Has anybody noticed the EX1 produces much cleaner images under blue light?
 

It seems the camera produces much cleaner images under blue light. I have been shooting with the EX1 almost exclusively under tungsten light. But I had to gel all my tungsten light with CTB for a shot recently and I was like, whoa!... when I looked at the monitor. It looked very clean, way cleaner than it has ever looked. Has anybody noticed that?

This is a very well known fact for nearly all video cameras. Panavision did a nice education series some time ago, maybe a year or more, on why this happens and I posted it here. Others have posted it at other forums. If you're interested in digging into the why, I'll post the link to the videos here so you can see them.

Hi Perrone. It would be certainly interesting to see the videos. I don't think I have seen them yet.
I was aware about the blue channel problem in a similar way that the RED ONE has a problem. But I was under the impression it worked differently. In the shot I mentioned the light was blue. We had tungsten with full CTB, but the intention was not to match daylight. So we didn't have the white balance set to daylight. We kept the white balance set for tungsten (3200K for the shot) so the lights came off really blue. But it looked the cleanest I ever seen from the EX1. I was under the impression that the advice was to shoot with HMIs or use full CTB for tungsten but the WB should also be daylight.
Unless the equation goes like this:

*Tungsten+CTO at 5600k = The worst ( I have used this for an extreme Mars red effect)
*Tungsten+CTO at 3200K= still noisy but better than the above.( I have used this for sunset effects)
*Tungsten at 3200k= acceptable (many people will say it looks as good as any other)
*Tungsten +CTB or HMI lights at 5600k = Clean
*Tungsten + CTB or HMI lights at 3200k= The cleanest.

I guess I just misunderstood how it works. I thought when using daylight sources you should white balance for day light if you want the cleanest image.

Hmmmm...this is VERY interesting. I've been wanting to ditch our tungsten hot lights for a while now for some Kino's. Do Kino's qualify as "blue lights"? I assume any daylight balanced light source would qualify.

Blue, in our visible light, is poorly represented. Our cameras are STARVED for blue. The more blue light you can feed the cameras, regardless of white balance, the cleaner they will be. The white balance merely increases gain in the blue circuit (hence the noise) so that it roughly equals out to the red and green channels.

You can see this effect easily if you have a program that can show you each of the RGB channels as grayscales. You'll note that green is lovely, red is decent, and blue is awful. Especially under tungsten lighting.

I'd shoot with 6500k lights if I could get them.

Have a gander at this...

Demystifying Digital Camera Specifications Part 7: Single Sensor Cameras Continued

Regardless of white balance? How can that be?

The amount of blue light hitting the sensors does not change whatsoever with a white balance. That happens AFTER the sensors convert the light to digital data. Which is why on cameras like the RED and Viper, that give you the RAW data, you can change white balance and ISO after the fact.

Yes, but if the white balance increases gain in the blue circuit to match the red and green channels, the white balance should make a difference. What am I missing?

Downloading the video now. Thanks!

The white balance is trying to make up for the fact that there isn't enough blue light. So it's faking it. When there is ample, or nearly ample light, the white balance circuit adds FAR less gain, making the image appear cleaner. So yes, the white balance matters, but not nearly as much as actually having adequate blue light.

This is why white balancing with filters is superior - the filters provide the sensor with an image which is colour balanced with the same gain on all three channels. The only time electronic white balance is better is if the light levels are so low that the gain has to be increased to compensate for the light lost in the filters.

N

The filters merely reduce the amount of red and green reaching the sensors. They do nothing to help increase the blue. This is akin to asserting that if you get a flat tire on the road, you should flatten the other three so the car will ride level.

The answer is to give the camera what it wants, which is MORE blue light. And no filter is going to do that. Going to HMI lighting will do that. Adding CTB to existing lights will do that.

Just finished the video. Very informative. It cracked me up when he said the JVC 4k camera prototype was a 12K camera using the "new math". What a dig at RED.
About the blue channel, it's a bit more clear now. But so if you have daylight, and you white balance for daylight or use a 5600k preset you add gain to the red channel, but it is a very little amount and way less than would be needed for the blue channel under tungsten. But if you have daylight and use a 3200k WB preset, wouldn't the WB be boosting the blue channel still? Wouldn't that still add noise? If not, because we would already have enough blue to begin with and the red channel is what need to be boosted but it won't be boosted, it means not extra red noise would be in the image. So although the image would look blue it would also technically have less noise than if you had white balanced to daylight and the red channel had to be boosted, even if just a little, causing noise in the red channel. If this is correct it could explain why I noticed the image was extra clean in my shot where I used blue light and a 3200k WB.
But starting from this same theory, if you use tungsten light but use a preset of 5600k, although your image will be orange, it should be cleaner than if you WB it to tungsten because at 5600k the WB won't be boosting the blue channel. So in both cases it would be interesting if shooting that way, with a blue or orange image and then correcting the color shift in post would be any better than white balancing on the set in order to avoid noise.
Also would mean that blue moon light scenes and sunset orange scenes would always look clean.
This would also show why one should NEVER use tungsten lights and white balance to them with the EX1.
Or my whole theory is wrong?

You're theory isn't necessarily wrong, but you are going to have to add that gain SOMEWHERE if you are going to get a usable picture. Whether it happens in the camera, or in post, you're going to have to boost the blues or the reds. And when you do that, you are GOING to get noise. The only way around the problem is to ensure that the camera is getting enough light in each channel. Hence white balance to tungsten if you are using tungsten, and then if you are willing to live with recording a "blue" image, gel the lights. I generally don't bother with this.

When I shoot at -3db on the EX1, and ensure I have enough light to record at F4, the blue channel is clean enough for nearly any purpose I can imagine. Note that this is not unique to digital filmmaking. Celluloid based film has exactly the same issues because they have to deal with the same light spectrum. HID or other light with a strong blue component improves matters for them too.

What is important is that the sensor gets sufficient blue (and green and red, for that matter) and that the amounts are in balance so it doesn't have to compensate by increasing the blue channel gain (and noise).

The answer is to give the camera more blue relative to red and green. HMI is close to daylight in colour temperature (ie, mix of the primaries) but the intensity is usually far lower. CTB adds nothing; it just reduces red and green.

N

Lets see if I can state this another way.

Let's assume that I want to record with my EX1. And for the settings I have in the camera, I need 100 lumens of light. If I provide 100 lumens with nothing but Tungsten, the blue channel is going to be starved, and the video will be more noisy than ideal because of the gain applied to blue IF I white balance and add the gain to blue.

If I could reduce amount of red and green light, and boost the blue so that the light going to the sensors was more proportional (like using HMI) then I would still have the 100 lumens I wanted, and I would have to add less gain to the blue channel.

If I had tungsten lights that could give me 200 lumens, and I used CTB on them and reduced their output to 100 lumens with a strong blue component, I have the light level I want, and the light color I want. But I have to use more powerful fixtures to get there because I am knocking down so much red and green.

There are a number of ways to skin this cat. But they all come down to the same thing. Reducing the gain that needs to be added to the blue channel, whether that be in post or in the camera.

I use Gyoury 56K florescent lights almost exclusively these days except when I need hard light then I use HMI's.

I find that the EX3 has a really good responce to the Gyoury's.

Lighting for film, video, photography by Gyoury Evolved

I have also noticed that it really helps on the IR problem people are talking about with the EX cams.

I still have a bunch of tungsten's but they stay in the cases most of the time. I do use a couple Arri 650 fresnels at times with dichroic filters and I carry blue bulbs for light fixtures on location, when I include table lamps or whatever in the shot.

It is just so much easier to have everything be daylight, no blue spill from windows.

The only exception is sunset shots with fill light, then I use my Arri 650's bare bulb or even with slight orange gels.

So my answer is yes the EX's and most other cams love lots of blue.

Can you explain how shooting tungstun at -3DB also solves this problem. Is that because the increased amount of light neccessary for -3DB insures an adequate level of blue light without needing to increase gain on the blue channel? If so that would imply that the -3DB may only actually be applied to the red and green channels. That's an interesting idea if I understood you correctly.

I wonder how this differs from the Red where its recommended to shoot with a blue filter in order to drop the red and green channels. In the Red of course you don't really white balance so the issue of overexposing red channels is critical when shooting under tungstun.

In the EX or any other "normal camera" when white balancing (esp under -3DB) might you really be reducing the amount of exposure on the Red and Green channels before it gets to later processing?

No, that's not where I was going.

My point was that -3db reduces the noise so much in ALL channels, that it is just less of a problem to deal with. The proportions of light off the sensor are still the same. Or at least I believe them to be.

Remember that the RED only has one sensor. It is not prism split like in most video cameras. Different rules, and the Panavision video speaks to that.

Well doesn't it stand to reason that if under tungstun, at 0DB it needs to add gain to the blue channel, then at -3DB it is probably still pushing the blue channel perhaps up to 0DB while pulling back only on the green and red channels?

Yeah, you're right. I'm glad the shot I mentioned in my original post was to be kept blue.

If you don't bother with this, then how do you do? Sorry, you lost me there in the end.

I generally don't bother because I am normally shooting in an environment where the blue channel is "clean enough" for me at -3db. When I do my color corrections in post, it's just a minor correction here or there, and then I am delivering web based video, or DVDs. I simply don't NEED the extra clean video the additional blue would give me.

Now, I did a shoot a couple weeks ago. Outdoors, existing light. I shot that very much like I'd shoot a movie. Light meter on the talent, working out the blocking, etc. And even with a 5600K pre-set, the image was VERY blue/cyan. I let it go because I knew it would help me later. The people with me were VERY worried about the colors, but I assured them everything would be just fine. :)

Very clean images. I've attached one for you to look at.

Thanks. Nice images.

But it's indoors that things get ugly. :)

What I didn't understand was this:
You white balance to tungsten and the put the CTB on? Shouldn't you be doing the WB after you put the CTB to avoid the noise?

This is why I need to stop thinking about RAW. Yes, you're right. In a RAW workflow, my comments are right. but in the EX1, it would raise the gain (and noise) even though we are adding more blue light.

I have a corporate shoot tomorrow, and we did a walk-through today. Mostly a podium delivery, live audience. I tested tungsten pre-set, with mostly 4500k fluo top-light and I added a Lowel DP light with diffusion. Didn't bother with CTB. I'll post an uncorrected still tomorrow night.

Perrone,

Was the corrected shot the final version you were going for? I still see a lot of blue. My concern is when allowing that much blue during the shoot whether you can correct to a balance that would be the same as if WB in camera.

So do you agree with Nick's last post on filter use?

I'm adding this to a list of tests to perform when I have some time.

The corrected shot was "corrected", not color graded. It is an accurate depiction of the scene as seen with our eyes. I typically work that way. I will do a color correction if necessary, and then let the director make choices on how to grade the material for effect.

As for Nick's post, technically, he is correct. No filter can ADD anything. And that is not their job. Their job is to remove things. And CTB's job is primarily to remove red light. That said, increasing overall light, while shifting color temps using filters is helpful.

But to be sure, when the scene is examined with the naked eye, it is going to look VERY blue. Hence tho pictures I posted. One that shows accurately how blue the scene was, and one that shows how much blue I let the camera record. They are not the same.

Your images look very clean and reflect the care you put into your work. What you're describing is Uniform White Balance in which all three components are equal, although the 5600K preset is probably very close. While this helps with metering, I don't see how this helps as much with noise in correct exposures. It can help with clipping since that information can be warped down before clipping, but the cost of protecting the highlights is loss of protection of the shadows.

An inaccurate white balance means that needed information is attenuated during video recording and amplified during color balancing in post processing. Such an exposure method makes more sense with equipment that records lossless information with more bits than the final output, such as Nikon D3/D700 DSLR's with their RAW format and nearly 12 bits/color channel of information that is usually distributed in a JPG file with 8 bits/color channel.
Does this method really work for you, or is it slightly better in situations because you typically color balance anyway?

Perrone,

I get what you are saying about putting the blue filters on the lights, but how are you handling things like stage lighting where you have no control over what they dish out?

How are you compensating for it? The EX1 seems much worse at handling that type of lighting than my Sony VX2000. Or don't you typically shoot stage work?

John

I don't often shoot "stage" work, but I do shoot conferences and other things where I hae little to no control over light color.

In my shoot last week, I had mostly fluo toplight which was at 4500k approximately. I added some tungsten with full CTB to try to get more blue light on the subject. I would have put a LOT more blue light on the subject because the shot needed it, but he was uncomfortable with the lighting. Consequently, the footage is very noisy. Probably as noisy as anything I've shot recently. I was on the -3db preset and I was at F4 - F3.4 so I wasn't starving the camera for light overall.

Sometimes there's just not a lot you can do.

I think you are exactly right. However, since I've been playing around with this, I've found that I tned to get some really clean images. When I export snapshots of my video as DPX files and pull them into my DPX viewer, I look at color separations and though the blue channel is more noisy than red, it's not the wide chasm it has been in the past.

Maybe it's just my mind, and perhaps the next time I take the camera out for playing around, I'll test a straight white balance against letting the image go more blue. Obviously this would only work in a scenario where I already had a surplus of blue light.

Hi Glints,

Can you explain beyond the shorthand references on that?

Was trying to follow you what you mean on the relation between those several areas. How more blue light helps metering, avoiding clipping of what information, and how it affects highlights and shadows?

Not so much more blue light, but all color channels balanced. Who knows how the luminance histogram is actually calculated given the shortcuts required in real-time processing, but usually, it is after the white balance operation.

The Uniform-Balance will allow you to set the exposure with greater respect for highlight or shadow clipping. Though, this is usually at the expense of about 1/3 stop of dynamic range, depending on the accuracy of your exposure parameters. This mild underexposure protects highlights at the expense of losing shadow information. Uniform white balance requires a white balance preset to be created and loaded. I've never done this on a video camera, but I've tried this on my Nikon DSLR.

Traditionally the different display modes of scopes have been used instead of histograms I believe. Not to say that the histogram isn't useful.

To create this uniform preset, called "scene file or camera setup file" in video cameras, you'd need to monitor the RGB channels live while attenuating each channel, yes? You should be able to do this with a scope although practically its not done since the luxury of modifying lighting to create RGB balance with a color cast instead of visually neutral white balance is forbidden in many client driven situations and of course when live.

I'm not clear on how uniform balance will affect dynamic range... highlight or shadow clipping. That's adjusted with knee, toe, black level, black gamma and with some cameras discrete knee controls for each color of the 3 channels. Agree?

so whats the ex1's native white balance, where it's not gaining or attenuating any of the channels? I would think it would be 5600k, but I've shot 5600k with -3db gain and still had very noticeable noise in the midtones

anyone know?

I have always thought it was 5600k.

I'm guessing closer to 6200-6500 actually.

Using a blue filter
 

Has anyone experimented with using a blue filter on the EX1 to correct for tungsten instead of doing it electronically?