Why not ISO 100?

[Jannard]

This is an often confused issue.

The RED Mysterium sensor is rated ISO 320-500. That means that if you raise or lower the ISO rating of the RED ONE, you are "stressing" the image (or bending curves) to get it brighter or darker. While you can certainly do this, you pay a price in either direction. And the price gets greater the further away you go in either direction from rated ISO 320-500.

If you raise the ISO to 1000, you will add noise. If you lower the ISO to 100, you will reduce resolution and get a "more sterile" looking image. Not much, but it should be noted.


RED is not film. But it does act in a similar way. If you take ISO 100 film and rate it at 1000, it will be noisier than if you shot it at 100. If you rate ISO 1000 film at 100, it will not have the resolution of ISO 100 film. The mistake in most people's thinking is they think shooting the RED ONE at ISO 100 should give the best image. It does not. Following the film example, if you replaced the Mysterium ISO 320-500 sensor with a native ISO 100 sensor, you would get the best image at ISO 100. But that is not what we are doing.

Shoot the RED ONE at ISO 320, 400 or 500 wherever possible. That is what this sensor does best. But also feel comfortable that, if you have a good reason to do so, good images can be obtained from other ISO settings. As long as you know the consequences.

Jim

 

[Patrick Tresch]

I do understand how noise is generated in higher iso ratings like 1000 iso.

But I didn't know about lower rating it (@ 100 iso) could decrease it's resolution (I never tried it). Could you explain what happens when the sensor is tressed @ 100 iso? What are the "physics" behind it?

Thanks.

Patrick

 

[Gunleik Groven]

Agreed, but there is a difference between shooting and posting ISO preferences (for me on pre-build 16...)

I haven't ever gone as low as 100, but for some lowlight and/or tungsten work an iso of 160-250 can give far superiour results (in post) IMHO.

 

[Jannard]

Agreed, but there is a difference between shooting and posting ISO preferences (for me on pre-build 16...)

I haven't ever gone as low as 100, but for some lowlight and/or tungsten work an iso of 160-250 can give far superiour results (in post) IMHO.

This is a really good point. The above info is for Daylight... which is native to our sensor.

We will probably have another set of ISO ranges for tungsten that recommend slightly lower ISOs. We'll do some more tests and post.

Jim

 

[Jannard]

I do understand how noise is generated in higher iso ratings like 1000 iso.

But I didn't know about lower rating it (@ 100 iso) could decrease it's resolution (I never tried it). Could you explain what happens when the sensor is tressed @ 100 iso? What are the "physics" behind it?

Thanks.

Patrick

Did you ever shoot ISO 1000 film and rate it at ISO 100? While this is an exaggerated example, it makes the point.

Jim

 

[Jan Reiff]

thanks for info.
okay, that´s for build 16. right? but now we have build 15 until the great 16 is released.
what´s here the best iso for 15? little more limitation maybe? 320 - 400? is 500 not more a great new fine iso of 16?

 

[Dj Joofa]

Could you explain what happens when the sensor is tressed @ 100 iso? What are the "physics" behind it?

Reducing gain by lowering ISO may reduce the (local) contrast in pixel values. Human eyes contrast sensitivity function is jointly dependent upon frequency and contrast. Lowering the contrast results in decreased human perception of even the same image frequencies. In other words those (higher) frequencies that are (more) visible in higher contrast are now difficult to see and hence a perception of decrease in resolution.

 

[Cüneyt Kaya]

This is a really good point. The above info is for Daylight... which is native to our sensor.

We will probably have another set of ISO ranges for tungsten that recommend slightly lower ISOs. We'll do some more tests and post.

Jim

this tungsten test would be very useful. please do it
a daylight guide
a tungsten guide
a redspace guide

 

[Raul Gonzo]

Thank you for explaining this stuff. I honestly don,t know what I do with out this place 8)

 

[Gunleik Groven]

Reducing gain by lowering ISO may reduce the (local) contrast in pixel values. Human eyes contrast sensitivity function is jointly dependent upon frequency and contrast. Lowering the contrast results in decreased human perception of even the same image frequencies. In other words those (higher) frequencies that are (more) visible in higher contrast are now difficult to see and hence a perception of decrease in resolution.

YUP, but there's a couple of BUT REALLY to add to that soup...

If you lower ISO (in my mentioned scenarios) and rise the Brightness and work with the curves, you can get the same contrast as with higher iso without the noise (more often than not). I even think you shouldn't care too much about WB in the developing stage of such shots, set it at sensor native (be it 5600 or 5000, people seem to disagree), compensate a bit with the color drawers and do the rest of the work in a real CC app.

Because of the high bitrate, there's really enough colorinfo there in each channel for a lot of correction, and it's so much more fun without the noise...

Just my 2c

 

[Brook Willard]

I'm having a difficult time understanding how the dynamic range does not change [considerably] anymore.

Lets say that the camera has 5 stops over and 6 stops under [at the native 320]. If I rate the camera at 640, it should now have 6 stops over and 5 stops under. Conversely, if I rate the camera at 100 ASA, it should now have 3-1/3 stops over and 7-2/3 stops over. By this logic, I don't see how any softening could come about... no resolution hit, just a dynamic range readjustment.

So to get to the same dynamic range at 100 ASA with the loss of some resolution detail makes it sound like some other mojo is happening there. Is it an actual gain-based change to the sensor or is it all somehow in the debayering algorithm?

 

[jbeale]

Could you explain what happens when the sensor is stressed @ 100 iso? What are the "physics" behind it?

This is pure speculation. But something nonlinear has to be happening if anything but brightness changes as ISO is shifted. Noise is nonlinear, and that process is well understood as ISO goes up. If the low-ISO image from the sensor is softer, it sounds like some captured charge on any given pixel is leaking into adjacent photosites. If it happens only in the brightest areas (which makes some intuitive sense) this is called "blooming" but normally CMOS sensors don't do this. Maybe they do if stressed by enough overexposure.

If the softening happens uniformly in bright and dark areas, I'm having a harder time imagining the physical mechanism, unless the low ISO setting changes some hardware settings, but we're told it's still just metadata.

 

[Jannard]

We are not saying the resolution hit is substantial... but it is there due to DSP, mojo, etc. I think it is harder to believe that moving away from rated ISO in either direction would NOT cause some degradation of some sort. Noise is easy to see and understand. No one has really been looking at the other direction. The resolution hit two stops less ISO is not as big as the noise hit two stops over, but we need to tell you what is going on.

Jim

 

[Dj Joofa]

If you lower ISO (in my mentioned scenarios) and rise the Brightness and work with the curves, you can get the same contrast as with higher iso without the noise (more often than not).

Gunleik, if you are going to do multiple transformations (change ISO, change brightness, change curves, etc.) at the same time, then it is very much dependent on the implementation. For e.g., is Red collapsing transformations in such cases or not (people working After Effects know very well the effect of collapsing transformations in spatial scaling; people working with Shake also know this on how to insert nodes without breaking the chain of collapsing transformations).

If you have found a secret sauce of getting good results in doing multiple transformations together, then it may not be taken to mean something inherent in the nature of transformations and it may be how the cumulative effect of those multiple transformations was "baked in" via a particular implementation by a particular manufacturer.

 

[Jan Reiff]

when i started to learn to exposure film, i shot days for days tests, tests, tests. i did my own experiences over the years, partly not with equal results that are standing in books. so i am looking forward to get my delivery to do the same with red. everyone will have his own settings and pros and contras with the "right" iso, but jims infos are always based on a great knowledge and will be my basic to explore the red.

 

[Jannard]

I am just posting our experience testing here at RED. It should be nothing more than informational and a starting place to work from. You get to do anything you want and maybe will tell us something we haven't discovered. But there are a lot of customers that are doing things and don't understand why.

The good news is that Build 16 does everything better.

Jim

 

[Jan Reiff]

it is a very good starting place. this whole forum is unique in support and knowledge. every day so much infos of you and the customers. that makes this camera so valuable.

 

[Dj Joofa]

everyone will have his own settings and pros and contras with the "right" iso, but jims infos are always based on a great knowledge and will be my basic to explore the red.

Red South West, not to hijack this thread, but I think a small comment is in order here. Developing a high-definition camera is not easy. There are a lot of "secret sauce" components here. What I personally find a little unsettling is that Red is rediscovering some problems and workarounds that are known to people in business of making such cameras for sometime. We have seen many of these issues before, and certainly people much before us have seen these issues before. However, the resolution of some of these issues is a closely guarded secret by companies. But some/many of these issues are certainly not new, though it may be new for Red.

What is comforting, though, is the openness and candor of sharing the issues and problems as they are found by Red with public, which is a healthy trend in my opinion, albeit at a great risk, however, it appears it has produced good results and certainly a feeling of goodwill by the online community towards them.

 

[number6]

This is an often confused issue.

The RED Mysterium sensor is rated ISO 320-500.

Shoot the RED ONE at ISO 320, 400 or 500 wherever possible. That is what this sensor does best. But also feel comfortable that, if you have a good reason to do so, good images can be obtained from other ISO settings. As long as you know the consequences.

Jim

I gather that 320-500 is the "sweet spot!" But does it further translate that ISO 400 is the "sweetest" sweet spot? Few if any will likely care to be that precise (certainly not me) but still I would like to know where it is for trivia's sake.

edit: never mind... Jarred just answered my question more or less, in another Build 16 thread.

 

[Bing Bailey]

Joofa also I would add that RED may have solved it in a better way compared to those other companies. given how much better the noise looks even at high ISO ratings