Further to the last post a little while ago, here's schematic overview of the system as it stands right now:

We have two photo diodes now, a single one to measure the laser output power via a 50/50 beam splitter, and the quad photo diode for the astigmatic focusing. A second, polarizing beam splitter plus a quarter wave plate allows us to only receive light coming back up from the sample under the objective in the quad PD.

The objective is still the voice coil assembly salvaged from a PS3 drive.

It allows us to have a dynamic focus range of several mm, paired with μm precision, all controlled by an DC voltage input to the VC driver board. The lens in the assembly is also pretty good, with a numerical aperture NA = 0.85. NA is important especially in microscopy, as it sets the resolving power: the smallest detail that can be resolved by the lens, and thus the microscope, is proportional to λ/2NA, where λ is the wavelength. So, in theory we should be able to separate details that are as little as 240nm apart. In theory...

We also have a high precision x-y-stage in the setup now. It can displace the sample under the objective at sub- μm precision.
From tests Andrew did in his amazing Michelson interferometer project, we know the amount of z-displacement per unit current (linear with VC voltage) and can now calibrate the focal depth of our setup: about 1 μm. We get this information by repeatedly scanning across a steep edge and stepping the focus (i.e. the VC voltage) for each line.

We could of course also use that procedure to focus on the sample, but the astigmatic approach makes it so much faster: All we need is to acquire the quad PD signal as function of VC displacement, as shown below.