Welcome to our community

Be a part of something great, join today!

Log in
  • Hey all, just changed over the backend after 15 years I figured time to give it a bit of an update, its probably gonna be a bit weird for most of you and i am sure there is a few bugs to work out but it should kinda work the same as before... hopefully :)

Quick Tungsten Test

Since the camera sensor is closer to daylight balance, any shot made under 3200K is basically orange in the RAW file. So if you want a neutral color, you have to remove the orange / increase the blue, which is where the noise is coming in.

The only question is whether some methods of doing that correction produce less blue noise than others. Just playing around in Photoshop with some 4K TIFF's from the RED camera of noisy footage shot at night under tungsten, I've found that I can avoid some noise simply by desaturating the blue channel and letting the image go a bit brown and monochromatic for a warmish look rather than try and get a perfectly neutral color by pumping up the underexposed blue channel. A more sophisticated color-corrector may have other tricks I suppose, maybe by applying noise reduction or softening to just the blue channel.

The problem is even worse if you want to take a shot made under 3200K and make it blue-ish -- I would tend to avoid that altogether and use blue gels or HMI's and other daylight sources for a blue look rather than try and make 3200K footage look blue. You're already pushing things just to make 3200K footage look white/neutral with a clean blue channel, so it's best to not push your luck even further.
 
jimhare said:
Buy why?
Click to expand...

Yeah, reading this thread makes me think Red Alert could be more effective at turning things cool. It currently seems to simply boost the blue channel and that's pretty ugly. It might be smarter if Red Alert added information from the stronger channels into the blue channel to make a cleaner cool tone. I think that logic is already being used in the DRX function.

The idea that FCP somehow boosts blue in a way that's superior to Red Alert is a little disconcerting.

I think I'd probably try adding a blue/cyan layer in a compositing app if I really wanted to turn a tungsten shot cool.
 
I think the 2k example speaks for itself. Processing your .r3d in Redcine or Red Alert and setting the White Balance to anything under 5000K and you're introducing noise. This noise does not reappear down the line when it's graded. I don't know if this is true with Scratch. If anyone is getting different results, please post examples.
 
Well, it's probably that the simplest correction is to just push the gain in the blue channel, whereas a more sophisticated color-corrector may try a mix of techniques to reduce the orange cast that is less prone to create noise in the blue channel. On the other hand, there is no way around the basic fact that you have an underexposed blue channel in tungsten light and any technique to correct that involves some image compromise especially in terms of noise.

In other words, sure, you can open up the files in a 5000K setting and fix the orange cast differently, but you're still trying to compensate for a weak blue channel no matter what you do, and I don't think all the other post tricks are necessarily going to ensure that the blue channel is noiseless.
 
Gregorios said:
I think the 2k example speaks for itself.
Click to expand...

Well, it basically shows that Red Alert is doing something different than whatever Final Cut is doing (and thanks for posting it because it is enlightening). This is all about the software and not really that much about what the camera actually did. The blue channel is underexposed because tungsten is blue deficient.

Are you simply using FCP's 3 way color corrector to turn the image cool?
 
12820_1221443648.jpg


This spectral power chart shows the relative color content of daylight on the left vs tungsten at the same level. The shift in relative exposure for red vs blue levels is about 4 stops between the two.

If you want to make best use of the overall dynamic range of the sensor, then like film it is best to filter the light for balanced spectral content, either at the light source or at the lens or some combination of the two. With a more balanced spectral content in the raw file, it will be easier to do post corrections without incurring noise problems.
 
David Rasberry said:
12820_1221443648.jpg


This spectral power chart shows the relative color content of daylight on the left vs tungsten at the same level. The shift in relative exposure for red vs blue levels is about 4 stops between the two.

If you want to make best use of the overall dynamic range of the sensor, then like film it is best to filter the light for balanced spectral content, either at the light source or at the lens or some combination of the two. With a more balanced spectral content in the raw file, it will be easier to do post corrections without incurring noise problems.
Click to expand...

You're talking theory (which I don't disagree with), but I'm all about practice. Tungsten lights are cheap and widely available. The post process I've outlined is very easy to implement. I will continue to use Tungsten lights. As always do your own testing and use your own eyes to judge the image. I will try to post some more examples later this week.
 
All I am saying, is if you shoot with tungsten, use a tungsten filter on the lens and take the two stop loss in ISO speed to get better meta data to work with in post. If you are starting out 4 stops under on the blue channel relative to red, it is hard to make it up without incurring some noise.
 
Yes, but that's not very practical if you want to light a night interior using mostly practical tungsten lamps at a low enough level to capture as natural a look as possible. Most dramatic movies shoot at f/2.8 at 400 to 500 ASA on tungsten-balanced film.

Using a full blue correction (80A) filter on a RED camera and turning a 320 ASA camera into a 64 ASA camera and still managing to light an intimate night interior to an f/2.8 and get an unlit look... is very difficult. Most of us would rather live with a little noise at that point rather than get into 1960's-era levels of interior lighting when 35mm movie film was 50 ASA tungsten-balance.

Personally, I think the noise level at 320 ASA in 3200K light on RED using Build 16 isn't bad. The problem is a bit overrated unless you are doing bluescreen work. Of course, I'd welcome even less noise.
 
David Mullen ASC said:
Yes, but that's not very practical if you want to light a night interior using mostly practical tungsten lamps at a low enough level to capture as natural a look as possible. Most dramatic movies shoot at f/2.8 at 400 to 500 ASA on tungsten-balanced film.

Using a full blue correction (80A) filter on a RED camera and turning a 320 ASA camera into a 64 ASA camera and still managing to light an intimate night interior to an f/2.8 and get an unlit look... is very difficult. Most of us would rather live with a little noise at that point rather than get into 1960's-era levels of interior lighting when 35mm movie film was 50 ASA tungsten-balance.

Personally, I think the noise level at 320 ASA in 3200K light on RED using Build 16 isn't bad. The problem is a bit overrated unless you are doing bluescreen work. Of course, I'd welcome even less noise.
Click to expand...

Good point. And it shouldn't be as much of a problem if the natural look desired is on the warm side anyway.
 
I've been testing the Schneider Sapphire Blue 1 and 2. The are roughly equivalent to 80c and 80d filters. They offer a nice compromise, the Sapphire one only loses you 2/3 of a stop and corrects about 1200 degree kelvin. The Sapphire 2 loses you 1.5 stops but corrects about 2000 deg. kelvin. These are estimates based on actual Red footage tests. Remember the Red One is a 5000K native sensor so it's not necessary to correct all the way to daylight.
 
This is a very good thread, one that can help people get better results from their footage.

I agree with David Mullen's post that shooting with daylight instruments or using blue correction filters is going to be problematic in many cases. For those situations, the question becomes what is the best way to convert and grade the footage. I have started to set the color temp in RC/RA at 4000 or so for 3200 footage, letting it be warm, and then taking it the rest of the way in other apps. Definitely less noise this way, which means that RC/RA could be better (but I am glad to have these programs for free anyway). I wonder how well Scratch itself deals with this.

On the other hand, I haven't found 3200 footage converted at 3200 in RC/RA to be a problem noisewise unless I have underexposed by two or more stops and I'm really pushing to get the levels back up. That's when going to 4000 or 4500 in RC/RA has made a dramatic improvement for me. Just my two cents. And yes, I have done shots that I knew were underexposed knowing how far back I could bring them, when it was a case of shooting underexposed or not getting the shot at all.
 
Anymore further developments on this? I'm wondering about interviews and whether I should consider adding an 80A or Saphire Blue to the kit. I will be using a decent amount of tungsten lighting, as I still prefer that to flourescents and LED in a lot of cases. The interviews need to be free of noise.
 
stevesherrick said:
Anymore further developments on this? I'm wondering about interviews and whether I should consider adding an 80A or Saphire Blue to the kit. I will be using a decent amount of tungsten lighting, as I still prefer that to flourescents and LED in a lot of cases. The interviews need to be free of noise.
Click to expand...

Use an 80C, The 80A is way too thick.

I only put blue on the lens if I have to (ie. greenscreen)...
I'd probably shoot interviews with straight tungsten anyways, no filter.
can always gel the lights!

...Anyways
 
stevesherrick said:
Me too, great video!

Anyone have a suggestion on the 3200>5000 conversion?
Click to expand...

I've outlined how I treat Tungsten above. I was hoping to have time last month to compare all the 80 series vs. post correction, but maybe this month instead. I will post when I do. As always, run you own tests.

Glad you enjoyed the excerpt of A Day At The Office. It was a fun shoot with a talented dance troupe called Corpus.
 
You must log in or register to reply here.
Back
Top