Details
- Affiliation
-
AffiliationFundacion Fulgor
- Country
Backpropagation is a machine learning algorithm used to train neural networks. This paper introduces a back-propagation based technique for the calibration of the mismatch errors of time-interleaved analog to digital converters (TI-ADCs). It is applicable to digital receivers such as those used in coherent optical communications. The error at the slicer of the receiver is processed by the backpropagation algorithm, and applied to compensate the TI-ADC mismatches using an adaptive equalizer. Alternatively, the back propagated error is used to estimate the mismatches and correct them with analog techniques. Implementation complexity is low. The main advantages of the technique proposed here compared to prior art are its robustness, its speed of convergence, and the fact that it always works in background mode, independently of the oversampling factor and the properties of the input signal, as long as the receiver converges. Simulations are presented to demonstrate the effectiveness of the technique.