Exposing Slog 3 [a7s ii]

[Vivek Venkatraman]

Hi,

 

I am about to embark on a shoot with the Sony A7s ii and have a couple of queries as I am trying to understand how the sensor performs considering I have never worked with Sony cameras..

 

1 - Slog 3 is numbered as which picture profile on this camera ? There are 2 slog picture profiles in this camera what are the differences ?

 

2 - People recommend exposing to the right now -

 

i - Should I expose to the right of the camera meter ?

ii - Or should I over expose the main thing in the scene that I am aiming to expose properly by a meter reading ?[ say the characters face]

iii - Does this approach work for shooting at night ? Is there a picture profile that works better for you for night exterior ?

iv - What is the logic of exposing in this manner ?

v - At what ISO does noise become visible in blacks, is it normal for noise to be present ? Is de-noising in post a common practise with this camera ?

 

3 - How does one approach exposing a scene with a 4 stop reading between background and subject ? Still over expose by 2 stops ? 2 stops of what ? A camera meter or a light meter ?

 

4- Any tips for day interior, day exterior,night interior and night exterior ?

 

5 - Any other best practises while exposing with this camera ? Any useful tips ?

 

Thanks

 

PS - I will be testing the camera soon as well :)

 

[David Mullen ASC]

"Expose to the right" does not just mean overexpose on top of your meter reading -- it means using the histogram to make sure your information is towards the right side of the range but that you aren't clipping anything, or anything important (a tiny glint off of a metallic surface can be clipped.)

 

Truth is that while it works for still photography, it's only a rough guiding principle in cinematography -- the principle being that it is better to not have all of your picture information on the left end of the histogram (underexposed) except for certain scenes or moments.

 

The trouble with the ETTR principle is that it ignores an essential element of cinematography, which is continuity between shots within a sequence. You don't expose every shot as if it were its own scene or like you would in still photography -- a shot has to be intercut with another shot and it has to have a consistent noise level, saturation, etc. The subject in the scene should look like it is in the same light level even if the camera goes wide or tight.

 

For example, imagine a wide master of a person sitting at a desk with a table lamp next to them on their right and a big daylight window on the left side of the frame. Then imagine the medium shot no longer has the bright window in the frame but still has a slightly less bright (compared to the window) table lamp in the shot. Then imagine a close-up of the actor where there is no window and no table lamp in the frame.

 

If you were using the ETTR principle, then your exposure on the wide shot would be determined by the bright window (and the face maybe many stops underexposed as a result), in the medium shot by the bright lamp, and in the close-up by the face. So the face in all three angles would be exposed differently -- in fact, the close-up might look overexposed if the face was the only object and it was pushed to the right side of the histogram. You'd end up creating a lot of work for yourself in post trying to make the shots match each other so they could be intercut.

 

So the principle applies but only loosely in that for the wide shot, you'd want to expose for how you want the actor to look but if the window is too bright, you'd work on bringing it down with ND gel maybe or even an ND grad on the side of the frame. You may also slightly stop down to hold more window information. But also atmospherically maybe the scene should look like the actor's face is a little dark because most of the light comes from behind them. So when you get to the close-up, you'd probably continue the underexposure... but maybe you'd slightly open up to get a little more exposure on the face. But in general your exposures are all within a close range so that the three shots are intercuttable.

 

Also remember that your final image won't be shown in log gamma, it will be converted to the display gamma you will be showing the image in. So the blacks will be black instead of milky grey as with log. I'm only saying this because some people looking at log images get freaked out that they are seeing noise in the shadows but when the image is shown with the correct display gamma, a lot of that noise is hard to see.

[Vivek Venkatraman]

Thanks David, the continuity aspect is one thing I have to consider.

 

What practise do you follow for exposing S-log ?

[David Mullen ASC]

You might find this article informative:

http://www.4kshooters.net/2015/10/26/sony-a7s-ii-vs-sony-fs7-s-log-3-comparison-test/

[Robin R Probyn]

I believe your only in 8bit Slog on that camera.. sorry I haven't used it but i shoot Slog3.cine all the time with an F5.. there is alot of info about Slog3 on XDCAM-USER.com .. a guy called Alister Chapman.. who knows all things Sony.. !

 

http://www.xdcam-user.com/2015/12/deeper-understanding-of-log-gamma-experiments-with-a-waveform-display/

 

Its on his site but he also recommends not to use Slog for night ext.. by the nature of Slog3. you will be only using a very small amount of your recording bucket .. you wont have a large Dynamic range so you,ll be wasting alot of data .. better to use actually REC709.. or one of the hyper gamma type curves.. also grading 8 bit log doesnt leave you that much leeway when you start pulling things about..

 

Its nearly always better to over exp Slog a bit..and correct in post for noise.. but this ETTR thing is coming from stills people now shooting video.. and will give you hell in post ..in 8 bit it will be even worse.. you can over expose the whole scene/film by say 1 stop,like rating film stock lower.. for correction in post but this is then done universally and easily corrected in post.. this is often done to minimize noise .. you lose 1 stop of highlight but in log you will have so much anyway.. people sacrifice that for a "cleaner" image

 

When I shoot Slog I use a 709 MLUT for the VF/monitor and pretty much go by eye,and use the WFM/zebra,s if in doubt.. judging off an Slog image is very hard to set exposure..let alone focus.. basically its not a viewing gamma .. its only a capture gama.. dont let the director ever see it :)

Edited January 5, 2016 by Robin R Probyn

[Daniel Meier]

I just made this test video, because I also wonder, what is the best way to expose Slog. In my case Slog 2. But I think it's applicable to Slog3 as well.

I find that exposing your white to 95% to 100% (which means correct exposure to an 18% grey card) gives you the best of both worlds. Acceptable noise and no white clipping.

 

 

Might be also helpful links:

http://www.xdcam-user.com/2014/08/exposing-and-using-slog2-on-the-sony-a7s-part-one-gamma-and-exposure/

[Kenny Keeler]

After watching the Wolf crow video above. Which was very informative. From my understanding 60-70 IRE is the happy zone for skin tones/important info in your shot for this cameras sensor. Because the cameras sensor responds better on the mid-tones to highlight end of the SLog curve instead of the blacks. Then based on your look, for an example say its a night bedroom scene and your going for a moonlight from the windows look. If you shot at 3200 ISO the cameras Native ISO. Placed your actor in the 60-70 IRE range on your waveform. You should be exposed correctly. Now if you know your final look is going to be darker then shot thats ok.Then once in post you can bring down the image in post to give you the night look, assuming your image before grading was a tad over to give you more information to make your blacks less muddy and give you more detail.

 

Is that what I'm getting from that video? I only picked the moonlight situation for an example. I know using the videos techniques can be applied for many different situations.

[Daniel Meier]

That's exactly what the Wolf Crow video was trying to say.

And I guess you can kinda see it in my sample footage. The shot being exposed correctly (with white being at 95% IRE) looks best in terms of noise and clipping compromise.

Although Sony recommends to put white at 59% IRE. Look how muddy this shot ended up.

Edited March 13, 2016 by Daniel Meier

[Phil Rhodes]

Here's the quick and easy approach.

 

Have a monitor available with an Slog to 709 curve in it (or a processing box between camera and monitor).

 

Light so that the 709 looks right.

 

Tell post that's how you did it.

 

Profit.

 

P

[Kenny Keeler]

Here's the quick and easy approach.

 

Have a monitor available with an Slog to 709 curve in it (or a processing box between camera and monitor).

 

Light so that the 709 looks right.

 

Tell post that's how you did it.

 

Profit.

 

P

Ha you nailed it there Phil!

[Kenny Keeler]

That's exactly what the Wolf Crow video was trying to say.

And I guess you can kinda see it in my sample footage. The shot being exposed correctly (with white being at 95% IRE) looks best in terms of noise and clipping compromise.

Although Sony recommends to put white at 59% IRE. Look how muddy this shot ended up.

Its incredible how much noise there when you set your white at 59%!!!!

 

Now in your test when you had your white set at +1. Where was your white at in terms of IRE 109+?

 

So setting your white point at 95% to 100% seems to give the best compromise with noise and clipping. Di you have your zebra on? if so what was it set at? Just curious how you were meter for your test.

 

Thank for putting in the time doing this test! Super informative!

[Robin R Probyn]

Your setting white to 95% in Slog ??? thats a REC709 level.. the point of white/grey being set low in Slog to get your DR in the highlights..

The noise that you see in Slog gamma will all go when corrected for 709.. don't judge anything off the capture gamma.. ideally you would never even see it.. its meant to look muddy ..

 

Not sure if that DSLR has MLUT,s.. but as Phil says.. and its what Ive been doing for a few years with out any problems .. set your MLUT (monitor) to REC709.800 and shoot.. forget the voodoo .. :).. but you have to know why those levels are set low in the first place.. XDCAM User site has all the info..

 

PS the sony 59% is reflected white.. (95% for REC709) not 109% white clipping level..

Edited March 14, 2016 by Robin R Probyn

[Phil Rhodes]

Very few DSLRs do. Some of the Atomos recorders will do it.

 

P

[Kenny Keeler]

Your setting white to 95% in Slog ??? thats a REC709 level.. the point of white/grey being set low in Slog to get your DR in the highlights..

The noise that you see in Slog gamma will all go when corrected for 709.. don't judge anything off the capture gamma.. ideally you would never even see it.. its meant to look muddy ..

 

Not sure if that DSLR has MLUT,s.. but as Phil says.. and its what Ive been doing for a few years with out any problems .. set your MLUT (monitor) to REC709.800 and shoot.. forget the voodoo .. :).. but you have to know why those levels are set low in the first place.. XDCAM User site has all the info..

 

PS the sony 59% is reflected white.. (95% for REC709) not 109% white clipping level..

 

Robin the AS7 does have its own MLUT in camera. There is a setting in the menu labeled "GAMMA DISP. ASSIST"

this tab in the menu has a four option.

 

1. Gamma Disp. OFF

2. Gamma Disp. AUTO

3. SLog-2 to 709 (800%)

4 SLog-3 to 709 (800%)

 

Im not sure what Sony means when it has (800%) after the 709 but they give you an in came MLUT so you can expose/light in 709 but still record your file in SLog-2 and SLog-3

[Phil Rhodes]

 

 

AS7

 

A7S?

[Robin R Probyn]

the 800% is the DR percentage.. over the standard 709.. Slog is 1300% I think.. I use an F5 with the 709.800 MLUT .. but I think its alot better than the A7 "gamma assist" which only corrects contrast I think .. not colour as well...

 

my comment was more for daniel ..Slog or any real log will look like crap.. till its bought back,hopefully skillfully .. to 709 for tv anyway..display gamma.. how people light,expose and judge focus off an Slog image I,ll never know..

[Daniel Meier]

Its incredible how much noise there when you set your white at 59%!!!!

 

Now in your test when you had your white set at +1. Where was your white at in terms of IRE 109+?

 

So setting your white point at 95% to 100% seems to give the best compromise with noise and clipping. Di you have your zebra on? if so what was it set at? Just curious how you were meter for your test.

 

Thank for putting in the time doing this test! Super informative!

 

I exposed with a Zebra being set to 95%. Then I just opened up or stoped down the lens.

This was shot with a Sony FS 700. Just FIY.

 

@ Robin:

I did a color correction in Premiere to make it look like REC 709. (contrast, black pedestal, white point and saturation adjustments)!

[Kenny Keeler]

 

A7S?

 

Sorry typo, yes the A7S II

 

 

 

I exposed with a Zebra being set to 95%. Then I just opened up or stoped down the lens.

This was shot with a Sony FS 700. Just FIY.

 

@ Robin:

I did a color correction in Premiere to make it look like REC 709. (contrast, black pedestal, white point and saturation adjustments)!

Oh good to know. thanks

[Vivek Venkatraman]

Anyhow, I ended up shooting my project on the A7s II and instead of blindly just over exposing I exposed using the histogram, however I did not end up using S-LOG 3 for night as my testing showed a ridiculous increase in noise in shadow areas. So I used a combination of S-LOG 3 for day and no picture profile for night scenes and it worked out fine. Anyway I am yet to head to the coloring phase of the film and thats were you really get to know how good or bad your footage has turned out. So lets see.

[Phil Rhodes]

The apparent increase in noise is simply because of the way it's being recorded. If you look at it on a properly-corrected monitor, it's fine.

 

P

[Vivek Venkatraman]

Phil Rodes,

 

I saw it on a calibrated Mac Book pro retina display, There is clearly noise in blacks because S-Log 3 does lean more towards recording data in the highlights, anyway I will be seeing it soon in a 4k projector at a DI suite as well.

[Phil Rhodes]

Calibrated to what standard?

[Robin R Probyn]

Phil Rodes,

 

I saw it on a calibrated Mac Book pro retina display, There is clearly noise in blacks because S-Log 3 does lean more towards recording data in the highlights, anyway I will be seeing it soon in a 4k projector at a DI suite as well.

 

 

Incorrect sir.. Slog3 actually gives more data in the shadow,s.. (Slog 2 gives more to highlights).. thats why there appears to be alot of noise before grading .. before the black levels are taken down..

[Sam Sheppard]

Hello

 

If you'd like some advice on the post-production end: Here's what I've found from grading Sony RAW .mxf. Although the same holds true for XAVC if you're using that.

 

I love the F55 and F65 (at least I love the files that I receive. I can understand why the F65 is not every cameraman's friend when it comes to handheld or Stereo3D rigs). I've graded (mostly RAW .mxf from the F55) on Resolve and Baselight for HD television, and on Mistika the same delivery and also for HDR (1/2/ and 4,000 nits). In all cases I got great results but, in the case of HDR deliverables, the Sony exterior images were much easier to grade for the brighter displays than the equivalent from Alexa and RED Epic. Arri and RED naturally try to fit everything nicely between 0 and 1 (in float. That would be 0 and 1024 in 10 bit). Sony records 'everything that it sees', which can cause 1st time colourists not used to working with Sony material to panic a bit as doing things like debayering a bright exterior directly to Rec709 will give them what will look like a very clipped image (all of the information is actually there. Sony just allow it to extend beyond 1/1024 where as RED and Arri round off the highlights to give a nice 'normalized' image). If you know what you're doing though then Sony material is great as it will give you 'everything' and let a colourist who knows what they're doing creatively roll the exposure to where every they want it to sit. Grading Sony files set either to Linear Light or SLog3 (underneath a final transform to PQ and then to the monitor) gives a result that barely needs any adjustment if you're simply looking for a 'natural look' for how the scene looked on the day. Adding a creative grade from there is relatively easy compared to most other camera formats as I'm not fighting an in-built transforms like soft-clips/toe-roll-off.

 

I would always recommend exposing 'correctly'. However, if you have a gun to your head and a maniac forcing you to choose between protecting your blacks and protecting your whites I would recommend protecting your blacks when it comes to Sony material (set to SLog3). All of the information will be captured (I can easily unpick the SLog3 curve/gamma and transform it to any other the other curve options. The Log curve is mainly useful for 'cheating' extra headroom when recorded to XAVC. If you shoot XAVC set to SLog3 I can usually extract just as much info from the file as I would be able to get out of the same image shot Sony RAW. The SLog3 is mostly acting like an 'Image Zip file'). The trick is to use a purely mathamatical transform, what they call a 'Shader' in compositing-land, and not a LUT which usually won't work with values above 1/1024. In Mistika the effect is called 'UniColour' and it contains all of the camera transforms, including SLog2, SLog3, as well as LogC, Cineon, etc. It makes matching different camera types much easier that grading them all in their native log flavours. It's the contrast/gamma that's hard to match, not the chroma.

 

Why save the blacks? Because in the end it's all down to light hitting the sensor. If no light hits the sensor then I have no image. I have to either leave those shadows clipped/solid or composite in some sort of replacement. If you have protected the blacks then (with the way that the camera captures everything linearly) I'll probably have most of the highlight information that I need as the dynamic range is very impressive. I can usually do something with the whites even with the brightest image. However - if the blacks are clipped then they're just gone. Nothing that I can do with that.

 

I've never really had much trouble with Sony noise levels, but I've most graded F55 material. If you want a cheap but brilliant denoise tool though then the NEAT VIDEO plugin is brilliant and gets consistently great results. I know for a fact that almost every frame of The Hobbit went through that Plugin in the end.

 

Hope that's helpful. I absolutely love it when I hear that I'll be grading SLog3 or RAW Sony material. There's usually more options and less limitations than I would usually expect.

[David Mullen ASC]

I have yet to use the Sony F55 but a number of people seem to set the ASA of the camera to its base 1250 ASA but then expose it for 800 or 1000 ASA, which suggests to me that Sony wants to hold more headroom in S-Log but people are finding it necessary to give the image more exposure -- either because their S-Log-to-Rec.709 viewing signal to the monitors doesn't push the whites back up enough so at 1250 ASA, it feels a little dim in the highlights, or because they want more shadow detail.