We got a sample of a TO-can laser, red, in a flat-window package. Our fast photodetectors are all fiber optic, so the problem is to get some of the light into the fiber so we can see the optical waveform. Sticking a fiber up against the laser window produced nothing measurable, as a little math will demonstrate. The area of a 9 micron single-mode fiber, or even a 60 micron multimode, is really small. The laser makes maybe 50 milliwatts or so and our fast photodetector needs maybe a milliwatt, so we need to capture at least 2% of the light.
So, the logical thing to do is ask Phil. He pointed us to some stuff at Thorlabs, and we wound up with this:
Our driver board is pulsing the laser on the left, and the laser is inside the expensive collimator on one side of the Thorlabs thing. That turns the sloppy laser light into a tight collimated beam, and the expensive thing on the right focusses that into the fiber.
Each end has 6 adjustments (roll, pitch, yaw pointing, X and Y position, focal length) so the probability of finding a usable bunch of settings is essentially zero. There are tricks, in Phil's book and a Youtube video, so we eventually got it to work.