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

[blackdogdylan]

Actually come to think of it, does anyone know if S-log2 and the REC709 800% gamma fix the primary colored highlight issue where bright reds turn yellow, blues turn cyan, etc?

I can't speak for the 800% since I haven't experimented with it yet, but SLog2 does at last fix the ugly primary (actually more like pastel) highlights and behaves very similarly to the F55 in color rendition. it's a hell of a camera. just frustrating that with a little work, it could be so much better.

 

[Gene Sung]

Fortunately the worst of this issue will almost definitely be avoided by accessing the RAW bayer stream using the CD7Q or R5. Should also fix the annoying primary colored highlight issue which is, in my mind, perhaps an even worse issue also present in the FS100, FS700, and F3. Actually come to think of it, does anyone know if S-log2 and the REC709 800% gamma fix the primary colored highlight issue where bright reds turn yellow, blues turn cyan, etc?

The Reds in S-Log seem very orange and the greens can be kind of weird. But seeing Filipos tests on the R5 and the 7Q, the colors looked great so not sure the recorders interpret the colors differently?

One thing I LOVE about S-Log 2 is that the highlights don't have that yellow color that sometimes happens. That really bugged me about Cine 4. Cine 1 seemed a bit less prone to it, but S-Log at least got rid of that

 

[paulcurtis]

This has been around since day one as we know, and it's not the first time that i've heard it's a sensor issue either. If that's the case then you can't fix it, but perhaps you can minimise it in post. The usual 4K would be a much better place to start from and working on that footage at that level before downsampling.

In the OP case it isn't an issue of a high contrast edge (unlike the much more obvious example later on). It appears to be a super bright issue on an edge. Now in this case it ought to be possible to search for pixels that display an uncharacteristic luma change and isolate them. However working from AVCHD or compressed footage makes that process more difficult. Again seeing a 4k pixel for pixel output would be useful and in fact the fix could be even be made in debayering potentially.

There's the grid issue with the F55 sensor at the moment which looks even worse than this (so the comment about buying a pro camera is somewhat negated, it can happen to them all).

But truth be told it's never been a massive issue with the footage i've shot, tends to be very specific circumstances.

I still don't believe the FS700 is downsampling 4K -> 1080p very well. I'd expect a much better quality 1080p if that was really the case.

Hopefully those O7s will ship soon and we can finally try

cheers
Paul

 

[Barry_Green]

This has been around since day one as we know, and it's not the first time that i've heard it's a sensor issue either. If that's the case then you can't fix it

I am certain that it is a sensor issue, not a debayer issue, and other manufacturers have had it and have been able to fix it (Red, Sony F5/F55, and BlackMagic are three examples). Whether it's possible to fix it on an FS700 is something that only the company knows; sometimes the code is accessible, sometimes it isn't; sometimes the camera is built with reprogrammable chips, sometimes they save costs by burning the code in non-changeable roms. So just because other companies have been able to fix it is no guarantee that Sony is able to fix it on the FS700, but if I had to bet, I'd bet that it is likely fixable, just based on the price range of the camera. However, we do have to acknowledge that it's an NEX, not a CineAlta, and that may influence either the ability to be fixed or their interest level in fixing it.

 

[nyvz]

I am certain that it is a sensor issue, not a debayer issue, and other manufacturers have had it and have been able to fix it (Red, Sony F5/F55, and BlackMagic are three examples).

Based on what? You're still talking about issue with bright sharp edges? I have never seen such an issue with any other camera, could you provide a link or specifics? I downloaded some of the sample FS700 RAW footage and none of it seems to have those issues with edges or bright colors. It looks like it could be some kind of pre-debayer oversharpening because has sort of a moire effect even when moire is not visible in the final debayered image, it also could be happening due to the debayer or post-debayer but in any case it doesn't appear to be present in RAW footage recorded using the Sony R5 and also should not therefore occur in RAW-originated recordings on the CD 7Q. However, if you have evidence to the contrary, I'd be interested to see it.

 

[epsom salt]

signed and sent to the 5 people i know who own this camera. now... paging evil thought to squelch the rebellion.

 

[Sheldon Charron]

Oh wow. That is really bad. Gulp.

yeah. it's definitely the ugliest of the FS700's characteristics. I had hoped that with the SLog2 upgrade, it might be addressed, but as Justin mentioned, it is obviously a bi-product of the debayer from 4K down to 1080p. check out this grab from a test I ran after I got the upgrade (trying to see the upper limits of contrast ratio the camera could handle.)

View attachment 77603

nasty.

I hope producers never see this thread.

 

[paulcurtis]

I also think that, in the case of the high contrast white back and black stand example, the problem is partly generated by the lens choices. It looks like a lens aberration that has been amplified by whatever this rendering problem is. However as Noah says, lets wait til we get to check the 4K RAW (or O7) with a problematic case and see how bad the original is - i have a feeling that some of these issues are part of the cameras post scaling/debayering.

cheers
Paul

 

[blackdogdylan]

I also think that, in the case of the high contrast white back and black stand example, the problem is partly generated by the lens choices. It looks like a lens aberration that has been amplified by whatever this rendering problem is.

I somewhat doubt this particular matter is related to the lens. this shot was using a Canon FD SSC 1.4 (a lens generally appreciated for its sharpness and lack of CA) shooting at a 2.8/4.0 split. and to be clear, what you are seeing here isn't some extreme pixel peeping blow up... this is just a crop of the right side of the frame (ie 1080 lines vertically) recorded externally at 280Mbps. you can see that the fringing and aliasing disappears in the areas of lower contrast, but reappears (on the same stand and card) when the ratio increases. notice the white card has an artificial black line at the transition point.
as I think you've stated before, Paul, the 1080 out of this camera could look better.

 

[Marino215]

Hi. I just got this email from a professional product support specialist regarding the color fringing...tell me what you think. Thanks

You and many others have experienced color fringing on the NEX-FS700, PMW-F3, PMW-F5, PMW-F55 as well as Canon C300, Red cameras, and many others. This is not something that you "fix", it something you learn how to use creatively or how to avoid. Unlike professional and broadcast cameras which have flair compensation and chromatic aberration circuits, digital cinema cameras rely on the lens and talent of the photographer to eliminate any side effects.

There are many things that cause color fringing, aliasing, flair and ghosting. Lens design, image sensor size, number of sensors, the number of pixels, and video processing all have an effect on picture quality.


Lenses are most common cause of these artifacts. The internal structure of the lens must be such that light traveling through the lens does not reflect off the sides of the lens and off the lens elements themselves. You want the light to pass through and not bounce off anything inside the lens. While a 50,000.00 lens may come close, most lenses have some imperfections. When light of different wavelengths travel different paths through the lens, you get "color fringing" or "chromatic aberration". This can be dealt with by coating the lens elements to try and keep all the wavelengths traveling the same path through the lens. In studio and broadcast cameras, you also have adjustments to compensate for the lens being used. When light starts bouncing off the inside of the lens, you get wavelengths of light that take a different path then the original light path causing a purple or green cast on the side of images. If the light entering the lens is bright enough and at an oblique angel, it will bounce off the inside of the lens barrel and reflect off the back of the different lens elements causing a row of stars or reflections of bright objects in the frame, creating a ghosting or flairs in the image.

Digital cinema cameras such as the NEX-FS700 and PMW-F55 are designed along the traditions of the 16mm and 35mm film cameras so they have no adjustments. They are letting the raw light from the lens enter and hit the sensor. So it is up to the lens design and the camera operator to insure that the image looks the best it can. That is why cinematographers prefer to use prime lenses rather than zoom lenses. The prime lens has an optimized light path and doesn't have to compensate for different focal lengths. Zoom lenses, no matter how well they are designed bend the light ever so slightly when you zoom in and out due to the mechanical play in the zoom mechanism. This in turn causes the light path in the lens to vary, leading to chromatic aberration and flair.

You can often see flare in movies shot with film as well as video. It does not happen because the film stock or the image sensor is reflective; the reflections are happening between optical groups within the lens. For this reason, matte boxes evolved to include barn doors, and French flags and why DP's spend hours lighting a set and carefully flagging specular light sources from striking the lens causing unwanted flares.

With any fixed pixel sensor, there is a point where the spacing of the pixels on the sensor and the spacing of the detailed objects are the same. When you zoom in on the surface such as a brick wall, at some point, the detail of the wall will align with the pixel pitch and that is when you will see a moiré pattern appear. You often see that as a moiré pattern on a building or someone's shirt. This "aliasing" may appear as jagged edges, interference pattern ("moiré")or false color in the image. On 3 imager cameras, you will see more "moiré" while a single sensor camera you see more jagged edges and false color appearing.

Single sensor cameras have to use a "color filter" to recreate the red, green, and blue to give you a color picture. This means that each pixel has to do double duty to create the color picture. Single sensor cameras use DSP (Digital Signal Processing) to "de-bayer" the pixels and create the color image. This processing can lead to false color on edges of an image or a color pattern appearing in the image. Unlike three sensor cameras, when aliasing occurs it appears on all three colors at the same time, it gives you a black-and-white pattern.

To minimize this, camera manufactures use different methods to try and reduce the chances of this happening. Using an optical low pass filter on the imager helps reduce the amount of high frequency detail that can cause problems when the signal is processed. This filter is tailored to the pixel count of the imager and the resolution of the format of the camera. In the case of a 4K capable camera, this filter is less capable of filtering out detail than an optical filter on a HD camera, and therefore more likely to show false colors than a HD camera would.


The signal processing has been designed to try and mitigate these artifacts under most conditions, but it is not the end all answer to these problems. The circuit has to balance picture quality with picture processing. Too much processing and your picture quality suffers or you create other problems with the image such as poor color, lower resolution, etc. The design of DSP processing on the FS700 has been optimized to provide the best possible resolution and color while minimizing any artifacts during this processes. While it is not perfect, it is the best compromise for the price point of this camera.

Sincerely,

Douglas Bierschenk
Sr. Product Support Specialist
Sony Electronics Inc.
Tel: (800)883-6817

 

[starcentral]

hm, blame the user and tell them to take some cinematography courses.

:huh:

 

[Marino215]

I also think that, in the case of the high contrast white back and black stand example, the problem is partly generated by the lens choices. It looks like a lens aberration that has been amplified by whatever this rendering problem is. However as Noah says, lets wait til we get to check the 4K RAW (or O7) with a problematic case and see how bad the original is - i have a feeling that some of these issues are part of the cameras post scaling/debayering.

cheers
Paul

+1

 

[Marino215]

Haha. Felt the same way, but I understand what he meant. A. At that price points it's the best Sony will offer.
B. Get better lenses and C. Make better decisions when shooting in less forgiving environments, be more creative.

Its was a funny email however did make senses regarding the lenses. Oh well

 

[paulcurtis]

I somewhat doubt this particular matter is related to the lens. this shot was using a Canon FD SSC 1.4 (a lens generally appreciated for its sharpness and lack of CA) shooting at a 2.8/4.0 split. and to be clear, what you are seeing here isn't some extreme pixel peeping blow up... this is just a crop of the right side of the frame (ie 1080 lines vertically) recorded externally at 280Mbps. you can see that the fringing and aliasing disappears in the areas of lower contrast, but reappears (on the same stand and card) when the ratio increases. notice the white card has an artificial black line at the transition point.
as I think you've stated before, Paul, the 1080 out of this camera could look better.

Weeelll, that's not a very modern lens, coatings aren't modern and quite a bit of flare (i tried the lens way back). Also 2.8/4 is still pretty large aperture for S35. Stopping down to f8 would be preferable for a test. Also a lens designed to be f1.4 has compromises made even if you stop it down. A lens with a f2.8 max aperture would generally be noticeably better at f4 than a f1.4 at f4.

So i still maintain that the core source of the fringing is in the lens but the debayer/sensor has emphasized it and made it a lot worse. If the debayer tries to intelligently recreate colour based on surrounding pixels then the bright colour fringe could fool it into generating false results. Debayering is a reconstruction effort and i'd imagine a firmware based one wouldn't be able to be quite as intelligent as a software based slower one. Perhaps Sony is also sharpening the debayer result as well, creating false edges in high contrast areas.

A good test would be to do that scene again and then take a JPG still with the fs700 and see the difference. The still is Quad HD and can you see the same fringing issues? So same sensor, same set up but one is full sensor debayer and the other the 1080p scaled...

I think the post from the sony product specialist is worth reading in depth, because it shows that a camera system is a system, made up from lots of components and design is a series of compromises. It sounds a little condescending and simplified though at first read through.

cheers
Paul

 

[NextWaveG]

All I can say is my 5D MkII never gave me these highlight issues.

 

[starcentral]

The FS700 has been out for over 1.5 years right? Was this issue only spotted now or have other owners seen this kind of thing, or posted any previous samples or threads about it?

I agree with NextWaveG, there are a ton of cameras I have never seen anything like this on, so is there any chance this was a one-off or specific situation with lens flare or backlight causing some breakup in the highlight areas?

 

[paulcurtis]

The FS700 has been out for over 1.5 years right? Was this issue only spotted now or have other owners seen this kind of thing, or posted any previous samples or threads about it?

I agree with NextWaveG, there are a ton of cameras I have never seen anything like this on, so is there any chance this was a one-off or specific situation with lens flare or backlight causing some breakup in the highlight areas?

It's always been there, but it's in the extreme cases usually. I would guess the design of the 700 sensor is a less refined version of the F5/F55. Perhaps some of these issues have been addressed in the 5/55 or perhaps the firmware side is more intelligent.

I've mentioned several times that there are aliasing issues in 1080 which is probably more connected to this, than actual aliasing. But then it's never 'ruined' a shot, it's an annoyance but only if you're pixel peeping. The only time i've had real issues is with a monitor replacement shot where the original (providing lighting) was on and high and the edges in front of it showed this problem.

Before going all crazy at Sony, i think we should wait a few weeks and test properly with the 4K output.

Or, as i mentioned above, next time someone has the problem take a 4K still with the 700 photo option as a double check...

cheers
Paul

 

[jcs]

The Sony rep did not address the bug (perhaps an = somewhere in the code instead of == ) and instead made general filmmaking comments which have nothing to do with the actual bug. I would prefer no answer and "we're looking into the issue"...

At launch, this would be a lower priority bug, now at > 1 year it's time to fix it (or state a valid reason such as hardware flaw or patent issue, etc.).

 

[stefanhuberfilms]

petition signed and shared - lets hope they will fix something!

 

[sunnyjimfilms]

Judging by the response from Sony this isn't considered to be a bug, more a limitation of a camera in this price range, which to be fair, is at the low end of (professional) cameras.