Someone help clarify or simplify

[DS Williams]

Hello Red community!

I read this on a website where I was researching the linear nature of digital exposure:

"Most cameras are likely to cut highlights at low ISO, and cut dark shadows at high ISO. There is footroom, but no headroom at low ISO and vice versa"

I know that each sensor has a native ISO equivalent sensitivity brought about by numerous engineering limitations ect, and I know that higher/lower ISO sensitivitys are attained by applying a curve after the signal has been processed. But that is the extent of my knowledge. How does the curve look that's applied to lower sensitivity? And how does it lessen your headroom?

 

[Chris Newman]

As more light hits a sensor pixel, it fills up more buckets in the Analog to Digital converters. Every bucket is the same size. If it is a 12-bit ADC, you have 2^12 or 4096 buckets.

Imagine a situation where you expose properly (with no headroom) so that something white just fills all 4096 buckets. If you reduce the light by 1 stop, it will now use 2048 buckets, and you still have a lot of information to work with. You can further reduce the light by 10 more stops and still fill 1 bucket (be able to tell if something is black or white). Back with your original lighting, if you now increase the light by 1 stop, it is trying to use 8192 buckets that don't exist, so everything above 4096 clips to solid white.

By the way, some cameras (not RED) have analog amplifiers to set ISO sensitivity, so they have less noise at high ISOs than if you just grade it brighter.

 

[Reinhold R.]

Here's my attempt at an answer:

Imagine each photosite as a bucket of a specific volume. And imagine the photons as marbles pouring into your buckets. The sensor then counts the marbles and outputs the number as the brightness of that specific photosite.
If you expose properly everything is fine. Some buckets might fill up completely but that's OK as long as it's only a few. And there might be some buckets with only one or two marbles in them which might introduce some noise.
If you now expose your camera for a lower ISO that means that you would have to open up the lens 1, 2 stops. Accordingly the amount of marbles pouring in doubles or quadruples respectively. This means that more buckets will fill up completely (i.e. "clip") while the buckets where you only had a few marbles before will offer more marbles to count making for a more accurate brightness value (i.e. less noise).
If you do the opposite - i.e. exposing for a higher ISO - you will have to stop your lens down reducing the flow of marbles into the buckets. Chances of buckets pouring over decrease while the buckets at the lower end are getting less and less marbles, thus introducing more and more noise.
The short answer to your question is thus: since the files are raw, there natively is no curve that is applied. This is something that you can do as a means of creating a "nice looking" output for the director, producer, etc. by using REC709 or REDSpace as your display colorspace. The clipping happens because you open up the lens when shooting at a lower ISO, allowing more light to enter the lens and thus increasing the risk of clipping. The lower end on the other side can't really be clipped at all. There is just an aesthetic limit as to how much noise you can accept in the final output.
The first and last thing you should always do either way is, take a look at the histogram or false color mode. And expose as far to the right as you can without clipping. RED isn't film and you can run into trouble if you don't recognize that fact.

-Reinhold

 

[DS Williams]

First of all, Thanks to both of you for your input.

It's more clear now, let me see if I can get this straight:

The Red has a native ISO 320 sensitivity that will determine how fast these photosites react to light, and when you underexpose the sensor (i.e. rate it at a higher native ISO) you send less signal to your A/D converter and your lower midtones and shadows fall into the noise floor somewhat, and to bring them out you have to raise the electronic gain, i.e. apply an iso 'curve', which increases noise as you stretch the sensor away from it's native response.

But when you want to shoot at a LOWER rated ISO you end up overexposing and losing tons of information to clipped photosites, right?

Then how do you apply an ISO curve to clipped highlights, when youre overexposed a stop or two? The information is gone..

Where do your highlights end up falling on the curve? It obviously restricts your available latitude above midgray right?

I'm just trying to get my head around this. How do you go about metering and exposing if you're shooting ISO 100, for example?

 

[Chris Newman]

You can't shoot RED One rated at ISO 100 unless your subject has very flat contrast (no bright highlights) because, as you said, it will clip. If you wanted to attempt it, you would need to carefully watch the RAW histogram or false color (which you should do anyway). This assumes you don't want an overexposed blown out look.