S-Log on the FS100

Is this a problem with digital sensors as a whole? Had the same issue (and know a couple other people who did as well) with both the 5D and 7D.

The DR for RGB can be a little different, you can just make up for it in the curves you use. The problem I've seen, is that for different light wavelengths, the sensitivity of the photodiode is different. So, if you compensate for red let's say at one wavelength, and then you're reading in a slightly warmer or cooler red, how sensitive the camera is changes to those other colors.

Have you ever taken a shot with a camera at a faster shutter speed and lots of gain, then did a long shutter speed and no gain, and saw a difference in the coloring? That's because the sensitivity of the colors are not linear.

I think this is what happens on a lot of cameras. You set it up for a particular condition and then when you are utilizing the sensor at different lengths of exposure, the ratios of the colors change.
 
The DR for RGB can be a little different, you can just make up for it in the curves you use. The problem I've seen, is that for different light wavelengths, the sensitivity of the photodiode is different. So, if you compensate for red let's say at one wavelength, and then you're reading in a slightly warmer or cooler red, how sensitive the camera is changes to those other colors.

Have you ever taken a shot with a camera at a faster shutter speed and lots of gain, then did a long shutter speed and no gain, and saw a difference in the coloring? That's because the sensitivity of the colors are not linear.

I think this is what happens on a lot of cameras. You set it up for a particular condition and then when you are utilizing the sensor at different lengths of exposure, the ratios of the colors change.

That makes a lot of sense. I have had issues with the color red on every video camera I have ever used, even back in the analog SD days, shooting 3/4" Umatic SP, High 8 or S-VHS (I missed shooting on Beta SP - went to DV instead).
 
No, that is not true, it will not always look more filmic. It's all in your ability to grade the footage. Shooting log flattens out the highlights and the shadows, and in an 8-bit space, robs those values from the mid tones. For scenes with really high dynamic range, if you then apply the S-curve to crush blacks and the highlights, you are going to expand the midtones back into where they would have been had you done it in camera. However, they're quantized now and so you are introducing larger steps which can produce banding depending on how much you crush the top and bottom. To get the BEST out of your camera, you need to figure out how you want to handle the highlights that you would in post, and set the knee as such to get it as close as possible. That way, you quantize out and shrink the range in 8-bit you'd do in post and keep all the other range in the mid-tones for information you want to keep.To get the advantage of what people call a LOG curve, you need to have a larger bit depth for storing that extra top and bottom information so you aren't throwing away information in the mid tones, and this is EXACTLY what you are doing.I also don't agree with AVCHD randomizing the gradations, because I see the opposite. If your remove most of the noise from an image, both the AF100 and FS100 have a tendency to band. It's because they DON'T do anything in the gradations, like add dithering noise, that they do this. Further, AVCHD actually cleans out the noise and makes things have a bit more of a plastic look as compared to externally recording footage. It's not extreme like on a Canon DSLR so I have no problem with it, but it's there. It's an H.264 issue.
Not true? Forthe most part...it's very very true! You and I seem to be saying the same thing, but yes, log and 10 bit very much provide more to work with in post to achieve the film look. In camera contrast applies it to the least latitude possible, especially when using certain settings. Flat shooting and post grading is best achieved in 10 bit footage as opposed to 8 bit for the reasons you state. I know why 8bit is limited for grading, but it doesn't mean one can't still do it like postmaster is working on and make a great image. Also I believe one can better achieve a more filmic look when shooting slog 10 bit rather than doing it in camera. There are many tools like selective exposure and contrast that are available in post that you cannot get in camera. There are many workflows available to us, nothing wrong with doing it this way as long as you know the limitations...but you first have to see the benefits. Theres a reason Technicolor created a log curve for an 8bit camera. Is it ideal? No, but in the right hands it can do wonders. If you choose not to use it...thats finetoo, but it doesnt make it wrong.I have in my experience seen randomizations in the gradients but I'm comparing to DSLR h264 footage. Perhaps I should have said that. :)
 
Last edited:
Okay, here comes a little update. This was shot with GlenColor1 PP for outdoor use (ungraded).View attachment 43555View attachment 43557This was shot with G-LOG (ungraded)View attachment 43554View attachment 43559 And this is G-LOG after grading.This matches what may eyes saw on that early morning as close as possible.The shot with the Gelncolor PP is to blueish, the actual mood was much warmer. View attachment 43556View attachment 43558 If you look at the scopes, you ca see what G-LOG does. It´s not really an S-LOG, but it gets a different distribution of values, that are quite handy in post.There is one strange thing, that I don´t understand yet. If you look at the scopes from the first shot on GlenColor1 PP, there are those empty lines, in the luma and YUV parade.Looks like there is information missing. Whereby the information in the G-LOG shots seems to be more complete and dense.Rick, do you have any idea what could cause that?Here you can download the uncompressed full HD TIFFs, straight from the timeline: http://www.file-upload.net/download-3891460/G-LOG.zip.html Frank
I happen to really like the log footage after grading. Looks much more filmic to me rather than coloring in camera. Like I've been saying...if you can do it, shoot flat and grade in post...just be careful with the 8bit footage in the gradients and skin tones. Looking forward to trying out these settings! :)
 
Not true? Forthe most part...it's very very true! You and I seem to be saying the same thing, but yes, log and 10 bit very much provide more to work with in post to achieve the film look. In camera contrast applies it to the least latitude possible, especially when using certain settings. Flat shooting and post grading is best achieved in 10 bit footage as opposed to 8 bit for the reasons you state. I know why 8bit is limited for grading, but it doesn't mean one can't still do it like postmaster is working on and make a great image. Also I believe one can better achieve a more filmic look when shooting slog 10 bit rather than doing it in camera. There are many tools like selective exposure and contrast that are available in post that you cannot get in camera. There are many workflows available to us, nothing wrong with doing it this way as long as you know the limitations...but you first have to see the benefits. Theres a reason Technicolor created a log curve for an 8bit camera. Is it ideal? No, but in the right hands it can do wonders. If you choose not to use it...thats finetoo, but it doesnt make it wrong.I have in my experience seen randomizations in the gradients but I'm comparing to DSLR h264 footage. Perhaps I should have said that. :)

Ahh, okay, didn't realize you meant 10-bit too, and I agree, at 10-bit definitely, it gives you more freedom to do whatever you want.

AVCHD uses H.264, which is what the FS100 and AF100 use, that's why I mentioned it. At low noise, the H.264 cleans out the tiny bit of noise and detail and makes smooth transitions that cause the posterization. With a little noise added in on both, it does cause the gradations to be MUCH nicer (as it's a dithering like effect).

With this particular example, I feel the LOG picture is grading a shorter distance to the final than the PP1 settings. Given it's 8-bit, I'd adjust the PP1 setting to not crush so hard, since in the end, you don't want it that crushed. I find with 8-bit, getting it CLOSEST to the grade is the best bet.

I still think the red is too much in both log final grades though, it almost has a red cast to both images.
 
Ahh, okay, didn't realize you meant 10-bit too, and I agree, at 10-bit definitely, it gives you more freedom to do whatever you want.AVCHD uses H.264, which is what the FS100 and AF100 use, that's why I mentioned it. At low noise, the H.264 cleans out the tiny bit of noise and detail and makes smooth transitions that cause the posterization. With a little noise added in on both, it does cause the gradations to be MUCH nicer (as it's a dithering like effect).With this particular example, I feel the LOG picture is grading a shorter distance to the final than the PP1 settings. Given it's 8-bit, I'd adjust the PP1 setting to not crush so hard, since in the end, you don't want it that crushed. I find with 8-bit, getting it CLOSEST to the grade is the best bet. I still think the red is too much in both log final grades though, it almost has a red cast to both images.
I agree, getting it close in 8 bit is the safest route. Yeah the red thing is weird, interested in hearing some solutions. Great explanation on the color shifts on different shutter speeds/ isos Rick, good to know.
 
To me the weakness of the camera is the oddities as the higlights blow

I can see the value of a profile that massivley boosts the mids/shadows so that one can always keep the highest high in the exposure range?

S
 
To me the weakness of the camera is the oddities as the higlights blow

I can see the value of a profile that massivley boosts the mids/shadows so that one can always keep the highest high in the exposure range?

S

I'm working on a utility to fix that. I think it's a problem with how Sony designed their 3DLUT at the top end. I see what they are trying to do, and technically, at least compared to the 7D highlights, it's saving more information, but it looks horrible.

If any of you have some good examples of PNG frame grabs of BAD blowouts, please send them to me. They will help me study it more. I really want to get a tool together quickly to help others.

And pertaining to the gain/iso thing, saw it first on the D90 from Nikon. Then I started studying more about sensors and started to understand why they did that. The worst thing is hot pixels in long exposures. I do astrophotography and when a camera has them, you get green, red and blue dots all over the place after a long exposure. They are called hot because they have a higher sensivity. Even if it's a little higher, over a long time, they really differ.
 
The best approach to improving blown-out highlights is the one colorists have used for years and which RED applies in their DRX slider. You take the best of the three channels – they normally don't blow-out at the same time - and copy it to the others, but only for the highlights. This will give you black-and-white there, but retain detail and normally look much better. The only exception are high color light sources, like Neon.
 
The best approach to improving blown-out highlights is the one colorists have used for years and which RED applies in their DRX slider. You take the best of the three channels – they normally don't blow-out at the same time - and copy it to the others, but only for the highlights. This will give you black-and-white there, but retain detail and normally look much better. The only exception are high color light sources, like Neon.

Is this possible to do using something like Magic bullet colorista II or looks?

On a side note, Do you guys have recommendations for better color grading software? Da Vinci Resolve Lite looks nice but as I'm on a PC I can't use it. Looking for something that's the next step up from Colorista/Looks

another side note: Also curious about the decision in levels for M/R on all of these profiles
 
I have tool set that allows me to add the luminance from many other channel to a cliped one, for example if blue is clipped out in a sky, i can add red and green luminance to the blue channel, it's still blue, but has the cloud detail back now... or add straight summed luminance to a channel or any combo of channels.

Davinci has somewhat less elegant method of getting to the same place tho, and it's affordable in relitave terms.. maybe 10K for a basic system. Resolve will be released for PC in a few weeks, i'd wait for that or what ever Adobe comes up with now that they own Iridas SpeedGrade


d
 
I think it's a problem with how Sony designed their 3DLUT at the top end. I see what they are trying to do, and technically, at least compared to the 7D highlights, it's saving more information, but it looks horrible.

Going back to an early test of mine we see the highlighs go wrong dont roll smoothly to blown with variious PPs mine and franks just seem to make this happen at different places on the curve (Fstops in this case)

To get a smooth roll one then needs to avoid that part of the exposrue - ie create a boosted blacks mids profile and avoid blowing


S
 
Hah, I used to do this on the HVX 200 with a very high pedestal value. This is the secret sauce.

Good work.
 
The best approach to improving blown-out highlights is the one colorists have used for years and which RED applies in their DRX slider. You take the best of the three channels – they normally don't blow-out at the same time - and copy it to the others, but only for the highlights. This will give you black-and-white there, but retain detail and normally look much better. The only exception are high color light sources, like Neon.

Yeah, I see that also with the 7D, it's doing the same thing on the highlights when they blow out. They ALWAYS go to white eventually.

The FS100/AF100 are actually trying to be smarter about it, although it looks worse. They are trying to not let all 3 channels go to white when one is, and that's what causes those complementary color bands. That's what I am eliminating.
 
Okay gentlemen, the cat comes out of the bag:



Before I release it on my Blog, I would really like to hear your opinions,
since I really appreciate your knowledge and expertise.

G-Log comes in two flavors

G-LOG A: Looks deep into shadows

G-LOG B: Is a bit more on the conservative side

G-LOG A:
Black Level: +15
Gamma Standard
Black Gamma: High, +7
Knee: 102,5, -1
Color Mode: Cinematone 2, +8
Color Level -5
Color Phase -7
Color Depth: R-1, G+2, B-3, C+5, M+1, Y+6
WB-Shift: LB-6, CC-5
Detail-7

Make sure you set your highlights between 85 and 90 IRE on G-LOG A.

G-LOG B:
Black Level: +11
Gamma Cinematone1
Black Gamma: High, +7
Knee: 75, -3
Color Mode: Standard, +8
Color Level -5
Color Phase -7
Color Depth: R-1, G+2, B-3, C+5, M+1, Y+3
WB-Shift: LB-7, CC+1
Detail-7


Have fun :cheesy:

Frank
 
Last edited:
Nice thinking

I had made my own 'log' before you posted this

came up with virtually the same settings on Blacks, Knee et al

I guess the threory goes like this..

At the top end the colours clipping looks like crap

therefore try to keep brightest component in the image at less than 100% to stop clipping

Now the image goes dark

Boost the shadow to compensate

It seems a smaller F stop is required

I guess this is noisier, but being brighter means less gain is needed

all maes sense to me

Judging from the onvboard monitor we get smoother highlights, my guess however is you have desaturated the colours a little too much (in the first one)

Good thinking

SMM
 
I now have 4 out of 6, of my picture profiles built up purely from your info Frank. I have just put these 2 in, will do some real world tests with them and report back.

Cheers, Liam
 
Looking forward for your input guys.

And yeah Sam, the colors maybe can use a little more saturation, but I wait till I got sunlight tomorrow.

Frank
 
Back
Top