Banding Issues

[David Cho]

Hey, I am shooting at recommended settings 5000K. Looks like I'm getting some banding issues at the right side of the wall. I've experienced this before and it looks terrible. Is this normal with the RED or is it just my iMac monitor?

Also, my camera seems to be adding a tint to the image. I open up my file in red alert and I see that plus 10 tint has been added to my file, even if I do not have any type of gain, curve, or anything modified in camera.

 

[Jamie Havill]

Surely the tint has something to do with this?

 

[Graeme Nattress]

Tint is part of white balance (kelvin and tint): Kelvin being the correlated colour temperature and the tint being a compensation for a light source not purely on the black-body radiator curve.

5000k is not a recommended setting, but as the sensor thinks white looks white around that colour temp, it's a good temp to have the lights at so that dynamic range is maximized. If not, not usually a problem, especially as if you have any clipped highlights DRX can fix their colour and put missing detail back into them.

Now, onto banding. What precisely do you mean by that as some people think of it as vertical or horizontal lines, and others posterization bands.

Graeme

 

[Imran Farouk]

Surely those lights are tungsten and isn't around 5,000K daylight settings so just switching the white balance in RED Alert or whatever program your using to 3,200K or around there would get rid of the banding on the right side wall...

 

[David Cho]

Now, onto banding. What precisely do you mean by that as some people think of it as vertical or horizontal lines, and others posterization bands.

Graeme

Graeme,

Do you think that my camera is automatically adding tint without my knowledge? I even set my "look" settings in my camera and made sure everything was set to normal. I'm still getting this tint.

As for banding, I'm talking about the shades not having a smooth gradient, like you can see from the right side of the wall.

- David

 

[David Cho]

Surely those lights are tungsten and isn't around 5,000K daylight settings so just switching the white balance in RED Alert or whatever program your using to 3,200K or around there would get rid of the banding on the right side wall...

I tried this and it didn't work. Thanks for the suggestion though!

More info: I'm using Zeiss Standard Speeds PL mount for 35mm format, focal length 16mm.

This occurs with all of my lenses.

 

[David Mullen ASC]

Have you tried converting an R3D frame in different settings? I don't mean different white balance or color settings, I mean to different formats (Log, Camera RGB, etc.) Often that sort of banding is due to a conversion error.

 

[Graeme Nattress]

Tint will normally come when you do an auto-white-balance. It's a necessary part of how WB works.

The banding is most likely a monitoring issue. Say you export a 16bit tiff from a RED App - your monitor is 8bit. You could very well see contouring / banding due to the decimation from 16bit to 8bit. What I'd suggest is to use photoshop, say, to convert from 16bit to 8bit which dithers down properly, then look at it.

I've examined many examples of banding in RED footage and it's always been a monitoring artifact or a lack of dither in creating an 8bit from a higher bit depth.

Graeme

 

[David Cho]

Have you tried converting an R3D frame in different settings? I don't mean different white balance or color settings, I mean to different formats (Log, Camera RGB, etc.) Often that sort of banding is due to a conversion error.

Hey David,

I'm not sure if this matters but I see the banding in my original R3D file when played to quicktime in 4k, in RED Alert, and also in my 5.6 inch Red LCD monitor when I am shooting. It's very clear when I am shooting a blank wall, it looks like a strange bullseye type of pattern that grows and shrinks when I tilt my camera up and down.

 

[Graeme Nattress]

It's most likely a display artifact. Most displays are 8bit or less. You get these effects if you don't properly dither down the high bit depth of the R3D to your display bit depth.

Graeme

 

[David Cho]

It's most likely a display artifact. Most displays are 8bit or less. You get these effects if you don't properly dither down the high bit depth of the R3D to your display bit depth.

Graeme

I apologize, how would I properly dither down the bit depth of the R3d to my display bit depth with a relatively new iMac?

 

[Milan Spasic]

I apologize, how would I properly dither down the bit depth of the R3d to my display bit depth with a relatively new iMac?

I doubt you will see banding in ProRes or when played back at 1080p on a proper video monitor.

Also, the banding usually appears in areas of the frame that are not properly exposed. So I'd say, use false color when shooting and avoid purple!

 

[David Cho]

I doubt you will see banding in ProRes or when played back at 1080p on a proper video monitor.

Also, the banding usually appears in areas of the frame that are not properly exposed. So I'd say, use false color when shooting and avoid purple!

Yeah, I've been seeing banding on my ProRes HQ Half High files. Is my Imac not a proper video monitor?

 

[David Cho]

Also, is it normal to see banding on areas exposed at 5-25 IRE with the Red 5.6 inch lcd?

 

[Daniel Browning]

Also, the banding usually appears in areas of the frame that are not properly exposed.

Personally, I find it's even more likely to appear in areas that are well exposed (like blue sky), because then there is not enough noise to dither.

 

[Adrian Hernandez]

Also, the banding usually appears in areas of the frame that are not properly exposed. So I'd say, use false color when shooting and avoid purple!

I've noticed some banding too, especially in scenes with opened skies and large portion where the light smoothly falls from the source in the image (i.e walls, halls).... the false color indicated a grey-green in that part, i don't think is due a WB issue, cause the scene was illuminated with HMI and the camera was balanced to 6000 K.

Then, in RedAlert, i exported some DPX and TIFFs and send them to Lustre.... the banding is still there. The same occurred when worked in After Effects, opening a R3D, and the same dpx / tiff sequence.

Is there any optimal way of rendering from RedAlert to dpx?
if the workflow includes a print ending (such Vision Premiere 2), what would be the RECOMENDED workflow with: RedAlert - After Effects - Lustre - Print Film... avoiding any kind of "compression".

 

[Graeme Nattress]

Yes, but how are you displaying the >8bit depth of the raw file on your most probably 8bit monitor? Is Lustre dithering it's output or not when it downconverts the bit depth for display purposes?

Graeme

 

[David Mullen ASC]

I doubt you will see banding in ProRes or when played back at 1080p on a proper video monitor.

Also, the banding usually appears in areas of the frame that are not properly exposed. So I'd say, use false color when shooting and avoid purple!

That's crazy -- what if someone is wearing a black T-shirt in a scene? There's nothing wrong with a frame that has light values that extend all the way to pure black thru near black, dark grey, etc. otherwise all you would have is flat, low-contrast images. I would almost say that a frame that doesn't have something "purple" in the false colors would be rather boring...

The solution to noise is not to avoid dark areas in the frame, the solution is to have enough overall signal so that your dark areas in the frame can stay dark, if not go darker, after color-correction.

 

[Adrian Hernandez]

Is Lustre dithering it's output or not when it downconverts the bit depth for display purposes?

Don't know for sure...need to ask the people @ the lab
But what intrigues me is: the banding appearing on the monitoring will appear on prints?

 

[Graeme Nattress]

Well, I've never seen banding in the RAW data, even under extreme levels. I have seen it in how that data is displayed, but never in the image itself unless you reduce too many bit levels - like save out an 8bit tiff, without first dithering down to that bit depth.

Will it appear on prints - don't know, but it shouldn't. You could take the dpx into photoshop, convert down to 8bit which will dither, and then examine.

Graeme