Maximum power handling, spike leakage, and failure mechanisms have been characterized for limiters based on the thermally triggered metal-insulator transition of vanadium dioxide. These attributes are determined by properties of the metal-insulator material such as on/off resistance ratio, geometric properties that determine the film resistance and the currentcarrying capability of the device, and thermal properties such as heatsinking and thermal coupling. A limiter with greater than 10 GHz of bandwidth demonstrated 0.5 dB loss, 27 dBm threshold power, 8 Watts blocking power, and 0.4 mJ spike leakage at frequencies near 2 GHz. A separate limiter optimized for high power blocked over 60 Watts of incident power with leakage less than 25 dBm after triggering. The power handling demonstrates promise for these limiter devices, and device optimization presents opportunities for additional improvement in spike leakage, response speed, and reliability.
This paper presents the findings of using convolutional neural networks (CNNs) to classify human activity from micro-Doppler features. An emphasis on activities involving potential security threats such as holding a gun are explored. An automotive 24 GHz radar on chip was used to collect the data and a CNN (normally applied to image classification) was trained on the resulting spectrograms. The CNN achieves an error rate of 1.65 % on classifying running vs. walking, 17.3 % error on armed walking vs. unarmed walking, and 22 % on classifying six different actions.