WO2023118834A1

WO2023118834A1 - Improved digital silicon photomultiplier

Info

Publication number: WO2023118834A1
Application number: PCT/GB2022/053294
Authority: WO
Inventors: Aravind Venugopalan Nair Jalajakumari; John Mullins; Robert Henderson; Edward Marsden
Original assignee: Kromek Limited; University Court Of The University Of Edinburgh
Priority date: 2021-12-22
Filing date: 2022-12-19
Publication date: 2023-06-29

Abstract

The present invention relates to a digital silicon photomultiplier, methods of operating a digital silicon photomultiplier and a device comprising one or more digital silicon photomultipliers. The digital silicon photomultiplier may comprise at least one least one pixel, wherein each pixel includes at least one single photon avalanche photodiode and a toggle operatively connected to each of the at least one single photon avalanche photodiodes. The toggle may output at least one transition wherein each transition represents a detection of a photon. A programmable divider may be operatively connected to each pixel, wherein each programmable divider is programmed with a divisor and each programmable divider receives a first set of transitions and outputs a second set of transitions, wherein each transition in the first set of transitions represents a detection of a photon by a single photon avalanche photodiode of a respective pixel and each transition of the second set of transitions represents a number of detected photons of the respective pixel equivalent to the divisor.

Description

Improved Digital Silicon Photomultiplier

The present invention relates to a Silicon Photomultiplier and, in particular to a Low Power Asynchronous Digital Silicon Photomultiplier.

Background

A silicon photomultiplier (SiPM) is a sensor for detecting low light intensities ranging in wavelength from the ultraviolet (UV) through visible wavelengths and into the infrared (IR). These sensors may be used, in conjunction with scintillator materials, in systems which detect ionising radiation such gamma and X-rays and alpha and beta particles. In that application, high energy photons or particles first interact with the scintillator producing many visible or UV photons which can then be detected by the SiPM. A SiPM typically comprises an array of single photon avalanche photodiodes (SPADs) wherein each SPAD in a SiPM outputs an electrical signal when a photon is detected.

Although detection of an incident photon by a SPAD is an essentially digital process SiPMs have generally been operated in an analogue mode with the electrical signals output by all SPADs in a SiPM being integrated to provide an output current.

More recently digital SiPMs (DSiPMs) have been developed in which the detection of a photon, as determined by the output of an electrical signal by an individual SPAD, is converted to a digital T and this bit transmitted to the top level of the device incorporating the DSiPM. This direct to digital conversion approach has a number of advantages:

- Noise and temperature sensitivity are reduced as the effects of variation in the collected charge are eliminated.

- Timing integrity of photon detection can be better maintained.

- Downstream processing can be purely digital eliminating the additional noise contributions from analogue processing.

- Digital processing can be undertaken on the DSiPM chip itself.

However, a disadvantage with the currently available DSiPMs is high power consumption, which is typically due to fast transitions of digital signals and clock pulses. In addition to increasing power consumption, which is particularly undesirable for portable instruments (such as hand held scintillation detectors), high power dissipation also increases the temperature of the DSiPM thereby increasing the dark count rate (DCR) which further degrades the device performance.

Accordingly, the present invention seeks to address, at least in part, the above described disadvantages and problems.

Statement of Invention

According to a first aspect of the present invention there is provided a digital silicon photomultiplier comprising: at least one least one pixel, wherein each pixel includes at least one single photon avalanche photodiode and a toggle operatively connected to each of the at least one single photon avalanche photodiodes, wherein the toggle outputs at least one transition and each transition represents a detection of a photon; a programmable divider operatively connected to each pixel, wherein each programmable divider is programmed with a divisor; and each programmable divider receives a first set of transitions and outputs a second set of transitions, wherein each transition in the first set of transitions represents a detection of a photon by a single photon avalanche photodiode of a respective pixel and each transition of the second set of transitions represents a number of detected photons of the respective pixel equivalent to the divisor.

The digital silicon photomultiplier may comprise at least two pixels and may further comprise a first stage comprising two or more first toggles, wherein each first toggle is operatively connected to one programmable divider; wherein each first toggle receives one or more transitions output from the respective programmable divider and outputs one or more transitions.

If a toggled output of the first toggle is selected then each output transition may represent two received transitions output from the respective programmable divider.

The first stage may further comprise one or more first XOR gates wherein each first XOR gate is operatively connected to two adjacent first toggles; and wherein each first XOR gate combines the output of the two adjacent first toggles to a single output from the first XOR gate.

The digital silicon photomultiplier may further comprise one or more further stages, wherein each further stage may comprise two or more further toggles, wherein each further toggle may be operatively connected to one XOR gate of the previous stage; and one or more further XOR gates, wherein each further XOR gate may be operatively connected to two adjacent further toggles.

If a toggled output of the further toggle is selected then each output transition of each further toggle may represent two received transitions output from the respective XOR gate of the previous stage.

The programmable divider may be a ripple counter based on flip-flops.

The divisor of each programmable divider may be programmable via a control input to each programmable divider.

The digital silicon photomultiplier may further comprise a rolling shift register operatively connected to an output of the silicon photomultiplier.

The rolling shift register may be divided into bins, wherein each bin may include a number of bits and may represent a period of time; and a total number of photons detected during each period of time may be stored in the respective bin.

The number of bits allocated to each bin and the time period of each bin may be programmable via a control input to the rolling shift register.

The digital silicon photomultiplier may further comprise one or more trigger mechanisms to determine one or more events of interest, wherein if triggered the trigger mechanisms may trigger a transfer of data from the silicon photomultiplier.

A first trigger mechanism may comprise a comparator, wherein the comparator compares a number of detected photons in one or more bins to a threshold; and if the number of detected photons in the one or more bins exceeds the threshold an event of interest may be determined and the transfer of data is triggered.

A second trigger mechanism may comprise one or more filter stages operatively connected to a comparator, wherein the comparator may compare the output of the one or more filter stages to a threshold; and if the output of the one or more filter stages exceeds the threshold an event of interest may be determined and the transfer of data is triggered.

A third trigger mechanism may comprise summing a first set of bins, wherein the first set of bins may represent a most recent set of bins; summing a second set of bins, wherein the second set of bins may represent a next most recent set of bins; a comparator, wherein the comparator may compare the sum of the first set of bins to the sum of the second set of bins; and if the sum of the first set of bins is greater than the sum of the second set of bins an event of interest may be determined and the transfer of data is triggered.

The digital silicon photomultiplier may further comprise shifting the sum of the second set of bins to the left by a number of bits prior to the comparison.

Each pixel may include two or more single photon avalanche photodiodes arranged in a two-dimensional array.

The digital silicon photomultiplier may further comprise identifying one or more damaged single photon avalanche photodiodes; and disabling the one or more identified damaged single photon avalanche photodiodes.

Identifying the one or more damaged single photon avalanche photodiodes may comprise comparing a count rate of transitions from the one or more single photon avalanche photodiodes to a predetermined threshold; and disabling the one or more identified damaged single photon avalanche photodiodes may comprise removing power from the one or more identified damaged single photon avalanche photodiodes.

According to a second aspect of the present invention there is provided a method of operating a digital silicon photomultiplier, wherein the digital silicon photomultiplier comprises: at least one least one pixel, wherein each pixel includes at least one single photon avalanche photodiode and a toggle operatively connected to each of the at least one single photon avalanche photodiodes, wherein the toggle outputs at least one transition and each transition represents a detection of a photon; and a programmable divider operatively connected to each pixel, wherein each programmable divider is programmed with a divisor; the method comprising: receiving at each programmable divider a first set of transitions wherein each transition in the first set of transitions represents a detection of a photon by a single photon avalanche photodiode of a respective pixel; and outputting a second set of transitions, wherein each transition of the second set of transitions represents a number of detected photons of the respective pixel equivalent to the divisor. The digital silicon photomultiplier may comprise at least two pixels and a first stage comprising two or more first toggles, wherein each first toggle may be operatively connected to one programmable divider; the method may further comprise receiving at each first toggle one or more transitions output from the respective programmable divider; and outputting one or more transitions.

The method may further comprise selecting a toggle output of the first toggle such that each output transition may represent two received transitions output from the respective programmable divider.

The first stage may further comprise one or more first XOR gates wherein each first XOR gate may be operatively connected to two adjacent first toggles; and the method may further comprise combining, by each first XOR gate, the output of the two adjacent first toggles to a single output from the first XOR gate.

The digital silicon photomultiplier may further comprise one or more further stages, wherein each further stage may comprise two or more further toggles, wherein each further toggle may be operatively connected to one XOR gate of the previous stage; and one or more further XOR gates, wherein each further XOR gate may be operatively connected to two adjacent further toggles; and the method further comprises combining, by each further XOR gate, the output of the two adjacent further toggles to a single output from the further XOR gate.

The method may further comprise selecting a toggled output of each further toggle such that each output transition of each further toggle may represent two received transitions output from the respective XOR gate of the previous stage.

The method may further comprise programming, via a control input of each programmable divider, the divisor of each programmable divider.

The digital silicon photomultiplier may further comprise a rolling shift register operatively connected to an output of the silicon photomultiplier wherein the rolling shift register may be divided into bins, the method may further comprise storing a total number of photons detected during each period of time, wherein each bin may represent the period of time and each bin may include a number of bits.

The method may further comprise programming, via a control input of the rolling shift register, the number of bits and time period for each bin. The method may further comprise determining one or more events of interest; and triggering, by one or more trigger mechanisms, a transfer of data from the silicon photomultiplier if an event of interest is determined.

A first trigger mechanism may comprise comparing a number of detected photons in one or more bins to a threshold; determining an event of interest based on the comparison; and triggering the transfer of data based on the determined event of interest.

The method may further comprise filtering an output of the silicon photomultiplier by one or more filter stages; comparing an output of the one or more filter stages to a threshold; and determining an event of interest based on the comparison; and triggering the transfer of data based on the determined event of interest.

The method may further comprise summing a first set of bins, wherein the first set of bins represents a most recent set of bins; summing a second set of bins, wherein the second set of bins represent a next most recent set of bins; comparing the sum of the first set of bins to the sum of the second set of bins; determining an event of interest based on the comparison; and triggering the transfer of data based on the determined event of interest.

The method may further comprise shifting the sum of the second set of bins to the left by a number of bits prior to the comparison.

The method may further comprise identifying one or more damaged single photon avalanche photodiodes; and disabling the one or more identified damaged single photon avalanche photodiodes.

According to a third aspect of the present invention there is provided a device comprising: a microcontroller; and one or more digital silicon photomultipliers according to any one or more of the aspects and features described above. Drawings

Embodiments of the invention will now be described, by way of example only, and with reference to the following drawings, in which:

Figure 1 shows a block diagram of a conventional digital silicon photomultiplier

Figure 2 shows a conventional single photon avalanche photodiode.

Figure 3 shows a conventional layout of a pixel on a DSiPM die.

Figure 4 shows a combination of a silicon photomultiplier optically coupled to a scintillator according to one or more embodiments of the present invention.

Figure 5A is a schematic of a pixel and divider arrangement according to one or more embodiments of the present invention.

Figure 5B is a timing diagram according to one or more embodiments of the present invention.

Figure 6 is a schematic of a silicon photomultiplier with four stages according to one or more embodiments of the present invention.

Figure 7 shows a combination and division of transitions at each of x stages according to one or more embodiments of the present invention.

Figure 8 shows a schematic of an arrangement of pixels and three stages according to one or more embodiments of the present invention.

Figure 9 shows a first trigger mechanism according to one or more embodiments of the present invention.

Figure 10 shows a second trigger mechanism according to one or more embodiments of the present invention.

Figure 11 shows a third trigger mechanism according to one or more embodiments of the present invention.

Figure 12 is a block diagram of parallel triggering with multiple silicon photomultipliers according to one or more embodiments of the present invention.

Figure 13 is a schematic of an arrangement to select a toggle output according to one or more embodiments of the present invention. Description

The present invention relates to an asynchronous circuit architecture for a digital silicon photomultiplier (DSiPM) which reduces the requirement to transmit digital transitions and/or signals a significant distance across a DSiPM die, thereby reducing power dissipation and/or improving power consumption. This is particularly advantageous for portable instruments (such as hand held scintillation detectors).

With reference to Figure 1 , a DSiPM 101 typically comprises a two-dimensional array of pixels 102 where each pixel respectively comprises a two-dimensional array of single photon avalanche photodiodes (SPADs) 103. A DSiPM 101 may comprise any number of pixels 102 in a two-dimensional array, wherein the number of pixels may be dependent on the application of the DSiPM. For example, the number of pixels may range from tens, to hundreds, to thousands of pixels. Each pixel may comprise any number of SPADs in a two-dimensional array wherein the number of SPADs may also be dependent on the application and purpose of the DSiPM device. For example, the number of SPADs may range from tens, to hundreds, to thousands of SPADs. For simplicity of the drawings, the DSiPM shown in Figure 1 includes two pixels 102 and each pixel includes a 6x6 two-dimensional array of SPADs. The SPADs of each pixel are typically connected via an OR tree 104 to a single digital output 105. The outputs of all pixels in a DSiPM are typically summed to give the output of the entire DSiPM 101 .

A SPAD, designed to be used in a DSiPM, is shown in Figure 2. SPAD 201 is a sensor that includes a photodiode 202 wherein an incident photon causes an avalanche of electrons within the diode 202 and causes an electrical signal to be output. A SPAD is typically formed of a semi-conductor p-n junction that is highly reverse biased at VBias such that an incident photon causes the SPAD 201 to output a short electrical pulse 203.

The output of each SPAD passes through an inverter, such as a Schmitt triggered inverter 204 in order to generate a short rectangular pulse 205 each time a SPAD 201 is triggered by an incident photon.

Figure 3 shows a conventional layout of a pixel 301 on a DSiPM die. The pixel 301 includes an MxN two-dimensional array of SPADS 302. The outputs of the array of SPADs are XOR’ed together by an XOR tree 306 and a single output train of transitions is transmitted via associated conventional control, processing and input/output electronics 303 on a DSiPM die. The output of each pixel 301 may then be combined and transmitted by a digital communication link 304 to an external system 305.

For the detection of ionizing radiation, with reference to Figure 4, a DSiPM 401 may be optically coupled 402 to a scintillator 403. High energy ionizing radiation 404 (e.g. high energy alpha, beta or X-ray photons) interacts within the scintillating material 405 with each photon or particle producing many UV or visible photons 406 which then impinge on the DSiPM 401 via the optical coupling 402 to produce a number of output transitions 407. The scintillator 403 may be partially surrounded by a reflector 408 in order to efficiently collect the UV or visible light photons 406.

In order to employ digital processing at the top level of the device incorporating the DSiPM, it is necessary in the conventional DSiPMs to transmit the pulses representing the detection of photons by the pixels a relatively significant distance across the DSiPM die of the device. In conventional DSiPMs this data transmission has been done synchronously to maintain accurate temporal and special information about the event. The synchronous data transmission across the DSiPM in conventional DSiPMs requires a significant high power dissipation and power consumption which is particularly disadvantageous, as discussed hereinabove.

The present invention seeks to address, at least in part, the disadvantages of the prior-art DSiPMs so as to at least minimise power dissipation and/or power consumption. The present invention also seeks to minimise the transmission of fast transitions across relatively large distances in the device incorporating the DSiPM by utilising asynchronous transmission of digital transitions. This is advantageous as it is significantly more efficient, since the transitions only occur when a photon is detected, and there is no requirement to distribute clock signals resulting in significant power savings.

For applications, such as the detection of gamma rays for isotope identification, it is not necessary to know information about the path of the incoming gamma ray photon with high spatial resolution. Therefore, a few large pixels can be used and, in some applications, even a single pixel the size of the whole detector may be employed. For applications including isotope identification and medical imaging, e.g. computed tomography (CT) and single photon computed tomography (SPECT), it is preferable to accurately identify the energy of every detected x-ray or gamma photon. In a scintillator-based radiation detector, the number of optical photons created by the gamma photon interaction is proportional to the energy of the gamma photon. It is therefore preferable to count the number of optical photons from a gamma ray event, and to do so with sufficient accuracy in the energy range of interest. For most scintillating materials the majority of optical photons from a single radiation event will be generated in a single burst lasting between 10ns and 10ps. In these applications it may also be beneficial to maintain relatively accurate temporal information relating to the rate at which photons are detected, e.g. to determine segregation of events occurring close together (e.g. pulse pile up rejection) or to identify the decay characteristics of scintillation events, which can sometimes be used to identify the nature of the incoming radioactive particle (e.g. alpha, beta, gamma, neutron).

The embodiments of the present invention provide a low power photon detection device which may encapsulate one or more of the following features to minimise power dissipation and/or power consumption, whilst maintaining sufficient energy resolution and temporal information:

- Distributed convolution of signals, for global photon counting across a DSiPM pixel, by combining the outputs of individual SPADs and using a divider at the output of each combination event.

- Programmable configuration of the divider structure to allow optimised energy resolution over a wide range of photon energies.

- On-chip detection and signal processing to reduce the distances the signals must travel, thus reducing the power consumption.

- Single shift-register to capture the number of photons in accurate time- stamped intervals.

- Triggering circuit to enable asynchronous capture and transfer of data, for example when a significant number of photons arrive in a short period of time.

Embodiments of the present invention may therefore achieve a power dissipation and/or a power consumption that is significantly reduced in comparison to the conventional DSiPMs, in particular, by the incorporation of one or more dividing stages.

Figure 5A shows an example pixel level architecture according to one or more embodiments of the present invention in which the pixels comprise two SPADs 501 .

The output 506 of each SPAD 501 , is passed through an inverter, such as a Schmitt triggered inverter, 502, to produce a short rectangular pulse 507, and then through a toggle, such as an edge triggered toggle, 503, which outputs a transition 508, alternately positive and negative going, for either the rising or falling edges in the inverter output 507. This has the advantageous effect of reducing the number of transitions in the output by a factor of two. Each transition, either positive or negative going represents the detection of one photon. The toggled outputs of all SPADs in a pixel are then combined by an XOR tree structure 509 before being passed through a further toggle, 503, followed by a programmable divider 504 which reduces the number of output transitions from the pixel by a factor of the divisor. The divisor of the programmable divider may be any suitable divisor integer value, for example, 1 , 2, 4, 6, 8, 16, 32, 64, 128, and so on. The programmable divider may be any suitable programmable divider such as an asynchronous divider. Incorporating an asynchronous divider in this way has the advantage that power consumption and/or power dissipation can be maintained at a minimum. Alternatively, a synchronous divider may be used, however, a clock would be required for a synchronous divider which will increase power consumption.

The programmable asynchronous divider 504 may be any suitable divider such as a ripple counter based on flip-flops. The toggle and flip-flops may be any suitable flipflop, such as a D-type flip-flop, JK-type flip-flop, and so on. The ripple type counter divider in this example is based on the D-type flip-flops and the ripple counter divider may include any number of D-type flip-flops required depending on the required maximum divisor, as each flip-flop in the ripple counter will effectively divide the output transitions of incident photons by 2. For example, a maximum divisor of 32 would require 5 D-type flip-flops, a maximum divisor of 128 would require 7 D-type flip-flops. It will be apparent that both the toggles and the dividers either separately or in combination reduce the number of transitions passed to the next stage by a factor equal to a power of two. The programmable divider 504, may be programmed via a control input 505 which enables the DSiPM device to set a divisor up to the maximum divisor of the implemented programmable divider. For example, a ripple counter that includes 7 D- type flip-flops may be programmed with a divisor of 1 , 2, 4, 8, 16, 32, 64 and 128. The DSiPM device may include a register which can operatively connect to one or more of the programmable dividers, e.g. via a data port in the DSiPM chip, in order to control and set the required divisor in the respective programmable dividers.

The output of the programmable divider 504 (e.g. the output of the pixel) may then be provided to a further XOR tree structure 510 in combination with one or more output signals from one or more further pixels 511 and the output of the XOR tree structure 510 may be provided to a subsequent programmable divider or toggle 512, which is repeated for a required number of stages depending on the number of pixels in the DSiPM.

Figure 5B shows a timing diagram corresponding to the arrangement shown in Figure 5A. As shown in Figure 5B, the outputs 556 of the SPADs 551 are passed through the associated Schmitt trigger inverters 552 to produce a train of short rectangular pulses 553. This pulsetrain is then passed through a toggle, such as an edge triggered toggle, 554, which produces a train of transitions 555 where each transition represents the detection of 1 photon. The train of transitions 555 is transmitted or passed through a programmable divider, 557, such that the output of the programmable divider 557 is a train of transitions 558s where each transition represents a number of output transitions from the pixel based on the divisor of the programmable divider.

For a DSiPM comprising two or more pixels, the pixel outputs may be combined by a further XOR tree and further toggling and/or division can be incorporated at each stage of the tree.

The divisor is advantageous as a number of output transitions each representing a detected photon from a respective pixel can be grouped together to a single output transition that corresponds to, or represents, the programmed divisor value number of detected photons, rather than transmitting each individual output transition of each detected photon from the pixel, as is the case in the traditional and conventional DSiPM. For example, if the divisor is programmed via the control input to be 16 then an output transition from the divider will only occur once 16 photons are detected by the pixel operatively connected to the programmable divider, rather than transmitting an output pulse from the pixel corresponding to each of the 16 detected photons, which advantageously reduces the power consumption and/or power dissipation of the DSiPM.

As a programmable divider is located close, or proximate, to the origin of the output signals from the pixel of the two-dimensional array of SPADs, power dissipation and/or power consumption is minimal as the distance the signals travel is significantly reduced.

As mentioned above, in many applications the DSiPM may include tens, hundreds or thousands of pixels where each pixel may include tens, hundreds, or thousands of SPADs in a two-dimensional array. Therefore, in order to combine all of the outputs of each pixel to a top-level output of the DSiPM, several stages of combining the outputs by an XOR tree to produce the top-level output of the DSiPM may be implemented.

In order to further reduce the transmission of transitions and thereby further reduce the power dissipation and/or power consumption, each XOR stage associated with the pixels may be followed by further stages wherein at each stage additional toggles and/or programmable dividers may further reduce the overall transition rate. The overall transition rate may be reduced by a factor x, where x is the number of stages that include a toggle and/or programmable divider. This is shown in Figures 6, 7 and 8. Figure 6 is a schematic of a DSiPM with four XOR stages following the output from the pixels, Figure 7 shows the combination and division of the transitions at each of x stages, and Figure 8 shows a schematic of the arrangement of pixels and three stages. As will be appreciated, any number of XOR stages may be implemented.

Referring to Figure 6, the DSiPM 601 includes a plurality of stages, where four stages 602, 603, 604, 605 are shown in Figure 6. Each stage includes one or more multiplexers 608, one or more XOR gates 606 and one or more toggles 607. In the example of Figure 6, the toggles 607 are implemented as D-type Flip-Flops. Stage 1 602 includes one or more multiplexers 608 and one or more XOR gates 606, where sixteen multiplexes 608 and eight XOR gates 606 are shown in stage 1 602 of Figure 6 for the arrangement shown.

Each multiplexer effectively controls the toggles 607 at each stage in order to select the toggled output or not, in which case the input to the toggle is selected as the output. This may be achieved in several ways and one such example is shown in Figure 13.

In Figure 13 each toggle 1301 is operatively connected to a multiplexer 1302, or data selector, which effectively selects between two inputs 1303, 1304 and transmits the selected input to the output 1305. The first input 1303 of the multiplexer 1302 is connected to the output of the toggle 1301 and the second input 1304 of the multiplexer is connected to the input 1306 of the toggle 1301.

Thus, in operation, the multiplexer 1302 can select between the toggled output and the input to the toggle, wherein the input to the toggle is the output of a programmable divider or the XOR gate of the previous stage.

The output of an adjacent pair of toggles is then transmitted to an XOR gate 1307.

The selection of the inputs to pass through to the output of the multiplexer may be controlled via control input 1308 from a control register in the DSiPM device.

The arrangement shown in Figure 13 may be replicated for each pair of toggles in each stage of the DSiPM device.

Returning to Figure 6, each XOR gate 606 of stage 1 602 receives input from a group of two multiplexers 608. As mentioned above, if the toggled output is selected then the toggle will effectively halve the data rate transmission of the transitions to the next stage compared to the input to the toggle. As well as reducing power consumption, this halving of the transition rate at each stage ensures that the maximum transition rate supported by each stage is not exceeded. This is particularly important at the higher levels of the XOR tree where the transition rate is potentially much higher. It should be noted that this toggling in the XOR tree has the same effect as the post pixel divider in that each transition represents multiple detected photons. Stage 2 603 similarly includes one or more multiplexers 608 and one or more XOR gates 606, where eight multiplexers 608 and four XOR gates 606 are shown in stage

2 603 of Figure 6. Each multiplexer 608 controls the output of the associated toggle 607, where each toggle 607 is operatively connected to the output of a respective XOR gate 606 from the previous stage, i.e. from stage 1 602. Each XOR gate 606 of stage 2 603 receives input from a group of two multiplexers 608, If the toggled output is selected by the associated multiplexer 608 then the data rate transmission of the transitions to the next stage is effectively halved compared to the input to the toggle.

Stage 3 604 similarly includes one or more multiplexers 608 and one or more XOR gates 606, where four multiplexers 608 and two XOR gates 606 are shown in stage

3 604 of Figure 6. Each multiplexer 608 controls the output of the associated toggle 607, where each toggle is operatively connected to the output of a respective XOR gate 606 from the previous stage, i.e. from stage 2 603. Each XOR gate 606 of stage 3 604 receives input from a group of two multiplexers 608. If the toggled output is selected by the associated multiplexer 608 then the data rate transmission of the transitions to the next stage is effectively halved compared to the input to the toggle.

Stage 4 605 similarly includes one or more multiplexers 608 and one or more XOR gates 606, where two multiplexers 608 and one XOR gate 606 is shown in stage 4 605 of Figure 6. Each multiplexer 608 controls the output of the associated toggle 607, where each toggle is operatively connected to the output of a respective XOR gate 606 from the previous stage, i.e. from stage 3 603. The XOR gate 606 of stage

4 605 receives input from a group of two multiplexers 608. If the toggled output is selected by the associated multiplexer 608 then the data rate transmission of the transitions to the next stage is effectively halved compared to the input to the toggle.

As the components (e.g. programmable divider, toggle, and/or XOR gate) are physically located close, or proximate, to the pixels or the previous XOR gate to which they are operatively connected then this advantageous arrangement significantly reduces the power dissipation and/or power consumption of the DSiPM device. Furthermore, as the divider is programmable, thereby enabling a divisor to be programmed or set for each divider, and as the multiplexer can be controlled to select or not select a toggle output, then the operation of the DSiPM may be optimised for a wide dynamic range of photon detection rates.

With reference to Figure 7, a schematic of the combination and division of signals is shown for a DSiPM with x stages.

The arrangement of Figure 7 shows a plurality of programmable dividers 701 wherein each programmable divider 701 is operatively connected to a respective pixel and each programmable divider outputs a train of transitions 702. Each output transition represents one or more photons detected by the respective pixel, where the number of photons represented by each output transition is dependent on the divisor programmed in the divider.

The output of each programmable divider is transmitted through a respective toggle, multiplexer and XOR combination of a first stage 703. The output of the toggle is a series of transitions and, if the toggled output is selected by the multiplexer, the number of transitions output 705 is effectively half the number of transitions input to the toggle, thereby reducing the data rate transmissions. This also means that each output transition represents double the number of detected photons that are represented by the input transitions to the toggle from the programmable dividers.

The output of each toggle, multiplexer and XOR gate combination from the first stage 703 is transmitted through a further respective toggle and multiplexer of the second stage 704, wherein, if the toggled output is selected by the multiplexer, the number of transitions output 706 is effectively half the number of transitions input to the toggle, thereby reducing the data rate transmissions further. This also means that each output transition represents double the number of detected photons that are represented by the input transitions to the toggle from the first stage. The output of two adjacent toggles is then transmitted through a respective XOR gate in order to combine the outputs of the two toggles in the second stage 704.

This is repeated for further stages in the x stage DSiPM until the final stage 70x is reached which is the output 707 of the DSiPM, indicating the number of photons detected within the DSiPM. As will be appreciated, the value of x, i.e. the total number of stages, is dependent on the arrangement and design of the DSiPM in respect of the number of pixels. With reference to Figure 8, a schematic of a DSiPM comprising an arrangement of eight pixels and three stages is shown. Each pixel 801 is comprised of a 5x5 array of SPADs 802 and the output from a pair of pixels 801 is combined and divided via the toggles, multiplexer and XOR gates at a first stage 803. The output of the first stage 803 is then combined and divided via the toggles, multiplexer and XOR gates at a second stage 804. The output of the second stage 804 is then combined and divided via the toggles, multiplexer and XOR gates at a third stage 805, and so on to further stages 806 as necessary for the design and application of the specific DSiPM.

As will be appreciated, there can be any number of stages within the DSiPM device in the examples of Figures 6 to 8, where the number of stages implemented may be dependent on the design and application of the DSiPM device. In the examples of Figures 6 to 8, each stage comprises a number of toggles, multiplexers and XOR gates, however, as mentioned above each stage may additionally include a programmable divider, or to replace one or more of the toggles with a programmable divider, to provide a greater flexibility in the reduction of the data transmission rates.

In order to maintain timing information of incoming events for applications that may require such timing information, the combined and divided overall output counts of the DSiPM may be fed into a rolling shift register which may be synchronised to an external clock. Thus, the shift register and its associated output logic may be the only synchronous part of the device which also reduces the power requirement of the DSiPM device. The rolling shift register may be divided into bins representing a suitable time period, for example, a time period in the range of 1 nanosecond to 10 microseconds depending on the external clock period, and the total number of photons detected during each time period, as represented by the output transitions of the DSiPM, can be added to the current bin of the rolling shift register. Each bin may include one or more bits, for example, 1 to 12 bits, to store the number of detected photons within the respective time period of each bin. The time period of each bin may be adjusted based on the external clock period, and the number of bits allocated to each bin may be selectable or programmable, for example, via the microcontroller, depending on the design of the DSiPM and/or the application of the DSiPM, which advantageously increases the flexibility of the DSiPM. A DSiPM may typically comprise more than 100 pixels, where each pixel comprises a two-dimensional array of SPADs. However, as will be appreciated, the DSiPM may comprise any number of pixels suitable for the intended application of the DSiPM. The random distribution of detected photons and dark counts among the pixels of the DSiPM may ensure that the magnitude and temporal distribution of incoming photons is accurately represented.

As DSiPMs may be employed in a wide range of applications, such as radiation detection applications with scintillating crystals, incoming X-ray or gamma photons interact with a scintillator to produce a pulse of visible light with an intensity that is proportional to the energy of the incoming radiation. In typical operation and in many applications the radiation events are relatively infrequent. Accordingly, logic built into the DSiPM capable of detecting the occurrence of an event of interest may also be used to trigger the transfer of data to a memory buffer for communication off-chip to other system components of the device that the DSiPM is integrated within, while advantageously rejecting spontaneous background noise generated in the device as such spontaneous background noise would not be sufficient to trigger the transfer of data.

By restricting data transfer to only events of interest, e.g. those events associated with a burst of photons, for example, from a gamma induced scintillation event which will typically last between 10ns and 10ps, further significant reduction in power consumption and/or power dissipation may also be achieved in both the DSiPM and other system components of the device.

Three triggering mechanisms are envisaged and proposed for event-based triggering where one or more of the triggering mechanisms may be employed by the DSiPM.

The first trigger mechanism is schematically shown in Figure 9. This utilises the rolling shift register 901 that is operatively connected to the output of the DSiPM which is received as an input 902 to the rolling shift register 901 . As discussed above, the output of the DSiPM is stored in bins 905 of the rolling shift register 901 where each bin 905 represents a time period T, for example, T may be set as 10ns.

Once the period of time T for a first bin of the rolling shift register 901 has elapsed the current detected combined and divided transition counts output from the DSiPM and provided as input 902 to the rolling shift register 901 are transferred to and stored in the first bin. Subsequently, all bins 905 may be shifted to the right such that after then next period of time T the output of the DSiPM and provided as input 902 to the rolling shift register 901 is stored in the next bin. This process is continually repeated such that the rolling shift register 901 effectively contains the time history of the detector output.

Adder 904 may sum a first set of the n most recent bins A. The value of n can be predetermined or controlled/programmed by a programmable register. In the example of Figure 9, the value of n is set to 5 but, as will be appreciated, the value of n could be any suitable value. By summing n bins the first trigger mechanism is advantageously relatively immune to noise due to the integration inherent in the summations.

The sum of the first set of the n most recent bins A may then be compared by a digital comparator 906 with a reference value contained in a register B order to determine the start of an event of interest and subsequently the data transfer. An event detected signal is produced when the digital number in A exceeds that in B and a trigger signal 911 is output by the comparator 906.

The first trigger mechanism shown in Figure 9 is therefore effectively a comparator that compares the number of detected photons output by the DSiPM to a threshold and if the threshold is exceeded an event of interest may be determined and the transfer of data can be triggered by an output signal 909 of the first trigger mechanism.

The threshold may be any suitable threshold for the design and application of the DSiPM. For example, each DSiPM may be subject to a certain level of noise or Dark Counts and therefore the threshold may be predetermined, or programmed by the microcontroller in use, to a value that is just above, or by a predetermined value above, the expected or determined noise of the DSiPM chip and/or above the expected Dark Count that may occur in the DSiPM. As will be appreciated, the threshold may be different per DSiPM as each DSiPM may be subject to different levels of noise and different expected Dark Counts depending on the application and implementation of the DSiPM. The first trigger mechanism may further include a trigger clock 903, a timer 907 which provides an enable signal 910, and an AND gate 908. The trigger clock has a period equal to n times the shift register period and triggers the summation and comparison operations. The timer 907 may be initiated to provide a negative going pulse, starting shortly after an event has been detected and ending after a time similar to or longer than the expected duration of the events being monitored. This signal provides an enable signal 910 that is combined by the AND gate 908 with the trigger signal from the comparator. No trigger is output 909 when the enable signal 910 is low. This ensures that the output 909 is a short single pulse and that only one trigger pulse occurs for each event.

A second trigger mechanism is shown in Figure 10 and applies a multi-pole analogue filter and comparator to achieve a similar result to the first trigger mechanism. With reference to Figure 10, the output of the DSiPM 1001 is transmitted through at least one filter stage 1002. The number of filter stages is dependent on the accuracy required by the DSiPM device and/or the application as the greater number of filter stages improves the Gaussian shape pulse output of the filters. As an example, the number of filter stages may range between 1 and 8. The filters may also include an amplifier to prevent or reduce attenuation of the signal through the filters.

Alternatively, filtering could be implemented using digital filtering techniques taking data from a shift register storing the counts as in the first trigger mechanism above.

The output transitions of the DSiPM are transmitted through the respective filter stages 1002 and provided to a first input 1004 of a comparator 1003. A threshold value is provided to a second input 1005 of the comparator 1003. The threshold may be any suitable threshold for the DSiPM design and/or the application. The threshold value may be set just above, or a predetermined value above, the expected or determined noise in the DSiPM and/or the expected Dark count in the DSiPM. The threshold value may be predetermined and set for the comparator or may be controlled/programmed by the microcontroller in use. The threshold value may be in millivolts.

The comparator 1003 compares the output of the filter stages 1002 to the threshold value and, if the output of the filter stages exceeds the threshold value, an event of interest may be determined and the transfer of data can be triggered. As an alternative to one or more of the filters, an integrator/differentiator pair may be implemented.

A third trigger mechanism is shown schematically in Figure 11 , which utilises the rolling shift register 1101 that is operatively connected to the output of the DSiPM which is received as an input 1102 to the rolling shift register 1102. As discussed above, the output of the DSiPM is stored in bins 1103 of the rolling shift register 1101 where each bin 1103 represents a time period T, for example, T may be set as 10ns.

Once the period of time T for a first bin of the rolling shift register 1101 has elapsed the current detected combined and divided transition counts output from the DSiPM, which is received as an input 1102 to the rolling shift register 1102, is transferred to and stored in the first bin. Subsequently, all bins may be shifted to the right such that after then next period of time T the output of the DSiPM, which is received as an input 1102 to the rolling shift register 1102, is stored in the next bin. This process is continually repeated such that the rolling shift register 1101 contains the time history of the detector output.

Adders 1106 may sum a first set of the n most recent bins A and a second set of the n next most recent bins B. The value of n can be predetermined or controlled/programmed by the microcontroller in use. In the example of Figure 11 , the value of n is set to 5 but, as will be appreciated, the value of n could be any suitable value. By summing n bins in the third trigger mechanism it is advantageously relatively immune to noise due to the integration inherent in the summations.

The sum of the first set of the n most recent bins A may then be compared by a comparator 1104 with the sum of a second set of the n next most recent bins B, in order to determine the start of an event of interest and subsequently the data transfer. The comparator produces a trigger output 1113 when the value in A exceeds that in B.

Additionally, the likelihood of false triggers due to noise within the DSiPM chip may be further reduced by left shifting 1105 (e.g. multiplying) the result of the summation B prior to the comparison, as shown in Figure 11. In Figure 11 , the result of the summation B is left shifted by 3, but as will be appreciated the result of the summation of B may be left shifted by any value, where the value may be predetermined or controlled/programmed by the microcontroller in operation.

The digital comparator 1104 compares the sum A with sum B and if the value of A is greater than B an event of interest may be determined and the transfer of data can be triggered by an output signal 1108 of the third trigger mechanism.

The third trigger mechanism may further include a trigger clock 1107, a timer 1109 which provides an enable signal 1111 , a delay 1110 which acts as a reset 1113 for the summations, and an AND gate 1112.

The trigger clock has a period equal to n times the shift register period and triggers the additions, left shifting and comparison operations. Following the detection of an event, a delayed version of the comparator output is used to reset A and B registers to remove the trigger signal after a brief delay. In addition, the timer 1109 may be initiated to provide a negative going pulse, starting shortly after an event has been detected and ending after a time similar to or longer than the expected duration of the events being monitored. This signal provides an enable signal 1111 that is combined by the AND gate 1112 with the trigger signal 1113 from the comparator 1104. No output is provided when the enable signal is low. This ensures that the output 1108 is a short single pulse and that only one output pulse 1108 occurs for each event.

Turning now to Figure 12, as multiple DSiPMs 1201 may be used and combined into arrays which are coupled to a single scintillation detector 1202, then the first, second and third triggering mechanisms described above may be applied to multiple DSiPMs 1201 coupled in parallel. To realise a single trigger for multiple DSiPMs the data generated in each device, or the raw train of transitions generated in each device, is preferably summed together prior to being fed into the first, second or third trigger mechanisms. Alternatively, or additionally, the trigger output from each of the DSiPMs may be weighted for determining when to trigger the subsequent data transfer for a detected event. Alternatively, an external Field Programmable Gate Array may receive the output triggers from each of the DSiPMs and determine when to trigger the subsequent data transfer for a detected event.

A further feature of the DSiPM is the capability to turn off individual SPADs or groups of SPADs, while this has previously been employed to reduce noise and/or dark counts in devices, it has now been identified as a potential additional benefit for applications where radiation damage can be a problem for traditional silicon based photo detectors. As radiation damage, e.g. due to high energy neutrons, photons, or protons, is initially likely to be localised to individual SPADs, the ability to dynamically disable individual SPADs, or groups of SPADs, could significantly extend the amount of radiation damage an DSiPM device can absorb before it becomes un-usable. It is envisaged that this function could be automatically achieved e.g. on device power up under the control of the microcontroller. For example, damaged SPADs (or groups of damaged SPADs) may be identified, for example, by a microcontroller of the DSiPM, by the high spontaneous count rate they generate (for example, with no incident light). SPADs, or groups of SPADS, that have a count rate above a predetermined rate, or above a predetermined fraction of the whole DSiPM device may be identified and turned off, for example, by the microcontroller of the DSiPM. Thus, by identifying a count rate of individual SPADs, or groups of SPADs, at power on, or at predetermined regular intervals it may be determined that one or more SPADs, or group of SPADs, are damaged and are not functioning as expected. SPADs, or groups of SPADs, may be turned off by means of a transistor that removes power from the SPAD, group of SPADs, or downstream circuitry (e.g. programmable divider).

In the foregoing embodiments, features described in relation to one embodiment may be combined, in any manner, with features of a different embodiment in order to provide efficient and effective DSiPM device. Note that, the above description is for illustration only and other embodiments and variations may be envisaged without departing from the scope of the invention as defined by the appended claims.

Claims

Claims:

1 . A digital silicon photomultiplier comprising: at least one least one pixel, wherein each pixel includes at least one single photon avalanche photodiode and a toggle operatively connected to each of the at least one single photon avalanche photodiodes, wherein the toggle outputs at least one transition and each transition represents a detection of a photon; a programmable divider operatively connected to each pixel, wherein each programmable divider is programmed with a divisor; and each programmable divider receives a first set of transitions and outputs a second set of transitions, wherein each transition in the first set of transitions represents a detection of a photon by a single photon avalanche photodiode of a respective pixel and each transition of the second set of transitions represents a number of detected photons of the respective pixel equivalent to the divisor.

2. The digital silicon photomultiplier according to claim 1 , wherein the digital silicon photomultiplier comprises at least two pixels and further comprises: a first stage comprising two or more first toggles, wherein each first toggle is operatively connected to one programmable divider; wherein each first toggle receives one or more transitions output from the respective programmable divider and outputs one or more transitions.

3. The digital silicon photomultiplier according to claim 2, in which if a toggled output of the first toggle is selected then each output transition represents two received transitions output from the respective programmable divider.

4. The digital silicon photomultiplier according to claim 2 or 3, in which the first stage further comprises: one or more first XOR gates wherein each first XOR gate is operatively connected to two adjacent first toggles; and wherein each first XOR gate combines the output of the two adjacent first toggles to a single output from the first XOR gate.

5. The digital silicon photomultiplier according to claim 4, further comprising:

24 one or more further stages, wherein each further stage comprises: two or more further toggles and/or programmable dividers, wherein each further toggle and/or programmable divider is operatively connected to one XOR gate of the previous stage; and one or more further XOR gates, wherein each further XOR gate is operatively connected to two adjacent further toggles.

6. The digital silicon photomultiplier according to claim 5, in which if a toggled output of the further toggle is selected then each output transition of each further toggle represents two received transitions output from the respective XOR gate of the previous stage.

7. The digital silicon photomultiplier according to any one of the previous claims in which the programmable divider is a ripple counter based on flip-flops.

8. The digital silicon photomultiplier according to any one of the previous claims in which the divisor of each programmable divider is programmable via a control input to each programmable divider.

9. The digital silicon photomultiplier according to any one of the previous claims, further comprising a rolling shift register operatively connected to an output of the silicon photomultiplier.

10. The digital silicon photomultiplier according to claim 9, in which the rolling shift register is divided into bins, wherein each bin includes a number of bits and represents a period of time; and a total number of photons detected during each period of time is stored in the respective bin.

11 . The digital silicon photomultiplier according to claim 9 or 10, in which the number of bits allocated to each bin and the time period of each bin are programmable via a control input to the rolling shift register.

12. The digital silicon photomultiplier according to any one of the preceding claims, further comprising: one or more trigger mechanisms to determine one or more events of interest, wherein if triggered the trigger mechanisms trigger a transfer of data from the digital silicon photomultiplier.

13. The digital silicon photomultiplier according to claim 12, when dependent on any one of claims 9 to 11 , in which a first trigger mechanism comprises: a comparator, wherein the comparator compares a number of detected photons in one or more bins to a threshold; and if the number of detected photons in the one or more bins exceeds the threshold an event of interest is determined and the transfer of data is triggered.

14. The digital silicon photomultiplier according to claim 12, in which a second trigger mechanism comprises: one or more filter stages operatively connected to a comparator, wherein the comparator compares the output of the one or more filter stages to a threshold; and if the output of the one or more filter stages exceeds the threshold an event of interest is determined and the transfer of data is triggered.

15. The digital silicon photomultiplier according to claim 12, when dependent on any one of claims 9 to 11 , in which a third trigger mechanism comprises: summing a first set of bins, wherein the first set of bins represents a most recent set of bins; summing a second set of bins, wherein the second set of bins represent a next most recent set of bins; a comparator, wherein the comparator compares the sum of the first set of bins to the sum of the second set of bins; and if the sum of the first set of bins is greater than the sum of the second set of bins an event of interest is determined and the transfer of data is triggered.

16. The digital silicon photomultiplier according to claim 15, further comprising: shifting the sum of the second set of bins to the left by a number of bits prior to the comparison.

17. The digital silicon photomultiplier according to any one of the preceding claims, in which each pixel includes two or more single photon avalanche photodiodes arranged in a two-dimensional array.

18. The digital silicon photomultiplier according to any one of the preceding claims, further comprising: identifying one or more damaged single photon avalanche photodiodes; and disabling the one or more identified damaged single photon avalanche photodiodes.

19. The digital silicon photomultiplier according to claim 18, in which identifying the one or more damaged single photon avalanche photodiodes comprises: comparing a count rate of transitions from the one or more single photon avalanche photodiodes to a predetermined threshold; and disabling the one or more identified damaged single photon avalanche photodiodes comprises: removing power from the one or more identified damaged single photon avalanche photodiodes.

20. A method of operating a digital silicon photomultiplier, wherein the digital silicon photomultiplier comprises: at least one least one pixel, wherein each pixel includes at least one single photon avalanche photodiode and a toggle operatively connected to each of the at least one single photon avalanche photodiodes, wherein the toggle outputs at least one transition and each transition represents a detection of a photon; and a programmable divider operatively connected to each pixel, wherein each programmable divider is programmed with a divisor; the method comprising: receiving at each programmable divider a first set of transitions wherein each transition in the first set of transitions represents a detection of a photon by a single photon avalanche photodiode of a respective pixel; and outputting a second set of transitions, wherein each transition of the second set of transitions represents a number of detected photons of the respective pixel equivalent to the divisor.

21 . The method according to claim 20, wherein the digital silicon photomultiplier comprises at least two pixels and a first stage comprising two or more first toggles,

27 wherein each first toggle is operatively connected to one programmable divider; the method further comprising: receiving at each first toggle one or more transitions output from the respective programmable divider; and outputting one or more transitions.

22. The method according to claim 21 , further comprising: selecting a toggle output of the first toggle such that each output transition represents two received transitions output from the respective programmable divider.

23. The method according to claim 21 or 22, in which the first stage further comprises one or more first XOR gates wherein each first XOR gate is operatively connected to two adjacent first toggles; and the method further comprises combining, by each first XOR gate, the output of the two adjacent first toggles to a single output from the first XOR gate.

24 The method according to claim 23, in which the digital silicon photomultiplier further comprises one or more further stages, wherein each further stage comprises two or more further toggles and/or programmable dividers, wherein each further toggle and/or programmable divider is operatively connected to one XOR gate of the previous stage; and one or more further XOR gates, wherein each further XOR gate is operatively connected to two adjacent further toggles; and combining, by each further XOR gate, the output of the two adjacent further toggles to a single output from the further XOR gate.

25. The method according to claim 24, further comprising: selecting a toggled output of each further toggle such that each output transition of each further toggle represents two received transitions output from the respective XOR gate of the previous stage.

26. The method according to any one of claims 20 to 25, further comprising: programming, via a control input of each programmable divider, the divisor of each programmable divider.

28

27. The method according to any one of claims 20 to 26, in which the digital silicon photomultiplier further comprises a rolling shift register operatively connected to an output of the digital silicon photomultiplier wherein the rolling shift register is divided into bins, the method further comprising: storing a total number of photons detected during each period of time, wherein each bin represents the period of time and each bin includes a number of bits.

28. The method according to claim 27, further comprising: programming, via a control input of the rolling shift register, the number of bits and time period for each bin.

29. The method according to any one of claims 20 to 28, further comprising: determining one or more events of interest; and triggering, by one or more trigger mechanisms, a transfer of data from the silicon photomultiplier if an event of interest is determined.

30. The method according to claim 29, when dependent on claim 27 or 28, in which a first trigger mechanism comprises: comparing a number of detected photons in one or more bins to a threshold; determining an event of interest based on the comparison; and triggering the transfer of data based on the determined event of interest.

31 . The method according to claim 29, further comprising: filtering an output of the silicon photomultiplier by one or more filter stages; comparing an output of the one or more filter stages to a threshold; and determining an event of interest based on the comparison; and triggering the transfer of data based on the determined event of interest.

32. The method according to claim 29, when dependent on claim 27 or 28, further comprising:

29 summing a first set of bins, wherein the first set of bins represents a most recent set of bins; summing a second set of bins, wherein the second set of bins represent a next most recent set of bins; comparing the sum of the first set of bins to the sum of the second set of bins; determining an event of interest based on the comparison; and triggering the transfer of data based on the determined event of interest.

33. The method according to claim 32, further comprising: shifting the sum of the second set of bins to the left by a number of bits prior to the comparison.

34. The method according to any one of the claims 20 to 33, further comprising: identifying one or more damaged single photon avalanche photodiodes; and disabling the one or more identified damaged single photon avalanche photodiodes.

35. The method according to claim 34, in which identifying the one or more damaged single photon avalanche photodiodes comprises: comparing a count rate of transitions from the one or more single photon avalanche photodiodes to a predetermined threshold; and disabling the one or more identified damaged single photon avalanche photodiodes comprises: removing power from the one or more identified damaged single photon avalanche photodiodes.

36. A device comprising: a microcontroller; and one or more digital silicon photomultipliers according to any one of claims 1 to

19.

30