Reliable and consistent laser diodes?

[The_Doctor]

Well, that's getting closer to it. 'Spot quality', is a bit vague though. Do you mean rays traced from points such that they cross each other in idealised tracing from points on source to points on image? That alone might screw up detail. Or is there more to it, like interaction with electric field that perturbs paths even after they leave the source? I understand that single mode diodes have greater power density, but if it were down to that alone it might be an easy choice to make. It clearly isn't, people have agonised over it for more than a decade. Multimode diodes like Sony's SLD3233 have been around for donkey's years, so it's not as if there hasn't been plenty of time to resolve this question, but every time I look, it's being thrashed as if it's brand new. Which implies that most people are missing something.

[kecked]

I wonder if you add a 45 degree optical flat with a polarizer in it if that would stop the back reflection. That way the chance of a reflection of strength coming back in phase...polarization would be really tiny.

[The_Doctor]

I remember discussing stuff about that with Sam (Goldwasser, on Usenet). What he said was that that a Faraday Isolator exists, but is expensive. Very. So I quit there, hoping there was a cheaper way, but apparently not. The quarter-wave plate and polariser idea occured to me, but was apparently very wave;length sensitive, and diode wavelengths aren't steady enough. What you're suggesting sounds different, but I'm not sure how.. My guess is it's the same, the total 90 degree shift made up by the time it tries a second pass through the plate. Which it does, and is them presumably stopped by the polariser..

Anyway, this business was why I later emphasised cheapness of diodes, and simple and durable mountings to allow easy and frequent replacement.

EDIT:
I remember also some talk of mirrors reflecting more than once, each time steering 90°, to align polarisation planes. I wondered for a while if that might be a key to a viable method, but gave that up after deciding it wasn't, because whatever I did would be reversed on return anyway.