Danne's crop_rec_4k experiments for EOS M

[masc]

Reverted to an older build. Worked. Installed again. Working now. WTF.

Tested all 6 MCM rewire combinations between AR and x3 zoom: all working with dfort's focus pixel file. GREAT!

Positive side effect with MCM rewire: AUTO-ISO is working really good here.

[Danne]

Great! Mcm touching prop regs. Seems stable enough though. One really have to abuse some code watch guards to soft brick. Believe me, been there. A1ex fixed it in half an hour though  .
Autoiso. Gotta test. Never used it.

[MADagain]

Is there a stable build for the EOS M?

[DeafEyeJedi]

Great! Mcm touching prop regs. Seems stable enough though.

Agreed. Been playing with this all afternoon and so far I've been impressed. Still trying to push it's limits.

Autoiso. Gotta test. Never used it.

Same here. Just tried and I can see what @masc is referring to... especially when utilizing the EF-M 18-55 STM. That lens is a fun toy to play with. So light!

Is there a stable build for the EOS M?

Why don't you just give a try at least with today's build from @Danne's bitbucket page? 

[dfort]

Maybe should have a talk with the "dotman" dfort...

I just finished going through all the sample files I could find for the EOSM and it looks like there's "only" a half-dozen of these new crop_rec settings that aren't quite working 100% yet with the new map files. I've been pushing to the repository so if you want to give it a try (in MLVFS) please do so.

Off topic: The 100D tests are 100% at this point.

[Danne]

Great!
By the way. Is it possible to do the same kind of testing but from Mlv App instead?

[dfort]

Not sure. Look in in MLV-App-master/src/mlv/llrawproc/pixelproc.c around line 318 and you'll see there's code there to use ".fpm" files but I'm not sure how it works. The plan is to re-write the code to create the map files as needed for these new crop_rec settings.

[masc]

Is it possible to do the same kind of testing but from Mlv App instead?

This is what I do.

Not sure. Look in in MLV-App-master/src/mlv/llrawproc/pixelproc.c around line 318 and you'll see there's code there to use ".fpm" files but I'm not sure how it works.

MLVApp searchs for fpm files first. If it doesn't find one, it uses internal code. If it finds a fpm file, it uses the file. So just copy your fpm next to the executable (.exe on Win, executable inside App package on OSX) and have fun. This will override internal code for this setting.

[Danne]

This is what I do.
MLVApp searchs for fpm files first. If it doesn't find one, it uses internal code. If it finds a fpm file, it uses the file. So just copy your fpm next to the executable (.exe on Win, executable inside App package on OSX) and have fun. This will override internal code for this setting.

Bloody ace! Great to hear and have.

[IDA_ML]

The 100D tests are 100% at this point.

Dfort,

I did test quite extensively your FP maps from February 20-th on the 100D and also performed several stress tests in all modes as you requested.  Did not find any problems with them.  They all seem to work quite well with MLVFS. 

I will do more tests with the latest reworked FP maps that you posted yesterday.  But this will take some time.  Please be patient.

[dfort]

Ok--all sample MLV's are working with the latest map files. Feel free to continue testing focus pixel free. Well, at least until you come up with yet another setting that needs attention.

[uy scuti]

Friends, tell me who knows, I can not set a long exposure at 25 fps, 1/60, it can be done but only at 25,002 fps, what am I doing wrong?
shooting mode:mv1080p 1736x1158
Sorry for my English

[Danne]

 Ah, handcrafted map files. Appreciated. Will run tests soon. Thanks.
Here are test files if anyone wants to have a look:
https://bitbucket.org/Dannephoto/magic-lantern/downloads/focus_pixel_MLV_all_presets_eosm.zip

There could be one or three files not working since I changed offsets a few times. Most files should work.

[dfort]

The only one that still has a problem is 01_anamorphic. Strange, worked on that one and thought I had it. The map file fits perfectly so something weird is going on.



[EDIT] Just tried it with with dcraw and the map file is working perfectly. Anyone care to try it out? From Danne's samples in the focus_pixel_MLV_all_presets_eosm.zip package located on his Bitbucket downloads page look in the first_batch folder and run this on the 01_anamorphic.MLV file:

Code Select Expand

mlv_dump --dng --no-fixfp 01_anamorphic.MLV file
./mlv2badpixels.sh -v 01_anamorphic.MLV
dcraw -T -P EOSM_anamorphic_1664x2126.txt 01_anamorphic_000000.dng

The mlv2badpixels.sh script is located in the ML Focus Pixels repository in the scripts_archive directory. It takes the full raw buffer map file and trims it to the same size as the image so it can be run through dcraw. Note that I modified that old script to recognize the new crop_rec settings as best as possible using what we know about raw buffer dimensions.

In any case, why isn't this map file working in MLVFS and MLV App?

[EDIT 2] Oops -- my mistake. I must have not updated the focus pixel map file in the apps before testing. All good, all settings are now focus pixel free. You can still try the dcraw exercise, interesting how different apps can use these map files.

BTW--Kudos for the MLV App developers for supporting .fpm files and properly handling 1x3 sampling.

[dfort]

Going through all these new crop_rec settings for the EOSM there is one really interesting setting @theBilalFakhouri put into his 700D version that uses the entire sensor area as the full raw buffer. I made a fpm file for this setting though I haven't seen a sample on the EOSM yet. This post just jogged my memory about that setting.

[MADagain]

Why don't you just give a try at least with today's build from @Danne's bitbucket page? 

Link... please?

I tried the current nightly and for what I need like magic focus or the overlays... don't seem to work at all.

[DeafEyeJedi]

Link... please?

Perhaps try scrolling up two post above yours.

and please bookmark his Bitbucket downloads page while you're at it.

[Danne]

@dfort
Very nice work dfort. Tested my files regarding focus pixels and all seemingly worked. Next up some real life testing.

[masc]

Link... please?

I tried the current nightly and for what I need like magic focus or the overlays... don't seem to work at all.

In version from Danne magic focus and overlays work fine. I use them every day.

[Walter Schulz]

Friends, tell me who knows, I can not set a long exposure at 25 fps, 1/60, it can be done but only at 25,002 fps, what am I doing wrong?
shooting mode:mv1080p 1736x1158
Sorry for my English

Build used?

[Danne]

New build:
https://www.magiclantern.fm/forum/index.php?topic=9741.msg208959#msg208959

Managed to eliminate a few stubborn hot/cold pixels on my screen by tweaking cmos 4.

Code Select Expand

/* all presets */
if (is_EOSM)
{
/* hot/cold pixels. Usually 0x2. 0x34 to be tested */
        cmos_new[4] = 0x34;
}

Not sure how this will affect other eosm cameras but please test and see. The change is applied to all presets. If you have hot/cold pixels on your screen they might dissappear now. If you don´t have them check so nothing breaks the recording(should be safe).

cmos 4 0x2


My three hot friends.


Not easy to see but looks like this:


After the change to cmos4 was applied:
cmos 4 0x34

[uy scuti]

Build used?

Used build Danne:
EOS M ML Build 4K Crop Rec 1x3 12th Feb

[Larouso]

Not directly. Download 80000331_1872x1190.fpm from dfort's repos and copy it next to executable (.exe on Win, on OSX executable inside app package). MLVApp will recognize this file and use it.

Hi everyone,
just received the eos M last weekend - and spent already countless hours here reading through the stuff You are doing such great work. Thanks a hundred times to everybody contributing so far!

My question is:
I downloaded the 80000331_1872x1190.fpm from dforts bitbucked, placed the file next to the exe. Now whenever I open up a 1080p MCM rewire clip, the video in MLV app stays black, as soon as I hit play it crashes.
Other crop modes are not affected - so it seems to me map file is recognized, but something is going wrong then. Am  doing something ovious wrong?  Used Dannes latest build, but it`s the same with the other ones that I already tried.

Thanks for any hints.

[Danne]

Could you upload a sample MLV that you can confirm is crashing also after you shorten it?

[dfort]

This reminds me, a few (or is it just one) of the samples will trigger a segfault on mlv_dump when creating dng files with or without the --no-fixfp option. Same file works fine on MLV App and MLVFS. Got to dig through the files to find the trouble file.

[EDIT] Here are the problem files:

Code Select Expand

MLV Dumper
-----------------

Mode of operation:
   - Input MLV file: '013_4K 4038x2558.MLV'
   - Using output path '013_4K 4038x2558_' for DNGs
   - Convert to DNG frames
   - Output into '013_4K 4038x2558_'
File 013_4K 4038x2558.MLV opened
File 013_4K 4038x2558.M00 not existing.
Processing...
Segmentation fault: 11

MLV Dumper
-----------------

Mode of operation:
   - Input MLV file: '015_4K 4038x2558_169.MLV'
   - Using output path '015_4K 4038x2558_169_' for DNGs
   - Convert to DNG frames
   - Output into '015_4K 4038x2558_169_'
File 015_4K 4038x2558_169.MLV opened
File 015_4K 4038x2558_169.M00 not existing.
Processing...
Segmentation fault: 11


I also wrote a post on how to get rid of your hot pixel "friends" in post -- by modifying the focus pixel map file.