5D2 RAW video Builds 14-Bit

[ted ramasola]

also, turning OFF or ON the Global draw in overlays tab also trigger it consistently.

[a1ex]

Narrowed down; CONFIG_TSKMON is the culprit.

(and yes, I had it enabled by default in my local builds, along with GDB)

edit: the tskmon hooks themselves are OK (they don't trigger the bug), so it must be something in the tskmon code.

[ted ramasola]

Thats great. I also just checked that even the feb 10 nightly has that sensor cleen bug.

also, I find that the FPS 1 and lower is still buggy, 1.5 FPS and upwards are ok.

[a1ex]

It seems to be a timing issue in tskmon's stack checker (if I do anything that takes too long, even just a longer bmp_printf, it locks up at shutdown).

It works for me, but it's not exactly a clean fix: https://bitbucket.org/hudson/magic-lantern/commits/89fdaf6a27be18cccc94348e3f1d18a137c79367

(it only patches the symptoms, but doesn't fix the underlying problem; maybe g3gg0 has some insights)

Updated the nightly build, can you try it?

[a1ex]

If anyone can reproduce this kind of corrupted frames, please let me know:



(I might have a fix)

[ted ramasola]

It seems to be a timing issue in tskmon's stack checker (if I do anything that takes too long, even just a longer bmp_printf, it locks up at shutdown).

It works for me, but it's not exactly a clean fix: https://bitbucket.org/hudson/magic-lantern/commits/89fdaf6a27be18cccc94348e3f1d18a137c79367

(it only patches the symptoms, but doesn't fix the underlying problem; maybe g3gg0 has some insights)

Updated the nightly build, can you try it?

yes, tried it feb12 nightly. the cleaning sensor freeze is gone. Will use it longer in my tests today to be absolutely sure.

[cmac]

If anyone can reproduce this kind of corrupted frames, please let me know: http://www.magiclantern.fm/forum/index.php?topic=5763

(I might have a fix)

Have You tried 1920x1080 with fixed framerate to 23,97 @ crop mode 10x? I just bumped on this issue yesterday doing some tests. It record MLV @ 5x with no problem. And on my test @ 10x it gave me a completely unusuable sequence - most of the video was a mess. I deleted it, but I can make another test if You can't reproduce it.

Just to add - Your work is amazing! I feel like entering a toy store each time I call the ML menu :-)

[bnvm]

Raw video does not, has never, and probably never will work properly in 10x mode, so don't use it while recording.

[a1ex]

I'm not talking about the 10x crop issue, I'm talking about the issue from that link.

[ted ramasola]

If anyone can reproduce this kind of corrupted frames, please let me know: http://www.magiclantern.fm/forum/index.php?topic=5763

(I might have a fix)

The link says the topic is either missing or off limits.

[piloui]

I confirm .wav files from .mlv are a bit longer than video files. Perfect sync, however it can be a problem when importing multiple sequences on a same time line, each audio file adding a slight shift. (5D2, 2014/01/14 2b804e8cdd3c )

[reddeercity]

Try the new mlv+audio feb 4th build from a.d.
I thinks the audio length is a little close to video length.

[ted ramasola]

Testing Feb 14 build by a.d.

1. no pink frames on a series of tests, including crop modes

2. Audio meters now reflects weather sound module is On or Off

3. Audio meters now shows in FPS override mode, when sound module is on.

4. tested to record full 64gig cf card with sound.

5. Tested audio sync with video on a full 32 gig CF card. Audio remain in sync with no drift in the 7' 20" clip.

6. Record start/stop delay.
Previous builds like feb 10 (did not test 12 for delay) Had almost instant start recording with 1X: 3 centiseconds, 3xcropmode: 15 centiseconds.

for this build Feb 14, both 1X and 3X crop mode has a longer delay of 1.30 seconds

7. Low FPS frozen frame is still not fixed.

[a1ex]

1. I didn't apply the pink frame fix for 5D2, were you getting pink frames before?

2..5: yay!

6. raw or mlv? any preallocation tricks enabled?

7. it's still a mystery

[ted ramasola]

1. I didn't apply the pink frame fix for 5D2, were you getting pink frames before?

2..5: yay!

6. raw or mlv? any preallocation tricks enabled?

7. it's still a mystery

1. I havent had problems with pink frames with the 5D2 in a while, a quick check on my test lists show it was Jan 13 I last encountered it only when HDMI is attached.
 I just included it since its problematic on the 7D.

6. prealloc tirck or the reserve card space was OFF.

7. hope this could be trouble shooted since I often use this for night cityscapes.

[reddeercity]

Test the Feb 14th build, from A.D. hack crop view mode & the non hack crop view mode.
We have Audio meter now with frame over ride on,
and a "reserve card space" feature so we don't loose audio on full card records to stop.
There's at lest a 3 second delay now when you push the set button to record raw, was not there
on the feb4th build. Is this a problem now ? Or is there a change in the way raw record works ?
Lost all informational overlays/text box on Lv when recording Raw.
Just a little icon with only time as information , I don't like that, there is no way of judging
how much video you recorded at what write speed, File size GB? so you don't know when raw will stop recording or how much you just recorded.
I Would like to see a option to switch back to the text information box with buffer bars instead of the little camera icon.
I switch on "show buffer graph" , this dose not work it just keep that little move icon, no buffer graph.
I have notice when you record a short clip then stop, the buffer bar shows for a spilt second .
The "Global Draw Off" only work in 2 preview mode, HDMI & ML Grayscale, in all other preview
there is a Black-box overlay.
I also notice if you use the hdmi Preview without a hdmi device attach you get wrong framing.
So if you do not want the black box for framing ( as I Do) the White Framing box is only correct for hdmi device.
If you press the delete button while recording raw you can remove the black box for framing to a dark Grey translucent overlay.
It would be nice to have that feature to cycle though overlays or to kill them completely ,if possible.
Checking files, and notice "mlv_rec.tmp" on CF card, is this the reserve Space ?

For some reason after ever video clip capture ML save a screen shot of the Lv with overlays "bmp"

With the Crop view mode half shutter HI- Res preview in ML Gray preview, work like it should until you start
recording raw then  the hi-res preview becomes all pinky overcast image, feb4th build was not like that it work with or without
raw record running.

Not Sure but I think we have a new frame size or I never notice before , 1792x968 1.85:1
This is continuous for me.
 


 


 

[a1ex]

Most of these look like mlv-specific problems, so I'd say the best place to report them is in the MLV thread.

The FPS issue is 5D2-specific.

I'm also interested in what of these issues are also present in raw_rec. Rationale: if the new builds have introduced issues in both raw_rec and mlv_rec, it's probably something broken in the backend. If it's only on mlv_rec but not on raw_rec, it's likely a mlv-specific issue.

[reddeercity]

Ok I will do that 

[ted ramasola]

Most of these look like mlv-specific problems, so I'd say the best place to report them is in the MLV thread.

The FPS issue is 5D2-specific.

I'm also interested in what of these issues are also present in raw_rec. Rationale: if the new builds have introduced issues in both raw_rec and mlv_rec, it's probably something broken in the backend. If it's only on mlv_rec but not on raw_rec, it's likely a mlv-specific issue.

The FPS frozen frame problem is present in both raw_rec and mlv_rec modes.

Also, as a counter test for redddeercity's start/delay time, I also tested just now with HDMI monitor attached and its similar to the one without. Its a delay of 1" and 9-10 centiseconds.

[reddeercity]

Yes it looks like we do have a new set of frame sizes with 1792x#### 

[piloui]

Try the new mlv+audio feb 4th build from a.d.
I thinks the audio length is a little close to video length.

Are those builds safe ? I'd rather wait a release ^^

( By the way do i have to confess how ML impress me each day ? )

[ted ramasola]

try the Feb 14 build from a.d. its stable enough for most reasonable requirements.

A.d. has missed to update his first post to indicate whats improved and to show he has a new build yet, but just download from the link on the first post.

And just keep checking back on this thread to keep posted with bug fixes and improvements.

[reddeercity]

Are those builds safe ? I'd rather wait a release ^^

( By the way do i have to confess how ML impress me each day ? )

I agree with Ted , try new feb14th build.
But yes , the last 2 builds from a.d. are very safe , in my tests
I push ML raw/mlv to there limits and then I push it over it's limits
To see where it fails, if it dose that is.
So normal usage with raw will never be a problem.
Don't wait jump in and play!

[a.d.]

...
Lost all informational overlays/text box on Lv when recording Raw.
Just a little icon with only time as information , I don't like that, there is no way of judging
how much video you recorded at what write speed, File size GB? so you don't know when raw will stop recording or how much you just recorded.
I Would like to see a option to switch back to the text information box with buffer bars instead of the little camera icon.
I switch on "show buffer graph" , this dose not work it just keep that little move icon, no buffer graph.
..

Me, neither! It doesn't fit with the interface and complicate the on-screen concept.(same at raw_rec). However ML is Opensource, it's an user request. Btw the crop mark is currently a compromise, because there is no crop mark for Global Off.

The fps bug is a quite complicated problem. I still don't fully understand how the algorithm from g3gg0's fps-engio works and  who is the culprit.

Under the hood: It's really cool with a1ex's new memory backend (more memory and less conflict with canon code). g3gg0 optimized mlv_rec+ mlv_snd for more stability.

[SpcCb]

(...)
The fps bug is a quite complicated problem. I still don't fully understand how the algorithm from g3gg0's fps-engio works and  who is the culprit.
(...)

Maybe a way for investigations: the fps override bug is only in raw video; in h264 it's working fine. And this bug was not present _even in raw video_ on builds made last summer (old builds, I agree).