The sensor (sony) is the same CCD sensor that is used in Flatbed scanners. It does the job well as it is a horizontal rectangle shaped sensor. looks like another project!
No, it couldn't be. The sensor in this is like 2 inches wide. The sensor in a flatbed scanner is typically at least 9 inches wide. They're probably both linear CCDs, but I'm inclined to think that this is where the similarities end. Though I think you could still use a linear CCD from a flatbed scanner and just make a really wide spectrometer
In fact, if you got a really old scanner, you'd probably have a better change of the circuitry in the scanner itself feeding the data out in an unencoded / raw / easily understood serial cable data format.
EDIT:
I've previously mentioned that I really don't see a DIY or hobby level spectrometer being cost or ability prohibitive. I'm going to echo that sentiment again. There's this, for $15:
http://www.ebay.com/itm/CCD-LINEAR-...aultDomain_0&hash=item4ab479b1ec#ht_734wt_905
Which is actually a pretty high pixel count (I think about 50% higher than the science surplus units). It's not prohibitively difficult to hook up either, even for a hobbyist. This guy used the same CCD:
http://www.youtube.com/watch?v=ut9LfQmaYkU
As well as an actively produced micro-controller that costs about $15:
http://components.arrow.com/part/search/TMS320F28069
I'm not saying that this would be easy, but there are people here with enough experience to pull it off. The big question is hardware. In this case, a small metal box, with a mounted fibre cable attachment, and an internal mirror and diffraction grating that were adjustable, would do it. I would imagine that the adjustable mounts could be re-purposed from the technology that projector guys use for beam combining, and sourcing a diffraction grating shouldn't be too difficult (though the science surplus units seem to use a grating+mirror combined, so that the reflection itself is split, as opposed to having to pass a beam through a grating).
At any rate though there is nothing here ^ that isn't accomplish-able. In the past, I've said that it might not be a money saving endeavor in light of the $200 science surplus models. My perspective had been that we might just be able to create more interesting options at around the same cost. Now, I'm not so sure we couldn't save a bit on price too. Parts wise, you're looking at ~$30 for the CCD and micro processor. Say another $10 for assorted electronics you'd need to run a USB or serial interface. Add $10 for a good metal box and fibre cable adapter. Who knows what optics would cost - the diffraction grating would be the pricey /difficult to source item. But suppose you could do that for $50. That might lead to a $100 component cost. This might be DIY-able for $100 give or take a bit. If it's a DIY unit, then you don't really need "software", other than a method of just capturing a frame once. You could do the math by hand (applying coefficients manually). If someone was going to sell them, they could probably charge ~$300 for a calibrated unit (that's still $200 cheaper than the science surplus calibrated units at $500). They'd make some decent money, have a better product (higher resolution, cooler looking), and maybe also offer USB connectivity, which the science surplus doesn't have.
Idea for any enterprising members with free time this summer
The best part, is that once you calibrate this with 4 known wavelengths, it's actually really easy for end users to verify that the calibrate is still intact, but using a 532. Unlike LPMs where almost nobody really has a "known" power that couldn't ever have changed. With spectrometers, almost everyone does in fact have a stable known wavelength (or two) to test with.
