FS700 Highlight Aliasing and Color Fringing - SONY NEEDS TO FIX THIS!

[paulcurtis]

I cannot speak for Sony. I can speak for Convergent Design.

Convergent Design is doing our own deBayer algorithm from the FS700 RAW output. We feel that it delivers an excellent image. We are aware of the issue outlined in this thread and have observed it on other camera systems. If we feel that it is not already sufficiently addressed in our current deBayer algorithm then we will actively look to modify it to better address this issue.

There are a few samples of FS700 RAW footage as captured on the Odyssey7Q. Has anyone seen this artifact in those videos?

I've looked at the only 4K ProRes footage i could find (from the O7) and the water droplets on the fountain might have caused this but i don't really see a problem (the lens aberrations are more obvious). Aside from those samples from your website, are there others floating around i've not seen?

2K Footage from others (from Filippos O7Q short) show very clearly:

- very poor aliasing on hair
- actual coloured noise on edges (the short with the women in the forest, on a wide shot even on vimeo you can see aliasing edges with speckled colour)
- mosquito type noise around contrast edges (shoulder and dark background), looking like wavelet compression artefacts

Running the RAW through the limited but different debayer options in resolve can minimise some of these but the process is really hiding the issues with softening and reducing contrast. However in terms of this thread, i've not seen black pixels.

Comparing the O7Q 4K ProRes downsampled to 2K with the 2K RAW examples shows a substantial difference in quality and resolution as you'd expect. That 4K file with a decent scale to 2K is stunning IMHO.

From the limited (and prerelease) examples of what i've seen, the 2K RAW seems lower quality than normal 1080 in terms of resolution and edging at normal frame rates.

I presume that the O7Q has to debayer the camera RAW, then rebayer into cinema DNG?

We're only a few weeks away from testing ourselves hopefully. I'm sure you guys are doing a better job with the debayering and the edge noise, but i don't know how you can get around the binning nature of 2K. (Aside from scaling 4K to 2K which is on your roadmap)

I also completely understand that CD are doing their best with the RAW signal from the 700 and i doubt any of my observations are anything other than camera artefacts, the O7Q sits in the middle of this and tries to fix what can be fixed before giving us the footage.

cheers
Paul

 

[Gene Sung]

2K Footage from others (from Filippos O7Q short) show very clearly:

Hey Paul,

I think that if anyone can make it work (at least as well as possible under the circumstances), it would be CD just because of how much they have been on this forum. Since the inception of the FS700, Sony has had almost ZERO communication with the community so I wouldn't really expect them to do a good job with the 2K RAW. But CD has been very active with us and knows all the issues and seem very dedicated to the FS700.

It probably will never be as perfect as you hope, but at least the camera is only $12K with the 7Q vs. the price of an F5 or F55 and they are getting exact same issues with the 2K RAW. To my eye (from Filipo and P. Bloom's videos), the 240fps 2K RAW definitely looks better than the current 240fps AVCHD which is very aliased as has noise all over.

I'm wondering if anyone can make an OLPF for the 2K RAW on the FS700 in-case you want to shoot RAW for an entire shoot, like what they did for the F5/55? There's seems to be a little space between the lens and the glass covering the sensor. Here's the article that I'm sure many of you have read.

http://nofilmschool.com/2013/10/sony-f5-f55-interchangeable-optical-low-pass-filters/

 

[anatusa]

Yes, this also happened to me on a recent shoot - here's a 200% crop of the affected area. Probably should have used extra lighting to decrease the contrast. But it was a run-and-gun shot. Shot with the Canon 24-105 with Speedbooster at F2.8 (F4 on the lens, but the speedbooster made it 2.8).
View attachment 77790

maybe its your speed booster?

 

[nyvz]

Hey Paul,

I think that if anyone can make it work (at least as well as possible under the circumstances), it would be CD just because of how much they have been on this forum. Since the inception of the FS700, Sony has had almost ZERO communication with the community so I wouldn't really expect them to do a good job with the 2K RAW. But CD has been very active with us and knows all the issues and seem very dedicated to the FS700.

It probably will never be as perfect as you hope, but at least the camera is only $12K with the 7Q vs. the price of an F5 or F55 and they are getting exact same issues with the 2K RAW. To my eye (from Filipo and P. Bloom's videos), the 240fps 2K RAW definitely looks better than the current 240fps AVCHD which is very aliased as has noise all over.

I'm wondering if anyone can make an OLPF for the 2K RAW on the FS700 in-case you want to shoot RAW for an entire shoot, like what they did for the F5/55? There's seems to be a little space between the lens and the glass covering the sensor. Here's the article that I'm sure many of you have read.

http://nofilmschool.com/2013/10/sony-f5-f55-interchangeable-optical-low-pass-filters/

Gotta leave some features for the F5 users that spent twice as much on their camera that doesn't even have any way of recording debayered 4k

Why would you ever want to shoot a whole project in 2k RAW HFR mode unless your whole project is in 240fps? Otherwise there is probably little reason to shoot a whole project in 2K RAW HFR mode that uses binning/skipping and has more noise and less sensitiity. 4K scaled to 2K should pretty much always look better in every circumstance.

 

[paulcurtis]

Sorry to dredge this thread up, but whilst working on something completely unrelated i think i found what the black ringing issue is. It's not an uncommon artefact when working with filters that have negative lobes, especially some scaling algorithms (so a scale which includes sharpening - the edge curve may go negative creating more contrast - like some forms of unsharp/lancos scales). When this is applied to high range imagery (and i guess the sensor source before treatment is considered high range) it can produce overflows (the black pixels next to the whites). I've seen this when dealing with some high range CGI work and thought that looks familiar...

This sounds exactly like what we see with the FS700.

The nice thing is that recording the source 4K data wouldn't show this up and scaling in post allows us to use better scaling algorithms that can get around this, which seems to match the observable case (that the 4K stills don't have the same problems)

Just thought i'd add this...

cheers
Paul