Color mode & Dynamic Range

[yannick_c]

I'm pleased to see so many folks here working on this to the betterment of everyone else -- we have some knowledgeable people and this helps. I have posted a number of videos on the P4P and been amazed the the range of comments I get. One of the ones that I've seen a number of times in the last few days is that with D-log you are supposed to over expose a bit. Of course, with 8 bit the DR is going to be on the low side so exposing to prevent the sky from blowing out at all would mean the mid tones and shadow detail would crush to black so with 8 bit you have to sacrifice a bit of highlight detail so that you can bring up the mids and shadows. If you have something more like 14 bits then you have more headroom to expose for the sky and then bring up the mids and shadow.

One other thing that's been commented on and I've seen this a number of times is that to counter banding you need to use a slow shutter speed. The mechanism for this isn't entirely obvious but it would appear that adding a little blurring can help with banding -- anyone care to comment on this?

I think when they say you should overexpose, we may understand overexpose the underexposed (It's kind of weird I admit lol).

After reading that about slog-2 from sony (Exposing and Using S-Log2 on the Sony A7s. Part One: Gamma and Exposure.), the correct exposure level for the white are at 59%.

Let's say DLog is quite similar to this (maybe, maybe not, but let's say it), if we use the correct exposure, let's put the white at 59% (the underxposed). To get less noise in the shadows, we can overexpose (the underexposed) by maybe one stop, at around 71%

For the banding, I think (but it's just a supposition, I'm not sure of nothing there) that we are overexposing the DLog way too much (like if we were using None or DCinelike) and the values near the white are more compressed than if we had shot with the whites lower (less compressed in the middle). To test that, just do the same test, but at different exposure and look if there is still banding in the sky.

 

[Doctorcobani]

Excellent work guys!
So, we can use D-LOG as usual? -1/0/0 it's the way to go?
Use D-LOG but trying to underexpose? I don't understand the point when D-LOG looks like has crushed highlights compared with None using the same exposure...

Enviado desde mi Nexus 6P mediante Tapatalk

 

[phynesse]

This is a nice comparison. Any chance you might repeat it with the latest P4P firmware update and the changes to DLOG? I'm curious as to how the P4P's usable dynamic range compares with the RX100 now.

I ll try to get to it this weekend, also really curious how dlog has improved. Been away on work 3 weeks straight...

Sent from my ONEPLUS A3003 using PhantomPilots mobile app

 

[Tomas Wangen]

Excellent work guys!
So, we can use D-LOG as usual? -1/0/0 it's the way to go?
Use D-LOG but trying to underexpose? I don't understand the point when D-LOG looks like has crushed highlights compared with None using the same exposure...

Enviado desde mi Nexus 6P mediante Tapatalk

There are a total of 256 shades for 8 bit video. What dlog does is use much more of the shades for the darker parts of the image leaving fewer for the highlights. You should still avoid underexposing.

For low contrast scenes none exposed to the right may very well give a better result than dlog.

 

[Raptorman0909]

There are a total of 256 shades for 8 bit video. What dlog does is use much more of the shades for the darker parts of the image leaving fewer for the highlights. You should still avoid underexposing.

For low contrast scenes none exposed to the right may very well give a better result than dlog.

Yeah, if D-log uses more levels to handle mid-tones and shadows then there will be fewer levels to code the highlights -- no way around that. Under the right circumstances D-log can work just fine, but the percentage of times when that's the case seems to be a bit on the low side as far as I'm concerned. If you can limit or eliminate bright areas and the scene has a lot of shadow detail then D-log may be the right choice, but if you have bright sky I don't see how D-log can work without very noticeable banding.


Brian

 

[Doctorcobani]

Thanks Wangen and Raptorman for all your support! This forum is amazing and your help too!

Enviado desde mi Nexus 6P mediante Tapatalk

 

[yannick_c]

Have done some quick test with DLog and exposure and can't really see banding (only if you really go too far on grading, or maybe a little if it's way too overexposed). I think that Raptoman, you really overexposed your scene, that's why there's banding in the source file.

The source file for my quick test (not the most scientific test, and I think also too static). I just bumped the exposure and moved a little. DLog (-1,0,0) H265 : https://we.tl/oIvgyGRvQK (730Mo)

 

[2nd2non]

Maybe a GRAD filter will help. I used one with my P3P and it helped keep sky exposure closer to ground, thus less chance to overexpose sky and contribute to banding


Sent from my iPhone using Tapatalk

 

[LazyEightFilms]

As promised, here's my exposure tests using "D-Cinelike", "None", and "Art".

To view this content we will need your consent to set third party cookies.
For more detailed information, see our cookies page.

Accept third party cookies

Brian

That is such a shame because the Sony ExmorR Sensor is such a great sensor (as seen in the Sony RX100, RX10, and others). This camera could be truly phenomenal if DJI could fix their image processor.

 

[Raptorman0909]

Have done some quick test with DLog and exposure and can't really see banding (only if you really go too far on grading, or maybe a little if it's way too overexposed). I think that Raptoman, you really overexposed your scene, that's why there's banding in the source file.

The source file for my quick test (not the most scientific test, and I think also too static). I just bumped the exposure and moved a little. DLog (-1,0,0) H265 : https://we.tl/oIvgyGRvQK (730Mo)

At the end of the clip, as I turn towards the Sun, the sky gets blown out, but before I turn towards the Sun there is nothing that I can see that's over exposed yet even at the beginning of the clip the banding is evident in all three versions -- including the raw footage.


Brian

 

[yannick_c]

I wasn't talking about clipping but about the overall brightness that may be too high in DLog giving less values for the highlights and that can cause banding.

 

[Tomas Wangen]

I did a small a/b test yesterday (dlog/none) and the results can be found here: DLOG

 

[Raptorman0909]

I wasn't talking about clipping but about the overall brightness that may be too high in DLog giving less values for the highlights and that can cause banding.

I can agree with that. Interestingly there are folks that say you have to overexpose when using D-log, but that doesn't make sense for the reason you mention. Allocating most of the levels for mid tones makes sense for studio work with cameras that have 12 bit or more, but with an 8 bit camera D-log appears to be a recipe for banding unless you under expose the highlights. But that then risks crushing the blacks.


Brian

 

[2nd2non]

I can agree with that. Interestingly there are folks that say you have to overexpose when using D-log, but that doesn't make sense for the reason you mention. Allocating most of the levels for mid tones makes sense for studio work with cameras that have 12 bit or more, but with an 8 bit camera D-log appears to be a recipe for banding unless you under expose the highlights. But that then risks crushing the blacks.


Brian

This is why a GRAD filter can be very helpful

 

[Tomas Wangen]

I can agree with that. Interestingly there are folks that say you have to overexpose when using D-log, but that doesn't make sense for the reason you mention. Allocating most of the levels for mid tones makes sense for studio work with cameras that have 12 bit or more, but with an 8 bit camera D-log appears to be a recipe for banding unless you under expose the highlights. But that then risks crushing the blacks.


Brian

More than 8 bit would be fantastic, but I think at the moment one would then have to go all the way to raw footage (and the inspire2).

There is, however, surprising amount of tone to be extracted also for 8 bit as long as the exposure is correct and you do not have a bland, uniform sky in which case you will get banding (as you have demonstrated).

This sample frame is from footage I am editing now, and has about 10 stops of range (estimated based on dlog histogram) and no clipping.

I have isolated both shadows and highlights to give an indication what is present, and also added one creative toning.

Notice that it is here impossible to find any banding or other artifacts caused by the 8 bit limitation, which is good news.

original dlog

shadows

highlights

creative tone

 

[Roamer105]

I can agree with that. Interestingly there are folks that say you have to overexpose when using D-log, but that doesn't make sense for the reason you mention. Allocating most of the levels for mid tones makes sense for studio work with cameras that have 12 bit or more, but with an 8 bit camera D-log appears to be a recipe for banding unless you under expose the highlights. But that then risks crushing the blacks.


Brian

I agree. I have been using h265 with Main10 at a relatively bitrate (40+) and it does minimize or even eliminate much of the banding after grading it at full color depth. Of course, the TV has to be able to lay wide color gamut.


Sent from my iPad using PhantomPilots

 

[Raptorman0909]

The thing with banding is that this is something the NLE's like Premiere Pro etc should be able to deal with. If you could tell the program what area of the image is sky and have the program track that from frame to frame it could then smooth the banding or just about eliminate it altogether. Of course, if PP etc could track the sky as above then you could do some local adjustments to the sky that isn't applied to the ground. Hopefully PP etc will add these smarts into the editors as that would greatly improve our videos. The major picture studios have the time/money/people to go through the video frame by frame and do local edits to the sky etc, but us mortals can't spend 100 hours of editing per hour of video doing that so hopefully automated tools will develop to make that possible.

Yes, PP does have crude tracking capabilities, but not of the kind that would really help here.


Brian

 

[yannick_c]

Raptorman, there is After Effects for that kind of things.

 

[Tom K]

Raptorman, there is After Effects for that kind of things.

DaVinci Resolve too. And it's free.

 

[Tomas Wangen]

DaVinci Resolve too. And it's free.

In deed, a good mini-tutorial here:

To view this content we will need your consent to set third party cookies.
For more detailed information, see our cookies page.

Accept third party cookies