ISO rating and Linear Capture

[Chris Nuzzaco]

As some of you here know, I currently shoot with an Andromeda DVX100. In short, I love it! Great quality, and I now have perfected my technique for almost totally noise free captures.

The journey I had to take to get myself to that quality led me down a road with many interesting lessons...

I think I now understand why it was stated earlier on this board that ISO rating for the RED One is somewhat tricky. If you ever have the opportunity to mess with a camera like the Andromeda, which can give you access to the cameras linear response, you'll quickly see what I mean. Linear images just look strange, very high on contrast and usually darker (steep fall off in tonality is another way to put it), but that doesn't mean plenty of stops of latitude are not being captured, you simply need to "remap" the tones, and viola! A more "normal" looking image. What I found was that if you base your ISO rating off of 18% gray, everything is up in the air, as this is one value that can be constantly remapped, so as you can imagine, that makes ISO rating just a tad unruly :tongue:

In the end I decided to base my ISO rating off of the highlight, just the before blow out while capturing linearly. If this piques your interest, check out this article I wrote and posted at the Reel Stream forum. It's based purely upon my own observations and experimentations with the sensors and various look up tables, as well as some prior common knowledge about digital sensors in general. Will the principles I discovered be applicable to the Mysterium sensor? I have no idea! Its a CMOS, DVX Andromeda uses 3 CCDs, and I've never messed with the RED One, but its a thought provoking read none the less.

Let me know what you guys think, it was a *very* enlightening experience!

http://forum.reel-stream.com/viewtopic.php?p=4215#4215

[Kevin Halverson]

Metering for and protecting the high lites would seem to be an idea that is well understood since this type of acquisition system's upper boundary is an a hard point (unless you choose to change the pre quantizer gain in some sort of dynamic way). The bottom end of the system is really about trading off signal to noise. The entire mid range, the area that we are actually interested in, is subject to manipulation and hence the ISO rating is also subject to interpretation.

Your observations of the characteristics of your "Andromidiazied DVX" makes for interesting reading. I am very curious about your pixel shift theory, this would make the evolution from the DVX to the HVX really an incremental development, not a ground up redesign.

Thanks for posting the link and for the obvious effort that went into complying the data.

[Chris Nuzzaco]

No problem! It was quite fun. At first I tried to decrease the contrast by lowering my highlights, but quickly found that I was merely blowing them out and recording zero textural detail at a lower luminance level. Its looks really weird. After messing with that, I realized the sensor needs to approached from a top end down approach. So I started applying LUTs that "opened" my shadows but bagged the approach because I thought it was better to just get the data "as is" and work from there in post, it also gives me more lee way to create better exposures because I can pour more light into the shadows without fear of making the image too flat, which is what happens if you use a LUT that opens the shadows more. You find yourself pulling back on the fill because it looks like "too much", but its really the remapping of the data that causes this.