I am starting this thread to see where it goes and maybe we all can learn something.
The way I got it figured, so far, is that you need a way to bring you sample of steel to incandescence. Methinks that fullering a small sample so that you end up with an 1/8th inch center section and hooking it to your buzz box would do that. Next you will need a narrow slit to let the light through to a prism or diffraction grating. Then you will need a way to (and here is where I get lost) photograph the spectrum as it comes from the diffraction grating.
Then (I guess) you will need to make sample spectrographs of known steels for comparison. And what else and how does it all fit together?
Warren

You also need a controlled atmosphere so that the different atmospheric gasses does not contaminate the spectral lines of the material you are testing, Helium is often used.
The spectrometer at my work does not need the sample to be hot, you just need a clean surface (from the belt grinder) Then you put the sample in the spectrometer and the machine will strike an arc between a electrode (Tungsten?) and the sample, the arc does two things, it pulls material off the sample and then the loose atoms are heated in the arc revealing their spectral lines.
A prism is then difracting the light onto an approx 1 meter long array of sensors.
The next step is making a database that can identify different elements based on a unique set of spectral lines and determine the concentration of that element based on the energy levels in the lines.

Thank you SCB,
The idea of starting this thread was to see if people with mechanical talents would be able to make a spectrograph that would give a pretty good idea what an unknown steel sample is. This would incorporate the fundamentals of spectrographand avoiding any electronic interfaces other than a digital camera.
A gas shielded tungsten arc is what tig (heliarc) is. Prisms and diffraction gratings are easy to come by. So we have that end of the device covered.
So what can we do with the diverging spectrum as it exits the prism? Can we pass it through a columnating lens and onto a ground glass in order to photograph the image of the spectrum? Here is what we are looking for: Iron, Carbon, Manganese, Chromium, Silicone, Nickle, Vanadium etc. and some tramp elements like sulfur and potassium. Knowing little other than basic physics, I ask, is this concept (with some modifications) workable?
Warren

Of course it can be done! It might not be easy but if they could do it in 1853..
The hardest part is not to identify the elements but to determine the concentration of each element. Sulphur, for example, is usually found in concentrations around 0.02% wich might be a bit tricky to measure reliably
To start, one would need pure samples of all the metals that should be deteced to document the spectral lines without interference, or look it up in a book or on the web
http://student.fizika.org/~nnctc/spectra.htm

Edited January 16, 200917 yr by SCB
more info added