Tomorrow I finish a project that has shot for three days already on the RED, build 16. Here’s what I’ve learned and discovered:
I’m viewing the image in REDSpace, with a button on the viewfinder set to toggle the image in and out of raw space, which (theoretically) shows what the sensor is really doing. REDSpace is basically “what you see is what you get” mode, although what I see on a calibrated computer monitor always looks better than it does on set using the on-camera LCD or an external 17″ Panasonic LCD monitor. Which is what you’d expect, as we’re not using a color accurate monitor to see what we’re doing, but as the RED’s output isn’t perfect that’s not a big deal. The important thing is to remember is that it always looks better than what you can see on set. Always.
I’ve set my zebras at 103-109 for best results. (Zebras require you to set a range, instead of just triggering above a certain level and staying on. That effect can be duplicated by setting a range that tops out at the maximum setting of 109.) At 103 I’ve still got .4 stops of overhead before hard clipping, in most situations. Saturated colors are a different matter.
I’m using the “traffic lights” a lot. I’m not sure why they trigger, or how (I seem to remember Adam Wilt saying they kick on when you’ve overdriven 2% or 4% of the photo sites on the sensor, but I could have been hallucinating), but it is handy to see how many channels I’m clipping at once. If I’m clipping one channel I don’t worry about it much. If I clip two or more channels I drop into raw mode and check again–and apply the same logic, as raw mode theoretically shows you exactly what’s going on. I’ve become comfortable with clipping in REDSpace mode as long as I’m not clipping more than one channel in raw mode; I’m not sure why, it just seems to be working. One clipped highlight generally has a fair bit of detail thanks to highlight recovery. Clipping more than one channel leads to white death.
When I set the stop using zebras, and then calculate the effective exposure index using my incident meter, I consistently come up with an EI of 160 with a couple of stops of overexposure latitude. That’s just the way I use it–I expose to the right and try to get the signal as far above the noise floor as possible. I’ve only had one shot with noisy shadows so far, and as it was intended to be a dark scene the noise was no big deal.
Most cameras seem to want exposure to be set properly within 1/4 of a stop; RED allows for 1/2 stop leeway, or such has been my experience.
We shot some green screen tests today and shot under both straight tungsten light and using an 80D blue filter (2/3’s stop loss, half correction from tungsten to daylight). We did some test composites on set by reducing the 4k images to 1920×1080 via REDRushes and comping the clips together via Keylight in After Effects and found no effective differences between the two keys. Zooming into the green screen and looking only at the blue channel in REDCine showed that under tungsten light the blue channel is a bit starved and shows noticeably more noise. As a result I will definitely be shooting our green screen work tomorrow using the 80D blue filter. BUT–our editor says it would work fine either way. I’m a perfectionist so I’m going to give him the least noisy footage possible. (I hope to have screen grabs to post later next week.)
Warm white flourescents go crazy green. See my previous article on RED’s green/magenta sensitivity.
There needs to be an indicator that shows what color space you are viewing at any given time. I set exposure for a shot in raw mode when I thought I was in REDSpace, and in REDSpace it looked overexposed, because REDSpace applies a gamma curve that boosts the midtones and highlights. I effectively overexposed the image according to REDSpace’s gamma curve, but since I didn’t overexpose it in raw mode it’s all fine. (Yes, I checked in REDCine.) This just means that that single shot will have to be graded in REDCine or REDAlert and not simply run through REDRushes using the Rec 709 preset for “one light” dailies because it will look overexposed with Rec 709’s gamma applied, even though all the raw highlight data is safe.
It’s too easy to format drives. There should be an extra step in the interface to prevent that from happening. Be careful.
It would be nice to be able to set reel numbers, and then make sure they stick. Apparently you have to set the reel numbers (A001, A002, etc.) before formatting the drive. After formatting the drive we’ve found that about 30% of the time we have a reel number we didn’t expect, for example jumping from A006 to A021 without obvious reason.
I make sure that we go to 1/40th shutter (216 degrees) if there’s any light in the shot, other than a Kino Flo, that flickers. That puts my shutter right in the middle of the HMI flicker-free window. (I’ve been asked why I don’t simply go to 1/60th of a second shutter (144 degrees) and the answer is that I like to have more light to work with than less, and the additional motion blur is minimal.) I’ve seen normal household flourescents interact with RED’s rolling shutter and cause roll bars at 1/48th (180 degree) shutter. They’re very subtle but if they happen they are most visible against evenly lit surfaces.
I use the IR filter, without fail, any time we have more than three stops of ND in front of the lens, to cut down far-red contamination in the blacks.
I always make sure the slate is in the shot before I roll so the thumbnail in Final Cut Pro shows the slate. We’re just writing the clip number on the slate, so A001-C041 ends up being roll A1 | C041 with no take number. That seems to be a reasonable take on the commercial style of numbering shots (101, 102, 103, etc.).
The image always looks better in REDCine than we thought it would when we shot it. ALWAYS. I’ve found that I can set fill levels by eye reliably. I also get more subtle light interactions with the RED than I do with other cameras: light reflecting off nearby objects, like tables, reads very nicely on the RED, much the way I’d expect from well-shot color negative.
I know that many of my cohorts in LA have toys that they consider clearly superior to the RED (Panavision Genesis, Dalsa Origin, etc.). I don’t have any of those up here in the San Francisco Bay Area, and we only have one Sony F23 that I haven’t worked with yet. I’m overjoyed to be working with the RED. It’s a funky camera with some oddities and interesting workarounds, but it sure does make pretty pictures.
More when I get my hands on the footage, probably next week.