carlosggarcia
commented
2 years ago @damonge thanks for having a look. Here it's the plot (the noise in the plot correspond to self.cmbk_nl, the one stored in beam_extra). Also, I fixed a bug with the data types for the tracer combinations that involved cmbk 
I have implemented the CMBk tracer for S4 using the noise from https://github.com/toshiyan/cmblensplus/blob/master/example/data/kappa_noise_cmbs4.tar as indicated in https://cmb-s4.uchicago.edu/wiki/index.php/Survey_Performance_Expectations#Lensing_reconstruction_noise_curve.
The noise used is the inverse of the sum of the inverse of the different contributions; i.e.
nl = 1 / np.sum(1 / file_loaded[1: 8], axis=0)