Soft Errors Research Articles

Effects of neutron radiation as cosmic radiation on food resources

One of the major challenges in deep space exploration are effects of cosmic rays on human body and soft error, but equally important are their effects on food resources. In this study, we focused on neutrons as the secondary particle in the deep space environment and investigated how they affect food resources. Fast neutron was irradiated by the RIKEN Accelerator-driven compact Neutron Source (RANS) on meat sample. Biochemical analysis was conducted to present concrete effects of cosmic rays on food in deep space environment. Oxidative and nitrative modifications of proteins were detected by the electrophoresis and Western blotting. The result shows that nitrative modification of tryptophan, which is an essential amino acid in protein, tended to increase with increasing neutron dose.

On FPGA SEU Mitigatiom under Irradiation Conditions

It is shown that in the case of soft errors even redundancy with casual choice for equiprobable configurations is equivalent to odd redundancy with less by one number of elements. The efficiencies of triple and quintuple modular redundancy are estimated.

Waveform-to-Waveform End-to-End Learning Framework in a Seamless Fiber-Terahertz Integrated Communication System

Seamless fiber-terahertz integrated communication has emerged as a promising technology in the special field of 6G, including mobile fronthaul and wireless bridges. Electronic terahertz communication systems have the advantages of high-level integration, small form factors, and potentially low costs, but their drawbacks are their small bandwidths and high harmonic interference levels. In this paper, an end-to-end learning-based waveform-to-waveform automatic equalization framework (W2WAEF) is proposed to overcome the above shortcomings in a seamless fiber-terahertz integrated communication system. An attention-based three-tributary heterogeneous neural network (ATTH) is designed to simulate the channel model of a seamless fiber-wireless communication system, taking fiber optics, optical-to-electrical conversion, and electrogenerated terahertz waves into account. With the proposed method, a data rate of 80.78 Gbps is experimentally demonstrated for discrete multitone modulation (DMT) signals over 5 km of fiber and 1 m of 209-GHz terahertz signals under a 20% soft decision-forward error correction (SD-FEC) threshold of 2.4 × 10 <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">−2</sup> . Compared with an approach without preprocessing, a receiver sensitivity gain exceeding 1.3 dB is successfully achieved at a data rate of 60 Gbps. The proposed method is a promising scheme for meeting the high speed and low-cost demands of future seamless fiber-terahertz integrated communication systems.

A Novel Low-Power and Soft Error Recovery 10T SRAM Cell.

In SRAM cells, as the size of transistors and the distance between transistors decrease rapidly, the critical charge of the sensitive node decreases, making SRAM cells more susceptible to soft errors. If radiation particles hit the sensitive nodes of a standard 6T SRAM cell, the data stored in the cell are flipped, resulting in a single event upset. Therefore, this paper proposes a low-power SRAM cell, called PP10T, for soft error recovery. To verify the performance of PP10T, the proposed cell is simulated by the 22 nm FDSOI process, and compared with the standard 6T cell and several 10T SRAM cells, such as Quatro-10T, PS10T, NS10T, and RHBD10T. The simulation results show that all of the sensitive nodes of PP10T can recover their data, even when S0 and S1 nodes flip at the same time. PP10T is also immune to read interference, because the change of the '0' storage node, directly accessed by the bit line during the read operation, does not affect other nodes. In addition, PP10T consumes very low-holding power due to the smaller leakage current of the circuit.

Impact of resampling interpolation FIR filter in the practical Kramers-Kronig receiver.

The practical Kramers-Kronig (KK) receiver has been a competitive receiving technique in the data-center, medium reach, and even long-haul metropolitan networks. Nevertheless, an extra digital resampling operation is required at both ends of the KK field reconstruction algorithm due to the spectrum broadening caused by adopting the nonlinear function. Generally, the digital resampling function can be implemented by using linear interpolation (LI-ITP), the Lagrange cubic interpolation (LC-ITP), the spline cubic interpolation (SC-ITP), time-domain anti-aliasing finite impulse response (FIR) filter method (TD-FRM) scheme, and fast Fourier transform (FFT)-based scheme. However, the performance and the computational complexity analysis of different resampling interpolation schemes in the KK receiver have not been thoroughly investigated yet. Different from the interpolation schemes of conventional coherent detection, the interpolation function of the KK system is followed by the nonlinear operation, which will broaden the spectrum significantly. Due to the frequency-domain transfer function of different interpolation schemes, the broadened spectrum will have a potential spectrum aliasing, which will cause serious inter-symbol interference (ISI) and further impair the KK phase retrieval performance. We experimentally investigate the performance of different interpolation schemes under different digital up-sampling rates (i.e. the computational complexity) as well as the cut-off frequency, the tap number of the anti-aliasing filter, and the shape factor of the TD-FRM scheme in a 112-Gbit/s SSB DD 16-QAM system over 1920-km Raman amplification (RFA)-based standard single-mode fiber (SSMF). The experimental results involve that the TD-FRM scheme outperforms other interpolation schemes and the complexity is reduced by at least 49.6%. In fiber transmission results, take 20% soft decision-forward error correction (SD-FEC) of 2×10-2 as the threshold, the LI-ITP and LC-ITP schemes only reach 720-km while others can reach up to 1440-km.

Fast and Accurate Error Simulation for CNNs Against Soft Errors

The great quest for adopting AI-based computation for safety-/mission-critical applications motivates the interest towards methods for assessing the robustness of the application w.r.t. not only its training/tuning but also errors due to faults, in particular soft errors, affecting the underlying hardware. Two strategies exist: architecture-level fault injection and application-level functional error simulation. We present a framework for the reliability analysis of Convolutional Neural Networks (CNNs) via an error simulation engine that exploits a set of validated error models extracted from a detailed fault injection campaign. These error models are defined based on the corruption patterns of the output of the CNN operators induced by faults and bridge the gap between fault injection and error simulation, exploiting the advantages of both approaches. We compared our methodology against SASSIFI for the accuracy of functional error simulation w.r.t. fault injection, and against TensorFI in terms of speedup for the error simulation strategy. Experimental results show that our methodology achieves about 99% accuracy of the fault effects w.r.t. SASSIFI, and a speedup ranging from 44x up to 63x w.r.t. TensorFI, that only implements a limited set of error models.

Radiation-Tolerant Deep Learning Processor Unit (DPU)-Based Platform Using Xilinx 20-nm Kintex UltraScale FPGA

This paper presents a platform and design approach for enabling radiation-tolerant deep learning acceleration on SRAM-based 20nm Kintex UltraScale™ FPGAs, for terrestrial and high-radiation environments. The presented platform is suitable for deep neural network (DNN) implementations with an emphasis on image classification and includes solutions to mitigate both radiation-induced Single Event Functional Interrupts (SEFIs) and network datapath corruptions. The radiation-tolerant deep learning platform combines Xilinx’s Deep Learning Processing Unit (DPU) IP, Triple Modular Redundancy (TMR) MicroBlaze soft processor IP and Soft Error Mitigation (SEM)-IP to mitigate SEFIs. Furthermore, a technique known as Fault Aware Training (FAT) was applied to effectively mitigate single event effects in the datapath. Test results from a high-energy proton beam (> 60 MeV) experiment using the ResNet-18 Convolutional Neural Network (CNN) for image classification are presented. The Single Event Upset (SEU) rate, system-level SEFI rate and neural network classification/datapath performance are compared between the radiation-tolerant platform and a standard, non-mitigated approach. Results show that datapath classification errors dominate the system response (90%) vs. SEFIs (10%). When compared to standard non-mitigated training techniques, the radiation-tolerant platform using fault aware training methods shows dramatic improvements in overall system response: the overall single event cross-section was reduced by half and 40% reduction in misclassification errors were observed. Also, datapath events with classification accuracy degradation larger than 5% were completely mitigated. The SEFI rate was reduced by 100X with implemented solutions and can be further reduced by optimizing the physical separation between TMR modules.

Energy and Reliability-Aware Task Scheduling for Cost Optimization of DVFS-Enabled Cloud Workflows

Due to the increasing complexity, the execution of workflow applications on cloud typically involves a large number of virtual machines (VMs), which makes the cost as well as energy consumption a great concern. To alleviate this issue, more and more cloud service providers introduce new pricing policies considering Dynamic Voltage and Frequency Scaling (DVFS), where users are charged on the basis of allocated CPU frequencies together with various combinations of VM configurations and prices. However, the customizable CPU frequencies make resource provisioning and scheduling harder to achieve a cost-optimal solution. The things become even worse, since lowering CPU voltages of VMs will increase their chance of suffering soft errors, which results in a high rate of completion time failures of workflow applications. To address the above problem, this paper proposes a novel task scheduling method for the purpose of cost optimization based on the genetic algorithm. By introducing new genetic operators and frequency scaling scheme for DVFS-enabled cloud workflows, our approach can quickly figure out cost-optimal resource provisioning and task scheduling solutions by allocating tasks to appropriate VMs with specific operating frequencies under energy, reliability, makespan and memory constraints. Extensive experiments on various well-known scientific workflow benchmarks validate the effectiveness of the proposed method. Comparing with state-of-the-art methods, our approach can significantly reduce the overall cost and energy consumption without violating the given constraints.

Radiation studies performed on the High Luminosity ATLAS TileCal link Daughterboard

The new electronics of the ATLAS Tile Calorimeter for the HL-LHC interfaces the on-detector and off-detector electronics by means of a Daughterboard. The Daughterboard is positioned on-detector featuring commercial SFPs+, CERN GBTx ASICs, ProASIC FPGAs and Kintex Ultrascale FPGAs. The design minimizes single points of failure and mitigates radiation damage by means of a double redundant scheme, Triple Mode Redundancy, Xilinx Soft Error Mitigation IP, CRC/FEC for link data transfer, and SEL protection circuitry. We present an updated summary of the TID, NIEL and SEE qualification tests, and performance studies of the Daughterboard revision 6 design.

Ternary SRAM circuit designs with CNTFETs

SummaryStatic random‐access memory (SRAM) is a cornerstone in modern microprocessors architecture, as it has high power consumption, large area, and high complexity. Also, the stability of the data in the SRAM against the noise and the performance under the radian exposure are main concern issues. To overcome these limitations in the quest for higher information density by memory element, the ternary logic system has been investigated, showing promising potential compared with the conventional binary base. Moreover, carbon nanotube field effect transistor (CNTFET) is a new alternative device with proper features like low power consumption and threshold voltage dependency on diameter. This paper proposes a new design for ternary SRAM using CNTFET and its evaluation by comparing it against two other designs in many aspects. Moreover, we investigated the static noise margin for the three designs to discuss their stability. Furthermore, we studied the reliability of the designs by evaluating the soft errors effect.

Single-Event Transient Study of 28 nm UTBB-FDSOI Technology Using Pulsed Laser Mapping

Single-event transient (SET)-induced soft errors are becoming a more significant threat to the reliability of electronic systems in space, especially for advanced technologies. The SET pulse width, which is vulnerable to SET propagation, is a critical parameter for developing SET mitigation techniques. This paper investigates the pulse-broadening effect in the process of SET propagation in logic circuits and the SET-sensitive region distribution in the layout using the pulsed-laser mapping technique in logic circuits implemented with 28 nm Ultra-Thin Body and BOX (UTBB) FDSOI technology. The experiments were carried out at the Naval Research Laboratory (NRL) to measure the SET-induced errors and map the SET-sensitive region distribution at various clock frequencies and laser energy levels. The results illustrate that the number of errors increases with the clock frequency and energy for combinational logic circuits and that the flip-flop SEU rate is less sensitive to clock frequency. The SET pulse-broadening effect was also observed using SET mapping for an OR gate chain at different laser energy levels. In addition, the simulation results revealed the mechanism of the SET pulse-broadening effect in an OR gate chain.

Vulnerability Estimation of DNN Model Parameters with Few Fault Injections

The reliability of deep neural networks (DNN) against hardware errors is essential as DNNs are increasingly employed in safety-critical applications such as automatic driving. Transient errors in memory, such as radiation-induced soft error, may propagate through the inference computation, resulting in unexpected output, which can adversely trigger catastrophic system failures. As a first step to tackle this problem, this paper proposes constructing a vulnerability model (VM) with a small number of fault injections to identify vulnerable model parameters in DNN. We reduce the number of bit locations for fault injection significantly and develop a flow to incrementally collect the training data, i.e., the fault injection results, for VM accuracy improvement. We enumerate key features (KF) that characterize the vulnerability of the parameters and use KF and the collected training data to construct VM. Experimental results show that VM can estimate vulnerabilities of all DNN model parameters only with 1/3490 computations compared with traditional fault injection-based vulnerability estimation.

Optimized reverse converters with multibit soft error correction support at 7nm technology

Residue number system (RNS) speeds up digital signal processing systems involving dominant addition and multiplication. Addition and multiplication are accelerated further and performed with a balanced performance on one-hot coded (OHC) residue digits. However, the high complexity of the RNS reverse converter (RC) may kill the performance gain. This paper proposes a high-speed and scalable RNS-RC for both regular and one-hot RNS. For redundant RNS (RRNS), an RRNS-RC is proposed, which based on majority-voting between OHC residue digits, corrects multibit soft errors occurring in a single residue channel. With pass-transistor logic and low-power FinFETs, the proposed RCs are optimized. The simulated RRNS-RC corrected 98.5% of soft errors, while consuming 7, 6.2 and 0.4% of the system's area, leakage power and dynamic power, respectively. As compared to the leading lookup table RC in the literature, RNS-RC exhibited 10.3X, 4.4X and 1.8X savings in area, average power, and delay, respectively.

1.57 Tb/s PDM-layered 80QAM-OFDM in super-channel transmission based on self-homodyne comb

In this paper, polarization division multiplexing-layered 80-ary quadrature amplitude modulation-orthogonal frequency division multiplexing (PDM-layered 80QAM-OFDM) is proposed and demonstrated in super-channel system based on self-homodyne comb. The composition of the super-channel is derived from the optical comb generated by a single laser-driven dual polarized Mach-Zehnder modulator (DP-MZM) modulated by odd sideband suppression. Since the optical comb generated by the transmitter is used as a local oscillator (LO) signal for self-homodyne coherent detection. This scheme effectively reduces the system cost and alleviates the performance constraints caused by phase noise. Moreover, the bit error rate (BER) and generalized mutual information (GMI) of layered 24QAM/48QAM/80QAM-OFDM are evaluated to obtain the highest transmission rate and the laser linewidth tolerance of the system. The results show that in Additive White Gaussian Noise (AWGN) channel, layered QAM signal realized by low-complexity constellation layering technique can significantly improve the BER performance at a small entropy cost compared with ordinary QAM signals. Meanwhile, the layered 80QAM can achieve a signal-to-noise ratio (SNR) gain of 2.7 dB. In addition, after 40 km standard single mode fiber (SSMF) transmission, the total data rate of the system can reach 1.57 Tbit/s and the corresponding spectral efficiency (SE) is 12 bit/s/Hz when the BER of the super-channel system meets soft decision-forward error correction (SD-FEC) threshold.

Soft error detection and correction for parallel digital filters using Hamming code

Soft error detection and correction for parallel digital filters using Hamming code

The potential of the ILC beam dump for high-intensity and large-area irradiation field with atmospheric-like neutrons and muons

We evaluate the neutron and muon fluxes produced in the ILC beam dumps by Monte Carlo simulations and discuss their potential use in irradiation fields. We show that the beam dumps can provide high-intensity neutron and muon fluxes with spectra quite similar to secondary cosmic rays, which are suitable for soft error studies. The beam dumps deliver neutrons about 1011 times the secondary cosmic rays on spaces perpendicular to the beam axis and muons 108 times downstream of the beam dumps in the initial phase of the ILC. Large-area irradiation of 1m2 or more is possible. Differences in the energy distribution of muons in electron and positron beam dumps are also discussed for particle physics experiments.

8192QAM Signal Transmission by an IM/DD System at W-Band Using Delta-Sigma Modulation

We present a low-cost intensity-modulated direct detection (IM/DD) system for the transmission of high-order QAM signals at millimeter-wave (MMW) band at 87 GHz. The system uses photon-assisted MMW generation techniques to transmit modulated signals and receive signals through low-cost envelope detection. Discrete multi-tone (DMT) signals up to 8192QAM are modulated into a one-bit quantized on/off keying (OOK) signal with delta-sigma modulation (DSM) to avoid envelope detector saturation effects and reduce nonlinear noise interference. Additionally, multi-level digital-to-analog converters (DACs) are not required. The 87 GHz W-band signal can transmit data at a baud rate of 12 G over a wireless propagation range of 1.2 m. The experimental results show that the bit error ratio (BER) of the 1024QAM signal can reach the hard decision forward error correction (HD-FEC) threshold, and the BER of the 8192QAM signal can reach the soft decision forward error correction (SD-FEC) threshold. To our knowledge, this is the first time that the DSM modulation format is used in the IM/DD W-band wireless transmission system.

A 30 MeV-cyclotron-based quasi-monoenergetic neutron source

This study developed a quasi-monoenergetic neutron source (QMN) for the semiconductor device's soft error rate test (SER). Quasi-monoenergetic neutrons are generated by Be(p,n)B99 nuclear reaction with a 1 mm beryllium target and 30 MeV protons from a cyclotron. An 8 mm water in the back of the beryllium target is used for avoiding proton penetration. The neutron spectra simulated by MCNP showed that the peak energy was around 26.5 MeV. The heat flow and mechanical properties are numerically analyzed, and the safe operating conditions are therefore determined.

A Checkpointing Recovery Approach for Soft Errors Based on Detector Locations

Soft errors are transient errors caused by single-event effects (SEEs) resulting from a strike by high-energy particles acting on sensitive areas of integrated circuits. Soft errors frequently occur in the space environment, adversely affecting the reliability of aerospace-based computing. A recovery process is launched to recover the program when soft errors are detected. A periodic checkpointing recovery approach is widely utilized to prevent soft errors. However, this approach does not consider the detector locations, resulting in a large time overhead. This paper proposes a checkpointing recovery approach for soft errors based on detector locations called DLCKPT. DLCKPT reduces the time overhead by considering detector locations. The experimental results show that the percentage decrease in the time overhead between the DLCKPT and the periodic checkpointing recovery approach is 13.4%. The average recovery rate and average space overhead are 99.3% and 44.4% for the periodic checkpointing recovery approach and 99.4% and 34.6% for the DLCKPT. These results show that the DLCKPT and the periodic checkpointing recovery approach produce comparable results for the recovery rate. The DLCKPT has a lower time overhead and a slightly lower space overhead than the periodic checkpointing recovery approach, demonstrating its effectiveness.

STRV — a radiation hard RISC-V microprocessor for high-energy physics applications

While microprocessors are used in various applications, they are precluded from the use in high-energy physics applications due to the harsh radiation present. To overcome this limitation a microprocessor design must withstand high doses of radiation and mitigate radiation induced soft errors. A TMR protection scheme is applied to protect a RISC-V microprocessor core against these faults. The protection of the integrated SRAM by an independent scrubbing algorithm is discussed. Initial irradiation results and power consumption measurements of the radiation-resistant RISC-V microprocessor implemented in 65 nm CMOS technology are presented.