8k r3d 5:1- 6 STOP UNDER AND OVER EXPOSE TEST

[rand thompson]

8k r3d 5:1- 6 STOP UNDER AND OVER EXPOSE TEST



By göker gören



He Shows what s what on the chart, which will be kinda small in the video here. However, the test starts with smaller side-by-side screens then goes to a single full Screen at about the 1:23 minute mark. So I added the YouTube link so it can be seen full-screen on there. This test is done with a Monstro.

https://www.youtube.com/watch?v=JrczSC4srXI

 

[Phil Holland]

The iris over/under test is good at revealing what happens at the point of clipping and what happens deep in the noise floor.

I just hope that people understand that with any digital system once you've hard clipped/over exposed into the heavens and hard crushed/under exposed into the noise floor there's nothing left to salvage.

Honestly, worked the same way for film. There is a point where nothing exists.

 

[rand thompson]

Phil,


Yeah I tried to recover .R3d files that were +5 and +6 respectively, however the color kinda goes to shit after +4 stops over. So, as you stated, its just not worth it after +4 stops over and maybe -2 or -3 stops under.

 

[Phil Holland]

You have a lot of workable latitude within a well exposed image. That's pretty much how this all works. There's some tricks to deploy when really strange things happen, but with any of these cameras it's really about exposing correctly.

Nobody is perfect, but if people are 6 stops over or under, they are getting fired. This is still a useful test for sure to see what is happening, but it emphasizes really why it's critical to expose well with whatever you're working with.

 

[rand thompson]

Phil,


I don't own the camera, but what in-camera tool do you feel gives the most accurate reading of proper exposure?

 

[Aaron Lochert]

I don't own the camera, but what in-camera tool do you feel gives the most accurate reading of proper exposure?

Gio Scope is incredibly useful to "zone system" your entire scene. But at a quick glance, the stop lights and the goal posts will save you from a lot of issues.

 

[Phil Holland]

I don't own the camera, but what in-camera tool do you feel gives the most accurate reading of proper exposure?

You've got "many" and I call them out in the DSMC Field Ops Guide.

As Aaron mentioned, GIOSCOPE sees what the sensor sees and it's incredibly useful for discerning what is noise and what is clipping and what's rolling off to clipping.

You also have the STOP LIGHTS, which will inform you as to what is clipping immediately.

And also the GOAL POSTS which fill up as more of the image is clipping.


Honestly, this is one of the areas where RED excels when it comes to modern digital cinema. The tools are right there available at a button press to ensure you aren't doing anything nuts.

 

[rand thompson]

Gio Scope is incredibly useful to "zone system" your entire scene. But at a quick glance, the stop lights and the goal posts will save you from a lot of issues.

You've got "many" and I call them out in the DSMC Field Ops Guide.

As Aaron mentioned, GIOSCOPE sees what the sensor sees and it's incredibly useful for discerning what is noise and what is clipping and what's rolling off to clipping.

You also have the STOP LIGHTS, which will inform you as to what is clipping immediately.

And also the GOAL POSTS which fill up as more of the image is clipping.


Honestly, this is one of the areas where RED excels when it comes to modern digital cinema. The tools are right there available at a button press to ensure you aren't doing anything nuts.

Thanks a lot both Aaron and Phil!!

 

[Todd S]

If there is one single message I could send to other camera manufacturers, it would be something like:
"OK cut the crap. Just show me the GIO Scope"

 

[Karim D. Ghantous]

These aren't the most important tests one can perform on a sensor, although they are indeed important. The main thing for me is how the sensor handles small, intense sources of light. Most sensors fail. The MX sensor didn't do very well but it was hardly the only or worst offender.

I have not seen all the evidence I want yet, but neither the Alexa nor the DSMC1/2 can handle tail lights properly. During the day when there is plenty of ambient light, they both do quite well. Not so at night. Notice, too, that some digital cameras also show flickering in automotive LEDs. Unacceptable for feature films.

Are there filters that can minimize the problems caused by point sources? For example, a Tiffen Ultracon. I sort of know what it is but I have never used one.

 

[Brandon Obey]

I concur. GioScopes help alot

 

[Patrick Tresch]

As always, the reference is taken too high. Iso 1600 should be native and the starting point.

Pat

 

[Marc Wielage]

The iris over/under test is good at revealing what happens at the point of clipping and what happens deep in the noise floor. I just hope that people understand that with any digital system once you've hard clipped/over exposed into the heavens and hard crushed/under exposed into the noise floor there's nothing left to salvage. Honestly, worked the same way for film. There is a point where nothing exists.

Actually, I think film was the opposite: negative was usually pretty kind to the highlights if you over-exposed a little bit, like 2/3 stop or 1 full stop. We often told people shooting for TV to do that, because that way we had a "thicker negative" that would yield more image.

With digital, we found the highlight range was more limited and you always had to protect the highlights whenever possible... because when it's clipped, it's gone forever. The histograms will tell you very quickly when the specular highlights have slammed into the digital ceiling. Ideally, the exposure is somewhere in the middle, but we get that real-world shooting conditions go all over the map, and that's just life (and weather and smoke and new shooting days and pickups and so on). A reasonable colorist can always match the material.

 

[Patrick Tresch]

Actually, I think film was the opposite: negative was usually pretty kind to the highlights if you over-exposed a little bit, like 2/3 stop or 1 full stop. We often told people shooting for TV to do that, because that way we had a "thicker negative" that would yield more image.

The reverse was also true for reversal film wich acts more like digital (over exposure are lost infos).

 

[Aris_Gavriilidis]

Rating Monstro at a 1600 base ISO seems like a good idea based on what I see here, really surprised to see it clip this bad at +5 over, even the Weapon and Epic Dragon did better on the ACS test from 2015. What OLPF was in the camera?

 

[Hrvoje Simic]

I have not seen all the evidence I want yet, but neither the Alexa nor the DSMC1/2 can handle tail lights properly.

As every other digital camera in this domain, where Alexa leads in that aspect.

 

[Hrvoje Simic]

Rating Monstro at a 1600 base ISO seems like a good idea based on what I see here, really surprised to see it clip this bad at +5 over, even the Weapon and Epic Dragon did better on the ACS test from 2015. What OLPF was in the camera?

Higher rating does to shadows what lower does to highlights. With additional price of a stop under thinner image overall, further degraded by effect of compression on low signal values.

You don't have any reference on face light values to draw comparisons.



Also, expecting marvelous lattitude and transition to clip in high(er) compression ("that which throws out data") is looking for the flowers with a lawnmower.

 

[Patrick Tresch]

Higher rating does to shadows what lower does to highlights. With additional price of a stop under thinner image overall, further degraded by effect of compression on low signal values.

You don't have any reference on face light values to draw comparisons.



Also, expecting marvelous lattitude and transition to clip in high(er) compression ("that which throws out data") is looking for the flowers with a lawnmower.

You don't see compression at 1600 iso level if you are below compression 1:8 , but above 1:12. So you don't degrade the low signal value if you expose for 1600 iso. We always speak about the highlight war but Alexa is crap in lowlight. Alexa is a 800 iso camera, Dragon and Monstro shine at 1600iso.

 

[Hrvoje Simic]

You don't see compression at 1600 iso level if you are below compression 1:8 , but above 1:12.

Yes I do.

And with guesswork that's also oversimplification.

So you don't degrade the low signal value if you expose for 1600 iso.

If you rate any camera a stop higher you degrade the signal density by underexposing one stop.
If the material is also compressed there is additional penalty in low signal values and lattitude properties. Penalty which depends on type of compression.

We always speak about the highlight war but Alexa is crap in lowlight. Alexa is a 800 iso camera, Dragon and Monstro shine at 1600iso.

It's not that simple.

Camera sensitivity is one thing.
DR utilization by the camera user is another.

Red cameras rated at 800 have superior shadow rendition than Alexa. Which also depends on transformation from log.
Rated at 1600...not really.

Pushing the signal to see more in the blacks has a price.
And that price primarily affects density.

 

[AndreasOberg]

I did a more detailed compression test and I can see the difference between
1:2, 1:5, 1:8, 1:12 and 1:18.

At 100% zoom already at 1:5 the quality of noise and smaller details are being reduced and at each step there is further reduction.
Especially at 1:12 the noise has clear compression patters and details are not there as much any more.

Andreas