- This topic has 17 replies, 5 voices, and was last updated 4 months, 3 weeks ago by
.
-
Topic
-
Sir Roger
I’ve attended a workshop about Exposure, and the DP mentioned that he uses High ISO in the ALEXA when shooting in a bright daylight to protect the highlightsand add more stops above the mid grey.
So that makes me wondering rather than using ISO 400 or lower, I should do the opposite to gain more stops of dynamic range for highlights.
Do you recommend doing such a thing ?
Thanks in advance.
-
Replies
-
Yeah, the dynamic range is redistributed toward underexposure when you lower ISO, which is often counterintuitive for the bright situations where you’d do so. The reasons to do it rather than maintaining more headroom at base ISO and using ND filters as needed would be that you want the highlights to clip sooner or you want less noise.
This is only valid when shooting raw, not when shooting a compressed codec.
You can do that, it will distribute more of the overall dynamic range to the highlights.
Personally I find it mostly unnecessary and stick to base ISO’s and expose from there.
I do often lower ISO in low light though to distribute more dynamic range to the shadows.
Depends on your day exterior situation — most of the time, the headroom on the Alexa at ISO 800 is fine. But I suppose you may have a scene where everyone is dressed in white on a sunny day or the camera is pointed into the ocean with hot sunlight glaring off of it where you may want a bit more headroom, but keep in mind that higher ISO means more noise.
I’m sorry, Stip, but you’re incorrect. Spatial compression does not influence the distribution of camera latitude.
The ideology behind this is a camera is a radiometrically linear device. You quantize (convert to digital) the signal linearly; once in this digital domain, the change of ISO by half or double is similar to a 2* or 1/2 operation of that linear quantized signal.
(In post, the manufacturer’s SDK typically linearises the logarithmic container in your NLE when dealing with RAW files to adjust ISO.)
Thus, raising your ISO is essentially ‘under-exposing’ the image. If you under-expose an image without this ISO adjustment (staying at native), it will look darker, with retained highlights. When you amplify this signal in the digital domain (over-expose with ISO), the image will look correct, with retained highlights.
It’s important to note that heavy amplification of finite digital values causes aggressive noise patterns. This is due to the finite values; if you have values of 2, 3, 4, and 5 as your linear quantised native signal and amplify them by a factor of 6 (ISO), your discrete steps become 12, 18, 24, and 30. Large value gaps are easily noticeable by our photopic vision because we perceive light logarithmically (each stop of linear light with the same amount of value).
As a final exciting thought, your photopic vision perceives light logarithmically in a linear realm; our brain does a linear to logarithmic transform, and a camera does a linear to logarithmic transform; this is why the logarithmic container looks incorrect to us, as we’re playing a logarithmic distribution of luminance before the linear to logarithmic conversion in our mind!
Infinityvision.tv
Gabriel Devereux - EngineerWhat do you mean (please in easy words).
Recording to a compressed codec is not the same as recording raw. If the ISO setting applies a gain, you will bake that in.
When using a compressed codec, if you rate down you’re still under-exposing the image and retaining more apparent highlight information.
The difference between recording to a compressed codec and recording a raw file is that in a RAW file, each pixel is stored as a single Y value.
This is because sensor RAW information is a single Y (luminance) value per pixel, and we assume values from surrounding pixels to make up the final RGB triplet.
You can still have spatial compression – block type or discrete cosine transform to this 2D plane of RAW Y values.
When changing ISO, you multiply this single Y value per pixel. When a pixel is demosaiced, it undergoes so many functions, such as OETF, WB, etc., that the original ISO function is no longer applicable to the RGB triplet.
Infinityvision.tv
Gabriel Devereux - EngineerWhen using a compressed codec, if you rate down you’re still under-exposing the image and retaining more apparent highlight information.
Yes but this is about rating ISO higher, not lower, aka baking gain into a compressed codec.
Sorry*
I made a mistake. If you rate *UP*
That was the only term that was incorrect. As far as I’m aware.
I hope this gives clarity.Infinityvision.tv
Gabriel Devereux - EngineerWith raw it’s just metadata. But when shooting to a codec, and raising ISO, you are now baking (the camera’s) gain into the file, which pushes the highlights closer to clipping point with every ISO / gain step added.
High dynamic range cameras may deal better with this but others won’t. I see more risk than benefit.
No, that’s incorrect.
Your codec and your codec’s bit depth do not dictate the amount of latitude it can handle. Extreme compression (outside of the world of ProRes, XAVC-I, etc.) has an impact, but again, it’s spatial compression. It doesn’t compress luminance values more than a bit-depth limiter.
The camera’s logarithmic container almost entirely does that process, which is why we shoot in log.
The goal of a logarithmic container is to encode linear values with an equal amount of logarithmic values past a linear threshold.
The 16-bit unsigned integer photo site value (a linear representation of the energy the photosite receives) is encoded to a 12-bit integer value in its Log-C space.
This compression allows us to record high-dynamic-range imagery visually losslessly in a codec with a bit depth less than our linear value.
A logarithmic container typically exceeds the camera’s requirement. Being able to store more logarithmic values than the camera can produce linearly.
All the previous explanation stands with this reasoning.
Infinityvision.tv
Gabriel Devereux - Engineer12-bit ARRIRAW stores no more dynamic range or latitude than 12-bit ProRes.
Both use logarithmic encoding to maximize dynamic range.
However, ARRIRAW’s primary advantage lies in its flexibility in post-processing, thanks to its retention of more of the sensor’s original data and metadata.
This has zero impact on rating at a higher ISO and retaining more highlight information due to the above-explained process.
Infinityvision.tv
Gabriel Devereux - EngineerI think we’re talking past each other. I don’t talk about bit depth or effects of compression.
My point is that changing ISO does nothing to the recorded raw file. But it does something when recording to a codec – if you raise ISO above base, the camera adds gain. That added gain is baked into the file – contrary to when shooting raw.
E.g. if you raise base ISO one stop, 1 stop of gain will be baked in, which will push the values 1 stop closer to clipping point. This does not happen when shooting raw, as it only changes metadata and your behaviour (exposure) when looking at a, now brighter, screen.
I’m not talking about highlight information, but clipping headroom.
Anyways, enough input from my side 🙂
P.S.: I’m a filmmaker, not a technician, also English is not my native language – so the words and terms I use may be too imprecise and cause confusion.
You are forgetting that if you select a higher ISO you are underexposing the sensor so there is no loss in highlight information recording ProRes compared to recording raw and underexposing, and then brightening the shot. You bake in noise and color temperature in ProRes Log C but there is no loss in dynamic range.
- You must be logged in to reply to this topic.