[550D] Silent pictures with higher resolution
Alex
(see this discussion: http://vimeo.com/groups/magiclantern/videos/18983824 )
Here's a (more or less) stable build which has this. To use it, press
Q on the silent picture option in ML menu, and then SET/DISP to adjust
the resolution.
How it works:
- It will take a matrix of small silent pics, in zoom x5 mode.
- All those pics are saved in the same 422 file
- You need to have the camera on a tripod and the scene should be
almost stationary (a pic is taken in a few seconds).
- Works with intervalometer and HDR bracketing modes.
- 422-jpg.py from this zip is able to decode them (and works
much faster on regular 422 files).
- 422-jpg.exe (6 MB) is here:
https://bitbucket.org/hudson/magic-lantern/downloads/422-jpg.exe
There may be some cache problems, so if you get the old file, try
another browser.
Other fixes in this build:
- HDMI support is a bit better (see "Ext monitor fix" thread for details)
- When you change a setting from Expo menu in LiveView, all the other
settings will disappear from the screen temporarily.
- Some minor stuff here and there (like spotmeter moved in the center,
silent pics working in almost all modes...)
JaKoB
Alex
New link for 422-jpg, without caching issues:
https://bitbucket.org/hudson/magic-lantern/downloads/422-jpg-v2.exe
On Fri, Jan 21, 2011 at 11:38 PM, JaKoB <jako...@gmail.com> wrote:
> everything works real nice but when the HDMI is plugin the Camera locksup
>
> --
> http://magiclantern.wikia.com/
>
> To post to this group, send email to ml-d...@googlegroups.com
> To unsubscribe from this group, send email to
> ml-devel+u...@googlegroups.com
> For more options, visit this group at
> http://groups.google.com/group/ml-devel?hl=en
JaKoB
Ba...@aol.com
na its not working its giving me error 80
--
Ba...@aol.com
Sorry... my trying to moving histogram to the right crashed the HDMI
mode. Fixed now.
Alex
is still buggy. Does it work if you turn them off and then enable them
one by one?
Alex
(something like this: http://people.rit.edu/andpph/text-slit-scan.html )
This is implemented as a third mode of silent pictures. Default
setting should be a good start.
With this, I also have some interesting findings about the video buffers:
- The two buffers we are using are refreshed at 12.5 fps.
- There is a variable which changes at 25fps, regardless of selected
video mode (not sure if it changes with pal/ntsc) at 0x1e24c (found
with mem-spy).
- A loop with msleep(1) will get 100fps.
- To test the refresh rate of a variable, call test_fps from zebra.c
and uncomment the FPS display.
So, one of these is true:
- The buffers we've found are not used for recording, nor for screen display
- test_fps is buggy
And also:
- There's no visual improvement in trying to optimize zebra and focus
peaking (this means they won't be any faster); instead, it will be
better to slow down the zebra in order to save battery power.
- There may be buffers which are refreshed which higher frequency.
Antony Newman
I agree. I reckoned (by sight) the screen was being display about 10-12 frames per second even when I get my task to run at 200 frames a second.
The Frames per second code has some comment to make you wary that it is suspect! (or at least mine has!)
What is missing was a call to the vsync_wait() call in the DryOs (there's 2 or 3 of them depending on how you want to be called).
My first thought on the reason that canon has two video banks is that one is updated, while the other one is displayed.
"The buffers we've found are not used for recording" Did you try and write data into the (uncached address) of the screen segment and see if it is recorded?
AJ
Alex
> updated, while the other one is displayed.
That could be a good hypothesis.
> "The buffers we've found are not used for recording" Did you try and write
> data into the (uncached address) of the screen segment and see if it is
> recorded?
No. But even at 50fps, the buffers are still refreshing at 12.5fps.
Maybe they have 4 buffers in this mode?!
Alex
AJ, you are really good at this!
{0x04000080, 0x0415407C,1360, 2048, 3744}, // seg 1
{0x10000080, 0x1015407C,1360, 2048, 3744}, // seg 2
{0x24000080, 0x2414FFFC,1335, 2048, 3744}, // seg 8
{0x34000080, 0x3415407C,1328, 2048, 3744}, // seg 10
So the conclusion is that my logic is buggy :)
Antony Newman
If code is written to obliterate part of all the vram segments while recording - we should get an answer very quickly.
My guess:
+) Whatever happens, all the data must be sent to the DIGIC for processing
+) And so the is not benefit in having 1 GB of Ram in the DIGIC itself (it just needs a high bandwidth DMA connection)
+) The DIGIC 4 uses a very expensive multi-layer chip fabrication (ie more likely to use it for Hardware Processing).
+) Which leads me to the guess that the ARM configures the DMA controller and the DIGIC.
+) and the DIGIC reads and writes use memory memory that is visible to both to other DMA devices, and the ARM.
AJ.
arm.indy
http://techon.nikkeibp.co.jp/english/NEWS_EN/20090218/165866/
Lionel Davey
> - 422-jpg.exe (6 MB) is here:https://bitbucket.org/hudson/magic-lantern/downloads/422-jpg.exe
Antony Newman
There is not (in my opinion) enough on Chip memory for a complete single frame of a 1080p image.
AJ
Alex
https://bitbucket.org/hudson/magic-lantern/downloads/422-jpg-v2.exe
Lionel Davey
> Link was changed due to bitbucket caching issues:https://bitbucket.org/hudson/magic-lantern/downloads/422-jpg-v2.exe
Next silly question. How do I change the compression with this converter?
Alex
Lionel Davey
Think I might have been missing a dependency.
Was to busy to look into it further though, at the time.
Alex
LiveView buffer (the small one) is updated at 30fps and has 3 banks
(each one updated at 10fps). A pointer to this buffer is at 0x246c
(updated at 30fps).
HD VRAM (the bigger one) is updated at 25 or 30fps (depending on
PAL/NTSC setting) and has 2 banks (each one updated at 12.5 or 15fps).