Search

It is challenging to apply deep learning in professional fields that lack big data support, especially in industrial nondestructive testings (NDT). To solve this problem, one feasible solution is to introduce the concept of NDT physics into a deep neural network to compensate for the network's poor predictive abilities when trained on small datasets. Therefore, we propose a physics-informed deep neural network, named GuwNet, using a unidirectional oblique-focusing (UOF) high-frequency, high-order shear horizontal guided wave electromagnetic acoustic transducer (EMAT) to quantify microcrack defects more accurately. We study the quantification principle of microcrack defects suitable for UOF-EMAT, and propose a network using physical knowledge regarding this theory including the network structure and loss functions design. Compared with traditional nonphysics-informed methods, the length, depth, and direction of the quantification errors are reduced to 0.127 mm, 0.279% dt, and 1.843', respectively, and the average quantification error is reduced by more than 80%

A theoretical analysis and contrast of silicon-on-insulator (SOI) floating-body (FB) and body-contacted (BC) metal–oxide–semiconductor field-effect transistors (FETs) including insertion loss (IL), isolation, power handling capability as well as harmonics for radio frequency switches applications, are firstly developed in detail. From the presented investigation, lower IL and smaller chip size can be obtained using FB FETs, while the BC devices are beneficial to isolation and harmonics improvement. As an experimental vehicle for the presented analysis, the high-power single-pole eight-throw (SP8T) antenna switches for multi-mode multi-band applications, are fabricated in a partially depleted 0.18 μm SOI process based on FB and BC FETs, respectively. At 1.9 GHz, the tested IL, TR × 1 to TR × 3 isolation, and P −0.1 dB of the FB and BC SP8T switches are approximately 0.53/0.65 dB, 27.6/29.3 dB and 38.0/38.4 dBm. For the two cases, the second and third harmonics are −70.6/−75.5 and −60.4/−79.9 dBc, respectively, with a +33 dBm input power at 1.9 GHz. The experimental results strongly support the analysis that the FB FET is a preferred device in low loss or low throw-count switching design while the BC FET is a better selection in the applications that demand high power and low harmonic.

A broadband class-F power amplifier for an S-band handset device is integrated on a 3 × 3 × 0.82 mm3 die using an InGaP/GaAs HBT process. With LC serial harmonic traps immersed into the broadband output matching circuit, good harmonic suppression performance can be achieved. A pure resistive impedance of the matching circuit, but near zero at second and infinite at third harmonic frequency, which enhances the efficiency, is obtained across 1.8–2.5 GHz. Tested with a continuous wave (CW) signal, the PA delivers an output power of 34 dBm and achieves a PAE of 57% at 2 GHz. In addition, excellent harmonic suppression levels of less than −53 dBc across the second to fifth harmonic are obtained.

A broadband class-F power amplifier for multiband LTE handsets applications is developed across 2.3–2.7 GHz. The power amplifier maintains constant fundamental impedance at the output matching circuit which is operating for broadband. The nearly zero of second harmonic impedance and nearly infinity of third harmonic impedance are found for highly efficient class-F PA. The harmonic control circuits are immersed into the broadband output matching for fundamental frequency. For demonstration, the PA is implemented in InGaP/GaAs HBT process, and tested across the frequency range of 2.3–2.7 GHz using a long-term evolution signal. The presented PA delivers good performance of high efficiency and high linearity, which shows that the broadband class-F PA supports the multiband LTE handsets applications.