FS7: SMM SLog2 Vs Slog3

broughtonfilm said:
That's exactly my problem with your test. By adjusting exposure this way you changed the shape of the curve (even by just using GAIN slider, not to mention you actually used curves to further tweak it...), and I can only guess that it was done individually for each shot. In other words, your skills in Davinci influence your opinion about the Slog2 and Slog3 curves more than the actual curves.
Click to expand...

The LUT/and or corrections only act on the 'core data' in the file - one cannot recover or play with what is not there so really what I did later is not of much relevance.

Im sure my commentary above will be of amusement to you :)
 
broughtonfilm said:
That's exactly my problem with your test. By adjusting exposure this way you changed the shape of the curve (even by just using GAIN slider, not to mention you actually used curves to further tweak it...), and I can only guess that it was done individually for each shot. In other words, your skills in Davinci influence your opinion about the Slog2 and Slog3 curves more than the actual curves.
Click to expand...

EXACTLY!!! You nailed it.
 
Doug Jensen said:
EXACTLY!!! You nailed it.
Click to expand...

I see this (the primary question) to be about the collection of core data.. aka what is recorded in different scenarios.

There are a multitude of LUTs (or manual, or Lutcalc or Aces) to manipulate that data once you have collected it.

It is of course a legitimate secondary question.. 'is this data fast easy, pleasant and consistent to manipulate?'

S
 
I don't see the point of seeing how much of a mess you can make . . . and then finding the best way to clean it up and salvage something out of it. When S-LOG2, S-LOG3, or RAW are expsoed properly (and even if they are 1-2 stops over-exposed) and graded correctly in post with Resolve they all can look great and are barely distinguishable from each other. But if someone wants to color outside the lines, apply unconventional LUTs, ignore standard operating procedures, and workflows in Resolve -- then all bets are off.

There is nothing that is as "fast easy, pleasant and consistent" to grade than RAW and S-LOG when you follow the workflow. My average time spent grading a clip that doesn't need secondaries is less than 10 seconds. How much faster do we need?
 
Doug Jensen said:
I don't see the point of seeing how much of a mess you can make . . . and then finding the best way to clean it up and salvage something out of it.
Click to expand...

Because life is a mess.. or the real world is often 14 maybe 20 stops of DR in a real scene which with a small package we can light/fill down to maybe 10 stops.

Midday sun and deep shade, pop concerts and their unlit audience, INT interviews where the client wants to see the EXT view but gives no budget for gelling windows or an HMI fill package.. or (the starter of all this stuff) Lens query about shooting in a forest with a small lights package.

You gotta know where you can (or cant) go.

You may note that Im not alone in stress testing cameras.. CML Geoff does it (well and consistently) all the time.
 
morgan_moore said:
CML Geoff does it (well and consistently) all the time.
Click to expand...

Stress testing is absolutely a valid thing to look at, and Geoff does a great job with lots of precise measurements, condtions that elimante variables, support and advice from other experts. But your tests are nothing like that. If they help YOU learn something about the camera then they have done their job for you. I encourage everyone to do their own testing as much as they can. But in my opinion, you are making things much more complicated and messy than they need to be and then posting conclusions that are based on faulty methods and workflows. I can't sit by and let that go by without comment. The very fact that you would mention you used the "Gain" control in Resolve to compensate for LOG exposure speaks volumes about your limited knowledge of working with LOG and RAW. If you feel your testing methods and conclusions meet professional standards why don't you post them on CML like Geoff does and see what people say? After all, if your conclusions are correct wouldn't that help an even greater number of people?
 
I completely agree that this is not hard science that I have performed.

Just a quick 10min looksee to investigate the graph issued by sony which suggests 2 being stronger for the first six stops then dropping off badly.

In terms of my comments to Len.. I said exactly to do your own tests.

Personally I don't like Geoff's tests because they don't have a specular highlights but there we go.

I like testing specular highlights because they often get horrid do-nuts as the colour channels clip and there are often specular highlights in real scenes like office ceilings kitchens and more.

And you are more than welcome to comment :)
 
Sam, your commentary confirms what I said in my previous comment perfectly. At 2:54 in your video you are applying a completely wrong LUT on the Slog2 footage, then try to manipulate it with secondary correction and then use it as a basic for your opinion about the curve? You applied SGamut3.cine/SLog3_to_SLog2-709 on the Slog2 footage, LUT designed for Slog3 ?!? I can't follow your logic at all.
 
Well - fook you are right.. not worked with Slog2 before.. I dont think the conclusions will change but absolutely Im gonna try a different LUT.. thanks for spotting that!

S
 
Yep had a dingle with a couple of Slog2 LUTS - the image goes to proper shadow.. but the recovery is still no good as the core data is simply not there.

What is in your opinion the best Slog2 -709 LUT from Sony?

Thanks

S
 
Why not make them in LUTCalc. Then you can make them precisely to the ISO for grading, and even play with the output gammas.
 
I didn't want to get into specific comments on the video, but since the subject has been broached . . .

1) You also wrongly apply the same S-LOG3 LUT to the S-LOG2 footage on the portrait -- thus invalidating all of your conclusions.

2) You should never be using the Primaries Bars sliders or wheels on the first node before the LUT. You should be using the LOG Color Wheels with S-LOG footage. Primaries Bars/Wheels can be used for secondary coloring on nodes 3 and up. If you need them at all.

3) Curves is the wrong tool to be using because it is imprecise. Most of the professional colorists I've spoken to rarely, if ever, use them. I never use them.

4) The Contrast control should never be touched.

5) At 8:40 you say you think you've exposed around 800-1000? No you haven't, I'll bet it is closer to 400. And either way, why don't you know how you have exposed if you're running a test?

6) At 1:38 you are basing your exposure on the hot spot. Why? Is the hot spot the subject of the shot? At 4:15 you say the client says the "blue bucket is extremely important" so if the bucket was the subject of the shot then the exposure should have been based on the blue bucket and the hot spot could go to hell (assuming you had no control over the lighting). The hot spot can't look any worse anyway no matter how blown out it gets. Blown out is blown out. But you have sacrificed the very important blue bucket by wrongly letting the hot spot dictate your exposure. Big mistake. Now the blue bucket is beyond recovery no matter what LUT or power windows you try. You have lost your client.

7) The best way to dig yourself out of a bad situation in post is to never allow yourself to get into that bad situation in the first place.

https://vimeo.com/ondemand/resolve
 
LennyLevy said:
Why not make them in LUTCalc. Then you can make them precisely to the ISO for grading, and even play with the output gammas.
Click to expand...

Waste of time. It is not necessary to have a different LUT for every ISO if you know how to grade LOG. Just one LUT with some offset adjustments will do the same thing.
 
Doug Jensen said:
I didn't want to get into specific comments on the video, but since the subject has been broached . . .

1) You also wrongly apply the same S-LOG3 LUT to the S-LOG2 footage on the portrait -- thus invalidating all of your conclusions.

2) You should never be using the Primaries Bars sliders or wheels on the first node before the LUT. You should be using the LOG Color Wheels with S-LOG footage. Primaries Bars/Wheels can be used for secondary coloring on nodes 3 and up. If you need them at all.

3) Curves is the wrong tool to be using because it is imprecise. Most of the professional colorists I've spoken to rarely, if ever, use them. I never use them.

4) The Contrast control should never be touched.

5) At 8:40 you say you think you've exposed around 800-1000? No you haven't, I'll bet it is closer to 400. And either way, why don't you know how you have exposed if you're running a test?

6) At 1:38 you are basing your exposure on the hot spot. Why? Is the hot spot the subject of the shot? At 4:15 you say the client says the "blue bucket is extremely important" so if the bucket was the subject of the shot then the exposure should have been based on the blue bucket and the hot spot could go to hell (assuming you had no control over the lighting). The hot spot can't look any worse anyway no matter how blown out it gets. Blown out is blown out. But you have sacrificed the very important blue bucket by wrongly letting the hot spot dictate your exposure. Big mistake. Now the blue bucket is beyond recovery no matter what LUT or power windows you try. You have lost your client.

7) The best way to dig yourself out of a bad situation in post is to never allow yourself to get into that bad situation in the first place.

https://vimeo.com/ondemand/resolve
Click to expand...

Overall - I threw this together - clearly an error. But I can also digest all the comments :)

I do not think they invalidate the conclusion because we can see in the untouched log that the signal is very weak in the shadows with Slog2 .. end of.

1) You also wrongly apply the same S-LOG3 LUT to the S-LOG2 footage on the portrait -- thus invalidating all of your conclusions.


Absolutely.

2) You should never be using the Primaries Bars sliders or wheels on the first node before the LUT. You should be using the LOG Color Wheels with S-LOG footage. Primaries Bars/Wheels can be used for secondary coloring on nodes 3 and up. If you need them at all.

With 32 float nothing matters if it gets you where you want.

I always work before the LUT as luts are 'destructive'.. any clipping induced by the LUT cannot be got back after the LUT.

3) Curves is the wrong tool to be using because it is imprecise. Most of the professional colorists I've spoken to rarely, if ever, use them. I never use them.


This is BS.. of course curves are used. Tip for you.. you can make these more precise by using them on a node of 50% opacity.


) At 8:40 you say you think you've exposed around 800-1000? No you haven't, I'll bet it is closer to 400. And either way, why don't you know how you have exposed if you're running a test?


Because the 'bucket test' was planned and I thought i would throw in a quick skin test - i idd not have a clear methodology for that skin setup.

6) At 1:38 you are basing your exposure on the hot spot. Why? Is the hot spot the subject of the shot? At 4:15 you say the client says the "blue bucket is extremely important" so if the bucket was the subject of the shot then the exposure should have been based on the blue bucket and the hot spot could go to hell (assuming you had no control over the lighting).

Sony data always refers to (and I paraphrase) 'data levels at stops below clip'.. it would seem therefore to test from clip back down the stop range.

The hotspot 'nails' a consistent point of full clip.

In 'real life' maybe the hotspot held something of value, be that clouds and ext view from an office, whatever, I did consider putting the pink bucket in the hot spot.
Consider this scene once it got a little darker.. https://twitter.com/WeArePlaster/status/879622729900716032
Letting the stage completely blow would not work.. one would have to 'nail' to the stage lights and pull the crowd/ambience back in post.

7) The best way to dig yourself out of a bad situation in post is to never allow yourself to get into that bad situation in the first place.


I completely agree. And forewarned is forearmed.
 
Last edited:
morgan_moore said:
With 32 float nothing matters if it gets you where you want.
I always work before the LUT as luts are 'destructive'.. any clipping induced by the LUT cannot be got back after the LUT.
Click to expand...

I never said you shouldn't adjust exposure BEFORE the LUT because you absolutely should. I said you shouldn't be using the Primaries sliders before the LUT. As for your attitude that anything goes "if it gets you where you want", that is total crappola and the lowest kind of argument in any discussion of techniques. I have tried to lead a horse to water, but I cannot make it drink. I wish you luck.
 
Doug Jensen said:
Waste of time. It is not necessary to have a different LUT for every ISO if you know how to grade LOG. Just one LUT with some offset adjustments will do the same thing.
Click to expand...

Hmmm... I am not an expert in this by any means but that sounds contrary to what I've read everywhere else.

So do you offset adjustments to get your levels in the same place as they would be for the Sony recommended EI 2000 (for Slog3 FS7 - I guess 60% for white) then apply LUT? If so using which controls to make sure that the relationships don't shift. Do you do other adjustments prior to LUT?

Thanks
 
Of course ive gone off and played with the Log Offset wheel in the first node to see if I could see value to what you said.

My head is not buried in the sand.

I cant actually see a lot of difference though between changing that and the offset slider.

S
 
You must log in or register to reply here.
Back
Top