Which are the ratios' methods?

[adrjork]

Hi everyone,

perhaps my question was covered in this forum (if so, sorry). This is the question:

I use a 5D3. In ML you can choose different ratios such 3:2, or 5:3, or 16:9 and so on.
I'd like to know "how" this ratios are obtained: i.e. I've read somewhere that 16:9 1920*1080 was obtained binning the entire sensor for an integer number. And this is good for me: using the entire sensor, and havig a "clean" integer binning.
So, my question is: which is the method (dividing for which number) to obtain all the other ratios? The sensor is still used entirely?

Thanks a lot.

[Walter Schulz]

Frame ratio and binning/skipping/1:1 crop are independent concepts. For 5D3 mlv recording in nightly builds: Native (un-cropped) is 3x3 binning, crop is 1:1.
For mlv recording in experimental 4k-crop: https://builds.magiclantern.fm/experiments.html -> 4K raw video recording; lossless compression

[adrjork]

So, if I understand correctly, EVERY NON-CROP sizes/ratios are 3x3 binning (even 5:3 1920*1152). While EVERY crop sizes are 1:1.
Is that right?

[a1ex]

Code Select Expand


mlv_dump M12-3456.MLV -v | grep RAWC -A 8 | grep sampling


Usual values: 1x1, 3x3, 5x3.

Possible values (always integer):

- vertical (first number):
   - 1 (read every line)
   - 3 (bin 3 lines) [only 5D3]
   - 3 (read 1 line, skip 2) [only models other than 5D3]
   - 5 (bin 5 lines) [only 5D3]
   - 5 (read 1 line, skip 4) [only models other than 5D3]
   - hardware allows many other values

- horizontal (second number):
   - 1 (read every column)
   - 3 (bin 3 columns)
   - no other values allowed by hardware

Theory: http://www.magiclantern.fm/forum/index.php?topic=16516.0
Spec: https://www.magiclantern.fm/forum/index.php?topic=17021.msg181639#msg181639