500d development

  • 2487 Replies
  • 415089 Views
*

alins

Re: 500d development
« Reply #130 on: 11 / February / 2011, 14:04:18 »
Advertisements

*

alins

Re: 500d development
« Reply #131 on: 11 / February / 2011, 14:08:26 »
RoaldFre: my dev environment is on ubuntu (whatever latest stable release is, i never remember those stupid names) and i set it up using exactly what Alex posted, the 550 build instruction. use "summon-arm-toolchain" and it should do everything for you. I remember that i had to modify the script a little bit tough, will try to dig that up tonight and post it here just in case.

Re: 500d development
« Reply #132 on: 11 / February / 2011, 14:13:35 »
Can you do two (or more) succesive raw dumps of a large piece of memory? If so: look for a large block of data that changes.
A quick and dirty way to do so: I assume you could just "hexdump -C" the binary dump and do a diff on that. If you see a large block that changes: that'll give you the approximate vram address.

Ok, I'll have a look at that summon-arm-toolchain, thx :)

*

alins

Re: 500d development
« Reply #133 on: 11 / February / 2011, 14:14:13 »
Any news on the flickering?

I believe Alex replied in a previous post that he looked at this problem for 550d and his fix is crashing the camera, so i just put this problem at the bottom of the priority list. I will like to fix the flickering,  kinda drives me nuts a little bit, but i got used or at least i pretend is not there. I'm focusing right now on figuring out the vram so we can get zebras working, a major milestone in my opinion.


*

alins

Re: 500d development
« Reply #134 on: 11 / February / 2011, 14:28:01 »
Can you do two (or more) succesive raw dumps of a large piece of memory? If so: look for a large block of data that changes.

my understanding of the whole memory dumping mechanics  is a little bit spotty, so my apologies in advance if i say something totally stupid (of course please point it out):

there's code in zebra.c that does this, takes two dumps of memory and write them on the card. it just needs to be activated. The code start taking the dumps beginning from  a certain address (i believe is 0x1000000??) and it also takes as parameter a segment number.

i got mad last night and i change the code to start dumping from 0x0 (ya' i know!) and the camera crashed with error 70 or something like this. a very interesting error message where the camera tells you something is screwed up, displays total shutter releases and a bunch of other thing and then reboots.

i kinda wonder if it would help in any way to dump the _whole_ memory on the card, just get the whole damn thing twice than start poking. 

*

Offline a1ex

  • *****
  • 671
  • ML dev
Re: 500d development
« Reply #135 on: 11 / February / 2011, 14:36:44 »
See the tutorial here:
http://magiclantern.wikia.com/wiki/VRAM/550D

There's a graph with some vertical bars: those tell you where to look. What you see is a plot of the difference between two successive dumps (where the image data changed a bit due to noise or camera movement).

I had trouble reading from 0xCxxxxxxx segment (camera freezes). 60D bypassed it without creating the file (strange).

Hint: 1xxxxxxx and 4xxxxxxx contain the same data (on 550D) but with different caching bits. Other segments are mirrored in this way too.

The parameter for that dump routine is the first digit of the address (0 to F). Each segment has 256M.
« Last Edit: 11 / February / 2011, 14:40:33 by a1ex »

Re: 500d development
« Reply #136 on: 11 / February / 2011, 14:45:49 »
Any news on the flickering?

I believe Alex replied in a previous post that he looked at this problem for 550d and his fix is crashing the camera, so i just put this problem at the bottom of the priority list. I will like to fix the flickering,  kinda drives me nuts a little bit, but i got used or at least i pretend is not there. I'm focusing right now on figuring out the vram so we can get zebras working, a major milestone in my opinion.

Yes i know, i'm not worried if it needs to be that way for a long time. Don't get me wrong, i appreciate all the hard work you guys are doing :)

*

alins

Re: 500d development
« Reply #137 on: 11 / February / 2011, 14:56:13 »
i followed those instructions alex, and yes sometimes i got the vertical bars. Sometimes the graph plotted was 3/4 solid filled...not sure why. I wasn't very methodical about what i was doing, just trying different things and see what works what doesn't. I will have to repeat the process in keep tracks of what i'm doing.

and yes, i always got two files of 256M.

i'm obviously doing something wrong here because no matter what i tried i still end up with some variant of that grey noise image.

question: i just need to be in live view (in any mode) before starting the dump right?. do i need to be in movies mode and/or start movie recording before the dump?.


*

Offline a1ex

  • *****
  • 671
  • ML dev
Re: 500d development
« Reply #138 on: 11 / February / 2011, 14:57:19 »
Just in LiveView.

*

alins

Re: 500d development
« Reply #139 on: 11 / February / 2011, 15:01:57 »
quick question (apologies, forgot to ask before):

in this step: s = readseg("./0.BIN", x, 1000000)

1000000 is this the value of the start segment or just const 1000000 ?.

 

Related Topics