The SEM came with a EDX setup for it.

What is EDX.

EDX (EDS) is Energy Dispersive X-Ray Spectroscopy. When an electron beam hits a sample x-rays are created. The energy of these x-rays is measured with a sensor and the unique pattern of the energy levels is indicative of the elements being scanned. This is very useful if you are trying to determine the alloy of a metal or composition of a solid object. The sensors are cooled, the most common ones are the cryogenically cooled Si(Li) detectors so you need to keep a dewar of liquid nitrogen around to keep it cool. This is very similar to the Germanium sensors used for Gamma Spectroscopy.

The detector mounts to the side of the specimen chamber and is mounted on a movable feedthrough that allows you to bring the sending end of the unit close to the sample. The detector is biased with high voltage (~-500v) and when a x-ray hits the sensor there is a pulse that is proportional to the energy of the x-ray outputted. This goes through a FET mounted right behind the sensor and that resulting pulse is passed though a preamp.

The pulse is then sent to the mainframe where there is filtering and discrimination and then passed to a MultiChannel Analyzer (MCA) where the pulses are binned in regards to the pulse height. This data is displayed as a histogram where you can determine energy peaks so you can figure out what is in there. Software databases can analyze the data and match the data to known materials and alloys.

So that's it in a nutshell, horribly simplified and quite boring. And my SEM has this, or well, I hope it will have it. There are issues, first the control mainframe and computer that came with it is freeking old, early 80's. It even has dual 8" floppy drives! There is a power supply issue, I could probably fix it but I would still be stuck with a 30 year old computer mainframe full of unobtainable parts and who knows if those floppys will even boot anymore.

Second problem, this may be a show stopper. EDX sensors do not like to be warmed up after they are brought up to temperature. Apparently what happen is the insulation around the sensor dewar is in a vacuum and these leak, as they leak the LN2 will liquify the gasses and absorb into the insulation. When the sensor warms up the absorbed gases are rereleased and a positive pressure is created in the dewar which results in the rupturing of the very thin beryllium window on the sensor tip rupturing. This is bad... very bad. You can get them repaired but that costs thousands.

Last problem, I hope. If the sensor was exposed to the high bias voltage while warm it is toast. It removes the lithium from the silicon in the sensor and it is all over with.

So I am just going to cross my fingers on the sensor unit. If it is bad I can get another old unit off ebay, it will be a crap shoot, but it is about the only option.

To replace the mainframe a friend gave me a Kevex Sigma Gold chassis to control it. It is basically a VME chassis with pulse processing and data acquisition cards built in that captures the data and sends it off to a PC over ethernet that controls the box and hosts the software that analyzes the output. Luckily the sensor and their preamps have hardly changed in the 10-15 years between the two units were made and the only real difference between the two interfaces was the gender of the DB9 connector and that the old preamp needs -24v for the opamp and the newer preamp uses -15v. No big deal, I pulled a isolated output 24v DC/DC converter out of the parts bin and used it to boost the -15 to -24.

I also got real lucky with getting data on the Kevex units. An Engineer at Thermo Fischer, who now owns Kevex, sent us a full set of schematics and a CD-ROM with the software for the box. The software is intended to run under Windows 98 but seems to be happy under XP though I cant get it to run on one PC with it. Probably something to do with the ethernet. The ethernet connection is kludged together with a 10Base2 to 10BaseT adapter on the Sigma, through a hub and to the PC. It works fine on an old Epson laptop I have but a desktop I tried with does not seem as happy. I will have to mess with that more.

The schematics have been very useful. They had the needed info for the connector for the preamp and other ports on the unit. They also showed an accessory which I did not have, the 4855 DBC, or Digital Beam Control. This is a box that is connected though a fiber optic serial link that is tied into he insides of the SEM. When off the box just passes signals though, when active it takes over the task of generating the X and Y scan signals through on board DACs and intercepts the "Video" signal from the Secondary Electron Detector.

This allows you do do a couple neat things. First, you can digitally capture images out of the scope. Second, you can map the results of the EDX over the image to give you an element map of the sample. So you can have one color represent copper, one silicon, and so on over the image of the sample. Pretty nifty.

I got one of the DBC boxes off eBay last week and the fiber optic cables showed up yesterday. It all seems to work. I hooked my scope to the outputs of the DBC and I got a nice Sawtooth pattern out on both channels and when I touched the Video input BNC it showed up on the computer.

Next step is to tie it all together. These boxes are designed to drive just about any Electron Microscope out there so they are very configurable and the schematics have a listing of all the jumper and trim pot settings. The ex-Hitachi service guy who I met online that sent me the filament bases thinks he has some information on how they wired these up in the day, if that fails I think I have figured out where I can tie in after the sweep generator on the SEM.

A few pics:

The back of the Kevex Sigma and the adapter for the old preamp:

The insides of the DBC Box, quite vintage (c1995)

And the resulting waveform out of the DBC Box, both X and Y are presented, you can see the slow Y over the fast X. Times like this I wish I had a color scope! There is some aliasing gun on because I am overlaying a significantly faster signal over the Y. I believe at this setting it is 512 vertical pixels, it will do up to a 2k x 2k image.