ETTR: Noise, Dynamic Range and Skin Tones

[jonpais]

23 minutes ago, jagnje said:

Wow, you guys are really taking this to the next level

I have since forever exposed my footage to the right, even now that I shoot with log profile of the fs7. The reasson, it's how the companys colorist wants it. 

That being said, it's just preference. The differences in this cheap 8 bit cameras are so minimal that it hardly makes a difference. These are toy cameras that can give great results in the right hands and in the right circumstances. Expose as you wish and film what you like. 

I like that! Cheap 8 bit toy cameras.  Would you mind sharing some of your FS7 footage so we can see what real pro stuff looks like?

[Fredrik Lyhne]

Here's the graded shot of the new shot which I would say is a normal exposure. With this shot and the one "ETTR"? from yesterday I had to lower the midtones, but on this shot I could also raise the highlights slightly without loosing to much detail on the shirt. 

I would say it's a very good exposure for skin tones, but the shirt is loosing some detail, so I think 1/3 stop less would be perfect. 

On the ETTR I had to lower the midtones, but can't raise the highlights more and lowering them doesn't help with the skin tones or the shirt. 

On the -2/3 I had to raise the highlights and because the skin tones and shirt was not overexposed it didn't affect the image in bad way. I still think this was the best overall exposure of the bunch, but the new shot from today is also good. 

To match the brightness of the ETTR and -2/3 I raised the highlights even more from +5 to +10 and even though it seemed to clip the red channel slightly it still look better then the other shots I think. 

@jagnje we're just trying to find the best exposure settings as these cameras can go from good to great if you nail the exposure. And looking at these test I would say it's worth the time and effort we put into it  I understand that ETTR is the way do to it with many cameras, but these cameras seems to perform best when they are slightly underexposed. 

[TwoScoops]

29 minutes ago, jonpais said:

I like that! Cheap 8 bit toy cameras.  Would you mind sharing some of your FS7 footage so we can see what real pro stuff looks like?

Well, random pictures of flowers and cats look "pro" on an FS7. Kind of the same way in stills, a Pentax 67 with Porta 400 will make anything look good. 

[jagnje]

this was the last video I worked on, but I only filmed the crane bits. it is what it is...I don`t like the grade, but thats not my problem, I only film it. But don`t get me wrong, I love these camera, I own 2 as well. 

 

[jonpais]

19 minutes ago, jagnje said:

this was the last video I worked on, but I only filmed the crane bits. it is what it is...I don`t like the grade, but thats not my problem, I only film it.

 

Nice work! So you filmed the pianist and the four singers?

21 minutes ago, Fredrik Lyhne said:

Here's the graded shot of the new shot which I would say is a normal exposure. With this shot and the one "ETTR"? from yesterday I had to lower the midtones, but on this shot I could also raise the highlights slightly without loosing to much detail on the shirt. 

I would say it's a very good exposure for skin tones, but the shirt is loosing some detail, so I think 1/3 stop less would be perfect. 

On the ETTR I had to lower the midtones, but can't raise the highlights more and lowering them doesn't help with the skin tones or the shirt. 

On the -2/3 I had to raise the highlights and because the skin tones and shirt was not overexposed it didn't affect the image in bad way. I still think this was the best overall exposure of the bunch, but the new shot from today is also good. 

To match the brightness of the ETTR and -2/3 I raised the highlights even more from +5 to +10 and even though it seemed to clip the red channel slightly it still look better then the other shots I think. 

@jagnje we're just trying to find the best exposure settings as these cameras can go from good to great if you nail the exposure. And looking at these test I would say it's worth the time and effort we put into it  I understand that ETTR is the way do to it with many cameras, but these cameras seems to perform best when they are slightly underexposed. 

Today I shot in a mix of tungsten and daylight, which I really enjoy. It adds a touch of warmth. I also sat back a little from the balcony, so the light wasn't as harsh as the one at the coffee shop yesterday. I agree that 1/3 of a stop less for today's shot might have been best in this situation, mostly because of the radioactive shirt I was wearing. I still believe that many are really underexposing, so I hope others join in the conversation and share their clips and screen shots of the waveform monitors so we can all benefit.

[jagnje]

8 minutes ago, jonpais said:

Nice work! So you filmed the pianist and the four singers?

yup

[Orangenz]

14 hours ago, jonpais said:

That particular shirt is pretty bright. I didn't try grading the clip, but I can tell just by looking at it that that particular shirt is ridiculously ETTR resistant.  For the record, up until only very recently, I'd been underexposing all my shots without being aware of it, so I confess to being a little gung-ho about ETTR. If only the G85 had just a touch more dynamic range! I'd really like to do a comparison between the Lumix and my X-T2 someday.

I thought there were two sort of definitions of ettr though; one having the whole waveform pushed to the right and the other, at the same time, that you use the highlight clip level to make sure you preserve highlights from clipping. That and the long essay about exposure I posted does not preclude lifting or dropping problematic areas. The whole point is still those I mentioned earlier: drop the whole image and lifting in post dramatically increases noise problems and reduces the dynamic range of the image. Over exposure clips highlights and under studio lighting can reduce contrast. With those basics in mind the actual art of finessing the final product has the best chance of getting somewhere. My first 10bit vlog test on the GH5 was shooting a person against a sky. Really had to underexpose the person to keep the sky in check and still looked ok lifting him back up. We're living in the blessed age for imaging for sure.

[jonpais]

V-Log on the GH5 has at least two stops more dynamic range than shooting Cinelike D on the G85, so you've got far more latitude for lifting the shadows in post. 

Edit: What I'd really be interested in seeing are some original clips shot with the GX80/85 or G80/85 (which few here seem to own!) underexposed, then with shadows lifted in post, along with the scopes, so we can compare to those with 'correct' exposure. A moderator here has said that great results can be had underexposing and lifting shadows with the lowly G7, that there's no banding or artifacts whatsoever, so I'd really appreciate if anyone can corroborate.

[mercer]

It seems like the sweet spot is a third under to a third over, depending on the circumstances. If you're filming in the shadows shoot a third over. If you're shooting in bright light then a third under. In mixed light and shadows, dead center is probably the way to go with cameras that have 10 or less stops of DR. 

I appreciate what the Leeming LUT is attempting, but you can only push and pull these cameras so much and a half to a stop more latitude isn't turning it into an Alexa. But for what it is, you guys are getting great images and that is all anybody can ask. 

[jonpais]

6 minutes ago, mercer said:

It seems like the sweet spot is a third under to a third over, depending on the circumstances. If you're filming in the shadows shoot a third over. If you're shooting in bright light then a third under. In mixed light and shadows, dead center is probably the way to go with cameras that have 10 or less stops of DR. 

I appreciate what the Leeming LUT is attempting, but you can only push and pull these cameras so much and a half to a stop more latitude isn't turning it into an Alexa. But for what it is, you guys are getting great images and that is all anybody can ask. 

Yes, they may be 'toy' cameras, but with the right lenses, lighting, exposure and WB, they can really rock! The next hurdle for me is probably going to have to be color correction and grading.

[TwoScoops]

1 hour ago, jagnje said:

this was the last video I worked on, but I only filmed the crane bits. it is what it is...I don`t like the grade, but thats not my problem, I only film it. But don`t get me wrong, I love these camera, I own 2 as well. 

 

Do you remember which lens was used for the wide shot of wheelchair kid at 0:20?

[Fredrik Lyhne]

3 hours ago, jonpais said:

Today I shot in a mix of tungsten and daylight, which I really enjoy. It adds a touch of warmth. I also sat back a little from the balcony, so the light wasn't as harsh as the one at the coffee shop yesterday. I agree that 1/3 of a stop less for today's shot might have been best in this situation, mostly because of the radioactive shirt I was wearing. I still believe that many are really underexposing, so I hope others join in the conversation and share their clips and screen shots of the waveform monitors so we can all benefit.

I tend to underexpose to much at times and that's is something I really need to get better at. It's different but the same as overexposing and very hard to make it look good in post. 

2 hours ago, jonpais said:

Yes, they may be 'toy' cameras, but with the right lenses, lighting, exposure and WB, they can really rock! The next hurdle for me is probably going to have to be color correction and grading.

Your videos look very nice and the exposure is very consistent so I wouldn't worry about cc and grading. I only had to do very slight correction on it which to me means you're doing things very right. I think your documentary/filmic style works very well for your videos. Grading looks however is off course a whole nother chapter.

[BTM_Pix]

Just a slightly different aspect to this but I was curious about how accurate the zebras are and how useful the histogram is with these cameras so I did a little test with the G7.

I set the zebras to 100 to flash the middle column where the colour checker is stuck on and framed it so the door was in shot to see the different shadow levels with different profiles.

Ignore the completely deliberately blown out right third portion of wall and sky as this is for a completely different test !

So that you could see what the camera was actually doing (and more importantly what it was saying it was doing), I recorded the output with the overlays on into a Ninja Inferno and did 3 passes using Standard, Cinelike D and Natural profiles, altering the shutter speed to change the exposure to show where the zebras came on for each.

Annoyingly, I can't record the scope overlays of the Ninja Inferno as it was interesting to monitor what the camera thought was 100 and what the Ninja Inferno did.

The good news is that the zebras are pretty accurate (I checked at 80, 70 and 60 as well) and you can see that reflected in the scopes in these grabs in FCPX.

You can also see that Cinelike D is reducing the peak and stretching other parts of the range compared to the same exposure using the different profiles (all profiles were set to 0 for every parameter so there is more to be had from tuning but that was outside of the scope if you pardon the pun of what I was doing.

What I found interesting is how there was quite a coarse change in level between the exposure combination of ISO/SS and F stop that would trip the zebras and the closest one that wouldn't, with the level going from 100 to around 92-ish. Using a variable ND, I was able to slightly tune that between the cracks as it were to eek a bit more out but if you've got the precision to be turning an ND like that all the time then you'd be better off being a safe cracker.

So, the takeaway for me from doing this is that the zebras are accurate enough to use at 100 if you want to really expose to the right and the histogram is, erm, a yellow graphic.

Download link for the video file if anyone wants to have a look in their own NLE is here

https://mega.nz/#!ZyYCgC6R!WTM3Nz2MXfSzKtwG0hPhOBAvSqNaOq4eOnIUycnPQ3I

 

[Fritz Pierre]

I think one exposes to the occasion...with any camera...period...and when confronted with too large a DR you make choices...I would not hesitate to blow a highlight, if the DR was to high, for whatever camera one's shooting on...Kubrick did it constantly in his films... (not comparing myself to Kubrick???, but we can learn a lot from the masters)

ETTR on the GH4 and 5 seems to give you a cleaner image in the shadows....if that detail is important, you expose accordingly and lower the shadows in your grade rather than raise them...unless you of course don't care about the noise...but it's always down to what you are trying to say with what you are shooting...I will say this much! IMO @jonpais exposes to as close to perfection as I see on most forums and on any camera...so whatever he's doing, he's doing right!...we are all lucky to have a source of knowledge like Jon's on the forum, tirelessly sharing with us all!!

[jonpais]

1 hour ago, BTM_Pix said:

Just a slightly different aspect to this but I was curious about how accurate the zebras are and how useful the histogram is with these cameras so I did a little test with the G7.

I set the zebras to 100 to flash the middle column where the colour checker is stuck on and framed it so the door was in shot to see the different shadow levels with different profiles.

Ignore the completely deliberately blown out right third portion of wall and sky as this is for a completely different test !

So that you could see what the camera was actually doing (and more importantly what it was saying it was doing), I recorded the output with the overlays on into a Ninja Inferno and did 3 passes using Standard, Cinelike D and Natural profiles, altering the shutter speed to change the exposure to show where the zebras came on for each.

Annoyingly, I can't record the scope overlays of the Ninja Inferno as it was interesting to monitor what the camera thought was 100 and what the Ninja Inferno did.

The good news is that the zebras are pretty accurate (I checked at 80, 70 and 60 as well) and you can see that reflected in the scopes in these grabs in FCPX.

You can also see that Cinelike D is reducing the peak and stretching other parts of the range compared to the same exposure using the different profiles (all profiles were set to 0 for every parameter so there is more to be had from tuning but that was outside of the scope if you pardon the pun of what I was doing.

What I found interesting is how there was quite a coarse change in level between the exposure combination of ISO/SS and F stop that would trip the zebras and the closest one that wouldn't, with the level going from 100 to around 92-ish. Using a variable ND, I was able to slightly tune that between the cracks as it were to eek a bit more out but if you've got the precision to be turning an ND like that all the time then you'd be better off being a safe cracker.

Download link for the video file if anyone wants to have a look in their own NLE is here

https://mega.nz/#!ZyYCgC6R!WTM3Nz2MXfSzKtwG0hPhOBAvSqNaOq4eOnIUycnPQ3I

 

Not sure I understand everything here, but after last night, I did begin to wonder just how accurate zebras actually are on these cameras, so good to know they're fairly reliable. 

[BTM_Pix]

1 hour ago, jonpais said:

Not sure I understand everything here, but after last night, I did begin to wonder just how accurate peaking actually is on these cameras, so good to know it's fairly reliable. 

Yeah, I was very interested in how much faith could be put in them if you wanted to really expose to the right and have them set at 100.

And my takeaway is that they can largely be relied on if you want to push that far. It was also interesting, for me at least (!), to see the interaction between the profiles and the levels at the same exposure settings. Considering how coarse the histogram display is then this is a good thing

What I've seen here might form the basis of a bit more testing for me at some point actually to look at the relationship with the profile, the contrast parameter within it and the additional global shadow and highlight adjusters in the camera. I think there is possibly some scope there to create profiles that incorporate all those parameters into single combination packages to create potentially finished OOC looks that I could make be deployed with a single button press.

There might actually be combinations of these parameters that you'd maybe describe as non-intuitive that you look at the actual image its producing and think "You know what, that shouldn't work but I like the way that that looks".

With analog synths for example, you pushed and prodded at the controls because they were there in front of you rather than having to deliberately go into menus and find them like you did with digital synths. It brought about happy accidents (as well as absolute abominations of noise) and I think it could be the same with these cameras. Its too fiddly to go in and change things in them through the menus so you become so methodical that you seldom create combinations that might just be really interesting.

If all those parameters are laid out in front of you for you to twiddle with and you can see the live results of them (because you don't have to go in and out of menus) then I think that might happen.

You'll probably end up with lots of rubbish but might just pan some gold.

I'm quite enthused about this now and I'm going off to hack together something to do it!

[jonpais]

@BTM_Pix A few weeks back, I did a direct comparison under studio lighting between the same three profiles, and Standard and Natural were so much more contrasty than Cinelike D that if I ever were to shoot with them again, I'd probably dial exposure back, and I'm not sure ETTR would be a good idea with them. Would you mind having a look and teling me what you think?

[BTM_Pix]

4 minutes ago, jonpais said:

@BTM_Pix A few weeks back, I did a direct comparison under studio lighting between the same three profiles, and Standard and Natural were so much more contrasty than Cinelike D that if I ever were to shoot with them again, I'd probably dial exposure back, and I'm not sure ETTR would be a good idea with them. Would you mind having a look and tell me what you think?

Will do.

Just let me finish prototyping this interactive picture profile creation tweaker 

 

[BTM_Pix]

1 hour ago, jonpais said:

@BTM_Pix A few weeks back, I did a direct comparison under studio lighting between the same three profiles, and Standard and Natural were so much more contrasty than Cinelike D that if I ever were to shoot with them again, I'd probably dial exposure back, and I'm not sure ETTR would be a good idea with them. Would you mind having a look and teling me what you think?

This is going to be in sections so bear with me, but loading them into Aperture, you can immediately see the distribution difference with Cinelike D and the Natural profile and this tallies with my test earlier on today.

Everything is condensed more with the Cinelike D but its interesting to see how high the blue channel is and how it is almost clipping.

 

 

*** NEXT PART ***

So I can get them to somewhat match visually and using the starting point of Aperture's Auto Exposure function (which is actually reasonably accurate based on if you underexpose in the camera then the correction it puts on more or less always corresponds to the amount of underexposure)

From that, it could be see that there was more to go with both of them but quite a bit more with the Cinelike D version.

Using the Auto Enhance function (which attempts to give a half decent balance in terms of it usually lifting the shadows) there was a bit more to be had from both and both could be pushed a bit further if needed.

But what was apparent was still this narrower band for the Cinelike D version which I had to tweak by adjusting the Green level and rolling off the top of the Blue channel.

 You can see most of the individual parameter changes for each image in the screenshots for each one.

In answer to whether its better to ETTR with the other profile, I'm not sure but there was certainly a smaller margin for error as it was closer to the edge than the Cinelike D one.

Is there something in the colour of the lights that was causing the raised levels in the blue channel with Cinelike D perhaps?

[jagnje]

5 hours ago, TwoScoops said:

Do you remember which lens was used for the wide shot of wheelchair kid at 0:20?

it must have been a canon 24-70, 2.8 on a speedbooster. I didn`t do those shots, but that is the lens we use most.