Barry_Green
Moderator
There's lots of stuff floating around about "native" ISOs, and whether you should only use certain ISOs and avoid others, skipping the ones inbetween, etc. I decided to get to the bottom of it.
Here's every ISO the AF100 has. Nothing changed between shots except the ISO itself; I used the iris and ND filters to compensate to make the exposure the same. Judge for yourself which ISOs you want to use and which you want to avoid.
Note that I didn't shrink anything down -- these are pixel-for-pixel extractions of what I shot. It seemed quite wasteful to shoot the chart full-screen; any noise that's gonna show up would show up on the bars at the size I shot them (each bar is about 28 x 66 pixels, so plenty of space for noise to show).
In this video I shot every ISO that the AF100 has, of the same subject (a DSC Labs ChromaDuMonde Billups VF/X28 test chart). The idea here was that the chart shows all the tonality ranges, so noise in any particular luminance range should be clearly visible. And the chart has 28 different color patches, so any indications of chroma noise in particular colors should show up. And, it has four skin tone patches (the colors just above/below the corners) representing a variety of human skintones.
In the first section of the video I shot every ISO of the chart, using a Zeiss Compact Prime 18mm lens, and all these shots are at 1:1 resolution. I framed up the shot very wide-angle, so that I could fit all of these onto one 1920x1080 frame with no scaling or stretching or downconverting. Download the original 1920x1080 file if you want to pixel-peep, as these are 1:1 files, not scaled.
In the second portion of the video I digitally zoomed the recorded footage by blowing it up to 250% horizontal size and 200% vertical size. This should aggravate and magnify any noise, which would show the effect that a higher ISO may have on the perceived noisiness of any particular luminance level or color. Note: if you see noise in the second half of the video, remember it's exaggerated! This is 5x larger than pixel-to-pixel, specifically to make the noise jump out. In the actual footage it would be much milder. Look at the first portion (with all 13 charts) for what the noise would really look like.
Settings for this video were reasonable settings for minimizing noise in an AF100's picture. They were not extreme, but were tailored to reducing the visibility of noise.
Detail and V-detail: -5
Coring: +2
Master Ped: -2
Gamma: HD NORM
Matrix: Norm2
Skin DTL: OFF
No noise reduction software or grading or any post-processing whatever was done to the footage, other than as mentioned -- it's been cropped and moved to different areas of the screen, but at a pixel-to-pixel accurate level, and then in the second part it was expanded/digitally zoomed in to exaggerate the appearance of noise, solely for the purpose of comparing one to another.
The ISO was changed for each shot, and the iris (and/or ND filter) were changed to compensate. I aimed for the white chip in the upper left to just barely, barely, barely register with 105 zebras.
I also repeated the test, using the "normal" F1 settings (all zeroes, HD NORM gamma, NORM1 matrix, etc). That file is available for download on MediaFire. As expected, it makes any noise more visible, as we already know. Not that it's unreasonable; it looks really clean at 200 ISO and noisiest at 3200, pretty much exactly what you'd expect..
For those who want to peep the pixels directly, here are the high-quality rendered MPEG4 files:
Low-Noise settings
Factory-stock settings
Here's every ISO the AF100 has. Nothing changed between shots except the ISO itself; I used the iris and ND filters to compensate to make the exposure the same. Judge for yourself which ISOs you want to use and which you want to avoid.
Note that I didn't shrink anything down -- these are pixel-for-pixel extractions of what I shot. It seemed quite wasteful to shoot the chart full-screen; any noise that's gonna show up would show up on the bars at the size I shot them (each bar is about 28 x 66 pixels, so plenty of space for noise to show).
In this video I shot every ISO that the AF100 has, of the same subject (a DSC Labs ChromaDuMonde Billups VF/X28 test chart). The idea here was that the chart shows all the tonality ranges, so noise in any particular luminance range should be clearly visible. And the chart has 28 different color patches, so any indications of chroma noise in particular colors should show up. And, it has four skin tone patches (the colors just above/below the corners) representing a variety of human skintones.
In the first section of the video I shot every ISO of the chart, using a Zeiss Compact Prime 18mm lens, and all these shots are at 1:1 resolution. I framed up the shot very wide-angle, so that I could fit all of these onto one 1920x1080 frame with no scaling or stretching or downconverting. Download the original 1920x1080 file if you want to pixel-peep, as these are 1:1 files, not scaled.
In the second portion of the video I digitally zoomed the recorded footage by blowing it up to 250% horizontal size and 200% vertical size. This should aggravate and magnify any noise, which would show the effect that a higher ISO may have on the perceived noisiness of any particular luminance level or color. Note: if you see noise in the second half of the video, remember it's exaggerated! This is 5x larger than pixel-to-pixel, specifically to make the noise jump out. In the actual footage it would be much milder. Look at the first portion (with all 13 charts) for what the noise would really look like.
Settings for this video were reasonable settings for minimizing noise in an AF100's picture. They were not extreme, but were tailored to reducing the visibility of noise.
Detail and V-detail: -5
Coring: +2
Master Ped: -2
Gamma: HD NORM
Matrix: Norm2
Skin DTL: OFF
No noise reduction software or grading or any post-processing whatever was done to the footage, other than as mentioned -- it's been cropped and moved to different areas of the screen, but at a pixel-to-pixel accurate level, and then in the second part it was expanded/digitally zoomed in to exaggerate the appearance of noise, solely for the purpose of comparing one to another.
The ISO was changed for each shot, and the iris (and/or ND filter) were changed to compensate. I aimed for the white chip in the upper left to just barely, barely, barely register with 105 zebras.
I also repeated the test, using the "normal" F1 settings (all zeroes, HD NORM gamma, NORM1 matrix, etc). That file is available for download on MediaFire. As expected, it makes any noise more visible, as we already know. Not that it's unreasonable; it looks really clean at 200 ISO and noisiest at 3200, pretty much exactly what you'd expect..
For those who want to peep the pixels directly, here are the high-quality rendered MPEG4 files:
Low-Noise settings
Factory-stock settings
I am glad you are seeing things my way 
