Now this might sound like a really spaced out question, but here goes:

Can a 2k camera be built from scratch, indie style - Like if someone can buy a CMOS sensor (maybe 12MP 4K size) and get some circuitry to take out RAW data - something like what an SI2k Mini does, etc?

I understand there are huge challenges here, however, is it possible? Are there any resources out there that can throw some light on this?

A number of people have tried this with various industrial cameras. Silicon Image, themselves, have done it, turning one of their industrial cameras into a digital cinema camera by integrating a recorder, software codec, etc. Dalsa tried to do it, but ultimately abandoned the project. Ikonoskop is trying to do it with a Kodak chip. Aaton is developing a digital mag for their Penelope camera using a Dalsa chip.

For someone without those resources, it might be better to adapt an off the shelf camera. The Elphel 353 looks promising, as does the Prosilica GX1910 (same Kodak chip as Ikonoskop). Elphel uses open source Linux based software, Prosilica recommends off the shelf Windows based software.

At the end of the day, though, you're going to be jumping through hoops to get an image into your edit system. These industrial cameras usually have to be tethered to a computer, and they don't typically shoot in edit friendly formats, and they don't ordinarily work at frame rates that are standard for film/video.

Not to say that it can't be done by an individual, but that it's difficult.

Of course it's possible. This is what Arri, Panavision, and of course Red do to get new camera designs. For example:

Panavision: Services & Systems: Camera Systems: SSR-1 - Solid State Recorder (http://www.panavision.com/product_detail.php?maincat=1&cat=36&id=384&node=c0,c136,c137)

ARRIFLEX D-21 | ARRI Digital (http://www.arridigital.com/)

RED / Cameras (http://www.red.com/cameras/)

Give some engineers a functional specification and a budget and turn them loose. This is how new cameras are designed, built, and brought to market.

The Apertus more or less describes one attempt to turn the Elphel into a "cinema" camera.

The Elphel is the most easily hackable in both hardware and software; it has a 3d version; and can do video to DCI specs 2048x1080 at 24fps rather than the hdtv's 1920x1080 or below. The Elphel *is* a computer and doesn't need to be tethered to another one. Its audio interface may need to be improved.

BTW: what stretching exercises do the people who use those other cameras do, and how bad are the stretch marks on the video after going from 1920x1080 to 2048x1080?

This seems to have turned up quietly and without fuss. It may have been heavily advertised, though here in Perth, one does not always get to know such things unless one actively searches and buys in all manner of industry newsletters.

P+S TECHNIK | Professional Cine Equipment Manufacture | SKATER Camera Dolly Product Family | SKATER Scope Snorkel Lens System (http://www.pstechnik.de/en/digitalfilm-16sr-magazine.php)

Interestingly, this is more or less how the SI2K got up if vague recollection serves me correctly. Given the connection with P+S my guess is there may be some commonality but this one has the added benefit of having HD-SDI out. Now if that could only be transplanted to the SI2K?

The digital mag was announced last year. Interesting to see how it does with the 35mm sensors now out there and RED's 3k 2/3" Scarlet coming out at some stage. I assume there's a cross over with the SI 2K to keep costs down.

The attraction of the digital backs may rest in the film camera owners who can have the option of shooting film if their client wants it, or digitallly and have the same familiar toolset.

My imagining is that some projects might even be shot as hybrids of film and digital, which should hurt Kodak and Fuji perhaps a little less.

Why would you? There yet remain some circumstances when only film will do.