Xinchao Shang

Xinchao Shang

Auto-generated profile, based on publicly available information

Xinchao Shang

National ASIC System Engineering Research Center, School of Electronic Science & engineering, Southeast University, Nanjing, China, 210096

Research Focus & Metrics

Academic Career Length

10Years (2017-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
28 March 2019 in IEEE Transactions on Circuits and Systems I: Regular Papers
Verified
Journal Article
Closed access
WSWeiwei Shan
Weiwei Shan
XSXinchao Shang
Xinchao Shang
XWXing Wan
Xing Wan
HCHao Cai
Hao Cai
CZChuan Zhang
Chuan Zhang
LHLuchi Hua
Luchi Hua
To reduce conservative timing margin in digital circuit designs, adaptive techniques based on timing-error detection were proposed to monitor the timing of selected critical paths. However, most of them can respond only to the detected error signals in the next, or a few, clock cycles. Then too, they usually have a large area overhead. In this paper, a low-overhead, 9-transistor transition-detector (TD) is proposed for use at a wide voltage range down to near-threshold voltage. These TDs are inserted at the half-path points of the critical paths as opposed to the conventional endpoints, so that the timing errors in the current clock cycle can be detected in advance to prevent actual timing errors at the endpoints. A half-path insertion point selection method is proposed here, with the middle points merging to reduce the number of inserted TDs. Test chips employing the proposed design approach are fabricated in a 40-nm CMOS process. Silicon measurements demonstrate that the whole design has achieved a 41.2% to 50.5% energy saving at near-threshold voltage as compared to conventional margined design at 0.56 V, and with a small area overhead. Thus, it is an effective method for a low-power and energy-efficient circuit design.

22 September 2020 in IEEE Transactions on Circuits and Systems I: Regular Papers
Verified
Journal Article
Closed access
XSXinchao Shang
Xinchao Shang
MLMinyi Lu
Minyi Lu
CWChengjun Wu
Chengjun Wu
YXYiming Xiang
Yiming Xiang
JXJiaming Xu
Jiaming Xu
WSWeiwei Shan
Weiwei Shan
Excessive timing margins are usually added in the wide-voltage-range design due to process, voltage and temperature (PVT) variations, which can be eliminated by adaptive voltage scaling (AVS). Some traditional Razor-based designs replace endpoint flip-flops with latches and ensure data sampled correctly. The duration of time-borrowing and the short path problem make the minimum-delay constraint worthy of consideration. To overcome the minimum-delay problem in the latch-based error detection and correction (EDAC) techniques, we propose a solution using a pulsed-latch and transition detector (TD). This method utilizes time-borrowing characteristics of the latch to ensure the correct function. To detect timing violations and minimize area overhead, we design a 15-transistor transition detector which is able to operate at a wide voltage range, from near-threshold voltage (NTV) to super-threshold voltage (STV). To minimize the overhead of pulse generators, a physical allocation-aware pulse generator insertion algorithm is presented to identify each desired group of a pulse generator and a pulsed-latch group. The proposed scheme is implemented in an 8-bit AES circuit through an automatic insertion flow and fabricated in a 28nm CMOS process. Chip measurements demonstrate that the whole design achieves up to 64.3% energy saving as compared to the conventional worst-case design at a small price of 4.3% area overhead.

22 August 2018 in IEEE Transactions on Circuits and Systems I: Regular Papers
Verified
Journal Article
Closed access
WDWentao Dai
Wentao Dai
WSWeiwei Shan
Weiwei Shan
XSXinchao Shang
Xinchao Shang
XLXinning Liu
Xinning Liu
HCHao Cai
Hao Cai
LHLuchi Hua
Luchi Hua
To mitigate the excess timing margin in conventional digital integrated circuit design, resilient circuits were proposed. The current techniques however either have a large error-detecting register or have difficulty in wide-voltage-range (down to near-threshold) operation. In this paper, we propose a light-weight holosymmetrical transition detector (HTD) that can work reliably at near-threshold to normal VDD with fast response time. Combined with a conventional latch, the HTD-latch is used to replace the end point flip-flop for timing error prediction with an area overhead of only 9.5% over a conventional flip-flop. Besides, we optimize the speculation window and design an adaptive duty-cycle control mechanism to reduce the short-path padding overhead in a resilient circuit. In addition, with voltage scaling, we keep the system working at the point before the first failure by using local detection and global clock gating, which needs neither architectural circuit change nor complicated error recovery mechanism. We apply the above-mentioned three techniques on an eighth-order filter test chip in a 40-nm CMOS process. The measurement results demonstrate that the proposed HTD can work robustly at 0.474-1.1 V. With 4.37% area overhead in total, the proposed resilient technique improves performance by 179.31%/28.2% and energy efficiency by 45.6%/28.1% in a near-V TH (0.474 V)/super-V TH (1.1 V) operation, as compared with the baseline design. Therefore, our proposed HTD-based resilient technique can improve the energy efficiency of circuits by almost eliminating all the timing margins.

14 June 2021 in IEEE Transactions on Circuits and Systems II: Express Briefs
Verified
Journal Article
Closed access
ZCZhikuang Cai
Zhikuang Cai
BCBoyang Cheng
Boyang Cheng
YDYuxuan Du
Yuxuan Du
XSXinchao Shang
Xinchao Shang
WSWeiwei Shan
Weiwei Shan
Due to the increasing demand of high energy-efficient processor for deep neural networks, traditional neural network engines with high-precision weights and activations that usually occupies huge on/off-chip resources with large power consumption are no longer suitable for internet-of-things applications. Binary neural networks (BNNs) reduce memory size and computation complexity, achieving drastically increased energy efficiency. In this paper, an energy-efficient time-domain binary neural network engine is optimized for image recognition, with time-domain accumulation (TD-MAC), timing error detection based adaptive voltage scaling design and the related approximate computing. The proposed key features are: 1) an error-tolerant adaptive voltage scaling system with TD-MAC chain truncation for aggressive power reduction, working from near-threshold to normal voltage; 2) architectural parallelism and data reuse with 100% TD-MAC utilization; 3) low power TD-MAC based on analog delay lines. Fabricated in a 28nm CMOS process, the whole system achieves a maximum 51.5TOPS/W energy efficiency at 0.42V and 25MHz, with 99.6% accuracy on MNIST dataset. When the length of the TD-MAC chain is truncated by configuration, with a 90% accuracy on MNIST and a 150MHz, the proposed BNN achieves a power-saving of 13.2% and a further energy efficiency increasing of 67.6%.

14 November 2017 in IEEE Transactions on Circuits and Systems II: Express Briefs
Verified
Journal Article
Closed access
WSWeiwei Shan
Weiwei Shan
LWLiang Wan
Liang Wan
XLXinning Liu
Xinning Liu
XSXinchao Shang
Xinchao Shang
WDWentao Dai
Wentao Dai
LSLongxing Shi
Longxing Shi
Adaptive voltage frequency scaling (AVFS) techniques based on in-situ timing monitoring can mitigate the excessive timing margin caused by process, voltage and temperature (PVT) variations. However, they usually adjust the frequency by clock gating, clock half-division, or PLL configuration, which cause large performance loss or need a long lock time. Thus, a compact structured adaptive clock stretching circuit is proposed here for very rapid frequency adjustment. It is very useful for AVFS system by stretching the clock cycle with only within-a-cycle response time. The proposed architecture generates multi-phase clocks through a series of delay elements and continuously picks an appropriate phase clock to form a stretched clock. It also uses a process voltage temperature monitor to adjust the phase clock selection and makes this circuit suitable under PVT variations. Fabricated in 40nm CMOS process, it only occupies a core area of 85*83μ². Measurements show that it can stretch the clock cycle immediately upon the stretch signal with within-a-cycle response time. It has a wide voltage range from 0.43V to 1.1V, with power consumptions of 38.2μW at 0.43V/10MHz and 4.55mW at 1.1V/1.2GHz. Plus, it can provide configurable stretch amounts for the AVFS system. Thus, it can provide a within-a-cycle, low overhead frequency adjustment solution for AVFS systems.

1 May 2019 in Chinese Journal of Electronics
Verified
Journal Article
Open access
PDF file
XSXinchao Shang
Xinchao Shang
WSWeiwei Shan
Weiwei Shan
YXYiming Xiang
Yiming Xiang
LWLiang Wan
Liang Wan
AFAo Fan
Ao Fan
Variation in on-chip power supply continues to be a major challenge to limit circuit performance. To mitigate the impact of high-frequency voltage droop, a novel all-digital adaptive clock system is proposed. The adaptive clock system is composed of a droop detector and an adaptive clock generator, which directly stretch the clock period when the circuit suffers from the supply voltage droop. The droop detector circuit detects the VDD droop and the digital adaptive clock generator circuit selects the stretched clock to prevent timing-margin failures. The response time for droop detection and clock stretch can be as fast as one cycle. The whole scheme is used on a test circuit under SMIC 40nm CMOS process with a layout area of 900um*1100um. Postlayout simulation results demonstrate power reductions are 7% at 1.1V and 15% at 0.7V for a 10% VDD droop, respectively.