Abstract Superconducting neural networks hold significant potential for future applications such as natural language processing and image recognition. To this end, we propose a binary neural computing unit implemented using a hybrid circuit of cryogenic CMOS and superconducting technologies. It offers two main advantages: firstly, we utilize current-mode computations for neural unit weight calculations, significantly reducing the unit’s footprint and enabling the potential for higher integration in the future. Secondly, all computations are performed in a low-temperature environment, which implies the possibility of on-chip learning in superconducting neural networks and the potential for achieving faster training rates in the future. We fabricated the chip using Nb 1 kA cm−2 process (1KP) technology and experimentally verified the correctness of the circuit logic. The margins for various control parameters of the circuit are approximately around 30%, and the superconducting circuit power consumption is estimated to be around 4 microwatts.
Abstract Superconducting neural networks hold significant potential for future applications such as natural language processing and image recognition. To this end, we propose a binary neural computing unit implemented using a hybrid circuit of cryogenic CMOS and superconducting technologies. It offers two main advantages: firstly, we utilize current-mode computations for neural unit weight calculations, significantly reducing the unit’s footprint and enabling the potential for higher integration in the future. Secondly, all computations are performed in a low-temperature environment, which implies the possibility of on-chip learning in superconducting neural networks and the potential for achieving faster training rates in the future. We fabricated the chip using Nb 1 kA cm−2 process (1KP) technology and experimentally verified the correctness of the circuit logic. The margins for various control parameters of the circuit are approximately around 30%, and the superconducting circuit power consumption is estimated to be around 4 microwatts.
Abstract With the increasing processor performance requirements of convolutional neural networks, we believe that research on superconducting neural networks based on single flux quantum (SFQ) circuits holds significant potential. However, the typical multiply and accumulate operations used in neural networks often lead to a significant increase in circuit area due to the limitations of the integration density of the current SFQ circuit fabrication process. In this work, we propose a binary neural processing element (PE) that utilizes time-domain signal expression. Time-domain signals express multi-bit binary data in time information, which reduces the number of data lines to decrease the circuit area. In the proposed binary neural PE, multiplication is achieved through exclusive-NOR gates and addition is realized by controlling the delay of time-domain signals. We designed and simulated a 3 × 3 convolution circuit using 10 Nb layer process with a critical current density of 10 kA cm−2. Compared to the convolution circuit implemented using conventional SFQ logic, the proposed circuit reduces the number of Josephson junctions by approximately 56% and decreases the area by about 60%.
Abstract : Lack of high-density, fast and memory has been a long-standing problem in superconducting digital electronics. Alternative Josephson-junction-based memory cells and peripheral circuits have intrinsic problems which impede the application of that kind of memory. The CMOS-Josephson hybrid memory idea was proposed in 1992 to circumvent this problem. In the present work we have designed and demonstrated a 64-kb CMOS-Josephson hybrid memory working at 1 GHz. In order to make the hybrid idea practical, a complete and comprehensive 4 K short-channel CMOS model for digital circuits was developed for this dissertation. It has been used in the design of the memory core and peripheral circuits. A fast hybrid interface circuit that transforms single-flux-quantum millivolt 2-ps pulse signals to CMOS volt-level signals has been studied and the optimized design has been completed. Using a nonvolatile cryogenic 3-T DRAM cell, the memory core is designed in such a way that the access time is much less than for room-temperature operation, not only because of the performance upgrade due to the temperature decrease but also because of a different reading mechanism. Fabricated by commercial submicron CMOS processes (0.25 microns and 0.18 microns) and a Nb/AlOx/Nb Josephson-junction process with 2.5 kA/cm2 tunneling current density, the memory chips were bonded together by direct wire-bonding and by flip-chip bump-bonding and tested at low frequencies, high frequencies, respectively. A subnanosecond access time is obtained both from simulations and experiments. The power for an interface circuit is measured to be 0.6 mW. The memory core consumes even less power. The total power consumption depends on the operation mode of the memory and is calculated to be 10 mW for reading and 28 mW for writing. Simulations also indicate that, with more advanced semiconductor and superconductor technologies, larger and faster hybrid memories are expected to be achievable in the future.
Abstract Single-flux-quantum (SFQ) circuits are promising candidates for post-Moore integrated circuits owing to their ultra-high-speed operation and low power consumption. However, limitations in current superconducting fabrication processes pose significant challenges to achieving area-efficient SFQ-based microprocessors, particularly in the design of compact arithmetic logic units (ALUs). Unary computing enables arithmetic operations using extremely simple logic, but typically suffers from long computation latency. This paper focuses on unary computing for superconducting circuits that exploits the complementary characteristics of SFQ logic and unary computing. The ultra-high-speed nature of SFQ circuits compensates for the inherent latency of unary computation, while unary computing enables an ultra-low hardware cost by substantially reducing circuit complexity. Based on this concept, we introduce a unary computing methodology tailored for SFQ circuits and design a superconducting unary ALU with ultra-low hardware cost. The proposed ALU requires only 183 Josephson junctions (JJs), achieving approximately 88.2% reduction in JJ count and 80.4% reduction in area compared with a conventional bit-serial SFQ ALU. The proposed design, including binary-to-unary and unary-to-binary converters, was fabricated using a 10 kA cm −2 Nb four-layer superconducting circuit fabrication process and experimentally verified to operate correctly at approximately 72 GHz. The total power consumption at this frequency is approximately 58.9 µ W. This work quantitatively demonstrates the effectiveness of combining unary computing with SFQ circuits for ultra-low-cost processing systems and provides superconducting circuits as a new design option for alternative computation paradigms, such as unary computing.
Most popular superconducting circuits operate on information carried by ps-wide, μV-tall, single flux quantum (SFQ) pulses. These circuits can operate at frequencies of hundreds of GHz with orders of magnitude lower switching energy than complementary-metal-oxide-semiconductors (CMOS). However, under the stringent area constraints of modern superconductor technologies, fully-fledged, CMOS-inspired superconducting architectures cannot be fabricated at large scales. Unary SFQ (U-SFQ) is an alternative computing paradigm that can address these area constraints. In U-SFQ, information is mapped to a combination of streams of SFQ pulses and in the temporal domain. In this work, we extend U-SFQ to introduce novel building blocks such as a multiplier and an accumulator. These blocks reduce area and power consumption by 2 × and 4 × compared with previously proposed U-SFQ building blocks and yield at least 97% area savings compared with binary approaches. Using these multiplier and adder, we propose a U-SFQ Convolutional Neural Network (CNN) hardware accelerator capable of comparable peak performance with state-of-the-art superconducting binary approach (B-SFQ) in 32 × less area. CNNs can operate with 5–8 bits of resolution with no significant degradation in classification accuracy. For 5 bits of resolution, our proposed accelerator yields 5 × to 63 × better performance than CMOS and 15 × to 173 × better area efficiency than B-SFQ.
This paper describes our programme to develop and demonstrate ultra-high performance single flux quantum (SFQ) VLSI technology that will enable superconducting digital processors for petaFLOPS-scale computing. In the hybrid technology, multi-threaded architecture, the computational engine to power a petaFLOPS machine at affordable power will consist of 4096 SFQ multi-chip processors, with 50 to 100 GHz clock frequency and associated cryogenic RAM. We present the superconducting technology requirements, progress to date and our plan to meet these requirements. We improved SFQ Nb VLSI by two generations, to a 8 kA cm −2 , 1.25 µm junction process, incorporated new CAD tools into our methodology, demonstrated methods for recycling the bias current and data communication at speeds up to 60 Gb s −1 , both on and between chips through passive transmission lines. FLUX-1 is the most ambitious project implemented in SFQ technology to date, a prototype general-purpose 8 bit microprocessor chip. We are testing the FLUX-1 chip (5K gates, 20 GHz clock) and designing a 32 bit floating-point SFQ multiplier with vector-register memory. We report correct operation of the complete stripline-connected gate library with large bias margins, as well as several larger functional units used in FLUX-1. The next stage will be an SFQ multi-processor machine. Important challenges include further reducing chip supply current and on-chip power dissipation, developing at least 64 kbit, sub-nanosecond cryogenic RAM chips, developing thermally and electrically efficient high data rate cryogenic-to-ambient input/output technology and improving Nb VLSI to increase gate density.
Inspired by recent interest in quantum computing and recent studies of cryo CMOS for control electronics, this paper presents a hybrid semiconductor-superconductor approach for engineering scalable computing systems that operate across the gradient between room temperature and the temperature of a cryogenic payload. Such a hybrid computer architecture would have unique suitability to quantum computers, scalable sensors, and the quantum internet. The approach is enabled by Cryogenic Adiabatic Transistor Circuits (CATCs), a novel way of using adiabatic circuits to substantially reduce cooling requirements. In a hybrid chip of CATCs and a second technology, such as Josephson junctions (JJs) or cryo CMOS, the CATCs complement the speed, power, and density of the second technology as well as becoming a longsought cryogenic memory. This paper describes higher-level design principles for CATC hybrids with a quantum computer control system that includes CATC memory, an FPGA-like logic module that uses CATC for dense configuration logic and JJs for fast configured logic, and I/O subsystems including microwave modulators and low frequency control signals.