Your ideal NX1 Settings

[caseywilsondp]

From looking at your originals, I think you have your Saturation turned down to far. 

My personal settings: 

Gamma DR

Saturation - 1

Sharpness - 10

Contrast - 6

Black Level + 5

 

But I do think banding is an issue when it comes to grade. This is the best settings I have found so far for the camera.

 

interesting. So perhaps I can try leaving some of the contrast and mbl low, and just keep my saturation more in check? That seems like a trade off I'd be happy to make.

[Geoff CB]

Yeah it seems that there is a balancing act to be made between Contrast/Black Level/Saturation. Push certain combo's to far and the color info starts falling apart.

 

[Pavel Mašek]

I would let contrast around "0". I know that pushing it to negative will much improve DR but you lost so much colour information...

Right now I prefer Vivid (MBL +8 and saturation on -2 or -3) or default Dynamic DR with MBL 0 - however it seems that DR of both is quite similar.

I like Vivid becuase I think that colours are most accurate - DynamicDR mode has still some green cast eg. on grass but with vivid there is more yellow. You can see the "problem" of green cast here below on Andrew's comparsion (where I believe he used Standard or DynamicDR settings). I believe it should be more close to Canon 1DC with Vivid settings.

[Marco Tecno]

You can lower the green component in settings...

[KarimNassar]

 

Gamma DR, Contrast -10, Sharpness -10, Saturation -4, MBL +10, range is 16-235

Please share your thoughts.

I just tested and 16-235 appears to have more banding than the full range. 

I've read some people say it captures the same amount of shades than the full range and is just a read information for premiere or your editing software but it appears it is not accurate or there is some other issue.

Anyways don't see the point of using 16-235 in first place? Nothing wrong with the full range to begin with.

save file and open in Photoshop, some color shift with internet browsers.

[Marco Tecno]

We need to get to a point with the 16-235 vs 0-255 debate. I've been shooting 16-235 since Andrew advised to, but your post makes me feel bad :'( 

[KarimNassar]

We need to get to a point with the 16-235 vs 0-255 debate. I've been shooting 16-235 since Andrew advised to, but your post makes me feel bad :'( 

I don't get it either so I looked at the waveform. What I did in the picture below is shoot a high contrast scene, trying to get the fullest histogram possible without clipping the highlights nor the blacks. I then simply offset the waveform so both their respective black points sit on the 0 line.

As you can see the 0-255 file covers a wider tonality.

Of course I can up the gain of the 16-235 to match, but it is my understanding that with an 8bit codec stretching the information degrades the image quality. Now regardless if it does, that is an unnecessary extra step for every single clip. 

Now this is in premiere with native h265 support, I assume the 16-235 recommendation had to do with the transcoding of the files at the time. So now with the native h265 support that recommendation is no longer valid?

Regarding the banding on the wall with the light, I purposely underexposed the shot to highlight the issue in my earlier post.

If you expose properly the banding is almost unnoticeable:

[Marco Tecno]

Ok for the black points set to 0, but shouldn't you also take up the whites (in the 16-235 image), in order to get a comparable waveform?

[KarimNassar]

Ok for the black point, but shouldn't you also up the whites (in the 16-235 image), in order to get a comparable waveform?

Yes and that is exactly why I think we should not use 16-235. 

Note that I did not stretch the waveform to move the black point to the 0 line so they are comparable.

I offset the entire waveform without stretching nor modifying the captured information and simply moved it down.

To match the 16-235 to the 0-255 you need to stretch the captured tonality. You are modifying what the sensor captured and stretching the exposure to match.

And with an 8bit codec there is not much room to work with without degrading the file quality.

Here is what the respective waveforms look like with no modifications:

I see no clipping neither in the highlights nor in the blacks in the 0-255 and I see a thicker waveform.

All I see from the 16-235 file is that I will need an extra step of grading every time to expand the tonality and that it might result in IQ loss.

It seems to me that the native h265 support in Premiere changes the 16-235 recommendation but I'm all ears if I'm missing something here :-)

[kidzrevil]

Yes and that is exactly why I think we should not use 16-235. 

Note that I did not stretch the waveform to move the black point to the 0 line so they are comparable.

I offset the entire waveform without stretching nor modifying the captured information and simply moved it down.

To match the 16-235 to the 0-255 you need to stretch the captured tonality. You are modifying what the sensor captured and stretching the exposure to match.

And with an 8bit codec there is not much room to work with without degrading the file quality.

Here is what the respective waveforms look like with no modifications:

I see no clipping neither in the highlights nor in the blacks in the 0-255 and I see a thicker waveform.

All I see from the 16-235 file is that I will need an extra step of grading every time to expand the tonality and that it might result in IQ loss.

It seems to me that the native h265 support in Premiere changes the 16-235 recommendation but I'm all ears if I'm missing something here :-)

if this is truly the case then we shouldn't raise the master pedestal because it takes bits from the low end of the codec. 15+ master pedestal is essentially 16-255

[Marco Tecno]

Ok, I see. I'd like to read Andrew's take on this...

[KarimNassar]

Ok, I see. I'd like to read Andrew's take on this...

same. I'm not telling anyone what to do. Just my observations and I would be happy for some input as well.

[Geoff CB]

if this is truly the case then we shouldn't raise the master pedestal because it takes bits from the low end of the codec. 15+ master pedestal is essentially 16-255

If I underexpose slightly I get horrific artifact in the blacks at a certain point while grading. If you expose perfectly yes, but you won't have any leeway. Personal preference here, trying to find a flexible image without introducing to many artifacts. The perfect middle ground

KarimNassar Have you been using my settings on these recent tests? If so awesome!

Also great thing I discovered, while Speedgrade doesn't have native H265 support currently, you can force it by using this method posted by Sekhar Ravinutala on the adobe forums:

OK, I compared the plugins folder between Premiere Pro and SpeedGrade, and it looks like for some reason SpeedGrade didn't get the right H.265 importer during the 2015.1 update. I copied the new plugin (ImporterMPEG.prm) from the Premiere Pro's Plug-Ins/Common to SpeedGrade's Plug-Ins/Common, and now SpeedGrade shows the video! This is on Windows 10, and FYI the original (nonworking) plugin size on SpeedGrade is 660 KB, whereas the new (working) plugin on Premiere Pro is 661 KB.

It works perfectly, and I'm getting some incredible results from it. It's extremely processor intensive though, cannot run playback in realtime in 4K even with quality at minimum. It plays back in real time in Premiere when I switch back with the grade on the clips.

[kidzrevil]

 

If I underexpose slightly I get horrific artifact in the blacks at a certain point while grading. If you expose perfectly yes, but you won't have any leeway. Personal preference here, trying to find a flexible image without introducing to many artifacts. The perfect middle ground

KarimNassar Have you been using my settings on these recent tests? If so awesome!

Also great thing I discovered, while Speedgrade doesn't have native H265 support currently, you can force it by using this method posted by Sekhar Ravinutala on the adobe forums:

It works perfectly, and I'm getting some incredible results from it. It's extremely processor intensive though, cannot run playback in realtime in 4K even with quality at minimum. It plays back in real time in Premiere when I switch back with the grade on the clips.

You gave me an idea when you reccommended using rocky mountains movie converter. I am going to use gamma dr or any profile where I can capture a scene completely  "filling" the histogram. I will put it in the movie converter with an impulz lut that converts the image to cineon and output the 4k file to prores 4444 at 1080. This should essentially create a 10bit file in a cineon log colorspace. Visioncolor's lut's does a good job of pulling info from the shadows and highlights that I personally couldn't see with my naked eye. Hopefully this creates that middle ground you spoke of

[kidzrevil]

I tried converting to prores 444 at 1080 and still getting horrific macroblocking. Idk I think this is just something we are gonna have to deal with when it comes to the h.265 codec unless someone makes some type of breakthrough with these picture profiles

[Marco Tecno]

A fw hack for nx1 could increase bitrate and even enabling 4:2:2 internal recording..

[kidzrevil]

A fw hack for nx1 could increase bitrate and even enabling 4:2:2 internal recording..

the camera could easily do 100+ mbps but it's capped at 80 for compatibility with their tv's smh. I know there is a way around this...we pretty much have to expose to reveal as much shadow detail as possible or crush the shadows in post (mainly the very deep blacks). I've already beat the camera's in camera sharpening/edge enhancement with the right diffusion filter. That's already a better use of the camera's bit rate now I just have to tackle those pesky shadows and come up with a suitable post production workflow

[caseywilsondp]

finally admitting to myself that GammaDR, Sat-4,Contrast-10,MBL+10,16-235 is just too flat for 95% of stuff that's shot with this camera (with just so much banding)... 

So...just shot the same scene with a bunch of different settings including various gammas, contrast/saturation/mbl amounts, and changed up between 16-235 and 0-255. Shot proress hq to the shogun for all of it. I was planning on uploading still tiffs which should contain all the editable info needed, but would anyone prefer to have the whole prores file to mess with? post coming shortly. It's not every conceivable setting, but it should give us generally a good idea of where to go from here, especially with the luminence discussion.

I maybe could have shot some internal stuff, but honestly I think h265 is actually a pretty good codec (for 8 bit stuff) and i dont think outputting over hdmi is going to give a significant amount of help even with 422... at least on this camera (i found it quite helpful shooting externally on the a7s though).

should show off highlights, a little bit of shadow (back chair with figures on it) and banding on the back wall.

[caseywilsondp]

LONG POST WARNING!

Ok team, here are the files. I'm really hoping some people with more skill and knowledge can chime in. If something else needs to be tested, let me know soon and I can do it sometime tomorrow.

i've attached uncompressed 32bit tiffs of each setting... so they should be just fine for any test-grading. ill supply a link with the 10gb dnxhr 10bit file that includes all the clips, as soon as i can find a place to host it. 

here are the camera settings that didn't change:

iso: 160
shutter 1/50
aperture was set at a T2 on a 20mm Sony CineAlta prime
everything was shot through a 1/4 hollywood blackmagic filter. in my earlier test i tried both with and without, and it doesn't seem affect anything thats really relevant to these tests enough to make a difference. i personally like this filter in front of this ultra sharp sensor, so thats why i left it on.

first impressions, right now looking at the images, its like anything 'flatter' than 'normal gamma' starts to break apart just a little bit. i'm feeling like the different gamma modes (as have been stated before) really aren't doing anything different in processing other than adjusting some saturation and contrast.

16-235 settings seem to keep black details better than 0-255 everything being equal.

previews:

Tiffs:

 NX1_3_Tiff.zip

[rimpamposh]

Some prores files would sure be nice. But my personal impression is that external vs internal won't be a big advantage because of overdone and poor processing done to the video in camera before output. Now I only have nx500, so maybe quite the same doesn't apply to nx1, but live view is heavily denoised, which leads to most of our problems like banding after encoding and blocking bigger than you'd want (60mbps for 1080p24 should be pretty damn great yet after iso100 everything crumbles). Btw, can you compare live view clips vs recorded clips in record mode? Might be a difference in manipulation.

As to color range, there are some very useful posts here about nx500's 0-255: http://***URL removed***/forums/thread/3933332?page=3

The conclusion seems to be that 0-255 for video is kind of a novelty and basically not properly supported yet by codecs and programs. Unfortunately I don't have an nx1 to make a test. What would be interesting is a clip 16-235 vs exact same scene with 0-255 remapped to 16-235 in ffmpeg or avisynth or wherever.

 

About denoising before grading, could be a good idea, but depends in what color depth premiere makes the calculations. In 8-bit it will enhance the banding easily, in 16-bit should be ok-ish.

Now a good workflow before grading on nx500 would be: avisynth convert to 16-235, dfttest in 16bit hack mode (denoising, select strength as needed), gradfun3 (debanding, use strength 0.8-1.2 as needed) with 16 bit input and 8 bit dithered output, save lossless in virtualdubmod. Grade in wherever like premiere, add some type of grain (like gorilla), this will cement the no-banding look. Export lossless, encode in x264 with proper settings for the grain (don't want to use premiere output if you care about quality).