8-bit 'log' profiles and colour grading - is it really worth bothering?

C

Colinelves

Well-known member
As a sort of partial follow-up to the Abel Cine profiles - I was jut wondering if people thought it actually worth trying to use a super-flat profile to 'maximise' the DR of the camera, given that it is only an 8-bit encoding.

i.e. Since you don't get much depth from 8-bit to grade with, would it be better to get the look on the day as close to the final look you actually want so as to minimise the amount of (8-bit) grading that's required? Or do people still think you'll get better results shooting flat and grading it fully?

Thanks,


Colin Elves
DoP and Steadicam Operator
Web: www.colinelves.com
 
Frankly LOG is quite useless unless you can monitor de-log'ed. So unless you have a LUT box with properly programmed LUTs to monitor and light from, you'll probably expose wrong and have no idea what you are looking at if you record and monitor in "log". That doesn't even cover the 8bit issue, which depends a *bit* on your content and workflow, pun intended. Shooting very flat might be of some use in the sense that erring on the side of flat is always better than the other way around. The closer you get in camera the better, but if you will be grading log-ish might be a bit better, not that there actually is any log curve in the FS700. Then again none of this really matters because FS700 doesnt have a log curve, and gamma corrected cine curves really arent far enough from log that there is much of a difference between 8bit 2.2 gamma and 8bit log (close to 2.6 gamma). Personally I think for a lot of stuff CINE 4 with Black level up a bit and black gamma up as far as +7 still looks fairly natural and certainly not log (maybe C-log ish, but thats not real log either) even though thats probably the closest thing to log as you get in the FS700, so I'd say this is a bit of a non issue.
 
That's sort of what I figured: unless you're shooting Raw, the whole low contrast 'log-ish' look thing is sort of pointless.

Since the FS700 isn't a raw camera (yet!) we shouldn't pretend it is.

However, I'm sure this isn't the whole story, so I'm interested in opening up a debate here...

Colin
 
My experience of the FS100 is that any grade of a 'fine tonal' image will band - makes sense - there are no bits beyond the delivery space

But not every image is 'fine tonal' and the final o/p compression will 'blur' the image too losing need for tonal resolution

Also the attribute of recording a very baked look gives you no room for exposure or CT error, which may be a trade off not worth making

So 420 I guess you can go maybe 5% flat.
I guess 422 10, 20%
and so on.

Those figures could probably be doubled if the final delivery is highly compressed

S
 
Sounds a bit six of one and half a dozen of the other to me. I suppose at the end of the day it depends upon a) how much time you actually have to get it right on the day (i.e. lots: shake and bake, none: medium rare) and b) how much you trust the person doing the grade. But, by the sounds of it, the difference in quality between shooting flat and getting a look in grade, vs shooting baked and tweaking the same look is somewhat negligible (or at least not something that will be noticed by your average producer!)

I'm shooting a pilot for a sitcom at the end of the week - so the discussion is not entirely academic. If I can find the time I'll try out a couple of looks to see what I like before the shoot, but I suspect I'll just nick someone else's 'log' custom PP and go with that and leave it all to the grade. A bit lazy I know, but time is tight and I'd rather spend more time and energy worrying the lighting than worrying about with camera settings.

Colin Elves
DoP and Steadicam Operator
Web: www.colinelves.com
 
Hi Colin I wonder how your shoot went? I wanted to add to this thread after reading your post.

I started shooting with the F3 last year before SLOG was free and learned to use the cinegammas to shift mid-tone levels (skin) relative to the rest of the scene so that less mid-shifting had to be done in post since I am recording to SXS cards in 8-bit 420. You will find some interesting posts about this under my name here on DVX or on my blog. (A good article was called "so you think you need SLOG")

I also recently did a real-world test (not charts, not looking at numbers on paper, not talking theories) that I plan to publish soon which compared the same scene shot with a gamma curve and then SLOG adjusted to match it. Both shot in 8-bit. There was negligible difference in the end and what's MORE important to note is that the shot using a gamma curve was properly exposed with skin tone at +1 stop over middle grey for dark skin, but with SLOG I was 2 stops higher over skin tone with whites kept under 90%. This attests to the "not having enough time to get it right on the day" as more and more I feel I can really treat SLOG as a "digital negative" when I use it.

A few things I will say now before I get a chance to write in more detail is that "pulling" (darkening) your footage in post will usually always improve your signal to noise ratio and decrease any grain (this is suitable for green screen work say). So although SLOG for example has a low middle grey point of 38% if you shoot your entire scene too low on the curve and "push" it (lighten) it in post you will increase the camera noise floor and introduce noise/grain.

Getting it right in camera is "always" the goal and establishing a good range of contrast in your scene between dark, mid and high tones will allow a good grade to occur in post. I am not pro or against SLOG and enjoy experimenting with these modes and learning all I can about using them. So far the biggest thing I've discovered is that shifting your mid-point relative to your scene is the single best thing you can do when shooting to minimize the level of pushing/pulling your mid-tones later which to me is the most important part of an image (the subjects skin and the subjects themselves).
 
Last edited:
Mike Most has the quote-0-the day on this subject, over on CML;
log is based on the characteristics and properties of film, which means it is not necessarily optimal for digital capture, nor is it optimized for the specific characteristics of the sensor being used. There's nothing "wrong" with it, and it is something of a standard.....
Click to expand...

A few years ago when we were trying to fit digital capture in to a LOG pipeline, for LOG delivery to film print - it made sense... in 2012 in the end days of film distb it really does not.. remember DCP = 2.6 gamma, Broadcast = 2.4, Web = 2.2. Rarely in 2012 is there a LOG anything to deliver anymore - and where there is, getting from P3 to LOG is pretty straightforward now.

Mike said it better than i did... and i don't use fauxlog on my FS100, i find it a real step backwards in many real world ways

I do like Clog / 8bit from the C300, but again find the AbleCine display LUT to be ummm.. well to be kind... somewhat suboptimal

A side note, my main gig is gradeing indie features and MOW's for US primetime, i have completed 10 shows in the first 8 months of the year, mostly from Epic, one from F3, one from film, two from Alexa, and a bunch of national TVC's for Canada, some of them have been shot on the C300... i have alot of seat time gradeing in a first class enviroment.. Christe projector on a 16 foot screen for the film work, Dolby PR4200 in a suite used for broadcast (the theatre also has a BVM-OLED for checking broadcast)... i do have a clue...

d
 
Last edited:
Dermot - can you provide a link to the original article you mentioned Mike Most had written? I'd like to get the full context of what he was saying.

I would quite disagree that LOG is not optimized for digital capture since the main reason it does exist is to maximize the data information from the sensor which effectively translates into an increase in overall sensor dynamic range as well as optimized signal to noise ratios.

As for how "well" SLOG ultimately translates into final delivery versus not using it, well I do agree that is debatable - especially with the advent of 10 & 12 bit recorders capable of recording 422 or 444, etc..
 
CML is a mailing list, they do have an archive online, you can join here;
http://www.cinematography.net/

But i would need to dig through days and days of emails to find all the notes on this topic.

LOG was designed to keep the film scans complete untill we plotted them out to film again, and allow us to work with non-linear color within the confines of highly limited monitoring... not much of that is relevant any more, actualy little to none in the days of P3

Fitting linear capture devices into that chain was the challenge that ViperLog/Slog/Clog/LogC/Si2Klog/RedLogFilm etc etc were designed to do...

Today we have linear capture devices, and linear display devices, and no NEED to use LOG encodeing... but i do use log encodeing every time with the raw i work with, and vastly prefer Slog/Clog/LogC over 709 for example

The crux of the matter is the sub par fauxlog picture profiles used in both the FS100 & FS700, these are not Slog, are not even close to Slog or any of the others... so let the camera be a linear device, use the full range of data needed for any given shot, an F3 the FS really is not, i'd love to see Sony implement Slog in the FS series, but until then.....

to recap.. as i understand it - the questions were
1) is LOG worth it when gradeing?
2) is 8bit a roadblock in implementing a LOG workflow?
IMHO - the answer to Q#1 is; "OH YEA" it's really worth it, and for Q#2? 8bit does not pose any issues with Clog, so no it's not an issue if implemented well.
where is all comes off the rails is when the FS100/700 enter the discussion.. then there's no option for log, and the fauxlog profiles are so far off the mark as to make a realistic discussion null & void.

To quote Colin who asked in the first place;
would it be better to get the look on the day as close to the final look you actually want so as to minimise the amount of (8-bit) grading that's required?
Click to expand...

with FS? yes.
with F3/709/420? yes
with F3/Slog/444? no.
with Epic? no.
with Alexa? no.

d
 
Last edited:
Don't worry about any digging, I was just curious to read someone elses perspective on it.

On the F3, shooting in SLOG adds an extra stop of dynamic range to the camera bringing it up to 13.5 stops. Is there any similar claim for the fauxlog on either the FS100 or FS700?
 
i love the F3, graded a film that the amazing Mark Irwin shot with F3/Slog/444;
http://www.youtube.com/watch?v=vbaqEpOwWPw

Sweet camera in Slog/444.. the pickups & splinter unit were shot on an Alexa, no one can tell the Alexa from the F3's

There is little to nothing to be gained with the fauxlog profiles on the FS, certainly nothing like switching a F3 from 709 to Slog, not even close.

There can be something lost tho

d
 
Last edited:
Dermot, since you grade as a profession can you advise any ideal or optimum overall scene contrast ratio or range that you feel grades the best? And would you recommend middle grey always be in the middle of the scale - or biased either way above or below again for "optimum" results?

Trailer looks great by the way, I really like the look.
 
morgan_moore said:
So 420 I guess you can go maybe 5% flat.
I guess 422 10, 20%
and so on.
Click to expand...

That is a complete misunderstanding of how 4:2:0, 4:2:2 and 4:4:4 differ. The only difference between 4:2:0 and 4:2:2 is the color resolution. This is most notable in the edges of clean mattes. In regular footage, it's quite difficult to spot. The thing is, you will not really gain anything in color correction from additional color edge information unless you are doing mattes from color not luminance.

That's why 4:4:4 is great for greenscreening. But after that has been done, everything goes back to 4:2:0 for delivery. When you go back to 4:2:0 you lose the advantage. That's why 4:4:4 does not give you really anything except for special circumstances. That's because all that advantage will be lost on delivery anyway.
 
Im not sure Im with you here - as far as I can see 8bit just falls apart tonally when graded

to me its all about tone and banding

If it is the space/depth or the compression or what I dont know,

But my understanding is that you need more colours in the input space than the output space

This enables you to move those colours around

In theory (as I understand it) there is NO space to grade if your in and out are 420, but the realities of compression mean you can grade a little (as per my original quote)

S
 
everything i do goes to either 444 or 422, nothing goes out my door as 420 beyond low rez client screeners....

to deal with banding i use anti-banding filters, particle noise & film grain usualy inside masks to get to regions of concern and leave the rest of the image alone

i don't see 8bit falling apart when graded, just needing a bit more work at times to get to the desired end game

i work in 32float/444 usualy, even when the footage orignated in a GoPro

Still the implementation of LOG in the FS does not exist, better to look at the C300's Clog footage for an example of 8bit LOG advantages

d
 
starcentral said:
Dermot, since you grade as a profession can you advise any ideal or optimum overall scene contrast ratio or range that you feel grades the best? And would you recommend middle grey always be in the middle of the scale - or biased either way above or below again for "optimum" results?

Trailer looks great by the way, I really like the look.
Click to expand...

Thanks! the trailer used the grades Mark & I did for the film, aside from the strange de-sat gals down the stairs at the end.. i've no idea where that came from, but i don't cut the trailers ;-)

I'd go for mid grey in the middle unless you have a reason to do otherwise.. there's so much latitude and so little noise in the F3+Slog combo that usualy it seems to deliver a great digital neg when sent down the middle of the road.

There might be a few times you want to hold highlights, and maybe a few times when there is close to zero exposure so you go WFO and cross your fingers, but aside from that i'd let it be what it wants to be.

d
 
Thanks for the insight.

I frequently use different gammas built into the F3 or even "SLOG" but place middle grey where I want to so that my skin tones are more "middle of the road". What this does for me (since I'm shooting a lot of 8-bit 420) is bake in the associated tonal curve that makes the scene look best in camera so in post I don't have to do as much push/pulling. I think this helps me maximize the quality of 8-bit 420 since so far I have not shot 10-bit 422 or higher.

I will go rematch the trailer looking for those girls on the stairs now...
 
You must log in or register to reply here.
Back
Top