Context: Adam shot some RED footage on Friday 7 March, as part of a camera comparison, and took the raw files home to play with. Art shot a quick test with the same camera six days later. The following email conversation ensued…
ART
…Sorry I couldn’t make it, I launched into testing a RED ONE at 10AM and didn’t come up for air until 2:30PM.
ADAM
Likewise: I had nightmares most of the night about color-balancing stuff with endless cycles between REDCINE and FCP, and woke up groggy to plunge back into it again…
ART
That’ll be me soon. I can’t wait to see what our dueling tests reveal together. I did it very differently from you but I suspect we’ll see similarities.
I’ve already got Graeme Nattress weighing in on my latest article.
ADAM
You should be proud that he’s reading you! That’s instant street cred with the redniks.
ART
Sweet! I’ve been warned not to post anything negative on the RED user forums or risk being flamed, so I’ll post here and see if I can stir up some turbulence.
ADAM
As far as I can tell, that would be the first time ANYTHING was ever deleted from reduser.net! I have a whole rant on the topic, coming soon…
And [referring to Graeme’s comment] if you’re limited already by the blue channel, then prefiltering red and green to pull them down to blue doesn’t lose you anything you don’t have to throw away in post anyway. But you (or he) could make a case that you only prefilter ’em down to the hottest color temperature you normally shoot in (like daylight; but if so, why pick 5000k?) so that you make the best image you can, in the optimum light for the sensor.
But that means you’re ISO 320 outdoors and maybe as low as ISO 80 indoors, assuming you filter (or even not, if you simply reduce ISO rating to keep the same headroom in your red channel!).
Either way, it seems that the Mysterium is really an ISO 80 chip under tungsten, that just happens to have a “negative 80A” filter on it for use in daylight–if you want more headroom than an EX1 gives you.
And you STILL have to fix it in post!
ART
It was pointed out to me that all the processing that cameras normally do is completely missing from the RED–including flare adjustment! Yikes. It would be very easy to end up with one color of black at the beginning of a shot and another color at the end.
ADAM
Yes, that’s RAW recording. You can still have flare comp (not that there’s a button marked as such now, but there could be), but you run it in post, whether as a filter pack inside REDCINE or as a separate process. The whole idea is to grab everything you can and NOT make any irreversible decisions in the camera.
ART
That would be great, but do we know who’s going to be doing all the post processing? Will producers pay for professional coloring or will it all be done by an editor somewhere? Editors typically don’t know how to color correct, although many will say they do.
It occurred to me that this is the only TRUE raw camera on the market. Every other camera records an uncompressed signal that has had some sort of log and knee work done to it.
ADAM
Silicon Image 2K? Dalsa Origin? And even the RED uses REDLOG. Maybe. I think.
ART
It’s just whatever comes off that chip with no processing until it hits REDLog. Currently the “highlight” control in REDCine is supposed to deal with unbalanced clipping of colors, but I can’t get it to work. That would solve some major issues.
Not sure how paying $17.5k for a camera and spending $65k on a post package to fix it works in favor of the independent filmmaker. Let’s hope for an improved version of REDCine in the near future!
ADAM
I’m just now seeing what I can do with rescaling the exposure down in REDCINE to avoid numerical overflow in any channel, desaturating highlights in FCP, pulling the gain back up, then using FCP’s Proc Amp to S-curve the highlights. I recover just about all the recoverable highlight detail and avoid the tonal inversions I saw without rescaling. It looks 100x better than the unprocessed image.
Yes, one would hope that this sort of thing could be wrapped into a future version of REDCINE, but at least it’s doable with affordable desktop tools.
You can’t think of the RED as either a video camera (which outputs nice pix, ready to use) OR a film camera (which outputs nice pix, ready to soup and print). It outputs a digi-neg which needs digi-souping, and part of that soup is adding back in all that fiddly stuff in the toe and shoulder that you get for free with film, and have knees and hypergammas and flare comps for in a “real” video camera.
I’m just not sure how many of the 4,000+ current buyers are aware of this, eh?
ART
What I’m trying to figure out is why it meters at 160 ASA (in 320 ASA mode!) under tungsten light and around 120 ASA under daylight. That makes very little sense to me. But then the red channel is the quietest channel in the camera, and that makes no sense either. It’s just different from other cameras, where green is the quietest channel.
ADAM
Can you explain how you’re getting these figures? I saw 320 pretty much dead on in my tungsten tests… with the same darned camera you used! I measured incident light at my chart at T4 at ISO 320, and at that exposure the chart midgrays are at about 50%.
ART
Hmm. I measured with my trusty Minolta Spotmeter-F.
Here’s what I think happened. In Rec 709, if you look at a waveform monitor and put gray at 50 units, it comes up as ASA 160 or 140. In REDLog it comes up as 320. There’s an extra stop in there, apparently, thanks to the way REDLog stores values. As long as you look at it in REDLog you’ve got more latitude, because what was Zone 4 in Rec 709 is Zone 5 in REDLog.
ADAM
I didn’t measure on a ‘scope during the shoot because we were using the onboard LCD. I just see that the shot I made at T4, processed through REDCINE into ProRES422 HQ, shows up on FCP’s ‘scopes with the midgray right around 48% or 49%. I wonder if there’s a gain or gamma difference in the monitor output? Or is this the result of the whole QuickTime RGB gamma 1.8 vs. YCrCb gamma 2.2 translation?
ART
I think that exposure could be set by meter and checked for clipping with a scope.
ADAM
The camera has that handy histogram, and David Macintosh’s superb false-color exposure meter (one fruit of the AccuScene acquisition). Those seemed to track exposure and clipping pretty well in my tests. And there’s the RGB “traffic light” clipping indicator–one thing RED has really done well.
ART
The blue filter on the camera definitely helps. What it looks like to me is that under tungsten light you get two stops reflected over 18% gray before red clips (one of the curves disappears at +3 stops) but with CTB added to the light the curves on the histogram all line up perfectly and you get three-and-a-half stops before any clipping happens.
ADAM
Makes sense to me.
ART
I’m sure I’m on a cyan list somewhere.
ADAM
An 85A will fix that. Somewhere.
ART
Yeah, it’ll be a 6×6 in my neck the way I’m going. : )
ADAM
Don’t go putting on airs. A 4×5.65 will suffice for the likes of you! ; – )