Fig. 2: Characterization of individual components.
From: Deep neural network inference on an integrated, reconfigurable photonic tensor processor
a, b Chip-level pictures of the Photonic Tensor Processor and high-Q Si3N4 microresonator used for microcomb generation. c Optical spectrum of the self-injection-locked microcomb with a FSR of 485 GHz. d Wavelength- and bias-voltage-dependent transmission of the EAMs. We operate the modulators around a bias voltage of 2 V in a linear regime. e Photocurrent vs optical input power for different bias voltages with the corresponding responsivity curve. We operate the photodiodes with a bias voltage of 3 V. f We measure transmission of a test structure before and after gluing of the fiber array. Due to the glueing and shrinkage, the peak transmission drops from −2.7 to −3.8 dB.
