Results from the iterative implementation of the unbiased map estimator. Top row: reconstructed maps. Middle row: difference between the reconstructed and the input maps. Bottom row: a study of convergence of our iterative solver. The left panel shows the standard residual, as defined in Eq. (55), which saturates and does not converge to our fiducial level of 10^-6 in as many as 100 iterations. The middle and right panels show that this lack of convergence is due to the largest angular modes as the fractional difference between the power spectrum of the input map and the power spectrum of the ith map estimate in the multipole range ℓ ∈ [500,2100] becomes quickly very small and reaches the level of better than 0.1% in fewer than ~100 iterations. This last observation has been used to set the convergence criterion used in the analysis of the first year POLARBEAR data set The POLARBEAR Collaboration (2014).