Menu

Show posts

This section allows you to view all posts made by this member. Note that you can only see posts made in areas you currently have access to.

Show posts Menu

Messages - VivaLaTip

#1
Camera-specific Development / Re: Canon 6D
March 06, 2018, 01:31:38 PM
Tested the following with @dfort latest build from his download page:

- 14 bit lossless mlv (1824x1026) video
- 12 bit lossless mlv (1824x1026) video
- 11-8 bit lossless mlv (1824x1026) video

Processed them with mlv_dump to get a single dng, and all seem to work well, no weird artifacts.

File size:
- 14 bit, 1,878 KB
- 12 bit, 1,630 KB
- 11-8 bit, 1,565 KB

Post-processing works with MLV APP

Sound works too.
#2
Great work! Do I hear 6D and lossless?  :D
#3
Feature Requests / Internal ND for RAW video
December 17, 2017, 07:35:21 PM
Internal ND for RAW video

New here to the forum and thank you all for the great out-of-the box inventions! That said, I have an idea that might be great or very stupid ??? I post it anyway, maybe we get somewhere. I have been searching around the forum a lot to find post that would indicate that the Canons are capable to calculate this. I think it might, but i'm no expert so let me know if it sucks or not.

So let's say for cinematic purposes we want to shoot 24 FPS and a shallow depth of field with something like a f/1.8 or f/2.8 lens. On a bright day even with ISO 100 the picture is blown out when we apply the 180 degrees rule (shutter == 1/twice the FPS). So we use ND filters to compensate for the wide aperture.
Great, but the filters are rather expensive when using different lenses and annoying, easy to break and so on. What if magic lantern could have another sophisticated solution to this problem.

The Idea
What if we set the shutter speed to 1/96, read two frames, take the average and write the resulting image to the sd card. This way we still have the blur we love with 24 FPS, because the average of two frames each being 1/96th of a second results in the same image as a 1/24th of a second image (I suppose). And we have half the light like a ND filter.

In theory I think possible in practice?

Normally we have this:

Read sensor(1/48th of a second) -> prepare data -> write to sd

Then it would be:

Read sensor(1/96th of a second) -> Read sensor(1/96th of a second) -> calculate average image -> prepare data -> write to sd

The only extra calculation is doing the averaging. The big question is, is the Canon CPU capable of doing this in real-time?

If yes, that would be great. Party time :D! To extend this we could do 3 times 1/144th of a second and average them, or 4 times 1/192nd of a second.

Another thing that might be interesting is that the noise floor is going down when averaging two frames. (Maybe this doesn't matter since we are at ISO 100 already).

Probably I am missing something completely, else it would already be in there, I'm very curious to read your answers!