Lowering exposure increases grain 35mm

[Nicky Rudolph]

Hi all. First post on the forum so thank you! I recently shot a project on 5219 rated at 320. I rated it lower to get a bit cleaner neg. Running into a bit of a stump though when grading. Common thought would tell me overexposing and pulling the image down in post will give a cleaner, less grain image. Yet as I decrease the exposure in davinci the grain becomes more prominent. Especially on these white walls I have in the shot. As I push the exposure back up, the grain gets brighter and less visible. A bit stumped and curious of your guys thoughts. thanks !

[Tyler Purcell]

Well, the film scan off the machine, isn't going to have a look applied, it should be pretty flat. Are you applying a look initially or grading it directly off the scanner? 

My process is to create a balanced node first, get the RGB values right on the nose for luminance using the color wheels (lift, gamma, gain). Then I'll start by working the image to a desired look in the subsequent nodes, mostly using the custom curves and eventually the color warper tool. 

What tool are you using to increase exposure? Are you checking your scopes? If you are working in the wrong color space or have a LUT applied, you could be pushing the levels in the wrong direction. 

[Nicky Rudolph]

4 hours ago, Tyler Purcell said:

Well, the film scan off the machine, isn't going to have a look applied, it should be pretty flat. Are you applying a look initially or grading it directly off the scanner? 

My process is to create a balanced node first, get the RGB values right on the nose for luminance using the color wheels (lift, gamma, gain). Then I'll start by working the image to a desired look in the subsequent nodes, mostly using the custom curves and eventually the color warper tool. 

What tool are you using to increase exposure? Are you checking your scopes? If you are working in the wrong color space or have a LUT applied, you could be pushing the levels in the wrong direction. 

Yeah its flat. It's a 4444 4k scan from fotokem. I've messed with it a few ways. One was applying a 2383 lut and then balancing. Another way I have done was just directly from the log image. I'm doing all the steps you say, balancing the image, correcting the temp. I just then get the image to a 2/3ish over look, which is what I exposed it as, and when I bring it down to a normal level, the grain just becomes more apparent. Mostly on the white/gray walls. The image looks great overall and I'm happy, I just find it odd that the more I decrease exposure (using offset wheel) the more visible the grain.

[Tyler Purcell]

6 hours ago, Nicky Rudolph said:

Yeah its flat. It's a 4444 4k scan from fotokem. I've messed with it a few ways. One was applying a 2383 lut and then balancing. Another way I have done was just directly from the log image. I'm doing all the steps you say, balancing the image, correcting the temp. I just then get the image to a 2/3ish over look, which is what I exposed it as, and when I bring it down to a normal level, the grain just becomes more apparent. Mostly on the white/gray walls. The image looks great overall and I'm happy, I just find it odd that the more I decrease exposure (using offset wheel) the more visible the grain.

Those LUT's are for finishing, not for grading. You don't apply them to the source material. They are only a "look" for a finished product. 

It's not a LOG image. LOG images are encoded and decoded. This image is not encoded OR decoded. You have to build the look yourself using the node color. Again, I build the look using the color wheels; lift, gamma, gain and of course, saturation. If you work the wheels, you can get the look your after with a single node and simply refine with a 2nd. 

The offset wheel, isn't really used for anything but correcting issues with single channels of color. I rarely touch it for anything unless I'm restoring heavily damaged film and need a way to change the gain on a single channel. If you're using the offset tool for grading, then that could be part of your problem as well.