Neural-network-assisted femtosecond laser pulse duration measurement using two-photon absorption

Abstract

In this work, a neural network (NN) method is developed for pulse duration inferring for an erbium-doped fiber laser at 1550 nm. Experimentally, the interferometric autocorrelation trace is observed clearly with the use of the two-photon absorption (TPA) effect in a GaAs photodiode. The intensity autocorrelation function is curve-fitted by the NN with an appropriate performance, and the measuring accuracy is consistent with a commercial autocorrelator. Compared with the Levenberg–Marquardt curve-fitting method, the NN can retrieve the intensity autocorrelation function more stably and has a certain noise reduction ability, simplifying the signal processing for a TPA photodiode-based autocorrelator.