A range reduction method for shift-and-add algorithms for exponential functions is proposed in this paper. An exponential function accelerator with this method and radix-16 shift-and-add algorithm has been implemented in SMIC 55 nm CMOS process. Compared with the existing method, the proposed method reduces the latency (cycles) by 33% and 20% for 16 and 32-bit precision results, respectively; thereby increasing the throughput to 50M exp/s and reducing the energy consumption to 4.6pJ/exp. In addition, this method saves die area since no arithmetic units are adopted. This exponential accelerator is supposed to be used in a neuromorphic chip for spiking neural network modeling.
We report on the modulation responses of an injection-locked vertical-cavity surface-emitting laser (VCSEL) measured using an optical spin-polarization modulation technique. We reveal that weak injection locking is suitable to maintain a strong resonance response at a high-frequency region depending on the linear birefringence of the VCSEL, a unique feature of spin-polarization modulation response. The measured results can be reasonably reproduced with simulated ones on the basis of spin-flip rate equations. Our findings support novel applications of spin-controlled VCSELs with injection locking.
This letter presents an integrated 5G NR n79-band power amplifier with high power and efficiency that was fabricated using WIN Semiconductors’ 0.25-μm GaN/SiC technology. The stability and linearity issues of the PA is thoroughly considered. High efficiency and broadband operation were achieved using a continuous Class-F mode output matching network. This two-stage PA had a power gain, 3-dB power bandwidth, saturation power, and peak power-added efficiency of 20.4dB, 3.6-5.4GHz, 39.1dBm, and 50.9%, respectively. Its average output power was 33.3dBm under an error vector magnitude requirement of 3.5% for a 5G NR FR1 256-QAM 100-MHz-bandwidth modulated signal with a frequency of 3.7-5.0GHz.
This paper proposes a broadband high-efficiency power amplifier using an improved continuous mode design. A tuning parameter is introduced into the current expression of the drain to purposely introduce more sine components. These sine components extend the impedance space of the traditional continuous mode, provide greater flexibility in the design of the broadband output matching network, thereby widening the operating bandwidth. Moreover, since no overlap between the current sine components and the voltage cosine components are produced, these sine components do not degrade the PA’s efficiency, unlike traditional continuous-mode PAs that often trade efficiency for bandwidth. To validate our approach, a broadband high-efficiency power amplifier is designed using GaN HEMT. Measurements indicate that output power is between 40.8dBm and 42.3dBm, and the drain efficiency varies from 72.7%-81.8% in 1.5-2.5GHz, while the gain is between 11.3dB and 13.0dB in the same frequency range.
This paper presents a broadband optical packet generation method by multiplexing low- and high-frequency signals using a bias tee. The cut-off frequency of the bias tee, fc is optimized in terms of the eye-opening ratio for optical packet trains with different guard times and intervals. As a result, the optimal condition is observed to be fGT<fc<f1, and fc should be closer to fGT. (where fGT is the guard frequency and f1 is the high-pass cut-off frequency of a low-frequency amplifier). We implemented optical packet generation using a bias tee based on the simulation results and confirmed the effectiveness of this method.