A problem with lighting

[Uli Plank]

We observed something quite strange on a recent shot, where lighting was done with HMI, so all cameras were set to daylight:

While the Red showed a very greenish tint from one of the lights, a Sony EX-1 present at the same set showed natural colors. To the naked eye the light was identical from all sources.

Anybody seen this before? Could it be a worn source or the infamous IR-problem? We didn't use a hot-mirror, but OTOH the light levels were low and no ND was needed.

Any ideas?

TIA

 

[Matthew Bennett]

I think david mullen also noticed something like this on his "Manure" shoot. Check that thread..

 

[Adrian Correia]

yes, David mentioned Red being particularly sensitive to green spikes in kino flos....

 

[David Mullen ASC]

It's sensitive to green spikes in all flos and HMI's. My gaffer measured one Kino and said it was less than a 1/8 Minus-Green correction off, but it was still noticeably green on the monitor. And when I walked into Bullocks Wilshire to shoot a scene, where I was not allowed to swap out the tubes in some display cases, they were SO green I had to shut them off -- I mean, green as in fluorescent candy lime green.

It reminded me of something Dave Stump said about the Viper -- it doesn't have a green "problem", it accurately shows you just how much green there is in nature, just that most cameras and eyeballs/brains correct the excess out to look neutral.

So the RED is accurately seeing what's there, a green spike. The question is whether it is simple to correct in post by reducing some green saturation, or whether eventually the camera needs to be adjusted or redesigned to be less sensitive to variations in green. In the meanwhile, my suggestion is to work harder at matching your light sources with gel, etc.

I doubt it's an IR problem.

 

[Jan Reiff]

hi david
what gels you are thinking of to reduce this?

 

[David Mullen ASC]

You'd want to match your lights, so it just depends on whether it is easier to add Plus Green to the lights & windows, or Minus Green to the greener lights.

Since this is digital, you can do this by monitor more or less, add gel until the light seems color-corrected.

But what I mainly do is avoid mixing types of light for the key -- i.e. it's OK to have Kinos for fill and HMI's for key, or Kinos for key, and tungsten or HMI for fill, whatever, but it's harder to match things when one person is in a Kino key and the other person is in a tungsten or HMI key, or the same person walks from one type of light as a key to another type of light. Same goes for backlights or edge lights, matching is less critical than when you are talking about a key light.

Also, I found that the problem is worse when you are working in 3200-3800K range with tungstens, which have no green spike in them, and add a Kino or HMI... as opposed to mixing daylight Kinos and HMI's in the 5600K range, where everything has a mix of green and blue in it to some extent, so you don't notice the green as much.

I didn't really have much of a problem with the daylight Kino 55 tubes in 5600K balance mixed with HMI's as I did with the Kino 32 tubes in 3200K balance mixed with tungsten.

I suppose there is a small chance that the way the RED corrects/processes the signal for lower color temp settings for its monitor output / metadata is somehow making the green stand out more on the monitors.

 

[Alexis Vanier]

This is actually both a result of a more accurate sensor and the ability to work in a larger colour space.

Also, since the EX-1 shoots baked video, it has to go through a matrix which might automatically eat the green spike out, either because the DSP considers the colour out of gamut and corrects it or the matrix is simply designed that way.

 

[Jan Reiff]

thanks david.

the place to be to learn.

 

[Brook Willard]

One thing to note is that the monitoring output is particularly sensitive to color spikes that are not fully present in the RAW file. I've had the monitor show some scary green and magenta before, only to find that it was acceptable/what I'd originally expected in the R3D.

 

[Uli Plank]

Thank you very much, everybody.
After further tests I'm sure you nailed it down, it's a question of the higher precision of the Red plus it's monitoring.
While we are at it: What do you guys (and gals) use for measuring color? Good old Minolta or the new meter from Sekonic that claims to be specifically corrected for digital?