D Log or D Cinelike??

[David_Cambridge]

Brian,


You points are well taken. I would expect banding to be a problem as well. However, in the testing I have done, it hasn't been a problem. If it had showed up, I would have switched to D-Cinelike, which is slightly higher contrast, but still flat.

It may be a difference in the test subjects we used. Mine were relatively high contrast to begin with, so little range expansion was required, and no visible banding occurred. The math behind an 8 bit image does work. The average eye can perceive about 100 different luminous levels in a video image. An 8 bit image capture has 256 luminous levels, so there is a fair amount of expansion space available before the average eye can pick up banding.

The problem that D-Log solves for me is shooting relatively high contrast scenes, like sunlight scenes where there are both deep shadows, and bright highlights, like clouds. In these situations, the dynamic range limits of the DJI sensor make it very difficult to hold detail in both the highlights and deep shadows. Digital clipping, particularly in the highlights, is very ugly to my eye.

 

[Raptorman0909]

For me, the banding problem is worst when the sky is of low contrast as is fairly common here in the high desert of Utah. When the luminosity range in the sky is low and you have only a few levels to code that then you get very obvious banding. I have less banding problems when the sky is more contrasty and has more detail (clouds etc). Here in the high desert of Utah, for much of the year between late April and October, there can be many days without so much as a single cloud and it's those low contrast skies where banding really rares it ugly head.


Brian

 

[David_Cambridge]

Brian,

Makes sense. What kind of post processing do you do? How much adjustment in post do you make? What kind of monitor do you use for evaluation?

All of this is a tradeoff between a limited dynamic range, and the bit depth of the sensor. I don't know what the dynamic range of the DJI sensors are, but they are a lot less than the Nikon's 14 stops. This for me means choosing dynamic range compression via either D-Log, or D-Cinema transfer functions. Unlike the high desert, in New England we are frequently shooting where the dynamic range of the scenes are pretty broad. I end up shooting to just preserve detail in the highlights, as measured by the zebras. That typically ranged from -.7 EV to - 2.5 EV, protecting the highlights while maintaining some shadow detail. From a perceptual point of view, this works well, as the bit mapping to luminosity is logarithmic, with more bits allocated to the highlights, and fewer to the shadows. If it is a little hot, I still have detail in the highlights, and can bring the shadows down to what I want, retaining the maximum number of bits for the low tones.

All in all, I am pretty satisfied with the 4 K results I get on a well calibrated 4K TV (Sony XBR). The material from the Phantom cuts well with 4K professional video from various other sources. I have yet had a chance to throw it up on a DCI theatrical projector,but I fully expect to be somewhat disappointed when I do.

 

[Raptorman0909]

I use Premiere Pro CC for editing and use a custom built PC that I built just over a year ago with an I7-5820K CPU that I over-clock to 4.4GHz as well as an EVGA nVidia 980 Ti Hydro and both the CPU and GPU are water cooled. In addition I have 32GB of DDR4 that is running at 3200 with 14-14-14-34 timings and a CAS of 14. My main video/image storage is two 6TB WD Black HD's that typically average about 200MB/sec. Lastly, my boot/OS (Win 10)/program drive is a Samsung 950 Pro PCIe SSD of 512GB. My visual output is to a BenQ PV3200PT 32 Inch 4K Monitor running at 60Hz.

My style of video is more natural and I don't tend to exaggerate the saturation or sharpening so my workflow is pretty simple and minimal. Generally a bit of contrast adjustments, often using 'curves', as well as a tad of saturation and sharpening in post. In camera I run with D-cinelike with -1/0/-1 as well as full manual exposure control with ISO100 and f/5 or f/5.6 most of the time. I have not experienced rolling shutter/jello issues in spite of the fact that my shutter speed is typically in the 1/640 to 1/1280 but can be higher or lower depending on lighting. Interestingly, my videos tend to be really sharp with more detail and I think the higher shutter speeds may be important in this.


Brian

 

[David_Cambridge]

Brian,

Nice setup. It is a real treat to have good systems to work with.

I'm on a Mac Pro, 8 cores,NVIDIA GeForce GTX 780 6144 MB, with 47.8 TB of disk storage, most of it level 0 RAID. Boot drive/applications is a Other World Computing SSD. 24 GB main memory. Apple Cinema display, so I don't get full 4K until I get to the 4K TV. Both Cinema Display and TV calibrated to Rec 709, gamma 2.2. Post software Final Cut Pro X 10.3.3, DaVinci Resolve 14b, Apple Motion & Compressor for VFX and finishing. I typically don't make many adjustments is post - just expanding the dynamic range, and adjusting the mid point to taste. Sometimes a little saturation with D-Log, none with normal

The higher shutter speed does make a huge difference. I think you are right there. I prefer the sharpness that comes from a higher shutters speed as well, and the drone footage changes so slowly, unless you are in the middle of a crash, so that the stuttering from the high shutter speed isn't noticeable and the results have a crystalline clarity. The Phantom 3 4K has a Sony sensor I think, and I like it pretty well. If it is the one I think, it has pretty good specs.

 

[Ade5]

Hey is it possible to obtain the LUT you used for this footage ? Really love the tones and I read that you could send them by email. Thanks in advance.

You better quote the guy your asking so you get a response..

 

[Juliencb]

Im going with Cinelike for now.. Some testing here:

Awesome footage mate. Really love the tones. I read that you could send the lut by email. Where can i find yours ?

 

[Ade5]

Awesome footage mate. Really love the tones. I read that you could send the lut by email. Where can i find yours ?

Check PM in this forum..

 

[Ben01]

Check PM in this forum..

Any chance you could PM my way as well mate ???

 

[Ade5]

Any chance you could PM my way as well mate ???

IWLTBAP Jon Olsson Look - Standard V2.cube

Meet Google Drive – One place for all your files

Here you go, two LUTS for that look

 

[davemex]

Brian,

Nice setup. It is a real treat to have good systems to work with.

I'm on a Mac Pro, 8 cores,NVIDIA GeForce GTX 780 6144 MB, with 47.8 TB of disk storage, most of it level 0 RAID. Boot drive/applications is a Other World Computing SSD. 24 GB main memory. Apple Cinema display, so I don't get full 4K until I get to the 4K TV. Both Cinema Display and TV calibrated to Rec 709, gamma 2.2. Post software Final Cut Pro X 10.3.3, DaVinci Resolve 14b, Apple Motion & Compressor for VFX and finishing. I typically don't make many adjustments is post - just expanding the dynamic range, and adjusting the mid point to taste. Sometimes a little saturation with D-Log, none with normal

The higher shutter speed does make a huge difference. I think you are right there. I prefer the sharpness that comes from a higher shutters speed as well, and the drone footage changes so slowly, unless you are in the middle of a crash, so that the stuttering from the high shutter speed isn't noticeable and the results have a crystalline clarity. The Phantom 3 4K has a Sony sensor I think, and I like it pretty well. If it is the one I think, it has pretty good specs.

I just want to chime in here too... I've tried both D-Log and D-Cinelike and found the D-Log to band really severely. I was using ND filters to keep my shutter speed to either 1/30 or 1/60 to match up with the 4K 30 FPS H.265 video I was shooting. For me, especially flying in places with a lot of sky and water both in the scene (very common for me), I saw lots of banding in both the sky and water even if the detail in the land/buildings was sharp. It was just unwatchable for me. For reference, I extract my H.265 video to DNxHD and import into DaVinci Resolve to grade/edit on a LG 4K OLED HDR TV (10 Bit). I'm doing everything I can to get as much data out of the video feed as possible. I just couldn't get D-Log to NOT band but Cinelike looks amazing by comparison.

If there's something else I can do to get D-Log to NOT band, I'm completely open to ideas.

 

[David_Cambridge]

Banding is typically caused by cranking up on the contrast in post. Did the original footage before adjustment show the banding? My testing with the 4k footage from my P3 and D-Log gamma is fine, and a normal adjustment didn't band noticeably. Did you need to make any exposure adjustments? D-Cine is very similar to the log curve, so if it works better for you use it. The point of using either is to give you something with more flexibility for adjustment in post.

 

[Raptorman0909]

You don;t have much choice but to stretch the contrast in post if you use D-log otherwise the video is unusable flat. You can get banding no matter what color profile you use owing to the 8-bit limitation of the camera, but using D-log makes it worse because you have to stretch the contrast in post. Put simply ... D-log allocates more levels for mid-tones leaving less for highlights so the banding will tend to be worse with D-log. D-Cinelike is and has been my standard color profile and I doubt I'll even bother to test D-log again...


Brian

 

[David_Cambridge]

All true, but the highlights get a disproportionate number of bits to begin with. The human eye perceives luminosity in a very non-linear fashion as well. Where you put the exposure makes a big difference as well. Again, this is a case of what works for your workflow, not an absolute right or wrong. 8 bits give you 256 luminous levels, the average human eye can only differentiate 100 levels, so there is a fair amount of opportunity to shift the gamma response around to get the results you want.

 

[Raptorman0909]

Yes, getting the exposure right is hugely important and is another reason D-log sucks as it makes getting the exposure right harder. It's harder because the video feed is more washed out so judging exposure from the video feed is more difficult. It's harder also because the histogram displays the compressed video info and not the raw video info so the histogram might look just fine while the video is significantly over exposed. But, in addition, when fewer levels are available for the highlights AND you have to stretch the contrast as you must with D-log THEN the bin values in the highlights increase to the point were even a single bit change is noticeable hence banding!


Brian

 

[David_Cambridge]

If banding is the biggest issue, use the standard setting. It makes the least effort to compress the tonal range. It's biggest danger is clipping, most likely the highlights.

Note also, that all of this discussion makes some assumptions about the tonal range of the subject. Subjects with a broad tonal range will get clipped with the standard, and not with either D-Log, or D-VCinema. Subjects with a narrow tonal range will exhibit banding after expansion of the range in post.

For bright sunny subjects, particularly partially cross lit D-Log with an exposure adjustment of - 1 stop works best. For overcast, low contrast scenes, normal works best. D-Cinema also works well with contrasty scenes, but is more likely to clip. Picking one "best" depends a lot on the subject, and making blanket "best" statements willed to sub-optimal results, depending on luminous range and exposure.

My working approach is:

Very bright contrasty scenes - D-log and -1 stop

Normal contrast scenes - D-Cine and -.7 stop

Low contrast scenes - Normal and -.7 stop

I very much appreciate DJI's providing these choices, as they provide maximum flexibility from an 8 bit sensor.

 

[Don Barar]

Hi David,

Thank you for the guidelines. To confirm as an example for Bright Contrasty scene you stated to use DLog and - 1 stop. For the minus one stop are you stating one stop of under or over exposure?

Look forward to your response.

Don Barar

 

[David_Cambridge]

Yes, - 1 stop. You may have to adjust that for your particular aircraft/combination. Every camera's exposure control is slightly different. They all perform the came on the gamma curves, but the exposure controls vary due to slight manufacturing differences. You should use either a waveform monitor in your post software, or a histogram display for still pictures.

For first class professional work, learning these tools will give you the best results possible for your a/c. Most professional videography is set up using waveform analysis. You can see the effects of exposure adjustments and the gamma curve interprttation for any given scene the camera is looking at.

 

[David_Cambridge]

Sorry if I wasn't clear in my response. -1 stop is 1 stop underexposed, and + stop is 1 stop over exposed.

 

[Don Barar]

Hi David,

Hmmm. I have reviewed a number of articles on the topic on how to expose videos recorded in some log format. Most claim that you should "expose to the right" ( ETTR) 1 to 2 stops, i.e. 1 - 2 stops over exposure.

My experience has been that when I adopted this practice of over exposing 1 - 2 stops with DLog that the result of color grading my videos improved dramatically. And, I have yet to experience any issues with posterization, i.e. banding.

Why do you believe that under exposing 1 stop is the correct thing to do with the DJI Phantom 4 Pro camera when recording in DLog?

Don