In fact I have been trying to model the coherent backscattering independently. While the model I tried to fit is working very well up to N=2048 above this it produces unreasonably low negative values for the backscattering ( few times stronger than the t-matrix results). I believe that this is firstly because the model is purely statistical and, while the goodness of the fit is very good the residuals are not normal at all which suggests that extrapolation is not a good idea with this model. Secondly, I found that the coherent backscattering effect starts when xm*m0 > 0.24. However I do not have a large enough data set to find a good purely statistical model. The actual dependence of the threshold for the coherent backscattering is probably more complex as N definitely have an effect too, but it is difficult to find an accurate cutoff values without more data.