Dialing in the AF100

[olof]

This is what I have settled on for my standard settings.

Here is a png direct exported from FCP:

Here is how my EX1R is set up, my standard PP for EX1 and EX3 cams:

Here is a bad photo of how I was doing the setup:

I have some more work to do. To come up with a PP for the EXcams to match the AF100.

I was using my Evaluation monitor (JVC DT) to view the scene direct HDSDI. This is the best compromise I have come up with for the AF100 to balance banding and noise. The lens on the AF100 is the Canon 24-70 f2.8 L at 50mm f4.5.

Bear in mind I was not using a computer monitor but a calibrated pro evaluation monitor, and hardware scopes.

I am sure I will be making more adjustments and other scene files as I shoot some projects, but this will be my starting point.

PS coring is +4 the rest is unchanged from the standard scene file.

 

[Jan_Crittenden]

I don't understand the -5 Iris setting. Why and which lens are you talking to for that setting to make sense?

Thanks,

Jan

 

[olof]

I don't understand the -5 Iris setting. Why and which lens are you talking to for that setting to make sense?

Thanks,

Jan

Jan, my understanding is that -5 on the auto iris means 1/2 stop under exposed. I just put that in there because on almost every camera I use I go 1/2 stop under when using cinegammas. It has no effect what so ever on the Canon I am using here.

The only way I normally use auto iris is push once to set, then when released it stays at that setting, this quickly gets you in the ballpark. It is how I have always done it, and it is the only way I use auto iris. I don't even have a lens for this cam yet that will auto, so I just threw 1/2 stop under in there, like I do on the rest of my cams.

 

[Jan_Crittenden]

The setting means nothing unless it is talking to a lens. Your lens has no communication with the camera and thus the setting means nothing.

Best,

Jan

 

[olof]

The setting means nothing unless it is talking to a lens. Your lens has no communication with the camera and thus the setting means nothing.

Best,

Jan

That is exactly what I was saying above. But if I was to put a Lumix or other lens that communicates, it would auto iris 1/2 stop under right?

Every camera I have I set auto iris 1/2 stop under. I thought that is what this setting does. I know at the moment all my glass is fully manual. That is how I like it. I virtually never use auto focus, auto iris or auto anything. The only exception is on my other cams you can push momentary auto iris., just to get you in the ballpark then you tweak it for fine adjustment.

 

[Fotojorno]

Hi Olof, quick question. Is Cine V your primary setting just for footage you do not plan on grading/adjusting in post, or for all your footage?

Thanks..

 

[olof]

Hi Olof, quick question. Is Cine V your primary setting just for footage you do not plan on grading/adjusting in post, or for all your footage?

Thanks..

At this point I am trying to find out what this camera can do. These settings seemed the best for this scene. I will probably change the saturation down a few notches. Cine B is a lot flatter. And I will definitely try that as well as some others on outdoor shots.

The image is more contrasty looking on my computer screen than my evaluation monitor. The banding is also more evident on the computer screen.

 

[Rick Burnett]

At this point I am trying to find out what this camera can do. These settings seemed the best for this scene. I will probably change the saturation down a few notches. Cine B is a lot flatter. And I will definitely try that as well as some others on outdoor shots.

The image is more contrasty looking on my computer screen than my evaluation monitor. The banding is also more evident on the computer screen.

I was going to comment on the banding as well. On my computer screen it is much more evident. I too have an AF100 and am going to begin working on settings to reduce the banding a bit more. Since I release web stuff as well, I kind of need the best of both worlds.

 

[Steve Kahn]

I was going to comment on the banding as well. .

That's what popped out to me as well. Why is it so much more evident in the AF100 when they are both 8bit cameras? In this shot the AF100 looks cleaner than the EX1/3 but maybe it would be wise to dial in a bit of noise?

 

[Rick Burnett]

That exactly what I am trying to do as well. In comparing shots with my 7D to the AF100, the AF100 I can get MUCH MUCH cleaner than the 7D with the settings I am using. However, it pops the banding right out like this when I use them in that way. So, I am trying to figure out how I want to push more noise back into the footage to reduce that banding a little more.

The problem I get is the banding isn't as smooth either. Look above how there are bands of lighter banding within darker banding.

What I am going to do is setup my Black Magic Intensity Pro card so I can get the footage on my monitor real time, then I can setup a scene with a gradient and play with the settings and see what enhances or reduces these effects.

 

[Steve Kahn]

Looking forward to see what you come up with though the strange thing is that he's using "cine-like" matrix and only a detail of -2 yet still shows very little noise.

Is cine-like V much cleaner than cine-like D? I thought most of the noise issue was in the matrix.

 

[Rick Burnett]

I've been using Cine-Like V as the noise was the cleanest of all of them for me. That said, I've felt the detail setting above -5 contributed more noise sharpening than anything else which was what I was trying to control before. I have been using FLUO for my matrix, so I have not tried other settings yet.

When looking at the amazing number of tests another user posted, I correlated the results and Cine-V always seemed to have less noise for that user and his tests as well. I haven't played with Norm1 or 2 yet. I figured the best next step was to setup a static scene like this one and start changing settings. The reason I haven't is I let another production borrow my lights and I need to get them back to really setup a good test with decent gradients.

 

[Alpha Beta Omega]

I did a commercial before christmas with my 5D mkII and I had sever banding in one take and had to add noise to make it smooth in post. During my last AF test I also noticed you end up with banding if you try to make the camera super clean - just like the 5D can be. I think it's better to preserve some fine noise in the recorded signal since it's 8-bit. So I tend to preserve more detail at maybe -2 and not use the hard killing noise settings.

 

[Rick Burnett]

I did a commercial before christmas with my 5D mkII and I had sever banding in one take and had to add noise to make it smooth in post. During my last AF test I also noticed you end up with banding if you try to make the camera super clean - just like the 5D can be. I think it's better to preserve some fine noise in the recorded signal since it's 8-bit. So I tend to preserve more detail at maybe -2 and not use the hard killing noise settings.

I'm trying to figure out what detail does specifically. I'd really like to know at what setting detail is leaving the signal alone that is coming off the sensor and not sharpening it because what I was finding at detail 0 was the noise looked like it was being sharpened, or something about the sharpening algorithm was enhancing that noise in a way that I didn't like. When I got to -7, it's REALLY clean, but of course, then has the problem with banding, which I don't want, so now I am floating in the middle to see what I want to go with.

What I will try and do is create a matrix of results with different detail settings. I will expose so that I get the noisy area I don't like at detail 0 so we can see how it moves between the settings. Also, I will probably try a few different matrices and gamma settings too and see what comes of it all.

 

[Alpha Beta Omega]

In certain combination of settings Like Cine B and Cinelike matrix - Detail, V Detail and Coring tames or enhances the noise very noticeable. Coring is a noise reducer and we all know the downside to turn it up to much. Same thing with the Norm2 matrix that's very noise free in blacks but also has the most smeared image of them all. I don't want smeared since it just looks awful. When looking at many other cameras there is noise just like on the AF and I actually think noise is a good thing to a certain extent. In 8-bit it's almost a must.

It's really time consuming trying out all combinations but I feel like I now do it bit by bit and it has a steady progress towards a 100% profiled camera. I have one or two jobs this week and it's essential to use it in real situations to get to know it.

 

[PaPa]

I'm trying to figure out what detail does specifically. I'd really like to know at what setting detail is leaving the signal alone that is coming off the sensor and not sharpening it because what I was finding at detail 0 was the noise looked like it was being sharpened, or something about the sharpening algorithm was enhancing that noise in a way that I didn't like. When I got to -7, it's REALLY clean, but of course, then has the problem with banding, which I don't want, so now I am floating in the middle to see what I want to go with.

What I will try and do is create a matrix of results with different detail settings. I will expose so that I get the noisy area I don't like at detail 0 so we can see how it moves between the settings. Also, I will probably try a few different matrices and gamma settings too and see what comes of it all.

I am in the same boat. Trying to figure out with my settings what Detail and what Coring i should be using to minimize noise without completely getting rid of it so that the Posterization around gradients doesn't appear to be so nasty.

I notice in your settings, with detail at -2 you are still getting some edge enhancement. I think a really good test should now be on how to find the best balance between noise and banding/posterizing to maximize the codec's abilities.

I'm almost finished dealing with the highlights issue, and im glad to say that I've found several solutions and now all that's left is the gradient posterization.

 

[PaPa]

Also, was the EX1 image recorded at 8 bit or 10 bit? I don't see any type of gradient posterization with it.

 

[olof]

Also, was the EX1 image recorded at 8 bit or 10 bit? I don't see any type of gradient posterization with it.

I was recording both to their native media. So the EX1R was also 8 bit.

I view the the output HDSDI when setting the cameras up. But then to get the real idea I capture to SxS and SD cards.

I will be using a NanoFlash a lot with the AF100, so the footage will be very similar to straight out SDI.

If I get time today I will make some more tests. Directly HDSDI into my Kona card. Much faster than reading cards.

The banding effect is much less direct out HDSDI than the SD card version. So this setup looks much better on my monitor than it does after compression (on the same monitor). So I will use this setup when shooting direct to Kona or to NanoFlash.

Like I said much more testing coming up between projects.

 

[PaPa]

For what it's worth, i just pointed my camera at a gradient along my wall at home and i noticed posterization become visible at:

Detail: anything lower than -2
with
Coring: anything higher than -6

So i guess it's really a trade between edge enhancement/noise/aliasing vs gradient posterization. So something with a little noise and detail is good to counter the banding effect, but finding the best middle ground is very difficult.

 

[ustein]

> i noticed posterization become visible at:
>Detail: anything lower than -2

You mean -4 gives more posterization? HOw can lower sharpening do that?