See also: http://microwiki.org
Since microscopes are a critical instrument this section provides information on both microscope techniques and equipment. There are already many good sites on microscope techniques so the focus will be more on discussing which ones are relevant for ICs and providing external links to provide more information.
TLDR:
When I first started seriously looking for equipment (Say around 2010) the Olympus BH / BH2 were pretty good options, with a complete BH costing maybe $300 and the BH2 costing maybe $600. I've heard the market has dried up a bit, but they still may be decent options. Note the BH is a 210 mm finite system while the BH2 uses infinity optics. Therefore, you should not mix parts from the two or image quality will suffer.
A lot of people go with AmScope but usually regret it after comparing with other microscopes. However, the objectives drive most of the image quality. So consider replacing them with higher quality objectives. Make sure you match optical parameters (ex: use objectives that also use 160 mm tube length).
Summary
Surplus stages designated for microscopes may cost $1000. However, if you have some basic machining skills, you should be able to hobble something together using other surplus stages. Generally steppers are easiest to drive.
Ideally your stage should be sufficient positioning resolution small relative to your image size. For example, if you have a 500 um FOV at highest resolution (100x), maybe get stages that can do at least 50 um steps. My stages can do something like 10 nm steps, which are simply overkill and just leads to overly long movement times. 1% tolerance at max FOV may be a good target.
Some non-CNC microscope chassis may be easier to mod for CNC control. For example, the Olympus BHMJ is a boom mount, which can be easy to re-mount with a stage.
Things to consider:
For Linux software, as of Nov 2017, I believe the MU800 is the only camera with V4L support in Linux (also noting there are MU300 and MD1900 drivers not in mainline). AmScope/TopuTek/ScopeTek does provide libusb drivers though, so you could make a custom application. Finally, note uManager supports misc microscope stages out of box.
Some people elect to use DSLRs. You can control these using gPhoto2 or similar solutions.
My standard system is something like:
TLDR:
A good entry level SEM is the RJ Lee PSEM. They are relatively compact and list for about $2-3k on eBay. For example, my PSEM cost about $2000, including shipping, to get operational. However, the instrument is relatively limited, providing an estimated resolution of only about 30 nm. The Super IIIA, for comparison, has a resolution of about 7 nm. This means a hit of about 18x as much information with a 2D image.
ISI SEMs can be found surplus but, mostly due to their age, can be hit or miss. A friend recently (as of Nov 2017) found not one, but two ISI DS130 SEMS free on Berkeley Craigslist. The first one was still in service and he was able to get it running no problem…after hiring riggers to move it. Mainly, if you know the SEM was still maintained and in operation until recently, it will probably be fine. If its fallen out of use for a long time, you will have to do a pretty serious overall, replacing o-rings and such. Ultimately I scrapped my Super IIIA because I was unable to fix a column vacuum leak (ie inaccessible o-ring without damaging the column, party due to old plastic)
Point and shoot cameras seem to work best as their apertures are better matched to the eyepieces found on microscopes. Some older microscopes, such as a Unitron N series, that were built to work with Polaroid style cameras might adapt to larger professional style lenses with appropriate modifications.
If your camera has a zoom setting, make sure to use it to get rid of the eyepiece ring.
A biological microscope works well enough for some initial tests on older technology. However, you will likely need something more advanced such as a confocal microscope or metallurgical microscope. If you have access to a SEM/TEM, those will likely have imaging equipment already built into them to capture digital images.
There seems to be no inherent difference in objectives between a metallurgical and a compound biological microscope. However, there are a number of fine points if you are trying to get a good image. An objective's threading and microscope's focusing system can also differ. See the following:
Infinity optical systems: http://www.microscopyu.com/articles/optics/objectivespecs.html Objective specifications: http://www.microscopyu.com/articles/optics/objectivespecs.html
Make sure the stage is very level with respect to the lens. This can be checked by moving the die from one side to another. You will also notice images that are in focus on one side and out of focus on the other. Levelling mechanisms:
Follow the Hugin tutorial for stitching scanner images. If you have an XY stage, you can take out the rotation parameter since it should be negligible.