Ocean Optics, and at least one competing firm, is selling remarkable little spectrometers based on CCD arrays. The S2000 model is a nice little unit, for the remarkable price of about $2000.
Modern electronics and production techniques have dropped the prices and raised the performance of a number of scientific instruments which used to be prohibitively expensive. One such instrument is the Ocean Optics S2000. Once, decent spectrometers cost twenty thousand dollars and more. Now, models such as the S2000 sell for under $2000. (Durn, almost the exact same wording as I used for the SRS RGA-100!)
At a former job, we purchased several of this model, and had excellent results with them. Ocean Optics helped us select the gratings, slits, fiber optic cables, and other accessories we would need. The units did most of what we had hoped, falling short only in light sensitivity.
The S2000 is a remarkably compact instrument, about the size of two packs of cigarettes. It is a diffraction-grating spectrometer, utilizing a sensitive CCD detector with 2048 elements. The actual resolution depends on the diffraction grating and slit chosen, but most peaks are at least several pixels wide. These units will not out-perform high-end lab instruments, but they're quite remarkably good for the price, and easy to use. They are really designed to be components used in a system, rather than stand-alone spectrometers, but with the right accessories they can be used as-is. Ocean Optics sells systems, including accessories such as light sources and reflectance standards, that let you configure a unit for most common spectrometry applications, from L* a* b* colorimetry, to filter transmission, to line emissions.
The model we chose plugs into a data acquisition card in the computer, while other models are available in which the spectrometer itself goes into the computer. One minor annoyance is that Ocean Optics sells the unit with a very short cable between the computer and spectrometer, and does not recommend longer cables. Instead, you're expected to get light to the unit via fiber optic cable. This isn't some cheap plastic fiber-optic cable for networking, this is high-quality broadband glass-fiber cable and it costs a bundle.
The fiber optic cable also is one of the limiters of the light collection ability of the unit. There's only so much light you can get thru a little fiber, and even more important, the end of that little fiber is not much of a target to hit with your collection optics. Ocean Optics does sell a little collection lens - I was unimpressed with its performance. The sensitivity of the unit is actually pretty good, something like 86 photons per count. If you can see the light being measured, the S2000 probably can see it with no trouble. Setting the unit for longer integration times, the sensitivity improves. However, one of our needs was to look for faint and rapidly-varying light at a particular wavelength, and it became obvious that we really needed an instrument with sensitivity on the order of a photon-counting PMT, at least for that use.
We were quite impressed with the resolution, however. We needed to measure Doppler broadening of emission peaks of certain highly excited ions, at energies of tens of keV. By selecting a holographic grating and a fairly narrow slit, we narrowed the total bandwidth of the instrument to about 90 nm, and found that we could not only measure Doppler broadening at these energies, but distinguish between isotopes. These aren't toy spectrometers from Edmund, they're serious little instruments!
The OOIBase software that comes with the unit is, well, functional. It does some useful things, but if you're planning on exploiting the S2000 for a special application, you'll want to purchase one of the higher packages. They do have available, if you ask sweetly, a "pseudocode" guideline telling programmers how to talk to the board that controls the unit, one of several A/D cards that have been employed for these units. I've made a living talking to such boards, and its not hard. We've also managed to get the S2000 to work with LabView software.
One limitation to OOIBase is that it does not allow integration times to be set below about 1 second in a mode in which a rapid sequence of spectra are recorded, however, using LabView, we found that the spectrometer will integrate much faster, providing it has sufficient light to achieve a usable spectrum. Another limitation is that the files it stores are entire spectra: if you are interested in only a small portion of the spectrum, you have to extract it yourself. The files can be read, with a little work, by ordinary spreadsheets, allowing some pretty sophisticated manipulations and analysis.
I don't sell the Ocean Optics S2000. I'm just a consultant, looking for work. If you think one of these little gems is just what you need, you'd like some help setting one up, and you live close to Northern Virginia, I'd be happy to come over and sell you a little of my time.