I have been wrestling with the knife edging of a few 445nm diodes that pass thru an anamorphic prism pair as the expansion "correction" optics. For convenience and to eliminate an extra reflection in the optical train, I set the pair to deviate the expanded beam about 10 degrees in order to catch the next mirror in the beam path. I noticed the typical lll generated l l l as expected, however the order was mixed, yet not for all the groupings ( I have four groups of three diodes). Furthermore, the total expansion width for each of these groupings was SUBSTANTIALLY different. Some barely looked expanded and others wouldn't fit the scanners. But, the far field beams were identical. Curious, I measured the wavelength of each of the diodes and that explained it. The range of wavelength was large, 10nm and the groupings that showed the enhanced performance had a steady and continuous change in wavelength from left to right while the worst reversed this trend and the mixed were, well mixed.
Adjusting the prism pair for zero deviation and the effect disappeared. So, by repositioning the diodes to take advantage of this wavelength effect I have been able to do what I found to be impractical with the red diodes and that is wavelength multiplex. The reds vary little in wavelength without temperature control. However, with a sufficient number of diodes to select from, ordering them correctly according to their wavelengths and then adjusting the deviation of the prism pairs to produce the necessary paralleling of the beams that are slowly converging onto the prism, I believe it might be possible to combine an almost limitless number of blues onto a fixed scanner size. I need to check the wavelength variation of the M series as this was studied in the A series and I don't know if this amount of variability is present in the more powerful diodes.
This is pretty significant.