine the amplitudes and distance values by processing the signal values of both signal paths in such a way that the signal values are exchanged crosswise.

5. 3D TOF sensor according to claim 3, wherein the optically sensitive surfaces of the pixels of the first signal path (pixels A) and of the pixels of the second signal path (pixels B) are connected by means of a diffuser that distributes the amount of light evenly across the optical surfaces of respectively one pixel from first signal path (pixel A) and one pixel from the second signal path (pixel B).

6. 3D TOF sensor according to at least one of claims 1 to 5 that comprises a comparator step designed to compare the signal values present in parallel at the outputs of the pixel matrix with each other and / or with a default value as comparator threshold value7. 3D TOF sensor according to at least one of claims 1 to 6 that comprises two analog-to-digital converters that are designed to digitize the analog signal values A and B in the respective first signal path A or second signal path B and that are respectively connected to the logic of first signal path A or the logic of second signal path B.

8. 3D TOF sensor according to at least one of claims 1 to 7 that is designed to test the functionality of the photo receivers by comparing the signal values A and the signal values B in an additional special signal processing step; it does so by activating the pixels of the first signal path A and the pixels of the second signal path B at the same time the transmitters are started.

9. 3D TOF sensor according to at least one of claims 1 to 8 that is designed to test the functionality of the transmitters by a further additional signal step; it does so by activating either the pixels of the first signal path A or the pixels of the second signal path B at the same time the transmitters in only one transmitter group are started.

10. 3D TOF sensor according to at least one of claims 1 to 9, characterized in that each signal path comprises a processor which is connected to the signal path of the logic of the channel, wherein the calculated amplitudes and distance values, which were transmitted by the respective logic, are transmitted to each processor11. 3D TOF sensor according to claim 10, wherein one of the two processors is a safety-related processor that has at least two independent reactionless computing cores and processes two inputs on one output (1oo2 - one out of two).

12. 3D TOF sensor with an image sensor that comprises an image recorder with a pixel matrix, wherein the image recorder can read out the image data in individual, at least two, separate independent blocks of the pixel matrix and transfer the data in at least two separate signal paths, the pixels of the pixel matrix of the image recorder comprise at least one photosensitive receiver and a storage cell for storing charge carriers, and the signal paths each have an analog-to-digital converter, a programmable logic, storage cells and a processor connected to them, wherein the image recorder comprises at least one global shutter for all blocks that can be released by a control signal of the programmable logic and can set a minimum shutter time of less than 100ns, and the signal paths are wired independently and reactionlessly from each other, and the signal paths are wired such that the data of the logics or the processors can be exchanged crosswise wherein each block of the image sensor is assigned its own imaging optics in such a way that the same image of the object side is imaged on the at least two separate in-dependent blocks of the pixel matrix13. 3D TOF sensor according to claim 12, characterized in that at least one light source is assigned to each block, which is synchronized with the global shutter via the logic14. 3D TOF sensor according to claim 12 or 13, characterized in that the sensor features a separate global shutter in each block that can be controlled by the logic of the assigned signal paths