Welcome to our community

Be a part of something great, join today!

  • Hey all, just changed over the backend after 15 years I figured time to give it a bit of an update, its probably gonna be a bit weird for most of you and i am sure there is a few bugs to work out but it should kinda work the same as before... hopefully :)

R3D sample: Folks with more RED experince than I please weigh in on noise performance

It does NOT look like that coming out of RCX pro on my computer....... Wow....

OK, so i gotta trouble shoot my computer perhaps?

sorry that is the shot at 800, I just updated the 2000ISO shot
 
SCREENSHOT.png


Hmm.... @ 2000 it looks the same as 800.

Open the image in a new tab to see a full size screen shot.
 
Thanks for taking the time Matt.

no problem man, I want to perfect this as well. The way I've been doing it is rating at 800 all the time, but in tungsten or low light I rate it at 500. I have been getting noise free images every time.

I know that going under 320 does nothing for the image being recorded. As a test you could go outside in daylight and start at 320 ISO, then go down to 250ISO, and then 100ISO and the meters will stay the same. I tried this the other day when we didn't have the proper ND for a shot (just a test, not a real shoot).
 
You want an interesting reality check ?

Take a DSLR with a crop sensor ( APSc ) and set the camera to RAW.
Set same ISO, same shutter speed.
Shoot subject in still photo mode, but continuous shutter. Hold down shutter and let it take as many shots as it can. This will be 8 to 20+ RAW pictures shot.
Take those photos ( lets call it a movie now, even if the frame rate is rather slow ) and develop in Adobe Camera Raw. Turn off sharpen, turn off all the noise reduction, etc.
Save out the little 'movie' as TIF 16 bits.

Now you can compare, after grading to similar contrast and look as the r3d footage.
Noise issues must be examined in motion. Play the movies.
Discuss ;)
 
Timur, although the shot doesn't look too bad noise-wise to me, I think it is underexposed.

What did the histogram look like on the camera at shoot time? There's an objective way to set the optimal exposure for the camera if you have the true histogram: expose to the right.

Your highest highlight should JUST fall short of 100% of the scale. RED gives us the super RGB traffic lights so we can see where we've gone a bit far and pixels have exceeded that maximum (they're clipping). The reason for this is that the camera has a certain range of light intensities it can capture, corresponding to a certain number of electron-hole pairs produced in the sensor, which corresponds to a certain readout signal.

The minimum signal it can register depends on the noise floor, which is what you do a black shade calibration to set. The camera subtracts off the average "black" level for each pixel. The minimum signal you can see depends on how noisy the chip is and how well you've done that subtraction.

The maximum signal it can register depends on the maximum well capacity of each pixel. Unlike film, this is a sharp cut-off. If any extra light falls on a sensor that's already "full", it doesn't register any extra because it can't. That's what clipping is.

Turning the ISO up or down doesn't change the physics of the sensor. The sensor's noise floor and maximum well capacity are the same whether you set ISO 100 or ISO 100,000.

If your histogram is all at the dark end and your highlights do not go to 100%, you've underexposed, you've got less signal that you'd ideally have, and have compressed your dynamic range because everything is crowded down in the darks at the left hand end. If you have tones crowded together, you need them crowded together at the highlight end, because fully half your available signal is in the final stop at the highlight end. Ironically a digital sensor is at its very finest, most discriminating, subtlest and most noise free just before it clips at the highlight end!

So the exposure routine I follow is to look what my highest highlights are doing, and turn the exposure up or down (either on camera or on lighting) to make sure they are near but not above 100%. Then I look at the blacks - the default setting on the 5" LCD bottom button brings up "red for clipping purple for danger of noise" overlay. If there's anything in the red, you need to deal with it. Either turn the lights down, stop down, or (in extremis like a specular highlight) choose to live with it. If there's stuff in the purple you care about, you need to reduce the contrast in your scene, probably by adding fill light, until you can see clearly into those shadows. It doesn't matter if that boosts them to 50%+ - all the better, in fact. You can bring them down again when you "develop" the image.

It is much better to get stuff crowding up towards the highlight end, where the best detail is recorded and can be brought out in development. Stuff crowding at the shadow end has a crappy signal-to-noise ratio and will never look as good. So the only stuff down near the 0%-10% end should be stuff you'd be happy to crush to absolute black.

This is very different from shooting for the look. You don't care about the apparent contrast or where the skin tones land or anything- that's the job of the RAW processing step to decide. What you need in camera is the "digital negative" with detail in all the places you need it. The ideal histogram is basically THE SAME regardless of subject: a nice flat range between 100% and 0%. Whether you're going to map those tones to reproduce the silky detailed blacks of a black cat in a coalmine or a polar bear on the ice pack is almost irrelevant - the camera will perform best with histogram "between the goalposts", shot to the right.

It is very similar to the process of shooting and printing a film negative (see Ansel Adams' "The Camera", "The Negative" and "The Print") or see Michael Reichmann's essays over at Luminous Landscape: http://www.luminous-landscape.com/tutorials/expose-right.shtml and http://www.luminous-landscape.com/tutorials/optimizing_exposure.shtml

You might be interested to know that the Scarlet's monitoring tools supports shooting exactly as Michael was calling for in that second essay- progress! :)

Cheers, Hywel
 
Last edited:
Yea, im figuring out that exposing a red is different than other digital systems. Im essentially trying to figure out where i should view (what ISO) to ensure that onset lighting descisions are giving e the most i can hold. Most other cameras tend to despise beiing pushed up that high, but it seems the red is solid right to the line which is great.

I suppose 800 is the place to be.
 
ISO800 is a great space to start with, and you'll not go too far wrong with. As you get the feel of how the camera works, it's possible to vary from that recommendation quite safely. If you go to a lower ISO, just get extra careful on highlight clipping (use the in-built tools here, traffic lights, goal posts and especially the exposure overlay). If you're going to a higher ISO, be careful that you're well exposed at that higher ISO as it's easy to not be as well exposed as you think you are as you go higher. As always testing is good.

Graeme
 
Timur, although the shot doesn't look too bad noise-wise to me, I think it is underexposed.

What did the histogram look like on the camera at shoot time? There's an objective way to set the optimal exposure for the camera if you have the true histogram: expose to the right.

Your highest highlight should JUST fall short of 100% of the scale. RED gives us the super RGB traffic lights so we can see where we've gone a bit far and pixels have exceeded that maximum (they're clipping). The reason for this is that the camera has a certain range of light intensities it can capture, corresponding to a certain number of electron-hole pairs produced in the sensor, which corresponds to a certain readout signal.

The minimum signal it can register depends on the noise floor, which is what you do a black shade calibration to set. The camera subtracts off the average "black" level for each pixel. The minimum signal you can see depends on how noisy the chip is and how well you've done that subtraction.

The maximum signal it can register depends on the maximum well capacity of each pixel. Unlike film, this is a sharp cut-off. If any extra light falls on a sensor that's already "full", it doesn't register any extra because it can't. That's what clipping is.

Turning the ISO up or down doesn't change the physics of the sensor. The sensor's noise floor and maximum well capacity are the same whether you set ISO 100 or ISO 100,000.

If your histogram is all at the dark end and your highlights do not go to 100%, you've underexposed, you've got less signal that you'd ideally have, and have compressed your dynamic range because everything is crowded down in the darks at the left hand end. If you have tones crowded together, you need them crowded together at the highlight end, because fully half your available signal is in the final stop at the highlight end. Ironically a digital sensor is at its very finest, most discriminating, subtlest and most noise free just before it clips at the highlight end!

So the exposure routine I follow is to look what my highest highlights are doing, and turn the exposure up or down (either on camera or on lighting) to make sure they are near but not above 100%. Then I look at the blacks - the default setting on the 5" LCD bottom button brings up "red for clipping purple for danger of noise" overlay. If there's anything in the red, you need to deal with it. Either turn the lights down, stop down, or (in extremis like a specular highlight) choose to live with it. If there's stuff in the purple you care about, you need to reduce the contrast in your scene, probably by adding fill light, until you can see clearly into those shadows. It doesn't matter if that boosts them to 50%+ - all the better, in fact. You can bring them down again when you "develop" the image.

It is much better to get stuff crowding up towards the highlight end, where the best detail is recorded and can be brought out in development. Stuff crowding at the shadow end has a crappy signal-to-noise ratio and will never look as good. So the only stuff down near the 0%-10% end should be stuff you'd be happy to crush to absolute black.

This is very different from shooting for the look. You don't care about the apparent contrast or where the skin tones land or anything- that's the job of the RAW processing step to decide. What you need in camera is the "digital negative" with detail in all the places you need it. The ideal histogram is basically THE SAME regardless of subject: a nice flat range between 100% and 0%. Whether you're going to map those tones to reproduce the silky detailed blacks of a black cat in a coalmine or a polar bear on the ice pack is almost irrelevant - the camera will perform best with histogram "between the goalposts", shot to the right.

It is very similar to the process of shooting and printing a film negative (see Ansel Adams' "The Camera", "The Negative" and "The Print") or see Michael Reichmann's essays over at Luminous Landscape: http://www.luminous-landscape.com/tutorials/expose-right.shtml and http://www.luminous-landscape.com/tutorials/optimizing_exposure.shtml

You might be interested to know that the Scarlet's monitoring tools supports shooting exactly as Michael was calling for in that second essay- progress! :)

Cheers, Hywel

Great post thanks :-)
 
Going from ISO250 to ISO800 is not the same as the other way around solely because your onlocation/physical exposure adjustments (shutter, f-stop, ND) will be different in both scenarios. In my (limited) experience, you get way more room shooting it high/bright/flat, reducing it down to ISO320 in post...which seems to confirm shooting to the right. Plus it's kind of awesome watching a bright sky go from washed-out white to a solid blue.

With this shot, I had trouble increasing exposure while keeping good dynamic-range/gradations in the skin/mid-tones. Most of the skin tones seemed to adjust in unison, rather than being able to fine-tune different shades of skin (highlight, mid, shadows of just the skins/mids), which is generally where I want to be. That said, it's still workable, especially with secondaries or with some of AE's effects.
 
There's an objective way to set the optimal exposure for the camera if you have the true histogram: expose to the right.

Your highest highlight should JUST fall short of 100% of the scale.
Just a side note: When you're monitoring RAW the histogram will not actually go to 100%. There'll be a gap on the right side even when your highlights are blown, so keep your eye on the stoplights and goalpost and use the exposure tool.
 
Hmm.... @ 2000 it looks the same as 800.

Open the image in a new tab to see a full size screen shot.

That's because your curve is bring down the exposure significantly.

There is an advantage to shooting 320ISO in that you sacrifice shadow information when it gets crushed down. But in this case you just exposed for 800. ;) Showing "2,000" ISO but with a curve that brings down the exposure by a stop isn't actually changing anything you're just push processing and then grading back down... which in 32 bit color doesn't actually harm the image.
 
Shoot for shadows! The histogram is too small and arbitrary to really have full control over your image. Use your meter. Fill the shadows with information. Overexpose by a stop to fill in the pores and corners! Drop into RED CINE X and shift your ISO to 400. Add some contrast and job done. No noise for me at all. Ever. This camera has really nice roll off in the clipped highlights, as well as latitude, so in RedLogFilm you don't have to be SO paranoid about your highs. It's not an AF100. There is some common sense involved for clipping, but overall it's easy peasy. Ansel's zoning system explains this, and it correlates perfectly to the Scarlet workflow.

I love this little beast...
 
Back
Top