Skin tones clip early on EX1?

[Paul Frederick]

I've only had my EX1 a few weeks but noticed the skin tones tend to bloom out WAY before anything else reaches 100 IRE Zebras! This is very odd and seems to happen on any of the Gammas. Before anything is clipping (zebras at 100), the skin tones compress and lose detail.

Is this a characteristic of the camera? I read the other post about the whites blooming, but this is different. I'm talking about the mid-skin tone region clipping (especially in sunlight) long before anything white is near 100.

Any solutions? Do I have a setting wrong somewhere?

If I drop the iris to where the skin tones aren't blooming, the IRE is WAY, WAY low for anything white in the frame.

[Paul Frederick]

I've noticed this effect on others work as well. Here it is on Philip Blooms film on the Letus Extreme:

Philip Bloom 35mm adaptor demo: The Letus Extreme

His face is blooming and clipping while the rest looks to be properly exposed. Sorry I don't have anything I shot to show this but this example is exactly what I'm talking about.

[Wacharapong Chiowanich]

Did the blooming out show up on the EX1's LCD screen or an HD monitor connected to the camera? Did you test the recorded raw footage or you only viewed it live on the screen(s)?

This is odds because the blooming should have been zebra ALONG WITH any clippings in the frame.

Wacharapong

[Serena Steuart]

Perhaps "blooming" isn't the best term to use in this case, being more appropriate to optical elements; probably "clipping" is the concern. I've had a look at the Phil Bloom example and in general I'd judge the skin tones to be fine. There are a couple of shots where the skin highlights look close to clipping but I'd have to look at them with a waveform display, and I see no evidence of skin highlights going before other highlights. When pushing close to maximum exposure you can clip blue before R & G and so get pinkish highlights. Perhaps you can post a frame grab with its corresponding illuminance waveform?

[Paul Frederick]

Actually I think the term "Blooming" is more accurate than "clipping". The signal is not clipping per se, it's not reaching 100% IRE but it is losing detail. Thats what I don't get. I'll try and post some frame grabs later today.

In that Philip Bloom example, to me, even on the freeze frame before you hit play, the highlights on the right side of his face are too hot. If you iris down, you will then see more color and detail in there, instead of just a white patch. I bet nothing was zebra striping at 100%. This is the phenomenon I'm talking about.

[Serena Steuart]

Yes. Why do you say that isn't clipping (close to)?

[Paul Frederick]

I say it isn't clipping because the signal is NOT at max white. It IS losing detail though. I guess you're right it would seem to be clipping but to me clipping has/should only happen when the signal is beyond a certain percent (110-120 IRE) depending on the camera.

Skin tones on this camera "clip" much sooner. They seem to be doing it at about 80% IRE. Hence, not really at the clipping stage, but more of a "blooming", though detail is being compressed as is color.

We're just using different terms for the same thing!

[Serena Steuart]

Anything over 109 will be clipped, and blue will reach that limit well before R&G. I'm not trying to refute your observation (although I don't see this in my shooting) and just want to be sure whether you're speaking subjectively or objectively. That is, by the look or with supporting measurement. Not much point in extracting a frame of Phil Bloom because that is well displaced from the camera original.

[Dean Sensui]

I set my zebras at 80 and 100, and try to avoid getting any zebras in the skin tones.

I wonder if one of the channels is clipping before the other two? That might be the reason you don't see it in the zebras but still get clipping.

Try taking a look at the "bad" footage via a waveform in "parade" mode, where the red, green and blue channels are displayed separately. You might see one of them maxing out before the others. If you look at that same signal on a waveform monitor set up in normal mode, that might not be apparent.

[Paul Frederick]

Bingo! That does seem to be what is happening Dean. The red channel is clipping out and compressing before the other channels do. What is causing this? Can it be the Auto Knee? I have 2 stills I've been trying to upload to show what I'm talking about but they just aren't going through. Not sure why? I've tried uploading on my PC, My MAC, nothing is working?

[Dean Sensui]

Maybe the camera is more sensitive to red and is therefore clipping that channel first?

I noticed that even with the zebras set at 80 there seemed to be a risk of clipping skin highlights.

However, I seem to also notice that when the histogram is centered in the box there is a lot less chance of clipping highlights or crushing shadows. The image tends to look a little dull, but I color correct and grade everything in post anyway.

[Leonard Levy]

Paul,
It's impossible top comment intelligently on what you're seeing unless you tell us what gamma settings you're using and what IRE you are seeing the so-called clipping at.

Re: skin zebra levels - 80 IRE is way too high to protect your skin tones. Traditionally in SD video we set them at 70% but for HD most people I know go lower. Mine are set at 65% and some people use 60.

Paul if you are using the cine gammas, they tend to compress highlights pretty radically and start that compression pretty low so traditional video exposure rules need to be changed a bit. This is actually why I avoid them, though maybe if I got used to them I would like them more.

With cine gammas my experience was that to get my whites to 100 I found myself overexposing the rest of the image because the compression was so radical it was just hard to get to 100. What you are calling "clipping" is more likely compression in the brighter parts of the skin tones which looks pretty lousy. I found the cine gammas tended to do this unless I underexposed flesh tones and really the image as a whole a bit. The idea is of course that you can bring it up in post if necessary and that your saving a wider swath of information.

Interestingly these Sony cine gammas look very different on a waveform than the Panasonic ones do at least the Cine D and the original cine gamma on the HVX and DVX. Those tend to be much flatter and even throughout the scale and frankly I liked them more because they didn't have so much compression.

I don't want to bad mouth the Sony cine gammas though - I just haven't spent the time to figure out how to use them, especially when using my normal exposure method of bringing my whites to 100. I will say that virtually all the experienced video engineers I work with avoid them on the professional 2/3" Panasonic and Sony HD cameras. But they are painting each picture.

- Lenny Levy

[Serena Steuart]

Originally Posted by Paul Frederick

Bingo! That does seem to be what is happening Dean. The red channel is clipping out and compressing before the other channels do. What is causing this?

I did say this a couple of times but hadn't appreciated that you thought clipping would be indicated by white. If you run the SAW test you'll see that RGB gain is equal on all channels at 3200K, but at 5600K the gain is less on G is less on B. As others have said, trying to put skin at 80 is inviting trouble.

[Greg Boston]

FWIW, I've had the same issue with the F350. I believe Nate Weaver also ran into some issues. The default value of 70 for zebra1 on skin tones will give them a washed out look with no detail.. almost like fake or plastic skin. I keep mine at 60 to 65 percent and I tend to use Cinegamma 4 a lot. It behaves more like a standard gamma except it rolls the highlights a bit more than standard gamma. However, I see the same problem shooting in DVCAM mode with the same gamma curve.

-gb-

[Greg Boston]

Originally Posted by Dean Sensui

Maybe the camera is more sensitive to red and is therefore clipping that channel first?

Dean you were spot on with that statement. I have some DVCAM video shot last week where the skin looks washed out under the stage light even though I was holding the skin tones at or below 70%. Sure enough, loaded the clip into FCP and turned on the scopes and the RGB parade has the Red channel clipping badly.

Good call there, Dean! This is one of the problems with all electronic white balance.

-gb-