Looking for Mad Genius to modify cameras

[Eki Halkka]

As far as i know, the signal one gets from the HDMI out of most good quality small cameras is uncompressed 4:2:2. The live output bypasses compression and chroma subsampling to 4:2:0.

I'm also virtually certain that the in-camera processing in most decent quality home cameras is done in higher than 8 bit. Even the HDMI out might be 10 bit depending on the model (though possibly containing only 8 bits of actual signal and two bits of padding).

PS: I work quite a bit with 4:2:0 8-bit footage (XDCAM-EX / HDV, even some AVCHD) and when processed properly, it can look pretty darn nice. Banding is rarely an issue when editing and color correcting is done in higher than 8-bit project, even if the source is 8 bit (the source usually breaks in other ways before banding becomes a problem).

This said, if banding occurs, simple remedy is to add a tiny bit of noise to the footage before color correction - essentially what this does is to dither the footage to higher bit depth. A fake, sure, but it works.

 

[JohnF]

J I've heard about Quantel DBR, you certain is not in use, even if under a different name, in other systems?

It's sometimes known as Dynamic Rounding the basic signal processing technique is known as dithering. As far as I know the only other user of this technique were Snell and Willcox vision mixers (DBR as an option I think)

As an (extreme) low-light shooter quantization is my enemy so I have some sympathy with your troubles. I do recommend doing a test with a Quantel to see if it produces the appropriate results.
If it works to your satisfaction you can edit on your standard system making all the decisions one would do for an online master then export the EDL to do the final online on a Quantel. By doing this one significantly reduces, the sometimes intimidating, costs associated with Quantels but the results are often worth it.

It's an often forgotten point but most deliverables end up in low data-rate 8bit format anyway. Many times I've seen (and even worked on) films that were shot with excellent kit only to have the picture reduced to a mess of banding, quantization noise and artifacts. My point being is that as long as the shots are good, the post is well handled it doesn't mean shooting in 8bit is a no go. DBR/Dithering can make all the difference!

All the best

JohnF

 

[Rudi Herbert]

Eki, John

I'm not against 8 bit per se, I do a lot of work on XDCAM EX myself, and gotten to a point where I can produce really outstanding results with my EX-1's, but alas, the sensor in the Canons is nowhere near as good as on the higher end Sony and it really shows rather quickly. I've been doing a lot of underwater 3D with a dual Canon set up using the cameras' AVCHD 24 mgbps for the last year and the codec breaks during post, keeping in mind that most exhibition formats only use half the resolution of every frame and you need to visit further punishment on the images with ghost busting passes, etc, etc. The footage looks quite good on smaller screens, but for the intended HD network TV distribution, I need something a little stronger for post. These cameras allow perfect IO for parallel rig shooting and now I have found a great set of wide angle ports made of superb optical glass that give me beautiful wide 3D imagery, something I cannot get with bigger cameras that have stronger codecs, so I'm really trying to squeeze the utmost quality out of these tiny things. I guess yes, I can always try and run my finished edit through a Quantel, but man, it would be beautiful if I could figure out a way to intercept the image path inside the image processor while it's still at 10 bit and take it out to a recorder...

 

[Jason Goodman]

21st Century 3D was very involved with the Andromeda project. We provided early funding, some hands on assistance and were the first users of the finished technology in our 3DVX stereo camera. Maybe you've given up on going this route, but just as a point of interest, 10 bit data was captured with this device. 12 bit would have been possible and 14 bit was planned for a similar mod for the HVX200 that was started but never completed. The signals were intercepted at the analog to digital converter, so it was true 10 bits and you could absolutely see the difference. This approach might be harder now with CMOS sensors. CCDs are analog, I don't think the same kind of A/D exists in a CMOS camera but somebody smarter than me would know for sure.

Technicalities aside, projects like Andromeda are TIME CONSUMING, EXPENSIVE and fraught with engineering challenges. I would go out on a limb and say if you started modding an off the shelf camera today, you'd have a six figure investment on your hands and probably would not see results before Scarlet comes out. There is a reason Reel Stream ceased operating. Pimping cameras ain't easy.


http://www.21c3d.com
jason@21stcentury3d.com

 

[Jeff Kilgroe]

Rent a couple EPICs and be happy. There should be plenty of them out there within the next 3~4 months. Faster than you could ever hope to mod cameras on your own and a heck of a lot cheaper.

When do you need to start shooting and how small do you need to go? There are quite a few options out there.

 

[Rudi Herbert]

Jeff,

I should have been more specific into why I want to use those cameras. My project starts whenever I'm ready, so if I had the desired setup right now, I could start shooting tomorrow. It is an underwater documentary series with an eye on conservation aimed at TV, so it will take me some time to accumulate a lot of the supporting images that do not require the human players to be in frame. Thus I would rather not wait for Epic for a few more months. More importantly, I do not want to shoot on S35, all that shallow DOF is really counterproductive for nature 3D, especially for run and gun underwater work where I need to hold focus throughout. I've tested a couple of REDs and the results were pristine as expected but the DOF killed the stereo effect that I was so easily achieving with the crappy camcorders. Through extensive experimentation, I determined the absolute biggest acceptable sensor to be 2/3", with 1/2" being the best compromise and 1/3" providing the easiest solution in terms of focus and stereo effect. YOU REALLY WANT DEEP FOCUS WHEN SHOOTING MARINE LIFE!!!. I would have preferred the Scarlet 2/3' but I'm sure that will not be really available until late next year, at a time where I need to start delivering material. The SI-2K is an attractive option, but you can read in my previous posts why I would prefer not to go with it. I've used a rig with dual EX-1s that has worked very well, but it is a big housing underwater that drastically reduces my mobility. Just as importantly, with the 1/3" cameras, I have the option of using wide angle ports with 80 deg FOV which is about the perfect width for underwater 3D, which still give me under 5" IO and produce beautiful rectilinear images thanks to their optic glass construction, something any other system does not offer, at least not without going through heavy and bulky wide angle lenses and dome port combinations. Anyway, I guess I was very mistaken in thinking/hoping that intercepting the image signal path somewhere before it comes to the AVCHD compression stage would not be as difficult as it seems to be, which shows why, ultimately, I am a shooter and not an engineer...