Messages

May I remind you to fill this chart?

http://www.magiclantern.fm/forum/index.php?topic=6215

I mean, with the new raw2dng (uploaded 5 minutes ago), is there any improvement?

Any improvement with this one? https://bitbucket.org/hudson/magic-lantern/downloads

Just type it without the -o

No

Still need some input regarding:

- links to confirm the numbers, especially max resolutions for continuous recording (24/30p) and max recording times at best quality (follow the example from Dario Paganini)
- double-check the entries with a question mark (and the others too)
- numbers for 6D, 650D, EOS-M, 1100D
- all other empty cells
- for 5D2/50D continuous 30p I expect identical figures
- for 6D/650D/EOSM 30p I expect identical figures

Thanks!

Right now it's the last one, but Thomas Worth found a test case where there's room for improvement.

Edit: this seems to be a lot better: https://bitbucket.org/hudson/magic-lantern/commits/d4b1e587cfab

(at least it covers all the DNGs that I've tried it on, with extremely strong curves)

Well, a very close approximation.

https://bitbucket.org/hudson/magic-lantern/commits/afa12e80e721

Now it's as easy as moving the focus box wherever you want, pressing zoom and recording.

Aliasing can't be removed in post without breaking the Nyquist theorem

(which is on the same level as breaking physics laws)

Aliasing can't be removed in post without breaking the Nyquist theorem

(which is on the same level as breaking physics laws)

If the subject is off-center, use the dolly mode and pan until you frame it correctly. Use grayscale preview for correct framing.

Absolute counting is variable, depends on image size. On the current histogram, overexposure warnings show percentage (1% clipped, 2% and so on; what's smaller than 1% is variable radius, without text).

I think it can reach 1728x972 or maybe 1856x1044 in theory (needs to be confirmed though).

https://docs.google.com/spreadsheet/ccc?key=0AgQ2MOkAZTFHdFFIcFp1d0R5TzVPTVJXOEVyUndteGc#gid=5

16:9 is 1728x972 on 60D. The mod32/64 restriction is no longer there.

Set 1280x720 from Canon menu and see what you get. The image is usually stretched, you need to find by how much.

The good part is that sensor window is larger in raw, so you need less jumps to cover a full picture.

https://docs.google.com/spreadsheet/ccc?key=0AgQ2MOkAZTFHdFFIcFp1d0R5TzVPTVJXOEVyUndteGc#gid=5

On 5D3 and also 6D IIRC, you would need only 2 columns and 3 lines to cover the entire frame if my theory is right.

Waveform is computed from jpeg YUV only. For raw video and photos, use raw histogram, raw zebras and maybe auto ETTR.

You can allocate a large contiguous buffer and fill it with edmac_copy_rectangle while moving the 5x crop window.

Lookup FEATURE_LV_FOCUS_BOX_SNAP_TO_X5_RAW - that math will help you figure out where you are (e.g. what part of sensor is covered at a given time and how much you need to move the box for the next position). If that code is portable (hopefully), this means silent pic matrix can be implemented without hardcoding the offsets for every camera.

Or it's just wishful thinking?

The values change depending on video mode; 1023 is just a rough starting point. On the same link, it says that black point is found by averaging the black borders from the image (which ML does).

On 60D I expect it to work quite well, because focus estimation updates quickly. It worked great on 5D3, I just move the focus ring back and forth a few times, and then it saves the sharpest pics only.

Post some samples.

Yeah, the HaCKeD sort is slowing it down. Feel free to optimize it a little.

QuickSort or HeapSort should be pretty good. Or some other algorithm with O(n log n) complexity.

What kind of keywords?

(I don't use Lightroom, do you know any open source app that uses these files?)

Here's my hack (in setup_buffers):

Code Select Expand


    PauseLiveView();
    msleep(1000);
   
    mem_suite = shoot_malloc_suite(0);
    if (!mem_suite) return 0;

    ResumeLiveView();
    msleep(1000);


=> got 32x6

on 6D I had to put bvram_mirror in shoot malloc before it would free it up

Makes sense, the 60D triggers ERR70 when you change NR, if something is allocated in shoot_malloc.