@Jason, it's a pretty involved list of stuff that's got to be done before imaging. Let's see...
First off, the sample needs to be conductive, so (if it's not already made of metal) it has to be glued to a metal stub using conductive paint, then put in a vacuum sputter coater where a microscopically thin layer of (typically) gold is applied.
There's whole checklists of things you can't image for various reasons. E.g. anything ferromagnetic that might have powdery residue is forbidden because it'd get sucked up into the magnetic coils in the microscope. Liquids are out too, as well as anything that'll vaporize in the vacuum.
Insulating objects (that aren't gold coated beforehand) aren't forbidden, but they can give extremely weird image artifacts. The typical example is trying to image a polystyrene target, which causes charge to build up on the surface, which repels the beam and effectively turns the polystrene into a mirror. Some cool pics of that happening are here:
http://www.optics.rochester.edu/workgroups/cml/opt307/spr11/greg/There's been more than one occasion where people thought they'd found new structure of things and it turned out they were just taking a picture of the inside of the microscope instead :-)
Anyway, once the coated specimen is ready it needs to be put in the sample chamber. The level of vacuum in the electron gun assembly (8e-10 Torr) is slightly better than the vacuum outside the international space station, but the sample chamber is allowed to be a bit worse (5e-6 Torr). Regardless opening the sample chamber is still something that's not done unless absolutely necessary. To avoid that, there's a dedicated sample airlock and a sample holder (stick) to allow loading the sample while keeping everything under vacuum.
Once the sample is in place and the vacuum is good, then the electron gun can be turned on and the process of alignment and focusing can begin. There's about six different power supplies controlling the behavior of the electron column (panning, zoom, focus, astigmatism, etc), so quite a few variables need to be worked out. But the nice thing is that a huge amount of the performance of the microscope can be controlled in 'software', rather than needing to move mechanical parts like an optical microscope.
Once that's set up you can choose which detector is giving you the best result from the sample. The classic detector is called an 'Everhart-Thornley' detector and it picks up secondary electrons thrown off by collisions the beam makes with the target. There's other types such as Backscatter Electron (BSE) detectors which detect the electrons from the original beam being backscattered by material. Different sensors show up different material properties, and it's also common to blend the various signals to give fake 'color' images too.
Things I didn't mention - compressed air supply, cooling water to the diffusion pump, baking out the electron column to get a good vacuum in the system, safety interlocks and preventing damage.
I can see this taking me a while to get to the stage where we can actually run samples through it, but it should be fun learning all the tricks :-)
Cheers,
Gavin
Oh, and Ben Krasnow's video's are simply awesome. He's what got me interested in SEMs in the first place, and I plan to reuse his Teensy code for making the microscope's acquisition system in the future.