Alpha Beta Omega
Veteran
I whipped out my AF101 and did a quick test. Sure enough less clipping in CineD and reduced chroma on a caucasian face in direct sunlight. Still plenty of chroma and looking pleasing.
Chroma Clipping
- Thread starter Chadac
- Start date
Cine-D held up good outside in harsh light editman? I have yet to test outside.
Alpha Beta Omega
Veteran
We have a low sun this time of year but it was blasting straight on the face and looked good.
Good to know. Thanks!
Barry_Green
Moderator
I'm working on comparing dynamic range, I just picked up an extremely precise, expensive, ultimate chart for that -- the best in the world, the DSC 17-stop backlit dynamic range chart. I will have absolutely definitive unassailable dynamic range tests sometime soon.
Cinema processes its color very differently than the other GH2 gammas. On the Cinema gamma, the color is consistent throughout -- no matter what brightness you set, the proportions of the color components will be the same. That's not true of the other gammas, they compress the range the brighter it gets (and they expand the range the darker it gets). That leads to an inaccurate presentation of color, but it also changes the way the highlights "roll off". Whether one is better than the other is a matter of personal taste. You may like the way one looks, and hate the way the other looks. That's okay -- there are reasons that both systems are in there, so you can choose which one you prefer.
However, if I was trying to match a GH2 and AF100, no doubt that I would use CINEMA on the GH2.
Barry_Green
Moderator
That's the funny thing about the Panasonics -- they have SO MUCH chroma to start with, that when you drop it to minimum, there's still a lot there. I don't have a Canon XHA1 to compare against, but I wouldn't be surprised if the AF101 at minimum chroma level, had as much chroma in its signal as an XHA1 at default chroma. I don't know that to be the case, I'm just theorizing, but I wouldn't be surprised if it were so.
So while it may sound drastic to chop the chroma so far, it really only brings it kind of more in line with how other cameras perform anyway.
The simplest takeaway from this I can say is -- if using default chroma levels on an AF100, don't let your skintones go above 80 IRE (which, I consider overexposure anyway).
If you need your skintones to go hotter, combine chroma level -7 with cine-D gamma and you can push the skintones to almost 90 IRE with no negative side effects.
Barry_Green
Moderator
Put the Chroma Level at -7, and put the gamma on cine-D, and you should be able to push the skin tone highlights up to 89 and still maintain accurate color with no chroma clipping. I don't know if it'll deliver the exact look you're going after, but try it and let us know what you think.
Tell me the exact scenario you want tested and I'll see if I can make it happen.
roei z
Veteran
thanks Barry, great great help.
i don't want to discuss GH2 here, but since you mention an upcoming DR test, it would be really helpful to examine DR stops between smooth/cinema (although you loose some color accuracy, i suspect you can get pretty suffocated in cinema DR. unless we could overexpose cinema 2-3 clicks and win both worlds. interesting tests ahead
)Alpha Beta Omega
Veteran
When using Cine D and chroma at -5 as I did when I tried it out you get a very flat looking video. No problem adding colour in grading if needed. I switched to HD normal and chroma at 0 which resulted in blown highlights and plain ugly using the same exposure. Using HD norm and stepping down to chroma -5 I got a good image when lowering exposure. So if I would like to add blown whites but still have decent face tones this is the way to go. It looks pretty washed but it's easy to adjust in post to have it more contrasty or colourful. Using CineD adds noise in blacks since it's so flat but lowering it gives about the same noiselevel as HD norm.
It seams so far I will create one preset for noisefree recording, one setting for sharp images and a special "skin tone under daylight" preset to shoot in direct sunlight.
Could it be Canons DSLR also compresses colour to avoid clipping?
It seams so far I will create one preset for noisefree recording, one setting for sharp images and a special "skin tone under daylight" preset to shoot in direct sunlight.
Could it be Canons DSLR also compresses colour to avoid clipping?
Last edited:
Barry_Green
Moderator
I'd like to find out, but don't really want to spend a couple thousand to pick up a 5D body to do some testing. Someone ship me a 5D for a couple of days and we'll get to the bottom of it... :thumbsup:
Steve Kahn
Well-known member
Wow Barry! Thanks so much! Amazing work. Great info!
I'd like to know if an external recorder has any effect on:
- chroma clipping
- dynamic range
- real world difference in color depth
- banding issues
Those are tests I'd love to see. This is a great camera. I'm very happy with it and quite thankful you are helping to make it that much better!
I'd like to know if an external recorder has any effect on:
- chroma clipping
- dynamic range
- real world difference in color depth
- banding issues
Those are tests I'd love to see. This is a great camera. I'm very happy with it and quite thankful you are helping to make it that much better!
Barry_Green
Moderator
Can't see how an external recorder would have any effect on chroma clipping or dynamic range, when dealing with an 8-bit output.
It has a noticeable to significant effect on color resolution, 4:2:2 is twice the color res of 4:2:0.
It can have a noticeable effect on banding but, as always, this is a two-edged sword. The Nano can show less banding than the onboard codec, but ... that's because the Nano, at super-high bitrates, will record *all* the noise. The AVCHD codec will discard some noise as part of its compression process. Banding shows up on fine gradients and the fix (in 8-bit codecs) is to sprinkle a little noise around to dither the image and avoid banding. Because the Nano (again, at high bitrates) will see all that noise, and compress it as part of the image, it will reduce the visibility of banding -- while increasing the appearance of noise. The footage will look noisier, with less banding. You can accomplish the same thing in-camera.
The main things the Nano can do for you is/are:
1) Provide 4:2:2 color, which is very nice and which is the primary reason that the Nano is required by the BBC when shooting with the XDCAM-EX cameras or the AF100.
2) Provide redundancy, as you're recording to two media simultaneously. Protects against card failures and gives you a copy you can hand off at the end of the shoot, if necessary.
3) At high-enough bitrates, the Nano can provide less-compressed footage that will retain the finest of details better. (Note -- you can also record in worse quality on the Nano than the AF100's native codec! If you choose the 18mbps or 35mbps modes, you'll actually be recording lesser-quality footage. So it's important that when you discuss an external recorder, that you're giving it enough bandwidth to do the job. Most folks seem to accept 100mbps long-GoP or 220mbps I-frame as the points where the image is clearly and always superior. Of course, you're also talking about 4x to 10x larger file sizes, so you have to factor that in.)
4) In IntraFrame mode, it can provide complete immunity from long-GoP motion/color/codec overload issues. This would be a huge benefit for an HDV camera, but AVCCAM is pretty resilient so the benefit here, while still there, isn't as pronounced.
5) Provides files that may be more edit-friendly depending on what NLE you use. The recorded MPEG-2 quicktime MOVs can be dropped right in the FCP timeline, whereas the .MTS files from AVCCAM would need conversion before FCP could use them. For other NLE's it's kind of a wash, although MPEG-2 (and especially I-frame-only) are easier on some NLEs/computer combos to edit than AVC.
It has a noticeable to significant effect on color resolution, 4:2:2 is twice the color res of 4:2:0.
It can have a noticeable effect on banding but, as always, this is a two-edged sword. The Nano can show less banding than the onboard codec, but ... that's because the Nano, at super-high bitrates, will record *all* the noise. The AVCHD codec will discard some noise as part of its compression process. Banding shows up on fine gradients and the fix (in 8-bit codecs) is to sprinkle a little noise around to dither the image and avoid banding. Because the Nano (again, at high bitrates) will see all that noise, and compress it as part of the image, it will reduce the visibility of banding -- while increasing the appearance of noise. The footage will look noisier, with less banding. You can accomplish the same thing in-camera.
The main things the Nano can do for you is/are:
1) Provide 4:2:2 color, which is very nice and which is the primary reason that the Nano is required by the BBC when shooting with the XDCAM-EX cameras or the AF100.
2) Provide redundancy, as you're recording to two media simultaneously. Protects against card failures and gives you a copy you can hand off at the end of the shoot, if necessary.
3) At high-enough bitrates, the Nano can provide less-compressed footage that will retain the finest of details better. (Note -- you can also record in worse quality on the Nano than the AF100's native codec! If you choose the 18mbps or 35mbps modes, you'll actually be recording lesser-quality footage. So it's important that when you discuss an external recorder, that you're giving it enough bandwidth to do the job. Most folks seem to accept 100mbps long-GoP or 220mbps I-frame as the points where the image is clearly and always superior. Of course, you're also talking about 4x to 10x larger file sizes, so you have to factor that in.)
4) In IntraFrame mode, it can provide complete immunity from long-GoP motion/color/codec overload issues. This would be a huge benefit for an HDV camera, but AVCCAM is pretty resilient so the benefit here, while still there, isn't as pronounced.
5) Provides files that may be more edit-friendly depending on what NLE you use. The recorded MPEG-2 quicktime MOVs can be dropped right in the FCP timeline, whereas the .MTS files from AVCCAM would need conversion before FCP could use them. For other NLE's it's kind of a wash, although MPEG-2 (and especially I-frame-only) are easier on some NLEs/computer combos to edit than AVC.
Steve Kahn
Well-known member
So when you say there is a noticeable effect on color are you saying you can see more shades of a color in a shallow gradient? This is where I thought an external recorder may help reduce banding by having less color subsampling and posterization.
On a side note: the message before seemed to be use Cine-Like D when you can control the environment, use a gamma cure with knee when you can't
Now the message seems to be always use D (it's my favorite anyway) kill the chroma when you have areas that you know will clip out of your control.
On a side note: the message before seemed to be use Cine-Like D when you can control the environment, use a gamma cure with knee when you can't
Now the message seems to be always use D (it's my favorite anyway) kill the chroma when you have areas that you know will clip out of your control.
Id say with a mild give or take, thats about right.
jburke14
Active member
Once one of the three channels clip, in this case 85 for red. Once it clips, it leaves just Blue and Green, and no red, leaving the image a yellow cast. Look at the sky after you send the highlights about 85, They turn a blueish-yellow-green color, that doesnt look like the sky color. So set the highest luminance values to 85 or below and you are fine, your colors will not shift in tone. It also makes for ugly skin tones because the red starts clipping and leaves faces yellowish.
Barry_Green
Moderator
Keep in mind that this test was only for skin tones, not anything else. The proportion of red to green to blue was based on a caucasian skin tone. If you were shooting Avatar, obviously the numbers would be very very different because blue would be a much higher proportion of the signal. Or if shooting traffic lights, or ... well, anything. Whatever the predominant primary color is of the object you're shooting, that's the channel that will clip first.
And if you're shooting different shades or colors, the clipping will happen differently relative to the other colors. What I found was that in caucasian skin, you can go to about 82 on a GH2, or 80 on a normal-gamma saturated-color AF100, with no clipping on the skin tones.
If you were shooting clouds or trees or something, the balance will be different because those are substantially different colors and the mix of red, green and blue will be very different for those.
And if you're shooting different shades or colors, the clipping will happen differently relative to the other colors. What I found was that in caucasian skin, you can go to about 82 on a GH2, or 80 on a normal-gamma saturated-color AF100, with no clipping on the skin tones.
If you were shooting clouds or trees or something, the balance will be different because those are substantially different colors and the mix of red, green and blue will be very different for those.
What are you running for your "sharp Image" scene file?
Alpha Beta Omega
Veteran
This is a sample I shoot yesterday in direct sunlight. No clipping even when hitting that face hard. I added a bit more colour in post. This one was with Detail -2. Chroma -5, HD Norm and Norm1 matrix. Daylight balanced WB.
View attachment 30016
View attachment 30016