abrupt highlights clipping

[Bob Grant]

Just opened one of those images in PS.

Question: What software did you process this in and what mode?

Just took a quick look with PS which shows what I'd have expected to happen in this situation. You have full digital clipping. The white areas of the sky read 255,255,255. The blue areas closest to the white areas read around 243,255,255. Surely you can see what is happening. All channels are clipped, when there's the slightest falloff in light one goes slightly below clipping and you see blue.

It looks odd because normally the first thing to clip are specular highlights, that they're clipped to white looks natural. In this case only certain parts of the sky are clipping and you're getting a normal color shift between clipping and just a tad less light. I'm pretty certain I've seen this happen with other cameras and I've certainly seen it with my EX1, I just went Oops, don't do that. It's perhaps more likely to happen in higher end camera with more dynamic range or different gamma curves where the onset of clipping is more gradual.

The problem is very easily fixed, 1/4 stop more or less exposure and you have either all blue or all white sky.

[Piotr Wozniacki]

Thanks Bob for your time and effort.

Yes you're right it's easy fixable, also in post - I'm not expert on this, but in Vegas it's enough to pull Highs towards 300deg a bit in the CC FX, and the blue sky is retrieved. However, after this experience, I think I'll also adopt the "Oops, don't do that" attitude while shooting! Even though it's not that easy - I didn't notice this happen on the LCD; but again I wasn't looking for it, either. Now that I know it might happen, I'll take additional care.

And yes I did exaggerate it by taking the snapshots, posted in this thread, from Vegas in 32bit mode - in 8bit it's not that much pronounced.

Good to know it's sort of normal with higher end cameras - I have just made a considerable step up my learning curve :)

[Bob Grant]

Originally Posted by Piotr Wozniacki

And yes I did exaggerate it by taking snapshot from Vegas in 32bit - in 8bit it's not that much pronounced.

In 32bit Vegas does some 'interesting' things with certain footage from the EX1. Nothing actually gets broken but as you've seen it can make a problem worse unless you process the output correctly. That most LCDs tend to clip highlights even more doesn't help either.

[Piotr Wozniacki]

OK so I recreated tests with STD3 and Hisat matrix, and the bottom line is that - with standard KNEE settings - it's no good for even slightly backlit situations. To avoid the higlights clipping unstability (i.e. staying saturated behind the "shadow" of e.g. trees, while blowing-away to pure white elsewhere), one must stay way into the left exposure side, which tends to badly oversaturate even slightly underexposed areas.

The PP in question might be of some use, but requires KNEE adjustment. Will be playing with that and report.

Thanks everyone for their response, and sorry if my post has been a little bit too much alarming.

[Sean Donnelly]

This will happen with an F23 or a Genesis if the exposure isn't correct (film too). Watch the Zebras and the histogram while you're shooting. Looks like later in the day to me, without much direct sun hitting the ground. The range of the scene is beyond that of the camera, so that's why your fourth frame where the sky is holding the foreground is underexposed. Watch the histogram and the zebras while you're shooting, and make a creative decision about which one to expose for. If you want to try to hold detail in a scene like this, try the knee function in the PP (manual, not auto). If you go too far you'll get some odd posterization, but it really does help most of the time. The other solution for a shot like this is a polarizer...

-Sean

[Mark David Williams]

Piotr. I think Skies are always something to be mindful of.

[Piotr Wozniacki]

Originally Posted by Mark David Williams

Piotr. I think Skies are always something to be mindful of.

Ditto, and memento ;)

[Bill Ravens]

There's a reason why manufacturers, like Sony, provide factory settings in their products that are extremely conservative. It keeps amateurs from getting themselves in too much trouble and blaming the equipment for malfunctioning. Problem is, the stock settings don't squeeze all the performance out of their equipment that they're capable of.

So, if you take settings designed to squeeze maximum latitude out of the EX1, you CAN expect the following to happen:

1-exposure has no headroom for error, sloppy camera work will show up without much effort.
2-mild exposure and gamma curve adjustment in post may be required, but, the rewards are more detail in the shadows and the highlights.
3-using fixed black and black gamma settings in conditions of widely varying exposure without adjusting the curve is doomed to failure. Sony put various presets in the EX1 that should be used with some knowledge of why they are used
4-use of the in-camera lightmeter (aka auto exposure)is doomed to failure. use of the in-camera zebra is a step in the right direction
5-extreme settings designed for controlled lighting situations should not be used in places where dyanmic range is all over the map(like outside)
6-weird effects when the assumptions used to generate the settings are violated
7-endless, and I do mean endless, discussions about what went wrong

In the end, the operator needs to think about what they are doing.

[Piotr Wozniacki]

Bill,

Do you think it's also possibble that the EX1 software doesn't consider the color matrix settings in zebra-trigerring calculations, only displaying the zebra basing on the luminance levels between the iris and the matrix circuitry - hence overexposing areas without even displaying zebra on them?

[Bill Ravens]

Piotr....

all the lightmeters I'm familiar with function by measuring luminance, only. They are all calibrated to place "proper exposure" in the middle of the gamma curve when looking at an 18% gray card. This is why I say you can't put 100% faith in what the light meter, or auto exposure system, is telling you. Real life isn't an 18% gray card. At least, not my life.Exactly what area of the image is the camera light meter deciding exposure? Is it an RMS meter, spot meter, or weighted meter? Sony has never told us. Expecting the auto exposure system to faithfully tell you the proper exposure 100% of the time is a fool's errand.

Despite the great exposure latitude of the EX1(I beleive Adam Wilt measured 10 f/stops)pointing at the ground for exposure then shooting the upper two thirds in the sky, will result in blown out skies.

On the subject of your specific question, I doubt the zebra function takes into account chroma values. It works strictly off of luma. Before you go running off crying that the color matrix changes have caused exposure faults, bear in mind that the saturation level can be greatly adjusted with a levels control without bad effects on the exposure readings. Sony even provides us with the HISAT preset for the purposes of increasing the gain in the color circuits.

The color matrix settings, themselves, function primarily as "hue" controls. Gosh I hate that word "hue". If you envision the colors on a wheel, red-orange-yellow-green-blue-violet---then back to red, the matrix adjustments provide for a means to "rotate" the colors on the wheel as individual R, G, and B channels. All I've done with the matrix settings are to reposition each channel on the wheel according to a calibrated chart telling me where the colors should lie, not where Sony says they should lie. This will NOT effect luminance readings with the lightmeter or zebra.

Having said that, keep in mind that luma and chroma are two different things. (Yes, I know, mathematically, one can relate the luma values to the chroma values.) The Truecolor preset I provided does two things. 1-It establishes luma settings(black, gamma, and 100%IRE) for maximum exposure latitude.
2-It establishes chroma adjustments(color matrix) for more lifelike colors. I'm tempted to say that this is somewhat of a subjective adjustment, but, it really isn't. A very carefully calibrated test chart was used to set the color matrix up. Now, I could have been irresponsible with my measurements, but, in the end, I'm very happy with my results. I decided to share those results. You can use them or not, entirely your choice. Latitude and headroom are mutually exclusive. You can't get one without compromising the other. Every photographer must find a place of balance with which they are comfortable. Manufacturers tell us, in their factory settings, where THEY think the compromises are. They have different priorities than users.

Amazingly, the color matrix values work very well at a wide range of luma settings. So, if you think the blacks are crushed, feel free(as if you ever weren't) to adjust the luma settings to give you more pleasing, more forgiving, or whatever, exposure values. My own objective was to maximize the amount of data I'm recording. Sometimes this isn't the most pleasing image when viewed without final grading in an NLE. A skilled and careful use of the Vegas Levels and Color Curve FX will yield things in your image you didn't realize were there. But, use them like a tasty spice, (red New Mexican chile comes to mind, not that stuff Chris calls Chili). Use sparingly and with finesse.

ahhh...one more thought...
I'm very happy to share the "science" of all this. After all, science is the public domain, anyone can do what I did. But, I won't share the "art" of what I do. That's proprietary and protected by the heavy arm of DRM....or it's supposed to be.

[Leonard Levy]

I have only looked at this shot in the stills on the web, but it looks like to me like something is wrong with a camera setting. It should not be neccessary for the operator to have to work around problems where the camera is treating the sky behind the trees differently than in the rest of the picture.
I don't know what it is, but telling the operator to underexpose slightly may be a fix in this particular shot, but its important to figure out what is going wrong because a professional camera should not do that in my opinion. I hope this is not a by product of the CAC lens function or anything else peculiar to the camera because it looks unacceptable to me . The knee should be creating a gentle roll off not a hard clip taht is different in different parts of the picture.

i neIed to learn a bit more about knee settings , but I noticed that the slope was "0" on your settings - what does that mean? Does that effectively get rid of a gentle roll off? I guess I'll look that up right now.

By the way in my mind the histogram is useless for this kind of exposure issue and useless for most exposure questions except the most general sense of whether you are weighted toward under or overexposure and you ought to know that anyway without looking at the histogram.

Has anyone every seen this on another professional camera.

- Lenny Levy

[Steven Thomas]

Leonard,
he was not using the cine gamma curves which roll into the highlights. He was using the std3 setting.

I've got plenty of sky images holding up decent latitude.

Here's an EX1 frame grab. It was shot about 11:00 AM.
I believe it was Cine3

[Piotr Wozniacki]

Leonard,

While I fully agree this behaviour is unacceptable, I'm slowly coming to terms with it. You see, we're trying here to squeeze maximum quality from a standard (i.e. non-cine) gamma curve, which has been used in conjuction with considerable colour saturation increase (with both Hisat matrix adoption and several primary pairs adjustmenst), but WITHOUT even trying to optimize the curve's KNEE settings...

As I said before - if you want to get saturated and rich colour, either use cinegamma (e.g. CINE1) which takes care of the proper KNEE setting itself, or adjust the KNEE with a standard gamma (like the STD3 used in this case).

Otherwise, don't expect good results with backlit scenery, cause you will either end up with this ugly patches I have shown, or you will severely overblow the backgroud, or you will get nice and rich blue sky but severely underexposed and over-saturated foreground.

[Leonard Levy]

Well maybe I've forgotten what to expect in this situation on an ordinary video camera, but i've been shooting professionally for over 20 years the vast majority with cameras in standard gammas - not cine gammas, and that looks wrong to me.
What looks wrong isn't that the camera can't handle the whites in the sky, but that its treating the sky behind the trees differently. You shouldn't have nasty abrupt clipping with areas in the sky abruptly having no color and others having color, that's just unnacceptable.

Maybe I need to look at those pictures more carefully again.

I am confused about one technical thing though.
I looked up point and slope and those settings don't look weird as "0" is a mid point on the slope, but when the auto knee is on does that mean your point and slope numbers are overridden?

[Piotr Wozniacki]

Originally Posted by Leonard Levy

I am confused about one technical thing though.
I looked up point and slope and those settings don't look weird as "0" is a mid point on the slope, but when the auto knee is on does that mean your point and slope numbers are overridden?

Yep. Leaving Knee on auto overrides all your fine-tuning to the point, slope and saturation.

Leonard, you say you've never saw such behaviour with your 20 years with cameras that didn't even have cine gammas; perhaps this is the clue? Now that the cine gammas are available, the standard curves are more "relaxed" and left to the (knowledgeable) users to shape. My experience is much less than 20 years, but I must admit on the V1 I never pushed the standard gamma so hard!

Anyway, this really isn't nice - that's for sure. This grab has not been exaggerated by the 32bit Vegas stretching; it's exactly how it looks at playback time on my monitor (and it's really difficult to rate it as "severely overexposed", either - as Bill is suggesting):