betaFeedback

WorksPhysics and Astronomy

Below 100 ps CTR using FastIC+, an ASIC including on-chip digitization for ToF-PET and beyond

Abstract

This work presents the 8-channel FastIC+, a low-power consumption and highly configurable multi-channel front-end ASIC with internal digitization, for the readout of photo-sensors with picosecond time resolution and intrinsic gain. This ASIC, manufactured in 65 nm CMOS technology, can readout positive or negative polarity sensors and provides a digitized measurement of the arrival time and energy of the detected events with a power consumption of 12.5 mW per channel. On-chip digitization is executed by a Time-to-Digital Converter (TDC) based on a Phase-Locked Loop (PLL) generating 16 phases at 1.28 GHz. The internal TDC introduces a jitter contribution of 31.3 ps FWHM, with minimal impact on timing measurements. When evaluating FastIC+ to readout 3×\times3 mm2^2 silicon photomultipliers (SiPMs) with a pulsed laser, we achieved a single-photon time resolution (SPTR) of (98 ±\pm 1) ps FWHM. We also performed time-of-flight positron emission tomography (ToF-PET) experiments using scintillator crystals of different sizes and materials. With LYSO:Ce,Ca crystals of 2.8×\times2.8×\times20 mm3^3 we obtained a coincidence time resolution (CTR) of (130 ±\pm 1) ps FWHM. With LGSO crystals of 2×\times2×\times3 mm3^3, a CTR of (85 ±\pm 1) ps FWHM. To the best of our knowledge, this is the first time that a CTR below 100 ps using on-chip digitization is reported.

Introduction

Time-of-flight Positron Emission Tomography (ToF-PET) has become an essential tool in medical imaging over the last years thanks to the improvement on the spatial resolution and the Signal-to-Noise Ratio (SNR) of PET scans compared to conventional PET systems [1]. One of the key parameters that influences the performance of these scans is the Coincidence Time Resolution (CTR), which determines the accuracy of the time measurement of two detected 511 keV photons with antiparallel momenta resulting from a positron-electron annihilation. This is used to estimate the position of the annihilation along the so-called Line-of-Response. Therefore, improving the CTR is critical to improve the image resolution [2, 3], specially for whole-body PET applications [4, 5].

Several factors influence the CTR performance, including the scintillation crystal, the photosensor and the readout electronics. Typically, fast scintillators based on Lutetium Oxyorthosilicate (LSO) and Lutetium Yttrium Orthosilicate (LYSO) have been used in ToF-PET detectors because of their excellent timing response [6], high stopping power and fast-decay time (∼40\sim40 ps). Regarding the photosensors, Silicon Photomultipliers (SiPMs) have become the detector of choice due to their compact size, high Photon Detection Efficiency (PDE) and insensitivity to magnetic fields [7]. For instance, Fondazione Bruno Kessler (FBK) has developed near-ultraviolet SiPMs with a PDE close to 70% at 420 nm with reduced dark noise and crosstalk probability. Furthermore, an intrinsic Single Photon Time Resolution (SPTR) below 50 ps for a 3×3 mm² SiPM has been achieved with this technology [8].

However, improving the time resolution requires not only optimizing scintillator and photodetector properties, but also the front-end electronics [9, 10, 11]. High-frequency (HF) amplifiers with low electronic jitter, (i.e., a large slew rate compared to the electronic noise) have been employed to study the fundamental limits of the photodetector-crystal pair [12]. This solution has reported so-far the best timing performance [6, 13]. However, the implementation of this readout circuit has been limited to small-scale prototypes due to their high power consumption (up to 150 mW/channel). In [14], the trade-off between power consumption and measured CTR has been studied for different amplifiers, showing that lower-power amplifiers can also achieve good timing results. Besides power consumption, this solution becomes impractical or suboptimal for scalability due to the physical dimensions of the electronics in large systems, where thousands of detectors are involved, and a more compact design is needed. The digitization of timing and energy information is another important aspect at the system level. The HF readout only includes analog signal processing, requiring an additional circuit to digitize the data. Analog-to-Digital Converters (ADCs) are a suitable option, especially for energy measurements, where timing is not an issue [15]. Regarding timing measurements, a Time-to-Digital Converter (TDC) is a more power-efficient circuit for obtaining precise timestamps with low power consumption. Two trends in TDC implementation already exist [16]: (1) Field Programmable Gate Array (FPGA) based TDCs that use the fastest delay elements in the device which can achieve good timing with moderate power consumption, and (2) Application Specific Integrated Circuits (ASIC) based TDCs that can be customized for specific applications, offering better overall performance particularly concerning the power consumption, and giving a similar timing resolution. Additionally, ASIC-based TDCs are more compact and scalable since the TDC can be integrated into the same ASIC as the analog readout.

Over the years, several ASICs have been proposed as compact and power-efficient readout circuits. ASICs with analog outputs include the NINO ASIC [17], which provides only timing information, and the Weeroc RADIOROC2 [18], which provides timing output as a binary signal and energy as an analog signal. Other ASICs that provide binary outputs include HRFlexToT and FastIC [19, 20, 21]. Among the ASICs with fully digital outputs (integrating internal ADCs or TDCs) are the Weeroc PETIROC2A [22] and the PETsys TOFPET series (TOFPET2, TOFPET3) [23].

In this work, we present the FastIC+, a low-power, multi-channel front-end ASIC with an integrated TDC. We describe its architecture, characterize the TDC performance using electric test signals, and evaluate the ASIC’s performance for the direct detection of optical photons using SiPMs, as well as for gamma-ray detection in PET applications using scintillators coupled to SiPMs.

FastIC+ Architecture

The analog processing of the FastIC+ is similar to that used in the FastIC [24], [25]. The digitization stage employs an event-driven TDC based on a Phase-Locked Loop (PLL) to achieve high-precision time tagging. Figure 1 illustrates the architecture of the FastIC+, including both the analog and digital signal processing stages.

Figure 1. FastIC+ block diagram, including both analog and digital signal processing.

Analog readout

The signals from sensors with intrinsic amplification, such as SiPMs, multi-anode photomultiplier tubes, or microchannel plates (MCPs), are read out by a current-mode input stage. The input stage can deal with positive or negative polarity signals. A current conveyor, based on a double feedback mechanism [26], [27], is employed to stabilize the input voltage and maintain a low input impedance within the desired bandwidth. Additionally, it creates three copies of the signals that are processed by three different branches: ”Time”, ”Energy” and ”Trigger”, as shown in Figure 1.

The time path compares a copy of the signal delivered by the sensor to a threshold using a leading-edge current comparator with non-linear feedback to ensure high-speed operation [28], [29]. This path provides (1) precise photon Time-of-Arrival (ToA) information and (2) energy measurement using a low-jitter, non-linear Time-over-Threshold (ToT) response.

The input stage trigger branch follows the same processing scheme but attenuates the input current, since its jitter requirements are less stringent, allowing for reduced power consumption. The trigger threshold can be adjusted independently of the time branch to validate events or serve as a second timing threshold for rise-time measurements [30]. The trigger signal is generated as a logical OR across all channels, meaning that only the fastest channel’s timing information is available. Additionally, the time signal can serve as a trigger for low-light-level detection, and an external trigger can be used for calibration purposes. Lastly, a high-level trigger is generated by summing the input signals from all channels in the analog domain before the comparator stage.

The dedicated signal for energy measurement is initially converted from current to voltage using a Trans-Impedance Amplifier (TIA). This path also employs a shaper with a first-order active integrator to control the peaking time and a pole-zero cancellation network to reduce signal tailing.

A peak-and-hold detector (PDH), based on [31], captures the pulse amplitude. The PDH can be adjusted to function as a track-and-hold circuit and includes a discharge mechanism to prevent inaccuracies from undesired pulses (e.g., dark pulses below the trigger threshold), as described in [26]. The detected peak is then compared to a linear ramp to emulate a single-slope ADC, producing a pulse whose duration is proportional to the input peak amplitude. The ramp starts with a lower DC voltage compared to the DC baseline of the energy signal to allow for the calibration of channel-to-channel variations. The energy pulse captured in the absence of an input signal is referred to as the pedestal and must be calibrated and then subtracted in the data analysis. The time and energy signals from all eight channels, along with the validation trigger, are then sent to the digitization stage.

Phase-Locked Loop (PLL)

Digitization requires low-jitter internal signals, achieved through a Phase-Locked Loop (PLL), which multiplies the 40 MHz reference clock by a factor of 32, resulting in a frequency of 1.28 GHz. The PLL comprises several key components: a Phase and Frequency Detector (PFD), a charge pump, a loop filter, a Voltage-Controlled Ring Oscillator (VCO) and a clock manager, which includes a frequency divider, and buffers.

The digital PFD compares the system reference clock at 40 MHz with a feedback clock generated from the VCO to determine whether the generated clock leads or lags the reference. This digital implementation allows control over how many cycles remain in the locked or unlocked state when detecting a phase difference, emulating hysteresis behavior. The charge pump charges and discharges the VCO’s control voltage based on the phase detector’s status. It is implemented using a regulated cascode current mirror with a programmable range to ensure an accurate current source. The VCO control voltage includes a buffer to mitigate charge-sharing effects. A loop filter with large capacitors is necessary to filter noise-induced fluctuations in the control voltage caused by the high VCO gain (7 GHz/V), ensuring a smooth control signal.

The VCO consists of 16 delay elements based on the current-steering differential delay cell [32], connected in a ring, with one cell output cross-coupled to ensure an odd number of inversions. The impedance of the diode-connected transistors, which affects the circuit’s RC constant and thus the delay, is regulated by the VCO’s control voltage. This implementation provides a wide tuning range, excellent power supply rejection, and decouples the control voltage from the power supply while achieving a time bin (delay of the cell) of 24.4 ps.

A buffer is added to each of the VCO phases before sending them to the next stage, forming the 5 Least Significant Bits (LSBs) of the TDC. These buffers serve two purposes: isolating the VCO from the Time Capture Registers and enabling compensation for time bin non-uniformities (e.g., for calibration to improve time bin linearity). Each buffer includes configurable bits to adjust both the rising and falling edge delays. Additionally, the odd-numbered stages can be disabled to reduce power consumption, at the cost of increased time bin width (49 ps), as only half of the VCO phases are used in subsequent stages.

The clock manager consists of a frequency divider. It generates reference clocks for the TDC blocks derived from the VCO’s 1.28 GHz clock. These clocks include the coarse counter, the 40 MHz feedback clock for the phase detector, and the 40 MHz synchronization clock. It also generates VCO counts for the Front-End Readout (FERO) block, contributing to the 5 most significant bits (MSBs) of the TDC.

Simulations considering variations in process (fast, slow, and typical corners), temperature (from -40 °C to 80 °C), and voltage (from 1.08 V to 1.32 V, with a typical value of 1.2 V) show that the frequency of the VCO can be locked at 1.28 GHz by automatically adjusting the VCO’s control voltage. The simulated jitter is below 5 ps, the phase noise is -80 dBc/Hz at 100 MHz, and the power consumption is 12 mW.

Front-End Readout (FERO) block

The digitization phase starts with the FERO block, which asynchronously receives binary pulses (hits) from the analog stage (time, energy, trigger, or both as consecutive pulses). The proposed architecture allows simultaneous rising and falling edge capture, as both are processed separately.

The FERO Control block, implemented using an Asynchronous Finite State Machine, ensures that multiple hits within the same 25 ns clock period are prevented. Pulses following an initial one are filtered to prevent the Time Capture Registers from firing twice, limiting the readout throughput to the 40 MHz reference clock.

The Ultra-Fast Time Capture Matrix (UF-TCM) includes 16 sampling elements (master-slave flip-flops optimized for timing) that store the internal state of the PLL’s VCO at the rising edge of the hit, achieving a time bin of 24.4 ps. Timing resolution for the falling edge is less stringent, as it is used only for pulse width measurement. The hit signal is synchronized with the PLL’s 1.28 GHz clock frequency. The Fast Time Capture Matrix (TCM) uses standard-cell flip-flops to capture the VCO’s 5-bit Gray counter, which encodes the number of oscillations within a 25 ns period, for both rising and falling edges of the hit. The time bin of the TCM is 390 ps, corresponding to half the VCO clock period.

Finally, a synchronization block resamples the data at 40 MHz to mitigate metastability issues. From this stage onward, all data becomes synchronous, eliminating the need for stringent timing constraints.

In total, one input hit generates 27 output bits for processing in subsequent stages. Rising edge information encodes 16 bits (thermometric encoding) from the UF-TCM, 5 bits from the Fast TCM. Falling edge information encodes 5 bits from the Fast TCM and 1 bit for the phase. The time bin for the rising and falling edges is 25 ps and 390 ps, respectively.

The Back-End Readout (BERO) block

The Back-End Readout (BERO) block processes the captured hit edges. The de-bubbling logic corrects TDC bubbles (errors in the digital output code that produce corrupted output codes after the sampling elements due to mismatches in delay elements, signal integrity issues or noise). In the FastIC+ architecture these may arise in the UF-TCMs, which sample VCO phases. After correction, the thermometer counter data is encoded into binary form to reduce the number of bits.

The Hit Pipeline (Rise, Fall) introduces a systematic latency of 75 ns to the acquired pulses by using three cascaded flip-flop stages operating at 40 MHz. This delay allows an external trigger system to perform event validation (such as a coincidence trigger in a PET system) and determine whether to store the already acquired data based on the validation signal. The external validation signal can be injected into the ASIC through the same external trigger input, while the energy acquisition trigger must rely on internal triggers.

The pulse factory combines rising and falling edges to construct a single data packet. Additionally, it can filter out pulses not validated by an external trigger or discard invalid pulses, such as those with consecutive rising or falling edges.

The Pulse Processor encodes the rising and falling edge counters into Timestamps (ToA) using the Ultra Fast, Fast, and Coarse Rise Counters, and Pulse Width (ToT) using the Fast and Coarse counters for both edges. It also provides the option to filter pulses by width. Note that the BERO block receives the coarse counter (12 bits), generated by a 40 MHz clock from the frequency divider. Finally, pulses are stored in the channel FIFO (First Input First Output) queue, awaiting the Arbiter’s grant.

Data transmission

The Arbiter and Multiplexer receive bus requests from the TDC channels and manage data transmission according to two arbitration policies: Round-Robin (RR) and Pulse Sorting by Timestamp (PST). In the RR mode, events are transmitted cyclically in the channel order, while in PST mode, events are sorted based on their arrival time, facilitating efficient data post-processing.

The multiplexed data from the Arbiter is stored in a global FIFO queue before being serialized and transmitted via a single Scalable Low-Voltage Signaling (SLVS) output. The data transmission protocol used at the link layer is Aurora 64/66 [33], supporting data rates ranging from 80 Mbps to 1.28 Gbps. This protocol enables the transmission of the information of up to 3 million pulses per second per channel (3 MHz/channel), ensuring high-speed data transfer.

Additionally, the TDC generates statistics, such as the number of filtered or discarded hits, at a low rate to aid in debugging and troubleshooting readout issues. To prevent overflow errors at low rates, the system also features an extended Coarse Counter (12 additional bits, for a total of 24 bits).

FastIC+ Configurability

Several internal registers are included in the ASIC to configure various parameters. These registers can be modified via the I2C communication protocol. The ASIC can operate in analog mode, functioning like its predecessor FastIC, where each channel provides either an analog or SLVS response. Alternatively, in digital mode, the Time-to-Digital Converter (TDC) is activated, and all information is transmitted through a single data link.

The TDC operates in four transmission modes, depending on the type of information encoded in the data link. In High Energy Resolution mode, both the Time signal (Time-of-Arrival, ToA) and Energy (Time-over-Threshold, ToT) are transmitted. The falling edge of the time pulse is generated arbitrarily to reduce the pulse width, thereby increasing the maximum allowable event rate when the ToT of the Time signal is not required. In High-Speed mode, only the ToA and the non-linear ToT of the Time signal are provided. Hybrid mode combines the ToA and non-linear ToT of the Time signal with the standard ToT energy measurement. Finally, in Single Pulse mode, either the ToA or the ToT of the Time signal is transmitted, depending on the configuration. Figure 2 illustrates how ToA and ToT (named as energy width) are generated from the time and energy paths, respectively, and transmitted in High Energy Resolution mode.

Response of the FastIC+ in High Energy Resolution mode

Figure 2. Response of the FastIC+ in High Energy Resolution mode, where Time (ToA) and Energy (ToT) are transmitted.

Materials and methods

In this section we describe the SiPMs and scintillators employed and the setups used to asses the SPTR and CTR measurements. All measurements were conducted in a light-tight box under room temperature conditions (∼20 ∘C\sim20\,^\circ\mathrm{C}).

Crystals and SiPMs

The scintillators used are summarized in Table I. All crystals were wrapped with at least 5 layers of Teflon and glued to the SiPMs using Cargille Melmount with a refractive index of nD=1.582n_D = 1.582 at 588 nm and ∼25 ∘C\sim25\,^\circ\mathrm{C}. The characteristics of the SiPMs are summarized in Table II.

FastIC+ PCB module

We tested the FastIC+ ASIC using the FastIC+ evaluation module, as illustrated in Figure 3. This module consists of two printed circuit boards (PCBs). The top board primarily houses two FastIC+ ASICs, allowing up to 16 SiPMs channels to be tested. It includes connectors to power both ASICs (which also incorporate voltage regulators to ensure a stable power supply) and the SiPMs. The TDC digitized output is sent to the bottom board, which hosts an FPGA. This FPGA transmits the acquired measurements to a computer for analysis via USB. Additionally, it manages the configuration of the ASICs via an I2C master.

FastIC+ module designed to test the FastIC+ ASIC

Figure 3. FastIC+ module designed to test the FastIC+ ASIC.

Electrical setup

The evaluation of the TDC inside the FastIC+ has been carried out using the experimental setups shown in Figure 4. The intrinsic linearity of the TDC is evaluated in terms of Differential Non-Linearity (DNL) and Integral Non-Linearity (INL) by performing a code density test using the setup shown in Figure 4 (a). The number of bins is computed using Eq. (1),

Schematic representation of the electrical setup employed to evaluate the FastIC+ TDC

Figure 4. Schematic representation of the electrical setup employed to evaluate the FastIC+ TDC. (a) Code density test; (b) Single-shot precision of the TDC.

MaterialGeometry (mm3^{3})Decay time (ns)ILYa^{a} (ph. · keV−1^{-1})Manufacturer
LYSO:Ce,Cab^{b}2×2×3; 2.8×2.8×20; 3×3×203530.0Taiwan Applied Crystal
LYSO:Ce,Cac^{c}2×2×3; 3.12×3.12×204045.0Crystal Photonics Inc.
LYSO:Ced^{d}3×3×203633.2Saint-Gobain
Fast-LGSO:Cee^{e}2×2×33390% of a NaI:Tl ref.Oxide Corporation

Table 1. Crystals employed in CTR measurements.

a^{a} Intrinsic Light Yield.

b^{b} From [13]

c,d^{c,d} From [19]

e^{e} Oxide Corporation datasheet: https://www.opt-oxide.com/en/products/fast-lgso/

NameArea (mm2^{2})SPAD size (μ\mum2^{2})VbreakaV_{\mathrm{break}}^{a} (V)
FBK NUV-HD-MT LF M04×4; 1×1; Single SPAD40×4032.5
FBK NUV-HD-MT LFv2 M03×350×5032.5
HPK S13360-3050VE3×350×5051.8

Table 2. SiPMs used in SPTR and CTR measurements.

a^{a} Breakdown voltage.

Nbins=2⋅foscPLLfclkref⋅Ncells=1024 ,(1)N_{\mathrm{bins}} = 2 \cdot\frac{f_{\mathrm{oscPLL}}}{f_{\mathrm{clkref}}} \cdot N_{\mathrm{cells}} = 1024 \,, \tag*{(1)}

where foscPLLf_{\mathrm{oscPLL}} is the PLL oscillation frequency (1.28 GHz), fclkreff_{\mathrm{clkref}} is the reference clock frequency (40 MHz), NcellsN_{\mathrm{cells}} is the number of delay cells in the VCO (16 in this case), and the factor of 2 accounts for the two transitions (rising and falling edges) required for the VCO to complete one full oscillation cycle.

The procedure to compute the DNL using a density code test analysis, as described in [34], is as follows. The TDC captures one million pulses generated by an external Pulse Pattern Generator (Agilent 81110A), with random arrival times following a uniform distribution relative to the TDC’s reference clock. The number of detected pulses (counts) is categorized into time bins to construct a histogram. Ideally, this distribution should be flat, as the input signal follows a uniform distribution and the TDC is expected to assign events equally across all time bins. However, non-linearities within the TDC distort this ideal outcome, resulting in a non-uniform distribution. The deviation of the actual bin counts from the ideal uniform value is used to determine the DNL.

In particular, the normalized number of counts per time bin, i.e., the number of counts in a given time bin (countsbini\mathrm{counts}_{\mathrm{bin}i}) divided by the average number of counts per time bin (countsaverage\mathrm{counts}_{\mathrm{average}}), corresponds to the actual width of time bin ii. Therefore, the DNL of time bin ii can be computed using Eq. (2),

DNLbini [ps]=(countsbinicountsaverage−1)⋅TDCLSB ,(2)DNL_{\mathrm{bin}i}\,[\mathrm{ps}] = \left(\frac{\mathrm{counts}_{\mathrm{bin}i}}{\mathrm{counts}_{\mathrm{average}}} - 1\right) \cdot TDC_{\mathrm{LSB}} \,, \tag*{(2)}

where the ideal time bin size is 1, and the TDC’s Least Significant Bit (LSB) corresponds to 24.4 ps. Finally, the Integral Non-Linearity (INL), which represents the cumulative deviation of the actual time bin widths (i.e., the integrated DNL), can be computed using Eq. (3),

INLbin_i [ps]=∑j=0i−1DNLbin_j.(3)INL_{bin\_i}\ [\mathrm{ps}] = \sum_{j=0}^{i-1} DNL_{bin\_j}. \tag*{(3)}

Figure 4b illustrates the measurement setup used to compute the single-shot precision of the TDC, which accounts for intrinsic jitter, linearity deviations, and quantization error. In this setup, it is important to highlight that the TDC input signal is sent to two different ASICs, each with its own independent PLL, to obtain two completely independent measurements. An external trigger signal from the same Pulse Pattern Generator as before is sent to two channels from different ASICs. The ToA is computed as the time difference between the timestamps obtained from each ASIC. Consequently, the standard deviation of the jitter—representing the single-shot precision of a single TDC—is obtained using Eq. (4),

jitterTDC [ps]=std(TimeASIC_0−TimeASIC_1)2.(4)jitter_{TDC}\ [\mathrm{ps}] = \frac{std\left(Time_{ASIC\_0}-Time_{ASIC\_1}\right)}{\sqrt{2}}. \tag*{(4)}

Single Photon Time Resolution

Figure 5 provides a schematic representation of the experimental setups employed for the SPTR measurements. The light source used was a 405 nm pico-second laser (PiLas) with a pulse width of 28 ps FWHM. The laser operated at a repetition rate of 100 kHz and was tuned at a intensity level of 50 %. Initially, the laser light passes through a collimator. Between the collimator and the SiPM, two different light attenuators were located to reduce the light intensity close to the single-photon level. First, a fixed optical attenuator decreased the light intensity down to a few photons. The second one was a liquid crystal attenuator (LCC1620/M from Thorlabs) in which the attenuation level can be modified by adjusting its operating voltage. Finally, the light reached the SiPM, which was connected to one channel of the FastIC+ ASIC module. The trigger signal generated by the laser driver was fed into the FastIC+ external trigger input.

Schematic representation of the experimental setup for SPTR measurements

Figure 5. Schematic representation of the experimental setup for SPTR measurements. (a) Digital mode; (b) Analog mode.

To evaluate the contribution of FastIC+ TDC to the SPTR we performed the measurements both in analog (Figure 5a) and digital (Figure 5b) mode (see Sec. II-F). In the analog mode, the FastIC+ binary signals (containing the information of the ToA and the energy width) along with the trigger signal provided by the laser driver, were sent to independent channels of an Agilent MS09254A oscilloscope (2.5 GHz, 20 GSa/s, 10 bits ADC). In the digital mode, the binary and trigger signals were processed and digitized by FastIC+. In both cases, the acquisition was triggered by the laser trigger signal.

For both measurements, the energy width was used for event selection. Therefore, a ToA distribution was constructed using the ToA values of these selected events. This distribution was fitted with a Gaussian function, and the time resolution was determined as the FWHM of the fit.

Coincidence Time Resolution

Figure 6 shows a schematic representation of the experimental setup used for the CTR measurements. A 22Na^{22}\mathrm{Na} source was positioned and aligned between the two scintillator detectors. Moreover, each SiPM from each FastIC+ module was connected to an ASIC channel.

Schematic representation of the experimental setup for CTR measurements

Figure 6. Schematic representation of the experimental setup for CTR measurements. (a) Digital mode; (b) Analog mode.

In this case, time and trigger comparators were used to acquire the events. The threshold for the time comparator was set to the single-SPAD signal level to achieve optimal timing performance. Conversely, the trigger comparator was set to its maximum value to reduce the number of undesired events caused by dark counts or low-energy interactions. The acquisition procedure was similar to those described in Sec. III-D, for both analog and digital signals. For the analog readout, a 25 ns coincidence time window was set on the oscilloscope between the two energy signals to ensure that both gamma photons originated from the same positron-electron annihilation, thus filtering out unwanted events. For digital measurements, an offline ToA condition was applied during analysis to classify events coming from different boards.

Performance evaluation

FastIC+ general performance

The FastIC+ has been developed using a CMOS 65 nm technology node, with a die size of approximately 2.9×2.92.9 \times2.9 mm2^2 and housed in a standard QFN88 package. The TDC consumes 3.3 mW per channel, contributing to an overall power consumption of 12.5 mW per channel for the FastIC+. This represents an increase of only ≈0.5\approx0.5 mW/channel compared to FastIC. This small difference arises from the fact that the SLVS drivers per channel, which were previously used to send binary signals (time and energy) outside the chip, are no longer required in the FastIC+, as all data is transmitted through a single SLVS driver.

Additionally, a low-power mode has been implemented by reducing the number of buffers driving the VCO phases to the TDC, thereby lowering switching activity. In this mode, the

FastIC+ general performance

TDC provides a time bin of approximately 49 ps (instead of 24.4 ps) while consuming just 1.8 mW per channel.

Figure 7 shows the DNL and INL results from a code density test. The standard deviation of the DNL is below 6 ps without calibration and improves to below 4 ps with time bin calibration, achieved by adjusting the drive strength of the PLL output buffers supplying the VCO phases. In both cases, the DNL remains lower than the LSB. The INL stays within ±22\pm22 ps. Finally, the single-shot precision is 13.3 ps (standard deviation) and 31.3 ps (FWHM).

DNL and INL results after performing the code density test of 1 million pulses

Figure 7. DNL and INL results after performing the code density test of 1 million pulses.

Digital vs analog performance for SPTR

Figure 8 and Figure 9 compares the results obtained in digital and analog mode, respectively. To obtain these plots we used an FBK NUV-HD-MT LFv2.0 SiPM (3×33 \times3 mm2^2, see Table II) and operated it at an overvoltage of 16 V. The lower plots (Figure 8b and Figure 9b) show the Energy distributions. We selected 1-phe events (denoted with a red area in the Energy plot) to build the ToA distributions of Figure 8a and Figure 9a to extract the SPTR. Both measurements yielded a similar SPTR value, with (98 ±\pm 1) ps and (99 ±\pm 1) ps for the digital and analog measurements, respectively. Therefore, the TDC does not degrade the time resolution compared to the analog results, indicating that the contribution of the jitter of the TDC is not significant.

Figure 8. SPTR measurement with FastIC+ digital mode for FBK NUV-HD-MT LFv2 M0 3×33 \times3 mm2^{2} at 16V of overvoltage.

Figure 9. SPTR measurement with FastIC+ analog mode for FBK NUV-HD-MT LFv2 M0 3×33 \times3 mm2^{2} at 16V of overvoltage.

Furthermore, Figure 8b shows that the energy readout has sufficient SNR to identify the different photons. Note that the distribution in Figure 8b is not Poissonian. This is because it combines acquisitions at different light intensities, which were taken to show FastIC+ capability to resolve events higher than 10 phe to evaluate the system’s jitter of FastIC+ TDC (see Sec. IV-C).

System's jitter FastIC+ TDC

Figure 10 shows the time resolution of the FastIC+ TDC for different numbers of fired SPADs. To build this plot, we constructed the ToA distributions and computed the time resolution (ΔT\Delta T) for each photoelectron peak (i.e. the different peaks observed in Figure 8b), as described in Sec. IV-B. As a first-order approximation, the time resolution for Np.e.N_{\mathrm{p.e.}} can be estimated by fitting the data to Eq. (5), as detailed in [6, 35],

Figure 10. Time resolution for different number of fired SPADs obtained in digital mode for different SiPMs. All measurements were taken at 16V of overvoltage. Data was fitted to Eq. (5).

ΔT(Np.e.)=22ln⁡(2)⋅σPJ2Np.e.+σsyst.2(5)\Delta T(N_{p.e.}) = 2\sqrt{2\ln(2)} \cdot\sqrt{\frac{\sigma_{PJ}^{2}}{N_{p.e.}}+\sigma_{\mathrm{syst.}}^{2}} \tag*{(5)}

where σPJ\sigma_{PJ} accounts for the single-photon electronic jitter and the intrinsic SPTR of the sensor. This parameter directly depends on the number of photons. The second parameter, σsyst.\sigma_{\mathrm{syst.}}, represents the constant jitter contribution of the system, which is independent of the number of photons. This term comprises the laser pulse shape, the laser trigger jitter, the electronic jitter at the maximum slew rate achievable by the electronics and the acquisition jitter.

The time resolution improves as the number of triggered SPADs increases. This improvement is mainly due to two factors. First, when multiple SPADs are triggered, the timing jitter associated with each individual detection (i.e., the intrinsic SPTR of the sensor) is averaged, resulting in a more accurate estimate of the photon’s arrival time. Second, the combined signal from multiple detections leads to a higher slew rate in the analog output, thereby reducing the electronic jitter contribution to the overall jitter.

The observed timing differences between sensors can be attributed to the sensors themselves. Since all measurements were performed under identical conditions (using the same laser setup and the same TDC), the time resolution differences can be attributed to the SiPM. In particular, smaller sensors (e.g., 1×11 \times1 mm2^{2} SiPMs) exhibit a lower jitter compared to larger ones. This is due to reduced electronic noise and optical noise associated with a smaller sensor area.

Digital vs analog performance for CTR

Figs. 11 and 12 compare the FastIC+ CTR measured with the digital and analog mode, respectively. Those plots were obtained using FBK NUV-HD-MT LFv2 3×33 \times3 mm2^{2} SiPMs

coupled to LYSO:Ce,Ca crystals of 2×2×32 \times2 \times3 mm3^{3} from TAC. Figs. 11a, 11b, 12a and 12b show the 22{}^{22}Na energy spectra recorded with the two detectors in both modes. As seen in the figures, the photopeaks are centered at similar energies when comparing analog and digital signal for both detectors. Figs 11c and 12c show the difference in the arrival time of the two detectors for 511 keV events (denoted as the red regions in the energy plots). Both measurements yielded a similar CTR value of (91 ± 1) ps and (87 ± 1) ps in digital and analog mode, respectively, achieving a sub-100 ps time resolution. As observed in Sec. IV-B for low-light-level signals, the TDC contribution to the time resolution is minimal, and only a small difference is observed when comparing with the analog output for large signals.

Figure 11. CTR measurement with FastIC+ digital mode for FBK NUV-HD-MT LFv2 M0 3×33 \times3 mm2^{2} SiPMs coupled to 2×2×32 \times2 \times3 mm3^{3} LYSO:Ce,Ca crystals from TAC at 10 V of overvoltage.

Figure 12. CTR measurement with FastIC+ analog mode for FBK NUV-HD-MT LFv2 M0 3×33 \times3 mm2^{2} SiPMs coupled to 2×2×32 \times2 \times3 mm3^{3} LYSO:Ce,Ca crystals from TAC at 10 V of overvoltage.

CTR performance using different scintillators

Table III presents the best CTR results achieved for a different combination of scintillators and SiPMs using the FastIC+ TDC. The lowest timing resolution achieved was (85 ± 1) ps for Fast-LGSO crystals from Oxide, coupled to FBK NUV-HD-MT LFv2 M0 SiPMs with a 3×33 \times3 mm2^{2} of active area, as seen in Figure 13. We also tested 2×3×32 \times3 \times3 mm3^{3} crystals from different manufacturers using the same SiPMs, obtaining sub-100 ps results for all of them. Additionally, we evaluated the same LYSO:Ce,Ca crystals from TAC with different SiPMs to study the contribution of the photosensor. For SiPMs based on the same technology but with different active areas and different SPAD size, i.e., FBK 3×33 \times3 mm2^{2} and 50 μ\mum and FBK 4×44 \times4 mm2^{2} and 40 μ\mum, we observed a 8 ps degradation in the timing resolution. However, the most significant difference was found when comparing HPK and FBK sensors using the same TAC crystal, with a ∼ 20 ps deterioration when using the HPK.

Figure 13. CTR measurement with FastIC+ TDC for FBK NUV-HD-MT LFv2 M0 3×3 mm² SiPMs coupled to Fast-LGSO 2×2×3 mm³ crystals from Oxide at 10 V of overvoltage.

CrystalSize (mm³)SiPM (mm²)CTR (± 1 ps FWHM)
Fast-LGSO (Oxide)2×2×3FBK NUV-HD-MT LFv2 M0 3×385
LYSO:Ce,Ca (CPI)2×2×3FBK NUV-HD-MT LFv2 M0 3×392
LYSO:Ce,Ca (TAC)2×2×3FBK NUV-HD-MT LFv2 M0 3×391
LYSO:Ce,Ca (TAC)2×2×3FBK NUV-HD-MT LF M0 4×499
LYSO:Ce,Ca (TAC)2×2×3HPK S13360-3050VE 3×3110
LYSO:Ce,Ca (TAC)2.8×2.8×20FBK NUV-HD-MT LF M0 4×4130
LYSO:Ce,Ca (TAC)3×3×20FBK NUV-HD-MT LF M0 4×4137
LYSO:Ce:Ca (Saint-Gobain)3×3×20FBK NUV-HD-MT LF M0 4×4134
LYSO:Ce:Ca (CPI)3.12×3.12×20FBK NUV-HD-MT LF M0 4×4135

Table III. Best CTR results performed with FastIC+ TDC.

We also evaluated the CTR performance of the FastIC+ TDC for longer crystals. The best result for 20 mm crystals was (130 ± 1) ps, achieved with a LYSO:Ce,Ca crystal from TAC measuring 2.8×2.8×202.8 \times2.8 \times20 mm3^{3}, coupled to an FBK NUV-HD-MT LF 4×44 \times4 mm2^{2} SiPM, as shown in Figure 14.

Figure 14. CTR measurement with FastIC+ TDC for FBK NUV-HD-MT LF M0 4×4 mm² SiPMs coupled to LYSO:Ce,Ca 2.8×2.8×20 mm³ crystals from TAC at 10 V of overvoltage. The peak around ∼ 200 ns observed in (a) is due to signal saturation at the configured energy gain in this readout channel.

Large crystals exhibited similar CTR values ranging from 130 to 140 ps. A degradation of ∼ 7 ps is observed when increasing the area of the crystal from 2.8×2.82.8 \times2.8 mm2^{2} to 3×33 \times3 mm2^{2} for the TAC manufacturer. For Saint-Gobain and CPI crystals, both measurements provided a similar CTR value of (134 ± 1) ps and (135 ± 1) ps, respectively, despite the larger area of the CPI crystal. All 20 mm crystals were coupled to the same 4×44 \times4 mm2^{2} SiPMs, which have a larger active area than the crystals to minimize light losses due to eventual misalignments.

Discussion

The new FastIC+ ASIC has demonstrated excellent timing resolution while introducing an on-chip digitization with only a 5% increase in power consumption compared to its predecessor, FastIC. This digitization, achieved through the internal TDC of the FastIC+ ASIC, does not introduce any significant jitter in the time resolution measurements, thereby simplifying data acquisition at the system level. Experimental results confirm that the timing resolution obtained from the analog output is comparable with the on-chip TDC results, both in SPTR and CTR measurements. Furthermore, the minimal contribution of the TDC can also be observed when analyzing the system’s jitter. As shown in Figure 10, the differences in time resolution clearly depend on the size of the SiPM, demonstrating that the measurement is dominated by SiPM-related jitter rather than acquisition jitter, which is independent of sensor size.

The best CTR obtained using the FastIC+ TDC was (85 ± 1) ps for Fast-LGSO crystals from Oxide measuring 2×2×32 \times2 \times3 mm3^{3}, coupled to FBK NUV-HD-MT LFv2 M0 SiPMs with a 3×33 \times3 mm2^{2} sensitive area. The performance obtained with this crystal was few ps better compared to the other crystals of the same size. When comparing the performance of crystals from different manufacturers, we observe an improvement of 6 ps and 7 ps relative to TAC and CPI, respectively. This improvement might be related to the faster decay time of Fast-LGSO crystals, although there are many factor affecting the time resolution such as scintillator light yield, efficiency of the crystal-SiPM coupling, etc.

We also studied the impact of the SiPM area on CTR measurements. When the same LYSO:Ce,Ca crystal of size 2×2×32 \times2 \times3 mm3^{3} from TAC is coupled to a SiPM with a larger sensitive area and smaller pixel size, an 8 ps degradation is observed. This deterioration is attributed to the increase in SPTR, as the 4×44 \times4 mm2^{2} SiPM presents a higher SPTR (approximately 60 ps worse) compared to the 3×3 mm² SiPM, as shown in Figure 10.

For 20 mm crystals, all measurements provided similar CTR values for different manufacturers and crystal sizes. The 3.12×3.12×20 mm³ crystals from CPI had been measured with FastIC using FBK NUV-HD LFv2 M0 3.12×3.12 mm² SiPMs, reaching a CTR value of 127 ps with analog signals [19]. We observed a deterioration of ∼ 8 ps when comparing with FBK NUV-HD-MT LF M0 4×4 mm², which can be attributed to a larger sensor area and smaller pixel size.

Table IV presents a comparison of various ASICs and discrete electronics studied in the literature for ToF-PET applications. To the best of the author’s knowledge, FastIC+ is the first ASIC with on-chip digitization, achieved through an internal TDC, to attain a sub-100 ps CTR for short crystals and sub-140 ps for 20 mm-long crystals.

ASICPower (mW/ch)Crystal [mm3^{3}]SiPM [mm2^{2}]CTR (ps FWHM)DAQa^{a}Reference
NINO27LSO:Ce,0.4%Ca [2 ×\times 2 ×\times 3]FBK NUV-HD [3 ×\times 3]73Osc.[36]
Radioroc3.3b^{b}LYSO:Ce,Ca [2 ×\times 2 ×\times 3]Broadcom NUV-MT [4 ×\times 4]83Osc.[18, 37]
Radioroc3.3b^{b}LYSO:Ce,Ca [3 ×\times 3 ×\times 20]Broadcom NUV-MT [4 ×\times 4]127Osc.[18, 37]
TOFPET2c^{c}8LYSO:Ce,Ca [3 ×\times 3 ×\times 19]Broadcom NUV-MT [3.8 ×\times 3.8]157TDC[13]
HF readout143LYSO:Ce,Ca [2 ×\times 2 ×\times 3]Broadcom NUV-MT [3.8 ×\times 3.8]56Osc. + IC[13]
HF readout143LYSO:Ce,Ca [3 ×\times 3 ×\times 19]Broadcom NUV-MT [3.8 ×\times 3.8]95Osc. + IC[13]
FastIC12LSO:Ce:0.2%:Ca [2 ×\times 2 ×\times 3]FBK NUV-HD LFv2 M0 [3.12 ×\times 3.2]76Osc.[19]
FastIC12LYSO:Ce,0.2%:Ca [3.13 ×\times 3.13 ×\times 20]FBK NUV-HD LFv2 M0 [3.12 ×\times 3.2]127Osc.[19]
FastIC+12.5Fast-LGSO [2 ×\times 2 ×\times 3]FBK NUV-HD-MT LFv2 M0 [3 ×\times 3]85TDCThis work
FastIC+12.5LYSO:Ce,Ca [2.8 ×\times 2.8 ×\times 20]FBK NUV-HD-MT LF M0 [4 ×\times 4]130TDCThis work

Table IV. Comparison of state-of-the-art electronics employed in TOF-PET applications.

Although FastIC+ has demonstrated excellent timing performance, its CTR results are still higher than those compared to high-power discrete electronics, indicating room for improvement. As shown in Table IV, the best results have been achieved with HF readouts. These systems provide lower electronic noise (σEN\sigma_{EN}) compared to FastIC+, therefore reducing the jitter contribution, since it scales with σEN/SR\sigma_{EN}/SR, where SRSR denotes the slew rate of the signal. Additionally, HF systems provide analog signals and they rely on fast oscilloscopes

combined with ideal comparators to extract the ToA using an offline leading-edge comparator algorithm applied to the analog data. In contrast, FastIC+ employs an integrated on-chip comparator which introduces additional jitter to the measurement, degrading the timing measurement. However, the power consumption of HF readouts makes them unsuitable for large-scale systems where hundreds of detectors are involved.

In terms of power consumption, FastIC+ has a higher power consumption than the TOFPET2c ASIC. It is important to note that the FastIC+ is a multipurpose ASIC designed to process not only SiPM signals but also those from MCPs or PMTs. These two types of sensors require a larger bandwidth for energy readout due to the short duration of their pulses. Therefore, the power consumption of the energy readout could be reduced if the FastIC energy readout were optimized specifically for SiPMs in PET applications.

The performance of FastIC with other types of scintillators, such as Cherenkov radiators and BGO crystals, has already been demonstrated in [19, 24, 38]. For instance, for BGO measurements it has already been shown that time-walk correction improves CTR [24]. A well-known method involves measuring the rise time of the input signal to better identify the events contributing to the time of arrival. FastIC+ can provide a second timestamp to measure the rise time using the trigger signal. Observe that the trigger signal is generated

as a fast-OR, meaning that only the timestamp of the fastest channel in the ASIC is available. Nevertheless, measuring the fastest channel may still provide sufficient information for CTR measurements with pixelated crystals.

FastIC+ capabilities for ToF-PET detectors can also be extended to dual-ended readout systems. These systems account for time-based DOI corrections and have already been implemented using high power consumption electronics, improving timing resolution not only for BGO crystals but also for LYSO [39]. However, this approach comes at the cost of doubling the number of readout channels and increasing the mechanical complexity of the system.

Conclusions

FastIC+ is a versatile read-out ASIC with internal digitization that exhibits low power consumption of 12.5 mW per channel, while achieving remarkable timing resolution. This ASIC is capable of measuring both positive and negative polarity sensors, providing digitized outputs for photon arrival time and energy. The internal TDC does not degrade the time resolution, as the measurements performed in digital and analog modes showed similar results.

A CTR of (85 ±\pm 1) ps was achieved with FastIC+ when coupling Fast-LGSO crystals measuring 2×2×32 \times2 \times3 mm3^{3} with FBK NUV-HD-MT-LFv2 M0 3×33 \times3 mm2^{2} SiPMs. For larger crystals, a CTR of (130 ±\pm 1) ps was obtained using LYSO:Ce,Ca crystals measuring 2.8×2.8×202.8 \times2.8 \times20 mm3^{3} with FBK NUV-HD-MT LF M0 4×44 \times4 mm2^{2} SiPMs. Furthermore, different 20 mm long crystal sizes from various manufacturers were tested, with all of them achieving sub-140 ps CTR results when coupled to the same 4×44 \times4 SiPMs.

Future avenues for FastIC+ include implementing a double-ended readout system when applied to long crystals, allowing for additional timing corrections [39]. Using these systems, it becomes possible to better identify the interaction point of the gamma in the crystal and apply depth-of-interaction corrections. Additionally, the use of BGO crystals for CTR measurements can also be explored. The combination of time and trigger comparators can help to determine the signal’s rise time and apply time walk corrections. Furthermore, FastIC+ is capable to digitize up to 8 different channels, making it versatile for working with arrays of SiPMs and enabling the simultaneous detection of multiple signals.

Thanks to the integration of the TDC in the chip, FastIC+ does not need external mechanisms like an FPGA to digitize the signals. This turns FastIC+ into a suitable chip to be employed in a complex system like a PET scanner.

Acknowledgments

This work was supported by HORIZON EUROPE European Research Council, Grant/Award Number: 101099896. This study was also supported by MICIIN with funding from the European Union NextGenerationEU fund (PRTR-C17.I1/AEI/10.13039/501100011033/Unión Europea NextGenerationEU/PRTR) and by the Generalitat de Catalunya.

References

  1. [1]P. Lecoq, “Pushing the limits in time-of-flight pet imaging,” IEEE Transactions on Radiation and Plasma Medical Sciences, vol. 1, no. 6, pp. 473–485, 2017.DOI
  2. [2]S. Vandenberghe, “[i180] recent developments in time-of-flight pet,” Physica Medica, vol. 52, p. 69, 2018, abstracts from the 2nd European Congress of Medical Physics. [Online]. Available: https://www.sciencedirect.com/science/article/pii/S1120179718307348DOI
  3. [3]D. R. Schaar, G. Schramm, J. Nuyts, and S. Surti, “Time of flight in perspective: Instrumental and computational aspects of time resolution in positron emission tomography,” IEEE Transactions on Radiation and Plasma Medical Sciences, vol. 5, no. 5, pp. 598–618, 2021.
  4. [4]M. Conti, “State of the art and challenges of time-of-flight pet,” Physica Medica, vol. 25, no. 1, pp. 1–11, 2009. [Online]. Available: https://www.sciencedirect.com/science/article/pii/S1120179708000975DOI
  5. [5]“Siemens Healthineers Debuts Biograph Vision.X PET/CT Scanner - Siemens Healthineers USA.” [Online]. Available: https://www.siemens-healthineers.com/en-us/press-room/press-releases/biograph-vision-x-pet-ct-scanner-debut
  6. [6]S. Gundacker, R. Martinez Turtos, N. Kratochwil, R. H. Pots, M. Paganoni, P. Lecoq, and E. Auffray, “Experimental time resolution limits of modern SiPMs and TOF-PET detectors exploring different scintillators and Cherenkov emission,” Physics in Medicine & Biology, vol. 65, no. 2, p. 025001, 2020. [Online]. Available: https://iopscience.iop.org/article/10.1088/1361-6560/ab63b4DOI
  7. [7]A. Gonzalez-Montoro, J. Barbera, D. Sanchez, A. Mondejar, M. Freire, K. Diaz, A. Lucero, S. Jimenez-Serrano, J. Alamo, C. Morera-Ballester, J. Barrio, N. Cucarella, V. Ilisie, L. Moliner, C. Valladares, A. J. Gonzalez, J. Prior, and J. M. Benlloch, “A new brain dedicated PET scanner with 4D detector information,” Bio-Algorithms and Med-Systems, vol. 18, no. 1, pp. 107–119, 12 2022. [Online]. Available: http://archive.sciendo.com/BIOAL/biol.2020.18.issue-1/biol-2022-0083/biol-2022-0083.pdf
  8. [8]S. Gundacker, G. Borghi, S. R. Cherry, A. Gola, D. Lee, S. Merzi, M. Penna, V. Schulz, and S. I. Kwon, “On timing-optimized sips for cherenkov detection to boost low cost time-of-flight pet,” Physics in Medicine & Biology, vol. 68, no. 16, p. 165016, aug 2023. [Online]. Available: https://dx.doi.org/10.1088/1361-6560/ace8ee
  9. [9]P. Lecoq, “Pushing the limits in time-of-flight pet imaging,” IEEE Transactions on Radiation and Plasma Medical Sciences, vol. 1, no. 6, pp. 473–485, 2017. [Online]. Available: https://dx.doi.org/10.1109/TRPMS.2017.2756674
  10. [10]S. Gundacker, E. Auffray, P. Jarron, T. Meyer, and P. Lecoq, “On the comparison of analog and digital sipm readout in terms of expected timing performance,” Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers, Detectors and Associated Equipment, vol. 787, pp. 6–11, 2015, new Developments in Photodetection NDIP14. [Online]. Available: https://www.sciencedirect.com/science/article/pii/S016890021401167XDOI
  11. [11]S. Seifert, H. T. van Dam, and D. R. Schaart, “The lower bound on the timing resolution of scintillation detectors,” Physics in Medicine & Biology, vol. 57, no. 7, p. 1797, mar 2012. [Online]. Available: https://dx.doi.org/10.1088/0031-9155/57/7/1797
  12. [12]S. Gundacker, R. M. Turtos, E. Auffray, M. Paganoni, and P. Lecoq, “High-frequency SiPM readout advances measured coincidence time resolution limits in TOF-PET,” Physics in Medicine & Biology, vol. 64, no. 5, p. 055012, 2019. [Online]. Available: https://iopscience.iop.org/article/10.1088/1361-6560/aafd52DOI
  13. [13]V. Nadig, K. Herweg, M. M. C. Chou, J. W. C. Lin, E. Chin, C.-A. Li, V. Schulz, and S. Gundacker, “Timing advances of commercial divalent-ion co-doped lyso:ce and sipms in sub-100 ps time-of-flight positron emission tomography,” Physics in Medicine & Biology, vol. 68, no. 7, p. 075002, mar 2023. [Online]. Available: https://dx.doi.org/10.1088/1361-6560/acbde4
  14. [14]J. W. Cates and W.-S. Choong, “Low power implementation of high frequency SiPM readout for Cherenkov and scintillation detectors in TOF-PET,” Physics in Medicine & Biology, vol. 67, no. 19, p. 195009, 10 2022. [Online]. Available: https://iopscience.iop.org/article/10.1088/1361-6560/ac8963DOI
  15. [15]A. Samnukh, S. Gómez, A. Comerma, J. Mauricio, R. Manera, A. Sanuy, D. Guberman, R. Catala, A. Espinya, M. Orta, O. de la Torre, D. Gascon, and on behalf of the HERD collaboration, “Low-power sipm readout beta asic for space applications,” Nuclear Science and Techniques, vol. 35, no. 3, p. 59, 2024. [Online]. Available: https://doi.org/10.1007/s41365-024-01919-z
  16. [16]J. Mauricio, L. Freixas, A. Sanuy, S. Gómez, R. Manera, J. Marín, J. M. Pérez, E. Picatoste, P. Rato, D. Sánchez, A. Sanmukh, O. Vela, and D. Gascon, “Matrix16: A 16-channel low-power tof asic with 8 ps time resolution,” Electronics, vol. 10, no. 15, 2021. [Online]. Available: https://www.mdpi.com/2079-9292/10/15/1816
  17. [17]F. Anghinolfi, P. Aspell, R. A. Ferrando-May, E. Usenko, H. Wenninger, M. Williams, and A. Zichichi, “NINO: an ultra-fast and low-power front-end amplifier/discriminator ASIC designed for the multigap resistive plate chamber,” Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers, Detectors and Associated Equipment, vol. 533, no. 1-2, pp. 183–187, 11 2004. [Online]. Available: https://linkinghub.elsevier.com/retrieve/pii/S0168900204014299DOI
  18. [18]T. Sanchez, S. Ahmad, J.-B. Cizerl, C. de la Taille, M. Morenas, V. Nadig, F. Perez, V. Schulz, S. Gundacker, and J. Fleury, “Study experimental time resolution limits of recent ASICs at Weeroc with different SiPMs and scintillators,” Journal of Instrumentation, vol. 18, no. 10, p. P10005, 10 2023. [Online]. Available: https://iopscience.iop.org/article/10.1088/1748-0221/18/10/P10005
  19. [19]A. Mariscal-Castilla, S. Gómez, R. Manera, J. M. Fernández-Tellado, J. Mauricio, N. Kratochwil, J. Alozy, M. Piller, S. Portero, A. Sanuy, D. Guberman, J. J. Silva, E. Auffray, R. Ballabriga, G. Ariño-Estrada, M. Campbell, and D. Gascon, “Toward sub-100 ps tof-pet systems employing the fastic asic with analog sipms,” IEEE Transactions on Radiation and Plasma Medical Sciences, vol. 8, no. 7, pp. 718–733, 2024. [Online]. Available: https://doi.org/10.1109/TRPMS.2024.3414578
  20. [20]S. Gómez, J. Alozy, M. Campbell, J. M. Fernandez-Tellado, R. Manera, J. Mauricio, C. Pujol, D. Sanchez, A. Sanmukh, A. Sanuy, R. Ballabriga, and D. Gascon, “Fastic: a fast integrated circuit for the readout of high performance detectors,” Journal of Instrumentation, vol. 17, May 2022. [Online]. Available: 10.1088/1748-0221/17/05/C05027
  21. [21]S. Gómez, J. M. Fernández-Tellado Arribas, J. Alozy, M. Campbell, R. Manera, J. Mauricio, A. Mariscal-Castilla, C. Pujol, D. Sanchez, A. Sanmukh, A. Sanuy, R. Ballabriga, and D. Gascon, “Fastic: A highly configurable asic for fast timing applications,” in 2021 IEEE Nuclear Science Symposium and Medical Imaging Conference (NSS/MIC), October 2021, pp. 1–4. [Online]. Available: https://doi.org/10.1109/NSS/MIC44867.2021.9875546
  22. [22]S. Ahmad, J. Fleury, J.-B. Cizerl, C. de la Taille, N. Seguin-Moreau, S. Gundacker, and E. Auffray-Hillemanns, “Petiroc2A: Characterization and Experimental Results,” in 2018 IEEE Nuclear Science Symposium and Medical Imaging Conference Proceedings (NSS/MIC). IEEE, 11 2018, pp. 1–4. [Online]. Available: https://ieeexplore.ieee.org/document/8824464
  23. [23]V. Nadig, B. Weissler, H. Radermacher, V. Schulz, and D. Schug, “Investigation of the Power Consumption of the PETsys TOFPET2 ASIC,” IEEE Transactions on Radiation and Plasma Medical Sciences, vol. 4, no. 3, pp. 378–388, 5 2020. [Online]. Available: https://ieeexplore.ieee.org/document/8908729/DOI
  24. [24]M. Piller, “Performance evaluation of the fastic readout asic with emphasis on cherenkov emission in tof-pet,” PMB, 4 2024. [Online]. Available: https://iopscience.iop.org/article/10.1088/1361-6560/ad42feDOI
  25. [25]S. Gómez, “Fastic: a fast integrated circuit for the readout of high performance detectors,” JINST, vol. 17, p. C05027, 5 2022. [Online]. Available: https://iopscience.iop.org/article/10.1088/1748-0221/17/05/C05027
  26. [26]D. Sánchez, S. Gómez, J. Mauricio, L. Freixas, A. Sanuy, G. Guixé, A. López, R. Manera, J. Marín, J. M. Pérez, E. Picatoste, C. Pujol, A. Sanmukh, P. Rato, D. Vela, and D. Gascon, “Hrflextot: A high dynamic range asic for time-of-flight positron emission tomography,” IEEE Transactions on Radiation and Plasma Medical Sciences, vol. 6, no. 1, pp. 51–67, 2022. [Online]. Available: https://doi.org/10.1109/TRPMS.2021.3066426
  27. [27]S. Gómez, D. Sánchez, J. Mauricio, E. Picatoste, A. Sanuy, A. Sanmukh, M. Ribó, and D. Gascón, “Multiple use sipm integrated circuit (music) for large area and high performance sensors,” Electronics, vol. 10, no. 8, 2021. [Online]. Available: https://www.mdpi.com/2079-9292/10/8/961
  28. [28]H. Traff, “Novel approach to high speed cmos current comparators,” Electronics Letters, vol. 28, pp. 310–312, 1992. [Online]. Available: https://api.semanticscholar.org/CorpusID:110982635DOI
  29. [29]R. del Rio-Fernandez, G. Linan-Cembrano, R. Dominguez-Castro, and A. Rodriguez-Vazquez, “A mismatch-insensitive high-accuracy high-speed continuous-time current comparator in low voltage cmos,” in 1997 2nd IEEE-CAS Region 8 Workshop on Analog and Mixed IC Design, 1997, pp. 111–116. [Online]. Available: https://doi.org/10.1109/AMICD.1997.637202
  30. [30]N. Kratochwil, S. Gundacker, and E. Auffray, “A roadmap for sole Cherenkov radiators with SiPMs in TOF-PET,” Physics in Medicine & Biology, vol. 66, no. 19, p. 195001, 10 2021. [Online]. Available: https://iopscience.iop.org/article/10.1088/1361-6560/ac212aDOI
  31. [31]G. De Geronimo, P. O’Connor, and A. Kandasamy, “Analog cmos peak detect and hold circuits. part 2. the two-phase offset-free and derandomizing configuration,” Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers, Detectors and Associated Equipment, vol. 484, no. 1, pp. 544–556, 2002. [Online]. Available: https://www.sciencedirect.com/science/article/pii/S0168900201020605DOI
  32. [32]J. G. M. O. Geronimo, “Precise delay generation using coupled oscillators,” IEEE JSSC, vol. 28, pp. 1273–1282, 1993. [Online]. Available: https://doi.org/10.1109/4.262000
  33. [33]“Aurora 64b/66b.” [Online]. Available: https://www.xilinx.com/products/intellectual-property/aurora64b66b.html
  34. [34]“Density code test.” [Online]. Available: https://www.allaboutcircuits.com/technical-articles/histogram-adc-testing-the-linear-ramp-histogram-test/
  35. [35]L. Chytka, M. Hrabovský, T. Komárek, V. Michálek, L. Nožka, T. Sýkora, and V. Urbášek, “Time resolution of the sipm-nuv3s,” Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers, Detectors and Associated Equipment, vol. 935, pp. 51–55, 2019. [Online]. Available: https://www.sciencedirect.com/science/article/pii/S0168900219305649
  36. [36]S. Gundacker, R. M. Turtos, E. Auffray, M. Paganoni, and P. Lecoq, “High-frequency SiPM readout advances measured coincidence time resolution limits in TOF-PET,” Physics in Medicine & Biology, vol. 64, no. 5, p. 055012, feb 2019. [Online]. Available: https://dx.doi.org/10.1088/1361-6560/aafd52
  37. [37]“Weeroc.” [Online]. Available: https://www.weeroc.com/read_out_chips/radiroc-2/
  38. [38]A. S. X. D. M. Pinto, N. Kratochwil, S. Gómez, D. Gascón, P. Correia, J. Veloso, E. Roncali, A. L. Silva, and G. A. No Estrada, “Improving ctr with the fastic asic for tof-pet by overcoming sipm noise with baseline correction,” IEEE Transactions on Radiation and Plasma Medical Sciences, pp. 1–1, 2025. [Online]. Available: https://doi.org/10.1109/TRPMS.2025.3532794
  39. [39]N. Kratochwil, E. Roncali, J. W. Cates, and G. Ariño-Estrada, “High-performance dual-ended sipm readout for tof-pet with bgo and lyso:ce,” IEEE Transactions on Radiation and Plasma Medical Sciences, pp. 1–1, 2025. [Online]. Available: https://doi.org/10.1109/TRPMS.2025.3539191

Paper details

vlnc:W_8bfqpqf2
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents