Messages

I've compiled adtg_log and it loads ok but I can't get it to save a log file on the 50D. Any ideas? 

Do you see the cmos values running on screen (not zeroes values) ?

Very good!

I have some toughts :

1 - these adtg registers are yet indirectly used for fps-shutter speed override and they have the same limitations (low fps still equals to low shutter speeds)

2 - resulutions can only be decreased but not increased (as for latest tests), and lower Canon in-camera standard resolutions of 720p and 480p are more than enough

So, the only real useful discovery (so far) is the CMOS16 register (only in newer cameras models) to obtain dual iso feature.
It seems we cannot squeeze nothing more out of them, at least for now, right?

This is just to know if the works for dumps on other camera models will be worthwile, or if it is better to first study the behavior of these registers on yet dumped models, until something useful will eventually arise.
Or if someone has yet some ideas to implement with these registers.

Try ISO 200, 24 (or 25) as fps, shutter = fps x 2, aperture = whatever to compensate the actual lighting.
The important is that you use a sd card where there is NO ML inside ! Let's test pure Canon mode.

Did you filmed with +4 Ev compensation ?

Like you can see, the noise is perfectly aligned with the normal values for 550D, so what you have in your videos is not "normal" noise, like I supposed first time I saw it. I will wait for the normal photos-video frames side-by-side to judge them...

C'mon boys... HDMI port on tablets and pc's are intended only for output (from the device to the external world), and never for input ! 

.. But i also know, that the Eos 550d is able to record in Very good quality.
But i want to ask : is it possible my Sensor Have a problem?

You can surely bet that, with your shooting parameters above, 550D is way better than your results shown! I can see a constant and uniform granularity that is definitely not coming from normal noise, this is for sure. In your tests try also a raw photo shoot and post the results also. Obviously along with all the parameters you used.
Definitely yes: it can be a sensor problem, or a side component as well... only tests will dictate something about... please: shoot also a photo as this and post the result: iris as closed ad possible, exposure as fast as possible, lens cap on, I wait for your images!

Allright! Ill check this that evening the same time.
Using main photo-function the quality will be perfect.
Do i have to reduce main picture resolution to the same as the video had?

As i saw ML has also an ISO-Value for analog ISO-configuration. Maybe some settings are wrong?

I checked a lot low-lighting-videos on youtube and was wondering how good the results were using the main video-recording-feauture (without ML). Or do you think in most case people used such denoise like neat-video to optimize their clips?

Resolution of photo and video doesn't matter, shoot it at whatever resolution you want. No problem.
For these tests, use Canon without ML, so ML is kept out from the variables for this time.
No, no denoise filters in post, and you must obtain much more quality with much less noise with the parameters you used !

It's impossible ! It's awful ! Far too much noise indeed ! Make this test: shoot a still pothograph with the same parameters and lens and subject and time of day and location, then post the side-by-side photos of the still and a frame of the video: lets see if the stills have the same noise of the video, so we can trace down the problem...

For 550D, the address for ADTG function is at 0xff27ee34, and for CMOS, 0xff27f028.

I didn't try these numbers, and I have no idea how Indy got these RAM addresses. Afaik, the 550D runs all the code from ROM (but there may be exceptions).

Great ! Many thanks A1ex ! It works !

So, the trick is to find the nearest point around the strings (not the exact string start address) where there is the following instruction :

[1] For CMOS (it is before the strings) :       STMFD      SP!, {R4-R8, LR}
[2] For ADTG (it is before the strings)  :       STMFD      SP!, {R4-R10, LR}

Do this is correct ?

And for CARTRIDGE registers and Others (like the first strange two lines found by Indy) do it will be similar ?

Do this procedure is correct to use log module ? (with all ML video and helper functions disabled) :

[1] Enter ML menu
[2] Enable adtg from MOVIE menu
[3] Exit ML menu
[4] Start video recording 1920x1080 25p
[5] Stop video recording
[6] Enter ML menu
[7] Disable adtg from MOVIE menu
[8] Shut off camera

I obtain a "ADTG.BIN" file inside the photos folder on SD.

Do this procedure is right ?

Whoa ! I managed to solve the problem ! Now module loads ok ! 

I'm searching for a fast method to transfer from virtual machine to usb memory, because everytime I insert the usb, it's immediately taken by Windows, and if I take it from virtual machine "devices" - "usb", it disconnects from Windows but when it riconnects, 99 times over 100 is re-taken by Windows again... it is quite frustrating... 

I think problem is inside the following function : "static unsigned int adtg_log_init()" where it blocks itself.

It calls "gdb_setup()" function and then sets a sequence of 4 watchpoints.
It could be the gdb_setup function or the watchpoints settings.

I traced it's inside "gdb_setup()" that all blocks.

I tried to eliminate all the declarations "short" by leaving them plain, because this hanged 550D on "boot_hack,c", but this didn't produced benefical effects, and the freezing still occour.

New clues indicates that the block happens BEFORE entering the function "static unsigned int adtg_log_init()".
This is odd: what is called BEFORE module initialization function ? 
I commented out completely all the init routine and placed inside her a well "NotifyBox", but it doesn't shows up at Runtime !
Do I'm choosing the wrong trace down method or do the problem is really before init routine ?

Now I tried the log module that come with Hudson unified: same exact problem. Even by getting rid of "short" declarations !

Now tried to eliminate the adtg_log module from the sd card : the trace module load flawlessy and then says OK near his name !
This is without any doubt a adtg_log module problem!

I'm currently investigating on this...

Side problem : I can see every camera model has different registers addresses, so... do 550D also have his own ? And how to discover them ? 

Ok ! I went to solve the 550D lock-up problem, now the problem is this :

Modules load, but there is no ADTG menu entry in movie menu to press "set" on, and the textual  modules loading log remains over the liveview even if i turn off global draw, and "show console" option in modules menu is off! I tried to change modes with main dial to photo then return to video mode. But no log written into the sd card. How can I use this log module ? 

Used this log module version : https://bitbucket.org/OtherOnePercent/tragic-lantern-6d/commits/all
The ML base is the last official Hudson one.
Followed same steps as user "vroem".
I'm on a 550D, ubuntu virtual machine on Windows.
All compilations are ok
I load manually the modules after entering ML menu the first time


The module loading log on screen is the following :

Scanning modules...
Scanning modules...
Load modules...
   [ i ] load: ADTG_LOG.MO
   [ i ] load: TRACE.MO
Linking..
Register modules...
Load configs...
Init modules...
   [ i ] Init: 'adtg_log'
_

It seems it blocks itself into initializing 'adtg_log' module (and this is the reason the log text remains on), what do you think ?

[11 BIS] Attenzione! Nelle ultime releases è cambiato un riferimento.
             Occorre modificare "makefile.user" dentro alla cartella magiclantern principale di compilazione facendo in modo che
             contenga la seguente linea così modificata :
       
             GCC_VERSION=-4.7.3
   
             Nella versione precedente era senza il segno meno (-) davanti. Ora ci va.

[11 BIS] Attention! Last releases changed a reference.
             Modify "makefile.user" inside magiclantern main folder so that
             you modify this line as follows :
       
             GCC_VERSION=-4.7.3
   
             On the previous version there was no minus sign (-) before the version number. Now it needs to be there.

http://www.magiclantern.fm/forum/index.php?topic=3072.msg60979;topicseen#msg60979

See the last few posts...

Many thanks, Stedda ! It seems I choosed the worst moment to start this work on 550D ! 

Used this version : https://bitbucket.org/OtherOnePercent/tragic-lantern-6d/commits/all
The ML base is the last official Hudson one.
Followed same steps as user "vroem".
I'm on a 550D, ubuntu virtual machine on Windows.
All compilations are ok, but when 550D is fired up, then red solid light and all is blocked. I must slip out the battery to turn off the light.

I need a big assistance, many thanks !   

Maybe here - https://bitbucket.org/OtherOnePercent/tragic-lantern-6d/commits/all

Please, be patient with me: by using this version I have some compiling errors.
I'm using ubuntu virtual machine in Windows, I already successful compile the full magic lantern, but when I launch "make" inside the adtg_log module above, I get this error :

[ CC    ] adtg_log.o
/bin/sh:  1:  /home/magiclantern/arm-toolchain/bin/arm-none-eabi-gcc4.7.3: not found
make: *** [adtg_log.o] Error 127

I made "make clean" before "make". I copied "makefile.user" inside the adtg module dir and also inside "modules" dir, but the error is still the same.

How can I do ? many thanks !

[EDIT] : Nevermind, solved ! The file which module makefilings refers to is  "/home/magiclantern/arm-toolchain/bin/arm-none-eabi-gcc4.7.3" while the existent file is "/home/magiclantern/arm-toolchain/bin/arm-none-eabi-gcc-4.7.3". The reason of this is unknown ! So, a CopyPasta + Renaming did the trick. Strange but true. 

1. I've been seeing talk about "line skipping" and shooting in "crop mode". Does this apply to the 550d. If so which is better to shoot with as far as the canon video settings (1920x1080 24p/30p, 1280x720 60p, or the 640 modes)? I'm experimenting with the RAW resolutions to find the best it can output to, I like 960x540, and 1280x426 @ 24fps.

Yes, it applies to 550D too. Crop mode (shooting by engaging first 5x zoom mode) is best, because there is no line skipping as in other modes, but it's very zoomed in, so there are pros and cons.

2. moire! I'm getting alot, any thoughts?

You can slightly defocus in post (worse method) or use a filter on the sensor (best method). You can search for "vaf filter" in internet, and on the post where all the cameras ML capabilities are listed: open the spreadsheet and read on the bottom right, near to the legend. Using crop mode (aid method) aids to mitigate the problem. Also, higher the resolution, less the moirè effect.

3. Kinda related to the first question. It seems people are zooming in (x5) in liveview and then recording. Does this apply to the 550d, or the other higher end canons?

Yes, this applies also to 550D and produces "crop mode" results as per answer [1]

I made the address write to the log so now its relatively easy.

Do we must consider using the G3GG0 debug module here : https://bitbucket.org/g3gg0/magic-lantern/src/6de0d9e9a4d4db70b61149f9c212ba02a7dfd18a/modules/adtg_log?at=unified
or do we must consider using the official one at : https://bitbucket.org/hudson/magic-lantern/src/tip/modules/adtg_log/

What is better ? And where are your latest modifications you are talking about ? Many thanks !

It would be of help for all if someone can explain better, with an example for instance, how to find what registers do what. For example by explaining how you have found the right registers and values to implement the dual iso feature.
This will enable all of us with a minimum programming abilities to help digging out this ADTG discovery.

My DNG files are 8 bit?  Is this normal for the t2i output?  Just a minor question and I'm not *really* looking for an answer in this thread. 

They are exactly 14 bit of real bit depth. You can treat it as 16 bit because there we go at powers of two, but the real depth is 14 bit

I imagine to avoid moire you would need to get an anti aliasing filter. the 5d mk3 already has one built in I believe, so the shots look sharp and there is no false color.

From what I understand, I could be wrong.

Not exactly. Every dslr has an antialias filter built in, but it serves only for photographic full resolution. The lower video resolution needs a more powerful antialias filter as an add-on. Mk3 has less moirè in raw video due to the fact that the max resolution is higher than all the Others cams can do (in raw video mode). More resolution = less moirè.

This thread will keep open in order to discuss 550D raw development, but... now that the nightly build is integrated with the raw_rec module, do the builds linked in the first post have still reason to stay ?