Exposure using VIPER filmstream (RAW)

[illia]

Hi, I'm working as a HD Technitian in a HD feature film in wich we're using a couple of viper filmstream cameras. As they feed a sort of Log curve the image is flat and greenish. We've been correcting the monitoring image by applying LUTs and would like to hear some opinions about exposure. Eventhough we know where our grey, black and white points are supposed to be (around 40, 10 and 70 IRE) the DP is using almost always the whole range the camera can take, which in some cases, reach the 109 IRE for specular highlights or 100 for bright whites (white backlights or white clothes). I'm used to check the raw footage by grabbing some frames using a cinetal monitor or checking out the footage, just to see if we've clipped any light. By adjusting the frames with photoshop I can see roughtly how the colour correction will apply to the image and because of that I'm confident about using the whole sensor range because it's easy to control highlights and so.
I was trained to try to expose images technically, using the waveform as a basic tool, but shot my last jobs with some DPs that like to expose the way they want the image to be at final point, rejecting the safety room I would recomend as a technitian)
My concern is about the final broadcast image: I know that the DP us exposing like we would do with a negative stock, but I'm not quite sure about how this increased latitude will output in TV, basically I'm concerned about the possibility of loosing too much detail in highlights (we're exposing until 109 IRE)
Would like to know opinions about if it would be better to expose using the Input-output levels recommendations (setting the grey point at 40IRE and the skin tones at 50-60IRE, which would feed an underexposed looking image) or if to expose by eye and then trusting colour correction...
I know this post can seem a little basic but I've heard so many opinions about that that I'm not sure about the best reasons to use one or another method.

 

[David Mullen ASC]

One of the main reasons for shooting in Log is that if you look at "white" on an 11-step DSC chip chart (Zone 10 generally, 5-stops over middle grey I believe) and the Log format makes that white patch hit 70% like it does with PanaLog, for example, then you have a couple of stops over 70% that you can still retain information in.

So just because you are shooting Log doesn't mean that you shouldn't have any bright information hitting 100% or a bit beyond, that's the whole point, that you can record information in that zone. Of course, whatever gets clipped beyond that is clipped period, but when making the Rec.709 version for broadcast later, you have that overexposure information to fit within the Rec.709 gamma window. Sometimes you may just lift the 70% white point to 90% or so and let some stuff beyond that get clipped in the final color-correction, other times you may apply some knee compression or other tricks to bring that super-white information within Rec.709.

So there is nothing wrong with your DP using the full range of Log to record information, the main problem with shooting in Log without a LUT to show you the more limited Rec.709 range for monitors is that one tends to expose so that white is at 100% instead of 70% (and it's even lower for Log-C in the Alexa, more like 60%), and once you do that and treat the top of the scale as white, you are losing your super-whites and placing your midpoint higher, your overexposure information that Log is supposed to be protecting. Of course, noise will be lower as a result but there will be more clippiness in the final image.

 

[illia]

Thankyou Dave,
What about REC709 conversion, when you have to lift the white levels to their proper point, isn't it like pushing the gamma curve? What would be the difference anyways? Is the bigger bit depth the key point in that conversion?
And what about the skin tones, actually, the dp is using the hole range of the ccds but that doesn't mean that he's overexposing the image, in fact, the image looks pretty edgy exposure wise... he like the image that's been recorded to look just like he would want it in post, so he underexposes faces (sometimes at 35 or 40%) because he wants that tones to be exposed this way according to the lighting or the mood he wants the footage to convey.
As we're monitoring the video stream with a cinetal display and a lut applied, and some serious graininess is appearing mainly in the low part of the curve (very noticeable at transitions between dark areas), I'm not sure if this noise is due to the lut (sometimes quite aggressive and contrasty), due to the cinetal, or if I have to expect that noise to be part of the final output...