Image sensor

Abstract

To provide a photon counting image sensor configured to reduce malfunctions.SOLUTION: An image sensor includes photon-counting detector elements (10) and a first group of pulse shapers (20) to convert signals (12) from the photon-counting detector elements (10) into pulse shaper output streams in which pulses have at minimum a first pulse length (t1). A connection unit (30) combines a plurality of the pulse shaper output streams into at least one macropixel stream. The macropixel stream is processed by at least one monostable / astable circuit (40) which outputs a processed stream (42) in which pulses have a second pulse length (t2). The second pulse length (t2) is shorter than the first pulse length (t1) by a time span t_PF and thus a pulse-free interval (PF) for at least the time span t_PF is created after each pulse in the processed stream (42).SELECTED DRAWING: Figure 11

Description

[Technical field]

[0001] The present invention relates to image sensors that include photon-counting detector elements and methods of operating image sensors. [Background technology]

[0002] Image sensors with photon-counting detector elements, such as single-photon avalanche diodes (hereinafter SPADs), are described, for example, in WO 2020 / 151838 A1 and WO 2020 / 207571 A1, as well as in the following prior art documents:

[0003] [1]: A. Carimatto et al., “Multipurpose, fully-integrated 128x128 event-driven MD-SiPM with 512 16-bit TDCs with 45 ps LSB and 20 ns gating in 40nm CMOS technology”, IEEE SOLID-STATE CIRCUITS LETTERS, Vol. 1, No. 12, December 2018, doi: 10.1109 / LSSC.2019.2911043.

[0004] [2]: LHC Braga et al., “A Fully Digital 8x16 SiPM Array for PET Applications With Per-Pixel TDCs and Real-Time Energy Output,” IEEE J. Solid-State Circuits, Vol. 49, no. 1, pp. 301-314, Jan. 2014, doi: 10.1109 / JSSC.2013.2284351.

[0005] [3]: M.-J. Lee et al., “High-Performance Back-Illuminated Three-Dimensional Stacked Single-Photon Avalanche Diode Implemented in 45-nm CMOS Technology,” IEEE J. Sel. Top. Quantum Electron., Vol. 24, no. 6, pp. 1-9, Nov. 2018, doi: 10.1109 / JSTQE.2018.2827669。

[0006] [4]: S. Lindner, C. Zhang, I. M. Antolovic, M. Wolf, and E. Charbon, “A 252 × 144 SPAD Pixel Flash Lidar with 1728 Dual-Clock 48.8 PS TDCs, Integrated Histogramming and 14.9-to-1 Compression in 180NM CMOS Technology,” in 2018 IEEE Symposium on VLSI Circuits, Jun. 2018, pp. 69-70, doi: 10.1109 / VLSIC.2018.8502386。

[0007] Such image sensors consist of an array of detector elements connected to electronics that measure and process the timing and intensity information of the detected photons. The processed information is transmitted to an external system. Raw data is rarely transmitted because the requirements of communication systems increase rapidly to the limit where it is impractical. Rapid advances in technology mean that the number of SPADs contained in a SPAD array is increasing more and more, making the interconnection between the SPADs and the electronics difficult to implement. In such a situation, it becomes impossible to provide electronic modules corresponding to each individual SPAD. Architectures that combine pulses from multiple SPADs into one single signal and use it as an input for electronics are very common. SPADs whose outputs are combined are said to form a group or macropixel with the aggregated information of its individual SPADs. In this type of architecture, the address or location of the SPAD within that group is usually lost, although the timing and intensity are preserved. In many applications, it is not necessary to distinguish which of the multiple SPADs in a macropixel generated a pulse. Electronics used for macropixels may include time-to-digital converters (TDCs), counters, level comparators, triggers, and scalers, etc. Regardless of the module required for an application, a glitch on the output line of the macropixel can cause the module to become unstable. The photons that trigger the SPAD arrive at the sensor randomly and asynchronously. The combination of pulses generated, usually but not limited to ORed, can create spurious pulses that are smaller in voltage or faster / shorter in time than the expected pulse from the electronics. In systems where it is common to have banks of electrons that are double-buffered and swapped over time, spurious pulses can also occur when this switching coincides with the arrival of a photon. These mismatched pulses or glitches do not meet the input rating of the next electronics and can cause failures of various severity that are unacceptable in the target field.

[0008] For ease of understanding, please refer to FIG. 1 which shows a related image sensor 1' comprising an array of photon-counting detector elements 10, i.e. SPADs. The signals 12 of the photon-counting detector elements 10 are processed by pulse shapers, i.e. monostables 20'. For consistency, the output of the pulse shapers / monostables is referred to as the pulse shaper output stream 22. A pulse 13 in one of the signals 12 represents a registered photon. The monostables 20' shorten the pulse length such that a pulse 23 in the pulse shaper output stream 22 has a first pulse length t1 that is shorter than the length of the original pulse 13.

[0009] A connection unit 30, such as an OR element, combines the pulse shaper output streams 22 originating from the monostables 20' into a macropixel stream 32. This is shown in more detail in FIG. 2, which shows two pulse shaper output streams 22 (originating from two monostables 20' and therefore from two photon-counting detector elements 10), each of which contains several pulses 23 indicative of detected photons. The macropixel stream 32 contains all the pulses 23 of the two pulse shaper output streams 22. FIG. 3 shows a similar case, but the pulses 23 originating from different SPADs are close to each other. In the macropixel stream 32, this results in a short duration t_short of the gap between two pulses 23. Since a photon can hit a SPAD at any time, the duration t_short can have any value. However, if t_short is too small, errors may occur in the following electronic module: Figure 4 shows the requirements for the rise and fall times of an electronic module such as a counter, a TDC, or a level detector. The pulses of the input signal 52 of the electronic module must have a minimum length tp and the gaps between the pulses must have a minimum length tn. The minimum length tn of the pulse-free interval is required as a recovery time to return the internal state of the electronic module to its original state so that further pulses can be correctly processed. Otherwise the electronic module may fail or output incorrect results. The exact values ​​of tp and tn depend on the specific electronic module. With reference to Figure 3, a pulse length t1 can be selected by appropriately designing the monostable 20', so that it is guaranteed that t1 ≥ tp is always satisfied and that the pulse 23 always meets the requirement of tp. However, the time width t_short of the gaps between the pulses 23 may have any value, and therefore t_short may be shorter than the minimum required time width tn, which may cause errors in the electronic module.

[0010] This is one example where a conventional image sensor may generate glitches or pulses that do not meet the requirements of the electronic module and therefore cause errors in subsequent processing steps, e.g., metastability of the electronic module, erroneous photon count values, or reduced signal-to-noise ratio. The present invention addresses the problems caused by irregular pulses and avoids malfunction of downstream electronics. Summary of the Invention

[0011] It is an object of the present invention to provide an image sensor and method that reduces or eliminates irregular pulses or glitches in the image sensor, in particular in its macropixel output lines, or in electronic modules connected to the macropixel output lines.

[0012] This object is achieved by an image sensor and a method comprising the features of the respective independent claims.

[0013] The image sensor of an embodiment of the present invention comprises photon-counting detector elements and a first group of pulse shapers (e.g. monostables) arranged to convert signals from the photon-counting detector elements into a pulse shaper output stream whose pulses have a minimum first pulse length. The pulse shapers may be arranged in parallel with each other, i.e. each pulse shaper in the first group receives a signal from another photon-counting detector element. The number of pulse shapers in the first group may match the number of photon-counting detector elements. Each output of one of the pulse shapers in the first group may be referred to as a pulse shaper output stream. A respective pulse shaper output stream may be generated for each photon-counting detector element. The image sensor may include The image sensor further comprises a connection unit configured to combine the plurality of pulse shaper output streams into at least one macropixel stream. The image sensor further comprises at least one monostable / astable circuit (i.e., at least one monostable circuit and / or at least one astable circuit) arranged to receive the at least one macropixel stream and to output a processed stream in which the pulses have a second pulse length. The at least one monostable / astable circuit is configured such that the second pulse length is shorter than the first pulse length by a time width t_PF to generate a pulse-free interval of at least the time width t_PF after each pulse in the processed stream.

[0014] In an embodiment of a method of operating an image sensor of the present invention, photons are received by photon-counting detector elements that output signals indicative of the received photons. A first group of pulse shapers converts the signals from the photon-counting detector elements into a pulse shaper output stream in which the pulses have a first pulse length. A connection unit combines the multiple pulse shaper output streams into at least one macropixel stream. At least one monostable / astable circuit processes / receives the at least one macropixel stream and outputs a processed stream in which the pulses have a second pulse length that is shorter than the first pulse length by a time width t_PF. This produces a pulse-free interval of at least a time width t_PF after each pulse in the processed stream.

[0015] The monostable / astable circuit guarantees a defined minimum distance from each pulse to the next. This allows to exclude two pulses in the processed stream that are too close to each other. The pulse-free interval is generated in particular by using a series arrangement of monostables, where the latter monostable (i.e. the monostable / astable circuit) takes less time to return to its stable state than the first monostable (i.e. the pulse shaper). A pulse-free interval of t_PF reduces the risk of errors in the subsequent electronics that require a certain minimum pause between pulses. If t_PF is selected according to the required minimum pause time, i.e. if t_PF is equal to or greater than the required minimum pause time, the risk of two successive pulses being too close to each other and not being processed correctly can be completely eliminated. For the sake of clarity, it is to be noted that the pulse-free interval is only generated if a monostable / astable circuit is placed downstream of the connection unit, which receives the macropixel stream. In practice, the arrangement with monostable / astable circuits creates dead times during which photon registration cannot be passed, but this dead time avoids or reduces the risk of malfunction of further electronics that may process the processed stream.

[0016] Optional Embodiments Advantageous variants of the inventive image sensor and the inventive method are the subject of the dependent claims and are explained in more detail in the following description.

[0017] Electronic module and length of pulse-free interval The image sensor may include an electronics module that processes a signal stream containing information from the photon-counting detector elements. The electronics module may be built on the same chip as the photon-counting detector elements.

[0018] In general, the electronic module may be any electronic device or circuit that performs a measurement of the strength or time of a received signal. The electronic module may consist of or include at least one of a time-to-digital converter (TDC), a digital counter, a level comparator, a trigger, a scaler, an inverter, a buffer, a pull-up, a latch, a level monostable, a delay line, a clock divider, a phase detector, a prescaler that reduces the frequency of an input signal of a downstream component, or any other digital circuit. The TDC outputs digital data indicating, for example, the arrival time of a detected photon. The digital counter outputs a count value indicating the number of detected photons. The level comparator may be configured to output data indicating whether or when the level of the received signal exceeds a particular threshold. The scaler may be configured to aggregate the pulses in the received signal into a single output pulse, which represents, among other things, a predetermined number of input pulses. Other types of scalers may be configured to convert the voltage level of the received signal to a different voltage level.

[0019] While a conventional image sensor may include an electronics module arranged such that it directly receives the combined output of the pulse shaper (referred to herein as the macropixel stream), in various embodiments of the invention the electronics module is arranged to receive the processed stream, i.e. the signal output by a monostable / astable circuit arranged downstream of the pulse shaper.

[0020] The electronic module may have a recovery time of tn, i.e. the electronic module requires a pause (a pulse-free interval) of at least tn length between two pulses in the processed stream in order to distinguish or correctly process a pulse in the processed stream. A "required" pause may be understood in the sense that if tn is not met, a pulse will be erroneously processed, or if the pause is shorter than tn, the risk of erroneous processing exceeds a predefined threshold.

[0021] The at least one monostable / astable circuit may be configured such that t_PF is at least as long as tn, i.e. t_PF≧tn. In other words, the second pulse length t2 generated by the at least one monostable / astable circuit is shorter than the first pulse length t1 of the pulse shaper by at least tn. In this case, it is always guaranteed that the two pulses have a sufficient distance so that the recovery time of the electronic module is met.

[0022] In order to avoid unnecessarily long dead times, an upper limit for the minimum pulse-free interval t_PF may be defined. In particular, the first group of pulse shapers and the at least one monostable / astable circuit may be configured such that t_PF is smaller than twice tn (or three times tn).

[0023] The electronic module may also require that the pulses in the processed stream have a minimum pulse length of tp. This requirement is met if at least one monostable / astable circuit is configured such that the second pulse length t2 is at least as long as tp. Optionally, the first group of pulse shapers may be configured such that the first pulse length t1 is at least as long as tp+tn, to allow sufficiently long pulse-free intervals. Optionally, t1 is smaller than 2*(tp+tn), to avoid unnecessarily long dead times.

[0024] If the processed streams are provided to electronic modules with different recovery times tn, the longest tn needs to be selected in the described embodiment. The image sensor may generate multiple processed streams that are input to various electronic modules (e.g., TDCs and counters) with different requirements for tp and tn. In this case, the monostable / astable circuits may be different from each other and generate different second pulse lengths t2 depending on the tp and tn of each of the following electronic modules. All pulse shapers of the first group of one macropixel (i.e., pulse shapers connected to the same monostable / astable circuit) may generate pulses of the same pulse length t1, but the macropixels may have different t1 for each of them, i.e., t1 of the pulse shapers of the first group connected to one monostable / astable circuit may be different from t1 of the pulse shapers of the first group connected to another monostable / astable circuit. In this way, t1 and t2 can be selected to suit the respective requirements for tp and tn. Optionally, the first group of pulse shapers and / or monostable / astable circuits may be configured to allow flexibly adjusting t1 and t2 during operation of the image sensor, whereby the t1 and t2 values ​​can be flexibly adjusted depending on the current connections set in the connection unit to adjustably generate and route processed streams to the various electronic modules, thereby allowing setting suitable t1 and t2 for the respective tn and tp of each electronic module taking into account the flexible instantaneous formation of the macropixel.

[0025] Pulse shapers, e.g. monostable and monostable / astable circuits A pulse shaper can be understood as an electronic component that outputs a pulse for each pulse received, the length of the output pulse being at least t1. The length of all output pulses may be the same or different lengths, greater than or equal to t1. A pulse shaper may be or include a monostable circuit, understood in particular as an electronic circuit or unit configured to receive an input signal and process the input signal to generate an output signal that changes between a stable state (no signal / no pulse) and an unstable state (a pulse when a photon is recorded), the monostable having a defined time for its output to change from an unstable state to a stable state. The defined time leads to a defined pulse length. The photons recorded on the detector elements lead to a pulse of a common length defined by the monostable / pulse shaper. A monostable is also called a monostable circuit or a monostable multivibrator.

[0026] The first group of pulse shapers are, for example, level monostables that are triggered when the level of the signal from the photon-counting detector element exceeds a predefined threshold. In principle, the pulse shaper can also be an edge-delay monostable (edge ​​detector monostable) that is triggered by an edge of the received signal, i.e. the start of a pulse representing the detection of a photon. It is also possible that some or all of the pulse shapers are integrated with the respective photon-counting detector element, i.e. the photon-counting detector element and the pulse shaper form one integral unit.

[0027] In general, a monostable / astable circuit may be understood as a monostable circuit or astable circuit. An astable circuit is a circuit or electronic component that produces an output that switches between at least two states, none of which is permanently stable, but which switches upon receiving the first side of the received pulse and switches back after a time shorter than t1 (or after a time t2, where t2 indicates a certain value or range of values, all of which are shorter than t1). Several monostable / astable circuits (each processing one of the macropixel streams) may include different types of monostables or astables, or some monostables and some astables. Monostable / astable circuits may be specifically level monostables or edge-delay monostables (EDM). However, level monostables usually have a recovery time required to reset their internal state after receiving a pulse. Thus, a level monostable requires a sufficient time difference between the first and second pulses received, otherwise only the first pulse will be properly processed to the desired pulse length (here t2), while the second pulse will be incorrectly processed to a pulse length shorter than t2. Such short spurious pulses can cause errors in subsequent electronic modules. This problem is avoided if at least one monostable / astable circuit is an EDM, which is triggered by the edge of each new pulse and outputs a pulse of constant length even if the first pulse is followed by a second pulse without a pause in between. Moreover, it may also happen that the second pulse starts while the first pulse has not yet completely finished, i.e. the rising edge of the second pulse may overlap to some extent with the falling edge of the first pulse. If the overlapping edges are still recognizable in the EDM, both of the two pulses with length t2 are output. If the first and second pulses overlap so strongly that the rising edge of the second pulse cannot be identified by the EDM as a new pulse, the EDM will output only one pulse instead of two, and will not output any spurious pulses that could cause errors in subsequent electronics.The suppression effect of approximately half the supply voltage creates an extra dead time effect that can be as short as 200ps.

[0028] Connection Unit The connection unit is configured to combine multiple asynchronous pulse shaper output streams (of the first group of pulse shapers) into at least one macropixel stream. A macropixel stream is thus a combination of two or more pulse shaper output streams and contains all pulses of those pulse shaper output streams. If one of those pulse shaper output streams contains a pulse at a particular time, the macropixel stream also contains a pulse at this particular time. If none of those pulse shaper output streams contains a pulse at a particular time, the macropixel stream also does not contain a pulse at this particular time. If pulses of different pulse shaper output streams overlap each other in time, the macropixel stream contains one pulse of a length defined by the overlapping pulses. The connection unit may be configured to output multiple macropixel streams with different combinations of the pulse shaper output streams. Since the connection unit combines the outputs of multiple pixels (=photon-counting detector elements), the connection unit can be considered as defining a macropixel consisting of photon-counting detector elements whose outputs are combined, and as a result the output of the connection unit is called a macropixel stream.

[0029] The connection unit may in particular be a connection matrix configured to flexibly combine any of the pulse shaper output streams to generate a reconfigurable macropixel. The connection matrix may include an OR matrix where a signal is output if any of the multiple inputs of the logical OR unit receive a signal. The connection matrix may also include logical units other than OR connections for combining multiple inputs.

[0030] The connection matrix can be configured to flexibly change which outputs are combined at run-time, for example using an internal memory. Any combination of outputs is possible, for which the connection matrix can be equipped with SRAM memory and / or LUTs (Look Up Tables). The connection matrix can be flexibly changed to change the number of macropixels. Furthermore, the number of photon-counting detector elements combined into one macropixel can be flexibly set. The connection matrix can be set in such a way that some of the pulse shaper output streams are simply forwarded without being combined with other pulse shaper output streams.

[0031] Bank switching support using artificial pulses (false photons) When bank switching splits the pulse in two, it can create one or two spurious pulses that can cause errors in the electronics, as we'll see below.

[0032] When an electronic component such as a counter is enabled or disabled and receives a pulse at the same time, spurious pulses can occur that can cause the counter to go into a metastable state. To avoid counter metastability, pulses during bank switching must be suppressed / rejected. This is achieved by introducing a photon-emulating signal (a "false photon") prior to the switch for a sufficient period of time until the switch is complete. Such techniques can be accompanied by anti-glitch circuitry as described, to prevent glitches from occurring in the unlikely event that the false photon signal overlaps in time with a real photon pulse. More information:

[0033] The electronic module may comprise at least two banks, e.g., two counters, one of which counts the input pulses and the other of which can be read. The image sensor may further comprise a bank switching unit configured to switch which of the at least two banks the processed stream is forwarded to. The bank switching unit is configured to switch at a specific bank switching time. Without further measures, the bank switching time may coincide with a pulse in the processed stream. This means that a first part of the pulse is sent to one of the banks and a second part of the pulse is sent to the other bank. Thus, the two banks receive only partial pulses. If both partial pulses are long enough to be processed as a correct pulse, the number of pulses will be inaccurate and, in the case of the counters, the counting number will be wrong. If the partial pulse is too short, it may not be processed correctly and cause a malfunction of the electronic module. Since a photon may strike a photon-counting detector element at any time, it is inevitable that the bank switching time coincides with the time when a photon strikes one of the photon-counting detector elements and a pulse is generated.

[0034] To address this issue, artificial pulses (sometimes called false photons) are generated that block the bank switching times, as will be described in more detail below. The image sensor comprises a pulse generator configured and arranged to introduce an artificial pulse (false photon) into at least one macropixel stream or one of the pulse shaper output streams. The length of the artificial pulse is denoted as t_artificial. The monostable / astable circuit processes the artificial pulse to shorten its pulse length to a second pulse length t2, followed by a pulse-free interval called an "artificial-pulse-free interval". The artificial-pulse-free interval has a time width of (at least) t_artificial minus the second pulse length t2. In this way, each artificial pulse in the processed stream is followed by a pulse-free interval. The pulse generator is configured to introduce the artificial pulses at a time such that the bank switching times occur during the artificial-pulse-free interval. Thus, the artificial pulses cause a pulse-free period during which the bank switching occurs.

[0035] In general, it is not necessary for all artificial pulses to be the same length, but instead different values ​​for t_artificial may be used.

[0036] The pulse generator can be configured to introduce one artificial pulse each time a bank switch occurs. The frequency at which the artificial pulse is generated may match the frequency of the bank switch. In particular, the pulse generator and the bank switch can operate at a common frequency such that the frequency of the bank switch is equal to the frequency of the artificial pulse. Each artificial pulse is timed such that its start time is shorter than t_artificial, which is one before the bank switch time, and its start time is longer than a second pulse length t2, which is one before the bank switch time. This ensures that the bank switch is performed during a pause when there is no pulse. The time width of t_artificial is larger than t2 and can be in the range of, for example, 1.3*t2 < t_artificial < 3.0*t2. This generates a pause with a length ranging from 0.3*t2 to 2.0*t2. In principle, t_artificial can also be longer, but this unduly increases the dead time. In contrast, within the aforementioned range, it is ensured that the bank switch is performed at a sufficient distance from any possible pulse while keeping the dead time relatively short.

[0037] General features The photon counting detector elements described can form at least one array (sensor array). A pixel is understood as one element of the sensor or sensor array, i.e., one photon counting detector element. The photon counting detector elements can be, for example, APD (avalanche photodiodes) or SPAD elements and can output one pulse for each recorded photon. The photon counting detector elements can be constructed in CMOS technology and / or can form one or more arrays, and each array can form a silicon photomultiplier (SiPM). Thus, the aforementioned "signal" from the photon counting detector elements includes the (electrical) pulse when a photon is detected.

[0038] Image sensors may be used in applications where precise counting and / or timing information of single photons is important, such as microscopes, LiDAR systems (LiDAR: Light Detection and Ranging), PET systems (PET: Position Emission Tomography), or particle detectors, among others. Thus, exemplary embodiments also include optical microscopes, LiDARs, PET systems, or particle detectors with image sensors as described herein.

[0039] The time periods t1 and t2 of the pulses may be defined as the time periods during which the pulse height exceeds a certain threshold. The threshold may be, for example, 50% of the respective pulse height ( / reference voltage defining the pulse height), or more generally a value between 10% and 100%. Similarly, the time width, which refers to the period without a pulse, may be defined as the time period during which the signal height falls below a certain threshold, which may be the same or different from the threshold used to define t1 and t2.

[0040] The intended uses of the described image sensor result in variants of the method of the invention.Similarly, the features described with respect to the method of the invention also form variants of the image sensor of the invention. [Brief description of the drawings]

[0041] A better understanding of the present invention and various other features and advantages will become readily apparent from the following description taken in conjunction with the schematic drawings given by way of example and not limitation, in which like reference numerals may refer to like or substantially similar components, in which:

[0042] [Figure 1] 1 illustrates diagrammatically the components of an image sensor; [Diagram 2] 1 shows a schematic representation of two signals originating from different sensor elements and being combined into one signal. [Diagram 3] 1 shows a schematic representation of signals originating from different sensor elements that are too close to each other to be correctly processed. [Figure 4]1 shows a schematic of the rise and fall times required for an electronic module; [Diagram 5] 1 illustrates generally the components of an image sensor of the present invention and exemplary signals processed by the image sensor; [Figure 6] 1 shows a schematic of a pulse being processed by a monostable / astable circuit that is a level monostable. [Figure 7] 1 illustrates a schematic of a pulse being processed by a monostable / astable circuit that is an edge-delay monostable. [Figure 8] Schematic showing strongly overlapping pulses that an edge-delay monostable would interpret as one pulse. [Figure 9] 13 illustrates diagrammatically bank switching for transferring signal intervals to a different bank; [Figure 10] 13 illustrates diagrammatically how bank switching splits a pulse. [Figure 11] 2 illustrates diagrammatically components of a further embodiment of an image sensor of the present invention; [Figure 12] 12A and 12B show schematic diagrams of signals of the image sensor of FIG. [Figure 13] 12 illustrates diagrammatically further signals of the image sensor of FIG. [Figure 14] 2 illustrates diagrammatically components of a further embodiment of an image sensor of the present invention; [Figure 15] 1 illustrates an exemplary monostable / astable circuit for use in various embodiments of the image sensor of the present invention. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0043] Detailed Description of the Invention Figure 5: Anti-glitch circuit with monostable / astable circuit Fig. 5 shows a schematic representation of an image sensor 1 according to an exemplary embodiment of the invention. The image sensor 1 comprises a number of photon-counting detector elements 10, such as SPADs, configured to output an electrical pulse 13 upon recording of a photon. Fig. 5 shows a schematic representation of signals 12 output by two different photon-counting detector elements 10. Each signal 12 comprises one pulse 13 indicating the detection of a respective photon. For every pulse, the horizontal axis indicates time and the vertical direction indicates the intensity of the electrical signal, e.g. voltage or current. Thus, the photons causing the two illustrated pulses 13 are received at different times and the two pulses 13 overlap.

[0044] The image sensor 1 includes a first group of pulse shapers 20 (e.g., monostable), and the output signal 12 of each photon-counting detector element 10 is sent to a respective pulse shaper 20. In this example, the pulse shapers 20 are monostable and process the signal 12, and the output of the pulse shaper 20 is called the pulse shaper output stream 22. The pulse shaper 20 modifies (here shortens) the length of the pulses 13, transforming the pulses 13 into pulses 23 with a certain length t1 that is common to all pulses 23 in the pulse shaper output stream 22. Thus, the pulse shaper 20 maintains the number and timing (start time) of the pulses 13, while the pulse length is modified.

[0045] The image sensor 1 further comprises a combining gate or connection unit 30 for combining the pulse shaper output stream 22 to a macropixel output / macropixel stream 32 for output via a macropixel output line 31. For ease of understanding, the schematic diagram shows only a small number of photon counting detector elements 10 whose outputs are all combined into one macropixel stream 32. However, the entire image sensor 1 may comprise a much larger number of photon counting detector elements 10, for example 100-10,000, each photon counting detector element 10 may be provided with its own pulse shaper 20. The connection unit 30 may comprise a number of macropixel output lines 31 for outputting a number of macropixel streams 32 (for example 10-1,000). The macropixel streams 32 depend on which outputs of the photon counting detector elements 10 are combined. The connection unit 30 may be fixed or flexible. In a flexible design, the connection unit 30 may be adjusted to change which pulse shaper output streams 22 are combined into one macropixel stream 32. In a fixed design, it is predefined which pulse shaper output streams 22 are combined into one macropixel stream 32. The connection unit 30 can combine the pulse shaper output streams 22 using a logical OR connection, i.e. if any one of the received pulse shaper output streams 22 is at a high level, a high level is output. Thus, the macropixel stream 32 contains all pulses 23 of the received pulse shaper output streams 22. The length of the pulses 23 is not changed by the connection unit 30.

[0046] As shown in this example and also explained in the introduction to figures 3 and 4, the gap or time width t_short may be too short for correct processing by a subsequent electronic module 50, e.g. a digital counter or a time-to-digital converter TDC. This is typically the case when t_short is smaller than the recovery time tn of the electronic module 50, which is required for the electronic module 50 to recover its original internal state after the end of one pulse. Since a photon may strike the photon-counting detector element 10 at any time, the gap t_short between two pulses 23 may be of any value.

[0047] The image sensor 1 further comprises a monostable or astable (referred to as a monostable / astable circuit 40) which receives the macropixel stream 32. In case of multiple macropixel output lines 31, an equal number of monostable / astable circuits 40 can be provided in parallel with each other. The monostable / astable circuit 40 is configured such that it shortens the pulse length of each pulse 23 of the macropixel stream 32. The monostable / astable circuit 40 thus outputs a signal, referred to as a processed stream 42, in which each pulse 43 has a second pulse length t2, which is shorter than t1. The number of pulses and their starting times are generally maintained by the monostable / astable circuit 40, only the length of the pulses is shortened from t1 to t2. As a result, a pulse-free interval PF is generated after each pulse 43. The time width t_PF of the generated pulse-free interval PF can be equal to t1 minus t2. The overall gap between two successive pulses 43 can be equal to the original gap t_short between the pulses 23 plus t_PF. The interval PF without additional pulses reduces the risk that the recovery time tn of the electronic module 50 is violated. In general, t_PF can take any value, and a small additional gap between the pulses helps to reduce the risk of errors in the electronic module. Please note that for clarity, the electronic module 50 is drawn with a dashed line, since it can optionally be considered as part of the image sensor 1 of the present invention. If the electronic module 50 is part of the image sensor 1, the pulse shaper 20 and / or the mono / astable circuit 40 can be configured according to the recovery time tn and the minimum pulse length tp required for the electronic module 50.

[0048] If t1 and t2 are set such that t1-t2≧tn, then the recovery time tn of the electronic module is always observed regardless of the timing of any photon impinging on the photon-counting detector element 10, i.e., the risk of violating tn is not only reduced but entirely eliminated. If t2 is set such that t2≧tp, then the required minimum pulse length is always met. In this case t1 is set such that t1≧tp++tn.

[0049] It should be noted that the risk of tn violation can only be reduced by using two consecutive monostable circuits (more typically, a pulse shaper 20 followed by a monostable / astable circuit 40), where the latter monostable produces a shorter pulse. If the monostable / astable circuit 40 is not used and the pulse shaper 20 produces pulses with a short length of t2, then any gap t_short between two consecutive pulses is again possible.

[0050] The monostable / astable circuit 40 may in principle be any kind of astable or monostable, such as a level monostable or an edge-delay monostable (EDM), however, EDM has some advantages, as shown in the following figure:

[0051] Figure 6-8: Output of monostable / astable circuit Figure 6 shows a macropixel stream 32 being fed into a monostable / astable circuit 40, here a level monostable. In this example, the two pulses 23 in the macropixel stream 32 are not clearly separated from each other. Level monostables will usually output a shorter pulse if their recovery time is not met. Thus, the monostable / astable circuit 40 in this example correctly processes only the first pulse and outputs a pulse 43 of the desired pulse length, while the second pulse becomes a shorter pulse 43B.

[0052] Figure 7 shows the macropixel stream 32 of the previous figure, but in this case it is fed into a monostable that is an EDM. In contrast to a level monostable, the pulse width produced by EDM is independent of the input pulse. This plays an important role in glitch suppression. Since EDM is input pulse independent there is no recovery time, therefore input pulses 23 that may not be separated or even overlapping will lead to an output pulse 43 that is the same width as in any other case. As shown, the EDM (monostable / astable circuit 40) outputs a processed stream 42 where the pulses 43 have the same length / width.

[0053] FIG. 8 shows a macropixel stream 32 in which two pulses 23 are too close together to be recognized as two pulses by the EDM. This is the case when the falling edge of the first pulse and the rising edge of the second pulse overlap so strongly that there is not enough of a drop in the signal between the two pulses 23. In this case, the EDM interprets the two pulses 23 as one pulse. The EDM rejects the incomplete edge between the pulses 23 and is not triggered, so the second pulse is suppressed. As an advantage, the EDM does not output corrupted pulses that could cause errors in subsequent electronics, in contrast to the level monostable described with respect to FIG. 6.

[0054] Figures 9 and 10: Bank switching The electronics used for the photon-counting detector elements may operate in a double-buffering mode. In this case, two banks of electronics with the same functionality are switched over after a predefined time. The reason is that one of the two banks is active and receives pulses from the photon-counting detector elements, while the other bank is inactive and can be read by an external system, thus achieving a service time close to 100%. The output signal is sent alternately to the first bank and to the second bank at each switchover. Figure 9 shows diagrammatically a macropixel stream 32 including pulses 23a and 23b. Up to the time of bank switch s1, the macropixel stream 32 is fed to the first bank, which therefore receives a first portion 33 of the macropixel stream 32 including only the pulse 23a before the bank switch s1 and not including the pulse 23b after the bank switch s1. After the time of bank switch s1, the macropixel stream 32 is fed to the second bank, which therefore receives a second portion 34 of the macropixel stream 32, which only contains pulse 23b and does not contain pulse 23a.

[0055] However, the pulses can occur at random times or they can occur exactly when a bank switch is taking place. Figure 10 shows the case where a pulse 23 in the macropixel stream 32 occurs exactly at the time of a bank switch s1. The pulse 23 is split into two pulse portions 24 and 25 which can have any ratio depending on the start time of the pulse 23 relative to the time of the bank switch s1. The first bank receives the first portion 33 of the macropixel stream 32 which includes the pulse portion 24, while the second bank receives the second portion 34 of the macropixel stream 32 which includes the pulse portion 25.

[0056] Pulse portions 23 and / or 24 constitute pulses that may not meet the aforementioned requirement of tp, i.e., pulse portions 23 and / or 24 are shorter than the minimum pulse time required by the subsequent electronic module. In particular, if pulse 23 is generated by a monostable circuit with a length that is not excessively long, for example the same length as tp or slightly longer than tp, pulse portions 23 and 24 will most likely be shorter than tp. Therefore, pulse portions 23 and 24 may cause malfunction of the electronic module.

[0057] This problem is solved by the design illustrated in the following diagram.

[0058] Figures 11 and 12: Bank switching assistant with pseudophotons 11 illustrates diagrammatically an image sensor 1 of a further exemplary embodiment of the present invention. The image sensor 1 comprises a plurality of photon-counting detector elements 10, a first group of pulse shapers 20 (e.g., monostable), a connection unit 30, at least one monostable / astable circuit 40, and an electronic module 50, all of which may be configured as described with respect to FIG.

[0059] The electronic module 50 includes two banks B1 and B2, e.g., counter banks used to count pulses / photons. One counter can be read while the other counter is active and counting input pulses. A bank switch S of the electronic module 50 determines whether the processed stream 42 is input to bank B1 or B2. Examples of the electronic module 50 are a ripple counter, a binary counter, an LFSR counter, a Gray counter, a prescaler, a TDC, or any other electronic circuit susceptible to metastability.

[0060] To overcome the problem caused by bank switches coinciding with pulses in the processed stream 42, an artificial pulse (false photon) AP is generated that causes a pulse-free interval at the bank switch. In detail: the pulse generator G provides an artificial pulse AP that is added to the macropixel stream of the connection unit 30. This can be realized by an OR gate 35 (or another component that combines the two inputs into one combined output) on the macropixel output line 31 between the connection unit 30 and the EDM / monostable / astable circuit 40. The monostable / astable circuit 40 shortens the artificial pulse AP and generates a pause (a pulse-free interval) after the shortened artificial pulse. The generation of the artificial pulse AP is timed according to the bank switch so that each bank switch occurs during a pulse-free interval generated by the monostable / astable circuit 40 immediately after the shortened artificial pulse.

[0061] The pulse-free interval for bank switching is generated by the monostable / astable circuit 40 regardless of the arrival time of any pulses generated by real photons, i.e. regardless of any pulses in the macropixel stream before the artificial pulse AP is added. If the macropixel stream contains a pulse that coincides with a bank switch, this pulse will overlap with the artificial pulse AP, resulting in only one pulse, which is shortened by the EDM, causing a pulse-free interval again at the bank switch.

[0062] The described design therefore allows to eliminate any spurious pulses or glitches that may occur when banks are swapped. Note that if a false photon pulse is on the line, the EDM will not accept the real pulse. This results in a loss of sensitivity equal to t_artificial / tb, where tb is the time the bank is active. In practical cases, this ratio amounts to about 1%.

[0063] The effect of the artificial pulse AP will be further explained with reference to FIG.

[0064] FIG. 12 shows a graph of the different signals, the horizontal direction indicates time. The bank-switching signal s indicates that a bank-switching has occurred at time s1. The pulse generator generates an output G_AP, which contains a pulse (artificial pulse AP) of time length t_artificial. The macropixel stream 32 output by the connection unit contains pulses 23, each having a length t1 (t1 is set by the pulse shaper of the first group). An OR gate adds the pulse generator output G_AP to the macropixel stream 32, generating a "macropixel stream with added artificial pulses" 32', which contains all the pulses 23 of the macropixel stream 32 and the artificial pulse(s) AP. The "macropixel stream with added artificial pulses" 32' is fed to a monostable / astable circuit 40, where all the pulses are shortened to a length of t2, resulting in a processed stream 42, shown diagrammatically. The processed stream 42 is fed to an electronic module, where it is divided by the bank-switching time s1. A first portion 42a of the processed stream 42, which contains only all pulses up to the bank switch time s1, is supplied to a first bank, and a second portion 42b of the processed stream 42, which contains only all pulses from the bank switch time s1 onwards, is supplied to a second bank.

[0065] The start time and length of the artificial pulse AP are selected so that no pulse occurs at the bank switch time s1. For this purpose, the start time of the artificial pulse AP is selected to be earlier than t2 before the bank switch time s1, and the pulse length t_artificial is selected to be greater than t2. The monostable / astable circuit 40 shortens the length of the artificial pulse AP to t2, thus generating a pulse-free interval of length t_pause=t_artificial-t2. Due to the shortened length t2 of the artificial pulse AP, the artificial pulse AP in the processed stream 42 ends before the bank switch s1 occurs. The start time of the artificial pulse AP is selected to be shorter than t_artificial before the bank switch time, so that when the bank switch s1 occurs, the artificial pulse-free interval t_pause (after the shortened artificial pulse) has not yet ended.

[0066] Furthermore, t2 and t1 may be selected as described with respect to the embodiment of Figure 5. In particular, t2 may be selected such that t2 > tp, and t1 may be selected such that t1 > tp + tn.

[0067] When a bank switch occurs, the artificial pulse AP is scaled down by the EDM so that the artificial pulse AP meets the requirements of the tp electronics. When the input of the EDM is set, any other real photons are rejected during this switching time. In this way, a bank switching assistant is provided that reliably excludes any pulse splitting due to bank switching.

[0068] Of course, the artificial pulse AP can be discarded or corrected by the electronic module, for example, in the case of a counter, the count number can be decreased by one every time a bank switch occurs (or an artificial pulse is generated), so that the count of the artificial pulse AP can be corrected.

[0069] Figure 13: Pulse-free period when switching banks 13 shows the case where bank switch s1 coincides with pulse 23c in macropixel stream 32. Since artificial pulse AP also coincides with bank switch s1, pulse 23c and artificial pulse AP result in only one pulse in the combined output 32'. In the example shown, artificial pulse AP completely covers the time of pulse 23c, i.e. it starts before pulse 23c and ends after pulse 23c, and therefore the combined output 32' contains only one pulse AP of length t_artificial, similar to the case described with respect to the previous figure.

[0070] This means that the pulse 23c that coincides with the bank switch s1 is filtered out. The artificial pulse therefore creates a dead time of t_artificial.

[0071] For completeness, if pulse 23c only overlaps with, but is not completely covered by, artificial pulse AP, these two pulses will result in only one pulse at the combined output 32', but with a length longer than t_artificial (at most t_artificial+t1). The monostable / astable circuit 40 shortens this pulse again to create a pulse-free interval during which the bank switch occurs.

[0072] Figure 14: Variations in the introduction of pseudophotons Figure 14 is a variant of Figure 11, differing from Figure 11 in that the pulse generator G is not connected to the macropixel output line 31 (i.e. downstream of the connection unit 30), but instead is connected to the input of the connection unit 30, and thus supplies the artificial pulse AP to the connection unit 30, which combines it with any pulses output by the pulse shaper 20. In both cases of Figure 11 and Figure 14, the artificial pulse AP is finally included in the macropixel stream on the macropixel output line 31. The OR gate 35 of Figure 11 is not necessary in the embodiment of Figure 14. As a further variant of the embodiment shown in Figure 14, the artificial pulse AP can also be added to the input of one or more pulse shapers 20.

[0073] Figure 15: Monostable / astable circuit with feedback loop Fig. 15 shows a schematic diagram of an EDM-like monostable / astable circuit 40 that can be used in the embodiments described with respect to the other figures. The macropixel output line 31 is connected to the input of the monostable / astable circuit 40. In this example, the macropixel stream 32 carried by the macropixel output line 31 contains two strongly overlapping pulses. The monostable / astable circuit 40 has at least two output ports, one of which is connected to a feedback loop 45 that returns to the inlet of the monostable / astable circuit 40, in this case the reset inlet R. The feedback loop 45 is equipped with a function f, i.e. an electronic component that modifies the signal in the feedback loop by at least modifying / shifting its phase, generating the feedback signal 32_feedback. As shown in the figure, in the case of strongly overlapping pulses, the latter pulse may not be recognized, so that the feedback signal 32_feedback contains only one pulse for two strongly overlapping pulses in the macropixel stream 32. Otherwise, each pulse in the macropixel stream 32 leads to one pulse in the feedback signal 32_feedback. On an output line 41, the monostable / astable circuit 40 outputs a processed stream 42, which is generated depending on the macropixel stream 32 and the feedback signal 32_feedback. The pulses in the processed stream 42 have a shorter pulse length than the pulses in the macropixel stream 32, due to the operation of the monostable / astable circuit 40 and its resetting after a specified delay.

[0074] The features described with respect to the different figures can also be combined. Furthermore, the described embodiments are less complex for ease of understanding. In particular, the image sensor can include multiple macropixel output lines. The described functionality refers to only one macropixel output line. A similar structure may be provided for each macropixel output line. In particular, for each macropixel output line, a monostable / astable circuit 40 and the introduction of an artificial pulse AP may be provided.

[0075] Reference Code List 1 Image sensor 1' Related image sensor 10 Photon-counting detector elements 12 Signal from photon counting detector element 10 13 Pulse from photon counting detector element 10 20 First group pulse shapers 20' Monostable 22 Pulse Shaper Output Stream 23, 23a-23c Pulses output by the first group of pulse shapers 20 / pulses in the pulse shaper output stream 22 24, 25 Pulse portion generated by dividing pulse 23 by bank switching 30 Connection unit for combining pulse shaper output streams 22 31 Macropixel output line(s) 32 Macropixel Stream / Macropixel Output 32' Macropixel stream with artificial pulse (false photon) added 32_feedback Feedback signal of monostable / astable circuit 40 33 First part of macropixel stream 32 before bank switching 34 Second part of macropixel stream 32 after bank switch 35 OR gate combining artificial pulses and macropixel streams 40 Monostable / Astable Circuits 41 Output line of monostable / astable circuit 40 42 Processed Stream (output of monostable / astable circuit 40) 42a First part of the processed stream (before bank switching) 42b Second part of the processed stream (after bank switch) 43 Pulse (in processed stream 42) 43B Spurious Pulse 45 Monostable / astable circuit 40 feedback loop 50 Electronic Module 52 Electronic module input signals AP Population Pulse B1, B2 banks, e.g., counters EDM Edge Delay Monostable f Functions in the feedback loop 45 (electronic components) G. Pulse generator that supplies artificial pulse AP G_AP Pulse generator output PF Pulse-free interval (after pulse 43 in the processed stream 42) R Reset entrance for monostable / astable circuits S Bank Switching Unit s Bank switching signal s1 Bank switching time t1 First pulse length (i.e., the length of pulse 23) t2 Second pulse length (i.e., the length of pulse 43) TDC Time to Digital Converter tn is the recovery time of the electronic module (i.e. the length of time required to pause between pulses) Minimum pulse length required for tp electronic module t_artificial Pulse length of artificial pulse AP t_pause Interval without artificial pulse t_PF Length of the interval without a pulse; the difference between the length of the first pulse, t1, and the length of the second pulse, t2 t_short The gap between two pulses in the macropixel stream.

Claims

1. An image sensor, a photon-counting detector element (10); a first group of pulse shapers (20) arranged to convert signals (12) from said photon counting detector elements (10) into a pulse shaper output stream (22) in which pulses (23) have a minimum first pulse length (t1); a connection unit (30) configured to combine a plurality of said pulse shaper output streams (22) into at least one macropixel stream (32); at least one monostable / astable circuit (40) arranged to receive the at least one macropixel stream (32) and output a processed stream (42) in which the pulses (43) have a second pulse length (t2); the at least one monostable / astable circuit (40) is configured such that the second pulse length (t2) is shorter than the first pulse length (t1) by a time duration t_PF to generate a pulse-free interval (PF) of at least the time duration t_PF after each pulse (43) in the processed stream (42).

2. 2. The image sensor of claim 1, wherein the first group of pulse shapers (20) comprises a group of monostables and the at least one monostable / astable circuit (40) is a further monostable.

3. Further comprising an electronic module (50) for processing the processed stream (42), the electronic module (50) requiring a pause of at least tn between two pulses (43) in the processed stream (42), 2. The image sensor of claim 1, wherein the at least one monostable / astable circuit is configured such that t_PF is at least as long as tn.

4. 4. The image sensor of claim 3, wherein the electronic module (50) comprises at least one of a time-to-digital converter (TDC), a digital counter, a level comparator, a trigger, a scaler, an inverter, a buffer, a pull-up, a latch, a level monostable, a clock divider, and a phase detector.

5. 4. The image sensor of claim 3, wherein the first group of pulse shapers (20) and the at least one monostable / astable circuit (40) are configured such that t_PF<2*tn to limit dead time of the image sensor.

6. the electronic module (50) requires that pulses (43) in the processed stream (42) have a minimum pulse length tp; the at least one monostable / astable circuit (40) is configured such that the second pulse length (t2) is at least as long as tp; 4. The image sensor of claim 3, wherein the first group of pulse shapers (20) are configured such that the first pulse length (t1) is at least as long as tp+tn.

7. 7. The image sensor of claim 6, wherein the first group of pulse shapers (20) are configured such that the first pulse length (t1) is shorter than 2*[tp+tn] to limit dead time of the image sensor.

8. 2. The image sensor of claim 1, wherein the at least one monostable / astable circuit (40) is an edge delayed monostable (EDM) rather than a level monostable.

9. The electronic module (50) comprises at least two banks (B1, B2), The image sensor further comprises a bank switching unit (S) configured to switch between the at least two banks to which the processed stream (42) is transferred, the bank switching unit (S) being configured to switch at a specific bank switching time (s1); The image sensor further comprises a pulse generator (G) constructed and arranged to introduce artificial pulses (AP) into one of the at least one macropixel stream (32) or the pulse shaper output stream (22), each artificial pulse (AP) having a length t_artificial; said at least one monostable / astable circuit (40) processes each artificial pulse (AP) such that its pulse length is shortened to said second pulse length (t2) followed by an artificial pulse-free interval (t_pause) of at least t_artificial minus said second pulse length (t2); 4. The image sensor of claim 3, wherein the pulse generator (G) is configured to introduce the artificial pulse (AP) at a time such that the bank switching time (s1) occurs during an interval (t_pause) without the artificial pulse.

10. 10. The image sensor of claim 9, wherein the pulse generator (G) is configured to introduce one artificial pulse (AP) per bank switch, and each artificial pulse (AP) is timed such that its start time is less than t_artificial before the bank switch time (s1) and its start time is longer than the second pulse length (t2) before the bank switch time (s1).

11. 10. The image sensor of claim 9, wherein the pulse generator (G) and the bank switching (S) operate at a common frequency, such that the frequency of the bank switching time (s1) is equal to the frequency of the artificial pulse (AP).

12. 12. A measurement system comprising an image sensor according to any one of the preceding claims, the measurement system being one of a microscope, a LIDAR system, a position emission tomography system, or a particle detector.

13. 1. A method of operating an image sensor, comprising: receiving photons at a photon-counting detector element (10) and outputting a signal (12) indicative of the photons received by the photon-counting detector element (10); converting the signals (12) of the photon-counting detector elements (10) using a first group of pulse shapers (20) into a pulse shaper output stream (22) in which pulses (23) have a first pulse length (t1); using a connection unit (30) for combining a plurality of said pulse shaper output streams (22) into at least one macropixel stream (32); using at least one monostable / astable circuit (40) for processing said at least one macropixel stream (32) to output a processed stream (42) in which pulses (43) have a second pulse length (t2) that is shorter than said first pulse length (t1) by a time width t_PF, thereby generating a pulse-free interval (PF) of at least said time width t_PF after each pulse (43) in said processed stream (42).

Images