Canon EOS M

[Teamsleepkid]

Probably because it's toshiba they make great shit only 30MBps though need 40 for video..

[Teamsleepkid]

Compressed raw video.. anything going on for the M?

[canoneer]

I updated my EOS M with Nightly Build, acquired a SD card to use the M-zoom-lense without shutterbug and it works.

My humble question now, probably because I have not used ML for several years - the Focus racking and stacking tools seems to have disappeared- or might they not be available on the M - had it on my 60d last I tried?

[Walter Schulz]

https://builds.magiclantern.fm/features.html -> Focus
Indeed focus features are not available for EOS M. Maybe dfort or a1ex will jump in and telling why.

[canoneer]

Thanks. Forgot to check the matrix.

[dfort]

The EOSM should have focus capabilities similar to the 650D and 700D but there is a problem controlling EF-M lenses.

https://bitbucket.org/hudson/magic-lantern/pull-requests/748/focus-backend-updates-precise-focus/diff

It does better with EF, EF-S lenses on an adapter.

I haven't looked into turning on those features. According to the source code:

features.h

Code: [Select]

/* Some Hope Yet */
#undef FEATURE_TRAP_FOCUS
#undef FEATURE_FOLLOW_FOCUS
#undef FEATURE_RACK_FOCUS
#undef FEATURE_FOCUS_STACKING


[cantsin]

Having made two videos in EOS-M 720p raw crop mode...

...I'm left with the following practical questions:

1) The video crop mode setting isn't remembered by the camera even after ML parameters have been saved. It needs to be manually reactivated in the ML video menu each time the camera has been turned on. Is there any way around this?

2) Sound recording with the mlv_snd module isn't 100% synchronous, because the audio track usually runs a bit longer than the video clip. (I tested this with two different copies of the EOS-M, and the behavior is consistent and reproducible.) While this shouldn't be an issue, it seems to throw off MLV-to-CinemaDNG converters. Raw2CDNG doesn't correctly decode the sound so that audio tracks end up being fragmentary and out-of-sync. This can be fixed by extracting .wav files from the MLV files via ffmpeg and manually aligning them to their corresponding video clips in the editing timeline. But is there a way to convert MLV created by the EOS-M to CinemaDNG without these issues? [I'm primarily a Windows user and couldn't get MLVFS to work.]

3) We keep reading great things about MagicLantern's new 10bit and 12bit raw recording options. Are they available (or in the pipeline) for the EOS-M at all?

[Teamsleepkid]

you can get 10 bit and 12 bit magic lantern under downloads downloads again then experiments

[Licaon_Kter]

> 1) The video crop mode setting isn't remembered by the camera

That's there for a reason: https://bitbucket.org/hudson/magic-lantern/pull-requests/776/disables-crop-mode-hack-when-turning/diff

[dfort]

There's an open pull request to save the "Movie crop mode" setting in a preference file and restore the last used state.

https://bitbucket.org/hudson/magic-lantern/pull-requests/787/crop-hack-config/diff

I haven't tried it in a while but it looks like maybe it doesn't work on the other cameras that can do that trick (650D 700D 100D).

[dfort]

So I'm searching for EOSM2 stubs using similar cameras as reference (700D 100D EOSM) and the msg_queue_create stub on the EOSM doesn't seem to be matching what the other cameras are using:

Code: [Select]

/** Message queues **/
NSTUB(0xFFA88414 - RAM_OFFSET,  msg_queue_count)            // 0x1EE84 Above "!!!!!!! QUE Overflo"
NSTUB(0xFFA73C84 - RAM_OFFSET,  msg_queue_create)
NSTUB(0xFFA883D4 - RAM_OFFSET,  msg_queue_post)             // 1EE44
NSTUB(0xFFA881E8 - RAM_OFFSET,  msg_queue_receive)          // 1EC58

Shouldn't it be?

Code: [Select]

NSTUB(0xFFA880D8 - RAM_OFFSET,  msg_queue_create)

Not sure how to check this.

[a1ex]

0xFFA73C84 is a wrapper for 0xFFA880D8 (the only difference being at error handling, which never happened within our usage - yet).

For consistency, it's best to use the same stub on all models. Probably a good candidate for automation (the algorithm is in ARM-console, but that one does have some weird dependencies...)

[dfort]

Also noticed that EOSM still has the old post_isr_hook/pre_isr_hook in unified. This was changed in the dm-spy-experiments branch with commit b1bd219.

EOSM.202/stubs.S

Code: [Select]

/** Interrupts **/
-NSTUB(0xFFAA775C - RAM_OFFSET,  post_isr_hook)
-NSTUB(0xFFAA7758 - RAM_OFFSET,  pre_isr_hook)
+NSTUB(0x3DD6C,     post_isr_hook)
+NSTUB(0x3DD68,     pre_isr_hook)


[dfort]

Pretty sure I caught another issue, this time with an EOSM constant.

magiclantern/platform/EOSM.202/const.h

Code: [Select]

#define YUV422_LV_BUFFER_DISPLAY_ADDR (*(uint32_t*)(0x3E650+0x11c))

Should be:

Code: [Select]

#define YUV422_LV_BUFFER_DISPLAY_ADDR (*(uint32_t*)(0x3E650+0x118))

Anyone care to double check? It is after *"call ChangeVramCallBack pFlipCBR=%x"

Compared it with the code for 100D, 700D and 550D.

[a1ex]

On 5D3 1.2.3's consts.h there's a better hint - near ImageEffect. That one would give 0x3E650 + 0x118.

Could this be what's missing for the so-called display filters? (you could try enabling them in internals.h, should be the same as on all other D5 models).

[Licaon_Kter]

Not a ML issue but..

How does one treat "Change battery pack" error? (I guess this is the message, I get the localized text)

Had it at 100%, shoot a bit yesterday (still fullbar), forgot to charge it last night, but I put it in the charger in the morning for 15 min, charger led still orange.

I had to go, put it in the camera and went outside.

After more than an hour of walking outside in sunny weather, trying to take a picture I get the surprise messag e with a red empty battery icon AND the ML message of "bad shutdown modules not loaded" (as expected).

This happens without SD card too.

I don't have a collapsible 15-45, just plain 18-55 and 22.

Original Canon battery that came in the box.

/LE: False alarm, got it charged up to green and now it starts fine. Got me scared there for a bit.

[dfort]

Could this be what's missing for the so-called display filters? (you could try enabling them in internals.h, should be the same as on all other D5 models).

Tried it, Display > Anamorphic and Display > Defishing are working. Is there anything else that should be checked?

Updated stubs pull request.

[a1ex]

Yay! This one was broken for a very long time.

Raw video previews also depend on this backend, and also HDR video should no longer flicker while recording.

There's also FEATURE_MAGIC_ZOOM_FULL_SCREEN, but that one doesn't work on 700D/650D, so I don't have high hopes for the M.

(We really need some tests for all these constants, to catch such mistakes early)

[Danne]

Raw video previews also depend on this backend, and also HDR video should no longer flicker while recording.

[dmilligan]

Nice work @dfort!

[dfort]

Thanks @dmilligan. You taught me many lessons.

There's also FEATURE_MAGIC_ZOOM_FULL_SCREEN, but that one doesn't work on 700D/650D, so I don't have high hopes for the M.

Just tried it--still doesn't work. Also didn't get rid of the shutter-bug, solved the mv1080 problem or suddenly started recording in 4K on the EOSM.

[Teamsleepkid]

nice dfort! always wondered what the black and white preview everyone was talking about was because id never seen it. nice to be able to frame in 5x crop mode. any chance this could be put into the old experimental builds? currently can only be used on 14bit. could it be put into the compressed raw and crop mode builds on experimental page for 10 bit and 12 bit?

[dfort]

@Teamsleepkid - If you really want to live on the cutting edge set up a development environment and learn how to merge branches. I just updated my Macintosh and Cygwin tutorials. Note that there is currently a gotcha on the Cygwin version that should get fixed upstream but it is easy enough to work around for now.

[Teamsleepkid]

yeah i figured something like that was coming. honestly I'm more into cameras then into coding. appreciate what you do a lot though. you're always killing it. same with the others. appreciate all your hard work on the eos m. its a great little camera now.

[Teamsleepkid]

looks like it made it into one of the new experimental builds. very cool. so being able to frame in 5x zoom mode is super useful. anyone ever notice? or am i way off base here? i have no scientific proof but...5x crop mode seems to have tighter grain? makes no sense. and also seems less noisy for sure. maybe a little brighter? it definitely looks better than regular "crop mode". try it. 5x crop mode is cleaner in low light i swear to god. I've done lots of tests. i do them for myself though and i always delete them. just looking for someone else to corroborate this claim..