Search

This paper presents an overview for low-power successive approximation register (SAR) analog-to-digital converters (ADCs). It covers the operation principle, error analysis, and practical design issues. Furthermore, this paper provides a comprehensive survey of state-of-the-art low-power design techniques for every circuit block in the SAR ADC, including comparator, capacitive digital-to-analog converter (DAC), and SAR logic. The goal of this paper is to provide a useful overview to SAR ADC designers who want to improve the energy efficiency targeting low-to-medium speed applications. [ABSTRACT FROM AUTHOR]

SRAM based computing-in-memory (SRAM-CIM) techniques have been widely studied for neural networks (NNs) to solve the “Von Neumann bottleneck”. However, as the scale of the NN model increasingly expands, the weight cannot be fully stored on-chip owing to the big device size (limited capacity) of SRAM. In this case, the NN weight data have to be frequently loaded from external memories, such as DRAM and Flash memory, which results in high energy consumption and low efficiency. In this paper, we propose a hybrid-device computing-in-memory (HD-CIM) architecture based on SRAM and MRAM (magnetic random-access memory). In our HD-CIM, the NN weight data are stored in on-chip MRAM and are loaded into SRAM-CIM core, significantly reducing energy and latency. Besides, in order to improve the data transfer efficiency between MRAM and SRAM, a high-speed pipelined MRAM readout structure is proposed to reduce the BL charging time. Our results show that the NN weight data loading energy in our design is only 0.242 pJ/bit, which is 289 $\times $ less in comparison with that from off-chip DRAM. Moreover, the energy breakdown and efficiency are analyzed based on different NN models, such as VGG19, ResNet18 and MobileNetV1. Our design can improve $\mathbf {58\times \,\,to\,\,124\times }$ energy efficiency. [ABSTRACT FROM AUTHOR]

Computing in memory (CIM) is a promising candidate for high throughput and energy-efficient data-driven applications, which mitigates the well-known memory bottleneck in Von Neumann architecture. In this paper, we present a reconfigurable bit-serial operation using toggle spin-orbit torque magnetic random access memory (TSOT-MRAM) to perform the computation completely in the bit-cell array instead of in a peripheral circuit. This bit-serial CIM (BSCIM) scheme achieves higher throughput and energy efficiency in CIM. First, basic Boolean logic operations are realized by utilizing the feature of TSOT device. A bit-cell array that implements the bit-serial operation is then built to provide the communication between column and row necessary for arithmetic operations, such as the carry propagation of addition and multiplication. Finally, we analyze the reliability of BSCIM scheme and demonstrate the performance advantage by performing convolution operations for $28\times 28$ handwritten digit images in a BSCIM architecture. The results show that the delay and energy of BSCIM architecture are respectively reduced by 1.16-5.49 times and 1.12-1.43 times compared with the existing digital CIM architectures. Besides, its throughput and energy efficiency are also enhanced to 51.2 GOPS and 9.9 TOPS/W respectively. [ABSTRACT FROM AUTHOR]

In recent years, fast computation, low power, and small footprint are the key motivations for building SNN hardware. The unique features of SNN hardware have not been fully exploited, where the computation speed and energy efficiency of the SNN hardware can be improved according to the sparse spiking and non-uniform traffic of SNN. In this paper, we propose a 5.28-mm $^{2}~4096$ -neuron 1M-synapse energy-efficient digital SNN hardware that can achieve ultra-low energy per synaptic operation of 4.5 pJ. The proposed neuron computing unit is implemented in pipeline architecture to achieve high synaptic processing speed. The proposed spike processing unit can significantly increase the processing speed of the neuron core by $1.9\times $ and $9.4\times $ when the spike injection rate is 50% and 10%, respectively. Besides, the increase in the processing speed of the neuron core leads to a reduction in energy consumption of up to 81.5%. An event-driven clock gating circuit that can reduce the power consumption of the proposed neuron block by more than 70% is proposed in this paper. This paper proposes a supervised STDP+ algorithm for SNN training, and the classification accuracy of the MNIST digits is 89.6% with 73.6% weight sparsity of the output layer. [ABSTRACT FROM AUTHOR]

By their very nature, Spin Waves (SWs) excited at the same frequency but different amplitudes, propagate through waveguides and interfere with each other at the expense of ultra-low energy consumption. In addition, all (part) of the SW energy can be moved from one waveguide to another by means of coupling effects. In this paper we make use of these SW features and introduce a novel non Boolean algebra based paradigm, which enables domain conversion free ultra-low energy consumption SW based computing. Subsequently, we leverage this computing paradigm by designing a non-binary spin wave adder, which we validate by means of micro-magnetic simulation. To get more inside on the proposed adder potential we assume a 2-bit adder implementation as discussion vehicle, evaluate its area, delay, and energy consumption, and compare it with conventional SW and 7 nm CMOS counterparts. The results indicate that our proposal diminishes the energy consumption by a factor of $3.14 \times $ and $6 \times $ , when compared with the conventional SW and 7 nm CMOS functionally equivalent designs, respectively. Furthermore, the proposed non-binary adder implementation requires the least number of devices, which indicates its potential for small chip real-estate realizations. [ABSTRACT FROM AUTHOR]

When deploying deep neural networks (DNNs), edge devices training is practical to improve model adaptivity for various user-specific scenarios while avoiding privacy disclosure. However, the training computation is intolerable for edge devices. It inspires sparse DNN training (SDT) into the limelight, which reduces training computation by dynamic weight pruning. Generally, SDT has two strategies based on the pruning granularity: the structured or the unstructured. Unfortunately, both of them suffer from limited training efficiency due to the gap between pruning granularity and hardware implementation. The former is hardware-friendly but has a low pruning ratio, indicating limited computation reduction. The latter has a high pruning ratio, but the unbalanced workload decreases utilization and irregular sparsity distribution causes considerable sparsity processing overhead. This paper proposes a software-hardware co- design to bridge the gap for improving the efficiency of SDT. On the algorithm side, a sub-structured pruning method, achieved with hybrid shape-wise and line-wise pruning, generates a high sparsity ratio and keeps the hardware-friendly property. On the hardware side, a sub-structured weight processing unit (SWPU) effectively handles the hybrid sparsity with three techniques. First, SWPU dynamically reorders the computation sequence with hamming-distance-based clustering, balancing the irregular workload. Second, SWPU performs runtime scheduling by exploiting the feature of sub-structured sparse convolution through a detect-before-load controller, which skips redundant memory access and sparsity processing. Third, SWPU performs sparse convolution by compressing operands with spatial disconnect log-based routing and recovers their location with bi-directional switching, avoiding the power-consumed routing logic. Synthesized with 28nm CMOS technology, SWPU can enable 0.56V-to-1.0V supply voltage with a maximum frequency of 675 MHz. It achieves a 50.1% higher pruning ratio than structured pruning and $1.53\times $ higher energy efficiency than unstructured pruning. The peak energy efficiency of SWPU is 126.04TFLOPS/W, outperforming the state-of-the-art training processor by $1.67\times $. When training a ResNet-18 model, SWPU reduces $3.72\times $ energy and offers $4.69\times $ speedup than previous sparse training processors. [ABSTRACT FROM AUTHOR]

Edge device deep neural network (DNN) training is practical to improve model adaptivity for unfamiliar datasets while avoiding privacy disclosure and huge communication cost. Nevertheless, apart from feed-forward (FF) as inference, DNN training still requires back-propagation (BP) and weight gradient (WG), introducing power-consuming floating-point computing requirements, hardware underutilization, and energy bottleneck from excessive memory access. This paper proposes a DNN training processor named PL-NPU to solve the above challenges with three innovations. First, a posit-based logarithm-domain processing element (PE) adapts to various training data requirements with a low bit-width format and reduces energy by transferring complicated arithmetics into simple logarithm domain operation. Second, a reconfigurable inter-intra-channel-reuse dataflow dynamically adjusts the PE mapping with a regrouping omega network to improve the operands reuse for higher hardware utilization. Third, a pointed-stake-shaped codec unit adaptively compresses small values to variable-length data format while compressing large values to fixed-length 8b posit format, reducing the memory access for breaking the training energy bottleneck. Simulated with 28nm CMOS technology, the proposed PL-NPU achieves a maximum frequency of 1040MHz with 343mW and 5.28mm $\mathbf {^{2}}$. The peak energy efficiency is 3.87TFLOPS/W for 0.6V at 60MHz. Compared with the state-of-the-art training processor, PL-NPU reaches $3.75\times $ higher energy efficiency and offers $1.68\times $ speedup when training ResNet18. [ABSTRACT FROM AUTHOR]

This paper presents an advanced offline scheduling scheme to improve the accelerator efficiency, especially for the highly-pruned convolutional neural networks (HP-CNNs). Based on the existing outlier-aware accelerator design, we demonstrate the efficiency drop of HP-CNN processing for the first time, and element-wise channel merging is proposed to make a dense processing sequence even for the highly-pruned model. The dedicated hardware architecture is also presented to process the merged channels with the minimum hardware-level overheads, improving the energy efficiency for handling HP-CNNs by preserving the hardware utilization. We further investigate the optimal size of accumulator and multiplexer, in addition to the number of merged channels, exploiting the attractive energy-performance trade-offs. As a result, unlike the practical HP-CNNs for the on-device solutions, the proposed method enhances the overall efficiency by up to 33% compared to the state-of-the-art schemes. [ABSTRACT FROM AUTHOR]

One key factor affecting the distributed Nash equilibrium (NE) seeking in aggregative games is the unbalanced communication structure for multiple players. Although some results on seeking NE over undirected or weight-balanced graphs were established, how to address the distributed NE seeking problem over general directed communication graphs is still an outstanding challenge. This paper addresses the NE seeking problem for a class of aggregative games with general directed communication graphs. To achieve this objective, two new kinds of distributed discrete-time NE seeking algorithms are developed for aggregative games over fixed digraphs and time-varying digraphs, respectively. In particular, motivated by the heavy-ball method in optimization studies, a momentum term is introduced to the update law of the players’ actions and it is numerically verified that this momentum term accelerates the convergence of the proposed algorithms. For both strongly connected fixed graph and $B$ -strongly connected time-varying graph, it is theoretically proved that the actions of players will converge to the NE of aggregative games for the case of decreasing step-size implemented by the proposed NE seeking algorithms if the cost functions and the aggregation of players satisfy some certain conditions. Finally, the developed NE seeking algorithms are applied to the energy consumption control of plug-in hybrid electric vehicles (PHEVs), which demonstrates the effectiveness of the theoretical results. [ABSTRACT FROM AUTHOR]

The complexity and size of recent deep neural network (DNN) models have increased significantly in pursuit of high inference accuracy. Chiplet-based accelerator is considered a viable scaling approach to provide substantial computation capability and on-chip memory for efficient process of such DNN models. However, communication using metallic interconnects in prior chiplet-based accelerators poses a major challenge to system performance, energy efficiency, and scalability. Photonic interconnects can adequately support communication across chiplets due to features such as distance-independent latency, high bandwidth density, and high energy efficiency. Furthermore, the salient ease of broadcast property makes photonic interconnects suitable for DNN inference which often incurs prevalent broadcast communication. In this paper, we propose a scalable chiplet-based DNN accelerator with photonic interconnects named ASCEND. ASCEND introduces (1) a novel photonic network that supports seamless intra- and inter- chiplet broadcast communication, and flexible mapping of diverse convolution layers, and (2) a tailored dataflow that exploits the ease of broadcast property and maximizes parallelism by simultaneously processing computations with shared input data. Simulation results using multiple DNN models show that ASCEND achieves 71% and 67% reduction in execution time and energy consumption, respectively, as compared to other state-of-the-art chiplet-based DNN accelerators with metallic or photonic interconnects. [ABSTRACT FROM AUTHOR]

Tree-based machine learning models, like random forests and decision trees, are low-complexity solutions that provide an efficient prediction for a wide range of applications. These models are particularly interesting for energy-constrained platforms since they can be implemented with simple logical operations. Tree-based accelerators are also intrinsically resilient to errors, and this can be leveraged to boost energy efficiency with approximate computing techniques. The key operations in these models are comparisons to constants, making comparators excellent candidates for approximation. This paper presents a technique to approximate comparisons to constants called C2PAx, which is capable of reducing the area and energy of tree-based accelerators. The method consists in finding alternative constants that reduce circuit area while keeping an efficient prediction performance. It is also shown that the selection of the constant parameters directly influences both hardware complexity and model performance, demanding cross-layer optimization. For that, we extend an existing framework that generates VLSI tree-based accelerators, inserting our approximation proposal that allows selecting the constant parameters that maximize energy efficiency at the cost of minor accuracy drops. Simulation results demonstrate that C2PAx outperforms the Don’t Care logic approximation technique when accuracy and energy are jointly considered. C2PAx trades accuracy for significant reductions in the VLSI area, power, delay, and energy consumption compared to precise models. [ABSTRACT FROM AUTHOR]

Spiking neural networks (SNNs), which are inspired by the human brain, have recently gained popularity due to their relatively simple and low-power hardware for transmitting binary spikes and highly sparse activation maps. However, because SNNs contain extra time dimension information, the SNN accelerator will require more buffers and take longer to infer, especially for the more difficult high-resolution object detection task. As a result, this paper proposes a sparse compressed spiking neural network accelerator that takes advantage of the high sparsity of activation maps and weights by utilizing the proposed gated one-to-all product for low power and highly parallel model execution. The experimental result of the neural network shows 71.5% mAP with mixed (1,3) time steps on the IVS 3cls dataset. The accelerator with the TSMC 28nm CMOS process can achieve $1024\times 576.29$ frames per second processing when running at 500MHz with 35.88TOPS/W energy efficiency and 1.05mJ energy consumption per frame. [ABSTRACT FROM AUTHOR]

The rising popularity of intelligent mobile devices and the computational cost of deep learning-based models call for efficient and accurate on-device inference schemes. We propose a novel model compression scheme that allows inference to be carried out using bit-level sparsity, which can be efficiently implemented using in-memory computing macros. In this paper, we introduce a method called BitS-Net to leverage the benefits of bit-sparsity (where the number of zeros are more than number of ones in binary representation of weight/activation values) when applied to compute-in-memory (CIM) with resistive RAM (RRAM) to develop energy efficient DNN accelerators operating in the inference mode. We demonstrate that BitS-Net improves the energy efficiency by up to 5x for ResNet models on the ImageNet dataset. [ABSTRACT FROM AUTHOR]

For the design of ring oscillator-based ADCs, little has been reported on how to optimally co-design the readout scheme and the ring oscillator core towards optimal energy efficiency. This paper describes a methodology to find this optimum for a target SQNR and signal bandwidth, for the case of 1st-order quantization noise shaping. In short, starting from initial assumptions on the VCO, the readout optimization boils down to finding the best VCO frequency. From this, the number of readout phases, the sampling frequency of the readout and the number of bits in the quantizers are derived. Then, it is explained how to adjust the VCO core toward the corresponding optimal VCO configuration. This is discussed for the case of a current-controlled ring oscillator, indicating the main constraints that define the design space for such a VCO optimization. If the actual VCO configuration is bounded by practical constraints, this approach can be re-iterated where a new readout optimization is performed taking these constraints into account. [ABSTRACT FROM AUTHOR]

Computing-in-memory (CIM) is a new architecture which is more energy-efficient than the Von Neumann architecture due to the fact that it performs calculation in the memory units which can reduce a large amount of data movement. Nowadays, CIM with non-volatile memory (nvCIM), such as resistive random access memory (RRAM), has become a research frontier to further improve computing performance. Recent works mainly explored how to improve the computing performance of nvCIM, but seldom paid attention to the problem of accuracy loss. In this paper, we propose the nvCIM framework, which can systematically analyze the relationship between the classification accuracy of network models and main analog factors. Based on the nvCIM framework, we further provide detailed optimization methods, including RRAM features, array properties, and ADC parameters. The adaptive voltage-controlled SET and pulse-controlled RESET (VSPR) program-verify scheme is proposed to achieve high-resolution RRAM. And the margin enhancement based current-mode sense amplifier (MECSA) and offset reduction based analog-to-digital converter (ORADC) are proposed to improve the accuracy of analog part computing. Experimental results show that the macro-level and system-level energy efficiency is 112.1 TOPS/W and 9.86 TOPS/W respectively with less than 3.51% loss in contrast to the ideal accuracy, which is $2.9\times $ - $25.9\times $ compared to the energy efficiency of existing RRAM based nvCIM accelerators. [ABSTRACT FROM AUTHOR]

In many sensor applications, a high-resolution analog-to-digital converter (ADC) is a key block. The use of an incremental delta-sigma ADC (IADC) is often well suited for such applications. While the energy-efficiency of IADCs has improved by several orders of magnitude over the past decade, the implementation of high performance IADCs, especially in battery-powered systems, is still challenging. This paper presents a tutorial review on energy-efficient IADCs and addresses the progress in this area. This paper describes the fundamentals of IADCs and energy-efficient hybrid IADC architectures. Various design techniques for improving the energy-efficiency of the IADCs are described. This paper is intended to serve as a starting point for the development of a new energy-efficient IADC. [ABSTRACT FROM AUTHOR]

This paper presents a nonlinear digital decoder (reconstruction filter) for incremental delta-sigma modulators. This decoder utilizes both the magnitude and pattern information of the modulator output to achieve accurate input estimation. Compared to the conventional linear filters with the same oversampling ratio (OSR), it can improve the converter’s signal-to-quantization noise ratio by a few dB to a few 10’s of dB with slight thermal noise performance degradation. Using the proposed decoder, the modulator’s OSR can be a few times less while achieving the same resolution and data rate, thus minimizing the modulator as well as its peripheral circuits’ energy consumption. In this paper, the proposed decoder is optimized for digital implementation, with its function being verified using a modulator prototype. This decoder is mainly designed for dc or near-dc signal conversions and it does not provide frequency notches. [ABSTRACT FROM AUTHOR]

This paper presents a buffer-embedded noise-shaping SAR ADC, whose input buffer separates the capacitive DAC (CDAC) and the sampling capacitor (CS) at the input and output of the input buffer. This compensates for the non-linearity of the input buffer and reduces the CS value, resulting in a significant power saving in the input buffer. This buffer-embedded architecture enables the effective implementation of the following passive loop filter and enhances energy efficiency. A bootstrapping switch in the feedback CDAC is coupled to the output of the buffer, thereby avoiding a signal dependency due to the parasitic capacitance of the switch. The buffer-embedded noise-shaping SAR ADC occupies 0.08mm2 in a 65 nm CMOS process and features a parasitic input capacitor of 0.2 pF. It achieves 73.8 dB SNDR, 77 dB DR and 87.3 dB SFDR in a 2 MHz bandwidth without any calibration. Including the power consumption of the input buffer, the ADC consumes only 2.1 mW. [ABSTRACT FROM AUTHOR]

Neuromorphic computing can solve computationally hard problems with energy efficiencies unattainable for von Neumann architectures. A locally-active memristor, which possesses the capability to amplify infinitesimal fluctuations in energy and can be used to generate neuromorphic behaviors, is a natural candidate for constructing an electronic equivalent of biological neurons. This paper identifies some unknown neuromorphic dynamics of the Chua corsage memristor (CCM), and shows that the CCM, when biased at the edge of chaos domain, can exhibit rich dynamics of biological neurons. Using Chua’s theories of local activity and edge of chaos, we demonstrate that under the destabilizing of the input voltage and the circuit parameters (inductance or capacitance), two CCM-based circuits can produce thirteen types of neuromorphic behaviors either on, or near the edge of chaos domain via supercritical or subcritical Hopf bifurcation. In addition, we give the conditions to test the edge of chaos of the CCM and the CCM-based circuit only by using the poles and the zero of their admittance functions. [ABSTRACT FROM AUTHOR]

Blockchain distributed ledger technology (DLT) has widespread applications in society 5.0 because it improves service efficiency and significantly reduces labor costs. However, employing blockchain DLT entails considerable energy consumption in the mining process. This paper proposes a blockchain accelerator (BCA) with ultralow power consumption and a high processing rate to address the problem. The BCA focuses on accelerating the double secure hash algorithm (SHA) 256 function required in the mining process at a system-on-chip (SoC) level. We propose three ideas, namely, multiple local memories (multimem), double-cell processing element (D-PE), and nonce autoupdate (NAU), to reduce the external data transfer time and improve the BCA hardware efficiency. We propose a cascaded multiple BCA chip model to enhance the system throughput by several-fold. Our experiments on an ASIC and FPGA prove that the proposed BCA successfully performs the mining process for multiple blockchain networks with much lower power consumption than that of the state-of-the-art CPUs and GPUs. The BCA is laid out with Renesas 65 nm technology with a chip area of $25~mm^{2}$ and consumes $530~mW$ at 100 MHz. The power efficiency of the layout chip is improved by 2428 and 143 times compared with that of the fastest CPU Intel i9-10940X and GPU RTX 3090, respectively. [ABSTRACT FROM AUTHOR]

This paper discusses finite/fixed-time (FNT/FXT) synchronization of multi-layer networks based on energy consumption estimation. Firstly, a novel FXT stability theorem of dynamical system is presented. Secondly, sufficient criteria for FNT/FXT synchronization in multi-layer networks are given, and the estimation of the energy consumption are also gained. Especially, the range of the number of layers of network is gained based on the least convergence time and the estimation of the least energy consumption. The results show that shorter synchronization time and less energy consumption does not mean less network layers. Finally, the validity of the theoretical method is verified by numerical simulation. [ABSTRACT FROM AUTHOR]

In this paper, a method for offline training of inverter-based memristive neural networks (IM-NNs), called ERIM, is presented. In this method, the output voltage of the inverter is modeled very accurately by considering the loading effect of the memristive crossbar. To properly choose the size of each inverter, its output load and the required slope of its voltage transfer characteristic (VTC) for an acceptable level of resiliency to the circuit element non-idealities are taken into account. The efficacy of ERIM is investigated by comparing its accuracy to those of two recently proposed offline training methods for IM-NNs (RIM and PHAX). The study is performed using IRIS, BCW, MNIST, and Fashion MNIST datasets. Simulation results show that 72% (56%) reduction in average energy consumption of the trained networks is achieved compared to RIM (PHAX) thanks to proper sizing of the inverters. In addition, due to the higher accuracy of the NN mathematical model, ERIM results in significant improvements in the match between the results of high-level modeling and HSPICE simulations while exhibiting lower sensitivity to circuit element variations. [ABSTRACT FROM AUTHOR]

In this paper, a two-stage reconfiguration strategy is proposed for a group of agents to find their specially designated formation, which can be seen as a transition from the current states to the predefined formations. The specially designated formation occurs during the inter-task idle time in order to minimize the reconfiguration time. The basic specially designated formation is designed for dotted agents based on the optimal mapping relationships and current probabilities of the predefined formations. Then, the reconfiguration schemes are developed with the agents modeled as circles or ellipses to account for their physical shapes, and the problem of agent overlapping during the formation is addressed. Examples and simulation results are presented to analyze and demonstrate the effectiveness and potential of the new design techniques. [ABSTRACT FROM AUTHOR]

This paper presents a charge-domain in-memory computing (IMC) macro for precision-scalable deep neural network accelerators. The proposed Dual-SRAM cell structure with coupling capacitors enables charge-domain multiply and accumulate (MAC) operation with variable-precision signed weights. Unlike prior charge-domain IMC macros that only support binary neural networks or digitally compute weighted sums for MAC operation with multi-bit weights, the proposed macro implements analog weighted sums for energy-efficient bit-scalable MAC operations with a novel series-coupled merging scheme. A test chip with a 16-kb SRAM macro is fabricated in 28-nm FDSOI process, and the measured macro throughput is 125.2-876.5 GOPS for weight bit-precision varying from 2 to 8. The macro also achieves energy efficiency ranging from 18.4 TOPS/W for 8–b weight to 119.2 TOPS/W for 2-b weight. [ABSTRACT FROM AUTHOR]

Data transfer across DRAM channels accounts for nearly a quarter of the total energy consumption of DDR4 DRAMs. Modern applications with high bandwidth requirements further increase channel energy consumption. However, channel energy consumption is dependent on data being transferred. Pseudo Open Drain (POD) asymmetric termination, used in current DDR4 systems, consumes energy only when 1’s are being transmitted over the channels. Many modern applications, including AI/ML ones are resilient to errors in data, and can work well with approximate data. This resilience can vary widely across and within applications, which provides a number of ways for exploiting these characteristics to save data transfer energy across the DRAM channel. However, all DRAM data encoding schemes have been targeted towards applications that require exact data and are not approximation resilient. In this paper, we propose Zero Aware Configurable Data Encoding by Skipping Transfer (ZAC-DEST), a data encoding scheme to reduce the energy consumption of DRAM channels, specifically targeted towards approximate computing and error resilient applications. ZAC-DEST exploits the similarity between recent data transfers across channels and information about error resilience behaviour of applications to reduce on-die termination and switching energy by reducing the number of 1’s transmitted over the channels. ZAC-DEST also provides a number of knobs for trading off application’s accuracy for energy savings, and vice versa, and can be applied to both training and inference. We apply ZAC-DEST to five machine learning applications. On average, across all applications and configurations, we observed a reduction of 40% in termination energy and 37% in switching energy as compared to the state of the art data encoding technique BD-Coder with an average output quality loss of 10%. We show that if both training and testing are done assuming the presence of ZAC-DEST, the output quality of the applications can be improved upto 9 × as compared to when ZAC-DEST is only applied during testing leading to energy savings during training and inference with increased output quality. [ABSTRACT FROM AUTHOR]

DRAM devices require periodic refresh operations to preserve data integrity. Slowing down the refresh rate can reduce the energy consumption; however, it may cause a loss of data stored in the DRAM cell. This paper proposes a new memory architecture of soft approximation for deep learning applications, which reduces the refresh energy consumption while maintaining accuracy and high performance. Utilizing the error-tolerant property of deep learning applications, the proposed memory architecture avoids the accuracy drop caused by data loss by flexibly controlling the refresh operation for different bits, depending on their criticality. For data storage, the approximate DRAM architecture reorganizes the data so that these data are sorted according to their bit significance. Critical bits are stored in more frequently refreshed devices while non-critical bits are stored in less frequently refreshed devices. In addition, for further reduction of the DRAM energy consumption, this paper combines hard approximation, which reduces the number of accesses to DRAM, with soft approximation. Simulation results show that the refresh energy consumption is reduced by 69.71%, and the total energy consumption is reduced by 26.0 % for the hybrid memory with a negligible drop in both training and testing phases on state-of-the-art deep networks. [ABSTRACT FROM AUTHOR]

This paper presents an energy-efficient spectral analysis method for the Internet of Things (IoT). The objective of this paper is to reduce the energy consumption of edge devices. The proposed method uses an autoregressive (AR) model for spectral analysis instead of the discrete Fourier transform, and its calculation process is distributed to the edge device and a base station by considering the energy consumption tradeoff of the data processing and the data communication. In this paper, the Yule–Walker method is employed for the AR coefficient calculation. The calculation process of Yule–Walker method can be divided into two parts: an autocorrelation calculation and an AR coefficient calculation. The autocorrelation calculation is implemented in the edge devices, and its dedicated hardware is designed using Verilog HDL. Meanwhile, the AR coefficient is calculated in the base station and is used for the spectral analysis. According to this distributed processing approach, the energy consumption of the edge device can be reduced compared with conventional DFT approaches using the fast Fourier transform (FFT). The system level energy consumption is evaluated assuming the IoT edge device, which has a wireless transceiver using Bluetooth low energy. The evaluation results show that the proposed method can reduce 79% of the edge device energy consumption for spectral analysis in a practical application. [ABSTRACT FROM AUTHOR]

This paper presents an integrated ultra-low-power (ULP) wireless sensor system-on-chip (SoC) that can be used for voltage sensing in both Internet of Things applications and bio-potential monitoring. In order to increase the energy efficiency of the analog front-end (AFE), we propose a noise and power efficient push-pull low noise instrumentation amplifier (LNIA) with a built-in ripple reduction loop based on capacitor reuse. A low-power ISM-band harmonic injection locked on-off-keying transmitter (OOK-TX) is also implemented for energy efficient wireless connectivity. Circuit implementations, design considerations, and detailed analysis are presented to improve the overall energy efficiencies of the SoC including the AFE, TX and, the power management unit. The proposed ULP-SoC is fabricated in 130 nm CMOS technology with a total area of 1.92 mm2. The total power consumption of the proposed system-on-chip is 197.1 μW which is one of the lowest among state-of-the-art wireless sensor SoC. [ABSTRACT FROM AUTHOR]

In emerging Spiking Neural Network (SNN) based neuromorphic hardware design, energy efficiency and on-line learning are attractive advantages mainly contributed by bio-inspired local learning with nonlinear dynamics and at the cost of associated hardware complexity. This paper presents a novel SNN design employing fast COordinate Rotation DIgital Computer (CORDIC) algorithm to achieve fast spike timing–dependent plasticity (STDP) learning with high hardware efficiency. In this study, a system design and evaluation method of CORDIC-based SNN is proposed for finding optimal CORDIC type and precision, from theoretical CORDIC-level error to application-level learning performance. From the proposed design and evaluation method, a reconfigurable SNN design based on fast-convergence CORDIC is designed to achieve high classification accuracy on MNIST, fast on-line learning and good energy efficiency. By utilizing SNN’s fault tolerance and time-division-multiplexing (TDM) strategy, the reconfigurable SNN design employs 8-bit fast-convergence CORDIC and TDM-based hardware accelerator for high efficiency. FPGA implementation results confirm that the proposed fast-convergence CORDIC SNN design outperforms the state-of-the-art CORDIC method by 38.5%−45.3% in terms of learning speed and energy efficiency, with the STDP learning of 30.2 ns/SOP, energy efficiency of 176.6 pJ/SOP, processing speed of 6.1 ms/image, and on-line learning convergence of 21.4 s (time to reach the final accuracy, on average), on MNIST benchmark. [ABSTRACT FROM AUTHOR]

Beyond 5G systems are expected to approach 1 Tb/s throughput. This poses a significant challenge to the channel decoder. In this paper, we propose a multi-core architecture based on full row parallel layered LDPC decoder with frame interleaving. Compared with conventional partially parallel layered architectures, the proposed architecture increases the throughput by applying frame interleaving into the pipeline architecture and by using multi-core architectures. Two high rate medium size QC LDPC codes are designed with fast decoding convergence speed for this architecture. Both codes are implemented with single core and multi-core architectures to explore different trade-offs between code design, communication performance and implementation. The four decoders are implemented in 16 nm CMOS FinFET technology with a clock rate of 1 GHz. The placement and routing implementation results show that the single core decoder for the LDPC (1027, 856) code is able to provide 114 Gb/s throughput at maximum 3 iterations with an area of 0.173 mm2 and energy efficiency of 1.56 pJ/bit; the multi-core decoder for the (1032, 860) code is able to provide 860 Gb/s throughput at maximum 2 iterations with an area of 1.48 mm2 and energy efficiency of 3.24 pJ/bit. The multi-core decoder achieves the highest throughput in the literature for medium size (1–2k) LDPC codes. When compared with other state-of-the-art fully parallel high speed architectures, the proposed architectures bring a significant gain both in area efficiency and energy efficiency while keeping the ability to offer flexibility in code rate, number of iterations and early stop. [ABSTRACT FROM AUTHOR]

Integrating Deep Neural Networks (DNNs) into the Internet of Thing (IoT) devices could result in the emergence of complex sensing and recognition tasks that support a new era of human interactions with surrounding environments. However, DNNs are power-hungry, performing billions of computations in terms of one inference. Spatial DNN accelerators in principle can support computation-pruning techniques compared to other common architectures such as systolic arrays. Energy-efficient DNN accelerators skip bit-wise or word-wise sparsity in the input feature maps (ifmaps) and filter weights which means ineffectual computations are skipped. However, there is still room for pruning the effectual computations without reducing the accuracy of DNNs. In this paper, we propose a novel real-time architecture and dataflow by decomposing multiplications down to the bit level and pruning identical computations in spatial designs while running benchmark networks. The proposed architecture prunes identical computations by identifying identical bit values available in both ifmaps and filter weights without changing the accuracy of benchmark networks. When compared to the reference design, our proposed design achieves an average per layer speedup of ×1.4 and an energy efficiency of × 1.21 per inference while maintaining the accuracy of benchmark networks. [ABSTRACT FROM AUTHOR]

Training of Deep Neural Network (DNN) at the edge faces the challenge of high energy consumption due to the requirements of a large number of memory accesses for gradient calculations. Therefore, it is necessary to minimize data fetches to perform training of a DNN model on the edge. In this paper, a novel technique has been proposed to reduce the memory access for the training of fully connected layers in transfer learning. By analyzing the memory access patterns in the backpropagation phase in fully connected layers, the memory access can be optimized. We introduce a new method to update the weights by introducing the delta term for every node of output and fully connected layer. Delta term aims to reduce memory access for the parameters which are required to access repeatedly during the training process of fully connected layers. The proposed technique shows 0.13x-13.93x energy savings for the training of fully connected layers for famous DNN architectures on multiple processor architectures. The proposed technique can be used to perform transfer learning on-chip to reduce energy consumption as well as memory access. [ABSTRACT FROM AUTHOR]

This paper proposes a high-performance bidirectional architecture for the quasi-comparison-free sorting algorithm. Our architecture improves the performance of the conventional unidirectional architecture by reducing the total number of sorting cycles via bidirectional sorting along with two auxiliary methods. Bidirectional sorting allows the sorting tasks to be conducted concurrently in the high- and low-index parts of our architecture. The first auxiliary method is boundary finding, which shortens the range for index searching by finding the boundaries of the range. The second auxiliary method is queue storing, which stores each useful index in a queue in advance to reduce the number of miss cycles during index searching. The performance of our architecture highly depends on the distribution of input data. For each set of input data to be sorted, five Gaussian distributions of the input data and four standard derivations for each distribution were adopted in our experiments. The results show that at the expense of some additional area cost, the number of sorting cycles and the energy consumption are significantly reduced by our method. [ABSTRACT FROM AUTHOR]

This paper presents fully-integrated analog neural network classifier architecture for low resolution image classification that eliminates memory access. We design custom activation functions using single-stage common-source amplifiers, and apply a hardware-software co-design methodology to incorporate knowledge of the custom activation functions into the training phase to achieve high accuracy. Performing all computations entirely in the analog domain eliminates energy cost associated with memory access and data movement. We demonstrate our classifier on multinomial classification task of recognizing down-sampled handwritten digits from MNIST dataset. Fabricated in 65nm CMOS process, the measured energy consumption for down-sampled MNIST dataset is 173pJ/classification, which is $3\times $ better than state-of-the-art. The prototype IC achieves mean classification accuracy of 81.3% even after down-sampling the original MNIST images by 96% from $28\times 28$ pixels to $5\times 5$ pixels. [ABSTRACT FROM AUTHOR]

A design methodology for making magnetic coupling wireless power transfer (MC-WPT) systems resonate at given frequencies is proposed in this paper. Proposing such a method not only can improve the energy efficiency against frequency splitting, but also can satisfy the requirement of frequency standards and then promote its industrial application. The method is based on eigenvalues configuration and comprehensively considers all electric parameters. The quantitative relationship between the performance indexes and electric parameters is derived, based on which the expected electric parameters can be quickly and accurately determined. The design theory and design flow are provided, and practical systems are designed, simulated and measured to show the effectiveness of the proposed method. The results show that the proposed method can: 1) eliminate frequency splitting to make MC-WPT systems resonate at a given frequency; 2) utilize frequency splitting to make MC-WPT systems resonate at two given frequencies. [ABSTRACT FROM AUTHOR]

This paper reports a half-rate bang-bang clock and data recovery (BBCDR) circuit supporting the trimodal (NRZ/PAM-4/PAM-8) operation. The observation of their crossover- points distribution at the transitions introduces the single-loop phase tracking technique. In addition, low-power techniques at both the architecture and circuit levels are employed to greatly improve the overall energy efficiency and multiply data throughput by increasing the number of levels on the magnitude. Fabricated in 28-nm CMOS, our BBCDR prototype scores a 0.29/0.17/0.14 pJ/bit efficiency at 14.4/28.8/43.2 Gb/s under NRZ/PAM-4/PAM-8 modes, respectively. The jitter is < 0.53 ps (integrated from 100 Hz to 1 GHz) with approximately-equivalent constant loop bandwidth, and we achieve at least 1-UIpp jitter tolerance up to 10 MHz for all the three modes. [ABSTRACT FROM AUTHOR]

Convolution is one of the key operations in signal processing and machine learning applications. In this paper, we propose a novel convolution computing paradigm based on the NOR Flash array (NFA) that is capable of executing the 2-D convolution computing in one clock cycle. In order to demonstrate the feasibility and efficiency of the proposed convolution computing paradigm, the feature extraction task on an image with the size of $20\times 20$ is executed using the NFA structure. We also prove the NOR Flash-driven convolution computing is capable of processing the image with a larger size. This paper presents a new approach to realize convolution computing with high speed and energy efficiency for the signal processing and convolution neural network. [ABSTRACT FROM AUTHOR]

This paper is concerned with global stabilization of discrete-time multiple integrators with bounded controls by utilizing the energy function based approach. First, a discrete-time double integrators system subject to input additive disturbances is stabilized by a bounded feedback whose formula involves a linear function of the state and a saturation function only. Next, this result is used to stabilize a discrete-time multiple integrators of arbitrary length by bounded control with the aid of a special canonical form. Compared with the existing results, the proposed controllers require fewer saturation functions, which allow a better use of the control energy. Moreover, some free parameters that are introduced into these controllers can help improve the transient performance of the closed-loop systems significantly. Finally, a numerical example is given to illustrate the effectiveness of the proposed method. [ABSTRACT FROM AUTHOR]

In this paper, an asynchronous design for soft error detection and correction in combinational and sequential circuits is presented. The proposed architecture is called Asynchronous Full Error Detection and Correction (AFEDC). A custom design flow with integrated commercial EDA tools generates the AFEDC using the asynchronous bundled-data design style. The AFEDC relies on an Error Detection Circuit (EDC) for protecting the combinational logic and fault-tolerant latches for protecting the sequential logic. The EDC can be implemented using different detection methods. For this work, two boundary variants are considered, the Full Duplication with Comparison (FDC) and the Partial Duplication with Parity Prediction (PDPP). The AFEDC architecture can handle single events and timing faults of arbitrarily long duration as well as the synchronous FEDC, but additionally can address known metastability issues of the FEDC and other similar synchronous architectures and provide a more practical solution for handling the error recovery process. Two case studies are developed, a carry look-ahead adder and a pipelined non-restoring array divider. Results show that the AFEDC provides equivalent fault coverage when compared to the FEDC while reducing area, ranging from 9.6% to 17.6%, and increasing energy efficiency, which can be up to 6.5%. [ABSTRACT FROM AUTHOR]

Aiming at reducing the complexity of minimum mean square error (MMSE) detection in massive multiple-input multiple-output (MIMO) systems, this paper proposes a detection algorithm with high convergence rate and an efficient hardware architecture based on second-order Richardson iteration (SORI). In the proposed algorithm, a pre-iteration-based initialization method is presented to accelerate the convergence without extra complexity. In addition, the approximation of relaxation factor and the log-likelihood ratio (LLR) is introduced to further reduce computing load. Theoretical analysis demonstrates the advantages of the proposed algorithm in fast convergence and low complexity, and simulation results show that the proposed algorithm can efficiently approach MMSE performance. Based on this algorithm, a flexible hardware architecture is designed, which is deeply pipelined to support $128\times U$ ($8\leq U\leq 32$) massive MIMO detection with the configurable number of iterations, and a folded dual-mode systolic array (DMSA) is fully utilized to achieve the flexibility with low hardware consumption. Implemented on Xilinx Virtex-7 FPGA and SMIC 40nm CMOS technology, the proposed detector is competitive in terms of energy and area efficiency compared to state-of-the-art iterative detectors, and it can adapt to the varied channel condition and the number of users in massive MIMO systems. [ABSTRACT FROM AUTHOR]

Single constant multiplication (SCM) and multiple constant multiplications (MCM) are among the most popular schemes used for low-complexity shift-add implementation of finite impulse response (FIR) filters. While SCM is used in the direct form realization of FIR filters, MCM is used in the transposed direct form structures. Very often, the hybrid form FIR filters where the sub-sections are implemented by fixed-size MCM blocks provide better area, time, and power efficiency than those of traditional MCM and SCM based implementations. To have an efficient hybrid form filter, in this paper, we have performed a detailed complexity analysis in terms of the hardware and time consumed by the hybrid form structures. We find that the existing hybrid form structures lead to an undesirable increase of complexity in the structural-adder block. Therefore, to have a more efficient implementation, a variable size partitioning approach is proposed in this paper. It is shown that the proposed approach consumes less area and provides nearly 11% reduction of critical path delay, 40% reduction of power consumption, 15% reduction of area-delay product, 52% reduction of energy-delay product, and 42% reduction of power-area product, on an average, over the state-of-the-art methods. [ABSTRACT FROM AUTHOR]

This paper introduces a fully integrated switched-capacitor (SC) DC-DC boost converter suitable for energy harvesting in miniature sensor systems with varying harvesting source and battery voltages. Unlike the conventional SC DC-DC boost converters, where efficiency is optimized only for a few topology-dependent voltage conversion ratios (VCRs), the proposed soft-charging-based SC converter achieved VCR-insensitive evenly high efficiency for a wide range of input and output voltages. A test chip was fabricated in 180nm process, and the measured peak efficiency and average efficiency are 85.3% and 83.6%, respectively, for 1.8V regulated output voltage and a wide range of input voltages (0.95–1.8 V). Also, a peak efficiency of 88.9% and average efficiency of 87.6% are achieved for a wide range of battery output voltages (3.0–4.2 V) and a 1.2V harvesting source input voltage. [ABSTRACT FROM AUTHOR]

Quantization is a promising technique to compress the size of Convolutional Neural Network (CNN) models. Recently, various precision-scalable designs have been presented to reduce the computational complexity in CNNs. However, most of them adopt straightforward calculation scheme to implement the CNN, which causes high bandwidth requirement and low hardware utilization efficiency. This paper proposes a new precision-scalable architecture which can fully reduce the computational complexity in CNN inference and meanwhile has a finely simplified calculation scheme. Based on the proposed scheme, a well-optimized multiplier called Compositional Processing Element (C-PE) is devised. Compared with the previous multipliers, the new C-PE requires less area and power. Furthermore, two levels of optimization are introduced to the design to relieve the bandwidth problem and increase the hardware utilization efficiency. Implemented under the TSMC 90nm CMOS technology, the whole design achieves 6-68.1 fps in various precisions on VGG16 benchmark and a 49.8TOPS/W energy efficiency at 500MHz when scaled to 28nm, which is much better than previous precision-scalable ones. [ABSTRACT FROM AUTHOR]

In this paper, we propose O4-DNN, a high-performance FPGA-based architecture for convolutional neural network (CNN) accelerators relying on operation packing and out-of-order (OoO) execution for DSP blocks augmented with LUT-based glue logic. The high-level architecture is comprised of a systolic array of processing elements (PEs), supporting output stationary dataflow. In this architecture, the computational unit of each PE is realized by using a DSP block as well as a small number of LUTs. Given the limited number of DSP blocks in FPGAs, the combination (DSP block and some LUTs) provides more computational power obtainable through each DSP block. The proposed computational unit performs eight convolutional operations on five input operands where one of them is an 8-bit weight and the others are four 8-bit input feature (IF) maps. In addition, to improve the energy efficiency of the proposed computational unit, we present an approximate form of the unit suitable for neural network applications. To reduce the memory bandwidth as well as increase the utilization of the computational units, a data reusing technique based on the weight sharing is also presented. To improve the performance of the proposed computational unit further, an addressing approach for computing the partial sums out-of-order is proposed. The efficacy of the architecture is assessed using two FPGA devices executing four state-of-the-art neural networks. Experimental results show that this architecture leads to, on average (up to), $2.5\times $ ($3.44\times$) higher throughput compared to a baseline structure. In addition, on average (maximum of), 12% (40%) energy efficiency improvement is achievable by employing the O4-DNN compared to the baseline structure. [ABSTRACT FROM AUTHOR]

‘In-memory computing’ is being widely explored as a novel computing paradigm to mitigate the well known memory bottleneck. This emerging paradigm aims at embedding some aspects of computations inside the memory array, thereby avoiding frequent and expensive movement of data between the compute unit and the storage memory. In-memory computing with respect to Silicon memories has been widely explored on various memory bit-cells. Embedding computation inside the 6 transistor (6T) SRAM array is of special interest since it is the most widely used on-chip memory. In this paper, we present a novel in-memory multiplication followed by accumulation operation capable of performing parallel dot products within 6T SRAM without any changes to the standard bitcell. We, further, study the effect of circuit non-idealities and process variations on the accuracy of the LeNet-5 and VGG neural network architectures against the MNIST and CIFAR-10 datasets, respectively. The proposed in-memory dot-product mechanism achieves 88.8% and 99% accuracy for the CIFAR-10 and MNIST, respectively. Compared to the standard von Neumann system, the proposed system is $6.24\times $ better in energy consumption and $9.42\times $ better in delay. [ABSTRACT FROM AUTHOR]

This paper presents a one-sided Schmitt-trigger-based 9T static random access memory cell with low energy consumption and high read stability, write ability, and hold stability yields in a bit-interleaving structure without write-back scheme. The proposed Schmitt-trigger-based 9T static random access memory cell obtains a high read stability yield by using a one-sided Schmitt-trigger inverter with a single bit-line structure. In addition, the write ability yield is improved by applying selective power gating and a Schmitt-trigger inverter write assist technique that controls the trip voltage of the Schmitt-trigger inverter. The proposed Schmitt-trigger-based 9T static random access memory cell has 0.79, 0.77, and 0.79 times the area, and consumes 0.31, 0.68, and 0.90 times the energy of Chang’s 10T, the Schmitt-trigger-based 10T, and MH’s 9T static random access memory cells, respectively, based on 22-nm FinFET technology. [ABSTRACT FROM AUTHOR]

The AOMedia Video 1 (AV1), developed by the Alliance for Open Media (AOM) industry consortium and released in 2018, is an open-source and royalty-free video coding format. The goal of AV1 is to deliver substantial compression gains over state-of-the-art codecs such as VP9 and HEVC, while keeping a practical decoding complexity, hardware feasibility and an open and free status. This paper presents a high-throughput hardware architecture for ten non-directional AV1 intra prediction modes: DC, Paeth, Smooth, Smooth Vertical, Smooth Horizontal and Recursive-filtering-based modes 0 to 4. The proposed design supports all 19 block sizes specified by AV1 and, in the worst-case scenario, is able to process every single combination of these blocks according to the 10-way partition tree, with a throughput of UHD 4K videos at up to 30 frames per second. When synthesized to the TSMC 40nm cell library targeting a frequency of 648MHz, the proposed design used 128.5K gates and showed a power dissipation and energy consumption per predicted sample of 65.5mW and 0.176pJ/sample, respectively. No other works were found in the literature describing hardware designs for AV1 intra prediction encoder. [ABSTRACT FROM AUTHOR]

Hardware acceleration substantially enhances both energy efficiency and performance, but raises major challenges for programmability. This is especially true in the domain of approximate computing, where energy-approximation tradeoffs at the hardware level are extremely difficult to encapsulate in interfaces to the software level. The programmability challenges have motivated co-design of accelerators with program-synthesis frameworks, where the structured computations resulting from synthesis are exploited towards hardware specialization. This paper proposes energy-aware code synthesis targeting heterogeneous architectures for approximate computing. A heterogeneous architecture for embedded sensor inference is employed, demonstrated in custom silicon, where programmable feature extraction is mapped to an accelerator via genetic programming. The high level of accelerator specialization and structured mapping of computations to the accelerator enable robust energy models, which are then employed in a genetic-programming algorithm to improve the energy-approximation Pareto frontier. The proposed algorithm is demonstrated in an electroencephalogram-based seizure-detection application and an electrocardiogram-based arrhythmia-detection application. At the same level of baseline inference performance, the energy consumption of genetic-programming models executed on the accelerator is 57.4% and 21.8% lower, respectively, with the proposed algorithm, compared to a conventional algorithm without incorporating energy models for execution on the accelerator. [ABSTRACT FROM AUTHOR]

Deep Neural Networks (DNNs) have recently shown state-of-the-art results on various applications, such as computer vision and recognition tasks. DNN inference engines can be implemented in hardware with high energy efficiency as the computation can be realized using a low-precision fixed point or even binary precision with sufficient cognition accuracies. On the other hand, training DNNs using the well-known back-propagation algorithm requires high-precision floating-point computations on a CPU and/or GPU causing significant power dissipation (more than hundreds of kW) and long training time (several days or more). In this paper, we demonstrate a training chip fabricated using a commercial 65-nm CMOS technology for machine learning. The chip performs training without back propagation by using invertible logic with stochastic computing that can directly obtain weight values using input/output training data with low precision suitable for inference. When training neurons that compute the weighted sum of all inputs and then apply a non-linear activation function, our chip demonstrates a reduction of power dissipation and latency by 99.98% and 99.95%, respectively, in comparison with a state-of-the-art software implementation. [ABSTRACT FROM AUTHOR]

This paper presents a high-voltage ultra-wideband (UWB) pulse generator. Shifting the UWB signal synthesis to digital domain using two consecutive trapezoidal waves, the design requires only a simple low-loss passive filter to conform to UWB spectrum regulations. A power amplifier stage is added to boost the output voltage above the supply voltage limit. To verify the design concept, the generator is designed in TSMC CMOS 65 nm technology with 1 V supply voltage occupying a die size of 0.34 mm2. The UWB generator achieves an efficiency of 21% while the total energy consumption is 11.67 pJ/pulse. On-chip measurement is in good agreement with the simulation and shows output peak-to-peak amplitude of 2.12 V after adding the cable and probe losses to the measured 1.62 V peak-to-peak voltage. The proposed UWB signal generator achieves the highest peak-to-peak to supply voltage ratio among the UWB signal generators reported to the date. [ABSTRACT FROM AUTHOR]