1: \begin{abstract}

2:   This paper introduces a backpropagation-based technique for the

3:   calibration of the mismatch errors of time-interleaved analog to

4:   digital converters (TI-ADCs). This technique is applicable to

5:   digital receivers such as those used in coherent optical

6:   communications. The error at the slicer of the receiver is processed

7:   using a modified version of the well known backpropagation algorithm

8:   from machine learning. The processed slicer error can be directly

9:   applied to compensate the TI-ADC mismatch errors with an adaptive

10:   equalizer, or it can be used to digitally estimate and correct said

11:   mismatch errors using analog techniques such as delay cells and

12:   programmable gain amplifiers (PGA). The main advantages of the

13:   technique proposed here compared to prior art are its robustness,

14:   its speed of convergence, and the fact that it always works in

15:   background mode, independently of the oversampling factor and the

16:   properties of the input signal, as long as the receiver

17:   converges. Moreover, this technique enables the joint compensation

18:   of impairments not addressed by traditional TI-ADC calibration

19:   techniques, such as I/Q skew in quadrature modulation

20:   receivers. Simulations are presented to demonstrate the

21:   effectiveness of the technique, and low complexity implementation

22:   options are discussed.

23: \end{abstract}

24: