I've recently purchased an a7iii and am noticing that i get color banding in some of my shots, especially when shooting sky. It fades in bands of color from dark blue to lighter shades. The untrained eye may not see it but crank up the contrast and its very obvious so I know I must do something. I shoot landscape photography for a resort so this is definitely a big issue. It is there when using different lenses so I believe it is a camera body issue and also in uncompressed RAW and ultra fine jpeg images so not a matter of quality. It was present at 1/800 a second & also at 1/8 second shutter speeds.
Any suggestions on what I can do to eliminate this would be greatly appreciated. It's there with and without the electronic front curtain shutter; at first I thought that may be the issue but it doesn't appear to be. Has anyone else dealt with color banding on mirrorless, particularly the a7iii? I switched form Canon so it's relatively new territory to me.
jackriley wrote:
I've recently purchased an a7iii and am noticing that i get color banding in some of my shots, especially when shooting sky. It fades in bands of color from dark blue to lighter shades. The untrained eye may not see it but crank up the contrast and its very obvious so I know I must do something. I shoot landscape photography for a resort so this is definitely a big issue. It is there when using different lenses so I believe it is a camera body issue and also in uncompressed RAW and ultra fine jpeg images so not a matter of quality. It was present at 1/800 a second & also at 1/8 second shutter speeds.
Any suggestions on what I can do to eliminate this would be greatly appreciated. It's there with and without the electronic front curtain shutter; at first I thought that may be the issue but it doesn't appear to be. Has anyone else dealt with color banding on mirrorless, particularly the a7iii? I switched form Canon so it's relatively new territory to me.
The shape of the banding/posterization is the same as the front element of the lens, which makes me suspect a filter problem or lens design issue.
Also, if not using a filter, maybe try a different lens.
If there is no filter, then there is possibly something seriously wrong with your lens. On the other hand, that lens has some wicked vignetting in the corners at 24mm, even stopped down. It may be a cause of your issue as well. In fact, I have never seen a lens with such extreme vignetting. It may not be a viable option for your work when used at 24mm.
Regarding the shape if the bands, they are same shape as the lens, however this is likely due to the shape of the sunset and fading to darkness in a circular pattern. It presents as vignette but is just the natural light. I get the banding on a 50 mm prime too so likely not the 24 mm lens.
jackriley wrote:
I've recently purchased an a7iii and am noticing that i get color banding in some of my shots, especially when shooting sky. It fades in bands of color from dark blue to lighter shades. The untrained eye may not see it but crank up the contrast and its very obvious so I know I must do something. I shoot landscape photography for a resort so this is definitely a big issue. It is there when using different lenses so I believe it is a camera body issue and also in uncompressed RAW and ultra fine jpeg images so not a matter of quality. It was present at 1/800 a second & also at 1/8 second shutter speeds.
Any suggestions on what I can do to eliminate this would be greatly appreciated. It's there with and without the electronic front curtain shutter; at first I thought that may be the issue but it doesn't appear to be. Has anyone else dealt with color banding on mirrorless, particularly the a7iii? I switched form Canon so it's relatively new territory to me.
The file is about a stop and a half underexposed, even if you want to retain in the very bright highlight area. See the histogram attached.
If we look at the upper left corner stats, we can see the blue channel average is about 900. The black point is 512, so the net blue average is about 400. That's more than five stops down from full scale. Because of the PDAF patterning, the a7II can get banding in smooth gradients. It is more apparent in the blue channel. It is more apparent when the image is underexposed.
Looks like you hit the trifecta.
Jim
PS. What shape are the bands you're complaining about? If they're arcs, it could be the lens vignetting correction.
Jim,
Thank you for the in depth look at that. The banding really does only show itself in the blues (skies) so that makes sense. I guess I'll just play around with it more and make sure to not underexpose too too much.
Thank you
-Jack
jackriley wrote:
Jim,
Thank you for the in depth look at that. The banding really does only show itself in the blues (skies) so that makes sense. I guess I'll just play around with it more and make sure to not underexpose too too much.
Thank you
-Jack
Try hitting it with a small median filter, followed by adding back in some Gaussian noise.
I examined the OP's A7III raw and found some interesting things. First, the camera had vignetting correction set to "Auto", which means on for the 24-105mm @ 24mm used in the photograph. However, the image was taken @ f/10, where one would expect the vignetting correction to be minor. To see if this is the case I took two sets of blackframes on my A7rII with the 24-105 @ 24mm - one set at f/4 and one at f/10, and graphed what the mean raw values looks like horizontally across the sensor. Here are the results:
The top row is for f/4, with vignetting correction on (Left) and off (Right). Notice how the mean raw values spike around the edges for the left graph - that is Sony applying the vignetting correction to the raw data (ie, increasing the raw values [brightness] at the edges of the frame)
The bottom row is for f/10, again with vignetting correction on (Left) and off (Right). Notice how there are no raw value spikes around the edges. However, the mean value of the raw data is elevated when vignetting correction is on, which probably isn't significant for this investigation. Based on this experiment I doubt the vignetting correction is materially influencing the raw data to product the visual posterization in the OP's image (more investigation needed though).
I then turned to examining the raw data itself. To my surprise it's 12-bit compressed, even though it was taken in single-shot mode with the mechanical shutter (EFCS). Jumping over to the dpreview Sony forum it looks like Jim already discovered this (link). More interesting however is the fact that the ADU gaps increase from 2-bits to 3-bits at exactly ADU 3200, which almost implies the bit depth drops from 12 bits to 11 bits (I need to look more into this though). Here's a snapshot of some of the ADU data from the raw, obtained via my rawshack utility:
The leftmost column is the ADU value, ie raw level. The next four columns list the number of pixels that are present in the image at that ADU value, for each of the four color channels.
It's not clear to me if this drop 11 bits is the result of Sony's known tonal compression scheme. I need to look into it.
Some work needs to be done to see if there's any correlation between the 12-bit ADU gaps and the visual posterization that is present in the image. Note that the sky ADU values are approx 1900 for the green channel, where the ADU gaps are 2-bits wide, so I don't think the 3-bit ADU gaps starting at ADU 3200 are uniquely correlated to the visual posterization.
I thought it may have been baked-in vignetting correction. Strange that they would do that to raw files. That lens actually still has a lot of vignetting at f10 24mm, so no surprise there. I've pushed a lit of files on various cameras for sky shots and never saw anything like this before.
Thanks for the detailed analysis. Very informative. Hopefully OP can try a similar shot with auto vignetting turned off and see how it turns out.
Btw, here is an Sony compression overlay image I created using my Sony compression overlay utility that shows where Sony's tonal compression is being employed in the OP's image. It's mostly at the boundary of the sky and ground and then content below that - there are plenty of > 3200 ADU pixels in the sky so I don't think the Sony compression is behind the drop in precision here.
Ok, just checked on my A7rII and I see the same drop in precision. Also determined the drop doesn't occur when using uncompressed raw. Here are the results:
A7rII Uncompressed, "14-bit" - no drop in precision
A7rII Compressed, "13-bit" (Mechanical shutter/EFCS): Precision drops to 12-bits at ADU 2000
A7rII Compressed, "12-bit" (electronic shutter): Precision drops 11-bits at ADU 3200
Just curious, has this issue been resolved? Would be interesting to know what was causing the problem and if the solutions of either using 14 bit uncompressed or vignetting correction disabled fixed it.
Yes it seems to have been fixed by switching to uncompressed 14 bit. The vignette on my 24-105 seems to be unavoidable at 24mm which is a shame but the banding is gone and that was the big concern. I'm relatively new to professional photography and really appreciate all the input and help from this community on my issue. Thank you to everyone for their advice and direction. Love the a7iii and will be sure to share more of my work as it comes. I could always use some helpful criticism to grow.
jackriley wrote:
Yes it seems to have been fixed by switching to uncompressed 14 bit. The vignette on my 24-105 seems to be unavoidable at 24mm which is a shame but the banding is gone and that was the big concern. I'm relatively new to professional photography and really appreciate all the input and help from this community on my issue. Thank you to everyone for their advice and direction. Love the a7iii and will be sure to share more of my work as it comes. I could always use some helpful criticism to grow.
Thanks for posting an update. If it's not too much trouble and when you have time could you photograph a scene that produces the banding using both uncompressed and compressed raws and then PM me the raws? I would like to definitively nail this down so that the information can be shared with others.
snapsy wrote:
Thanks for posting an update. If it's not too much trouble and when you have time could you photograph a scene that produces the banding using both uncompressed and compressed raws and then PM me the raws? I would like to definitively nail this down so that the information can be shared with others.
Absolutely yes I will do that. Thank you again for the guidance.
Cool. I wonder now if I could reproduce the same effect on my a6300 since it has no option for uncompressed RAW. In my research for info on Sony lenses, I found that some lenses have automatic vignetting correction applied regardless of the settings on the camera.
As much as I love what Sony is doing for the photo industry, I wish they would just play it straight with corrections and file formats instead of the complex minefield of caveats and block-box alterations of the RAW data. Lossless compression and 14 bit RAW in all modes seriously can't be that hard of a task considering how awesome they are at the tech and firmware side of many other aspects.
JohanEickmeyer wrote:
Cool. I wonder now if I could reproduce the same effect on my a6300 since it has no option for uncompressed RAW. In my research for info on Sony lenses, I found that some lenses have automatic vignetting correction applied regardless of the settings on the camera.
As much as I love what Sony is doing for the photo industry, I wish they would just play it straight with corrections and file formats instead of the complex minefield of caveats and block-box alterations of the RAW data. Lossless compression and 14 bit RAW in all modes seriously can't be that hard of a task considering how awesome they are at the tech and firmware side of many other aspects. ...Show more →
I believe they'd have to reduce some of the frame rates to do 14-bit precision all the time.
JimKasson wrote:
I believe they'd have to reduce some of the frame rates to do 14-bit precision all the time.
Jim
Makes sense I guess if they could otherwise not deliver such high FPS. Also makes sense that the faster readout noise and usual use of higher ISO during fast bursts would be fine to cut down below 14 bit.
Just as a Canon-only shooter for so many years, the concept of lowered bit depth, cooked RAW files, and lossy compression are rather strange to me. My nearly decade-old Rebel does 14 bit lossless, not that it can really make use of all those 14 bits though.
