Danne's crop_rec_4k experiments for EOS M

[dfort]

@baladev - Great, I've got three of these puppies too. All exhibit the shutter-bug. Here's a video I made a while back:

If you don't mind doing some further testing--

Do you have another EF-M lens with IS that you can test? There's a slight chance that maybe the IS isn't working properly on your lens.

Check the capacity of your 128GB card. There were some "fake" cards sold that weren't labeled properly. If it is less than 32GB it might not show the bug.

Finally--the shutter-bug is fickle. At one point it wasn't showing up when the card was freshly formatted in the camera but it eventually came back.

[baladev]

@dfort
Oh, a red one - very sexy! I have a white and two black ones. The white is very pretty (even my wife who's not into cameras said that).

I've got x3 18-55 kit lenses and x2 22mm f2. IS is working properly on this one, I've just tested - tried to turn it off and on and see a very distinct difference in shakiness of picture on the screen. But, I will test two other lenses tomorrow and report back. So, you want me to disable all modules, start in photo mode and take photos?

All my cards are genuine SanDisk stock - I only buy from authorized distributors, plus I get proper speed with them (close to 60 mb/s) and the capacity is what it should be - filled them to their capacity a few times.

Right, I wasn't even aware of this bug until today, when this discussion started, I will keep an aye on it. However, Danny said above "I got the same set up and it´s not happening here." So, I'm not the only one?

[DarkCAL]

@Danne

I am running ML build May16
Below is a picture of my run & gun rig (if not allowed please remove)
And here is a YouTube link to a test video I shot yesterday

Settings used are :
mv1080 MCM rewire
Bitdepth 12
Ratios 16:9
RAW video ON
Resolution 1736x976
Aspect ratio 16:9
14 bit lossless
Preview Realtime
Global Draw LiveView
Magic Zoom HalfS, Med, TL, 2:1
All modules activated
FPS override 24 fps
Image fine-tuning as close as possible to 1/48 sec
Picture style Faithfull

Footage shot using a Rokinon 35mm T1.5 Cine EF
Viltrox M2
T8.0
ISO 400
Available light

All works fine
Footage is awesome !

Only issue is when I use my F570 monitor and if I turn Kill GD OFF using the above setting it gives me pink frames

I know it might or is just a small detail
But when this would be fixed / resolved ML will be running 100% on my EOS-M
And I suppose on many others as well

Thanks again for all the effort you're putting in
Just recently got my EOS-M but really loving it as you can see !
.

[kuchka94]

Using build 16 may, and I can say that sd uhs module doesn't work for me. Speed is near 30 mb. Can't test exactly, there is no bench module.
Jul03 2018 with module + script give me near 57mb.
When I load old module + script on 16 may build, it start to work fine, so I can get continuous footage in RAW.

[DarkCAL]

@kuchka94

Rookie here on a crash course
How can I test the speed of the sd uhs module please ?

Just test running ML build May18
Runs fine
Except for the same problem when using my monitor
Kill GD OFF as posted above in ML build May16 (and the previous builds) gives pink frames
About 4 per minute of footage
With Kill GD ON not one pink frame !

[kuchka94]

I tested that way:
Choose raw footage size and proportions and see demand of the speed in the bottom. For example 36mb.
After start record and see at the top right color of timer, if red - speed is not enough for 36mb/s footage, if yellow - almost enough, green - continuous footage.
So for my case on jul 03 2018 build I had settings for 55-60mb/s footage, and green, times yellow, timer, that allowed me get continuous shoots.
But now I have red and yellow timer, and footage stops on settings 30mb/s.
For me that means that sd uhs module doesn't work.

[Danne]

You can build your own sd_uhs module which likely will fix your issue. Here you go:
https://www.magiclantern.fm/forum/index.php?topic=16040.msg216464#msg216464

[lightspeed]

for the anamorphic mode, would it be possible to have a stretch of .5 instead of the .33 so it doesn't upsample as much.

[masc]

for the anamorphic mode, would it be possible to have a stretch of .5 instead of the .33 so it doesn't upsample as much.

Physically impossible. You can read each column or each third column. Otherwise you'll never read in some colors. Look at a bayer matrix... This is also why there is a 1x1 and a 3x3 mode, but no 2x2.

[lightspeed]

but it is already stretched to 16:9. it just uses a lower resolution. Can we not stretch it to 16:9 but using the resolution in the 2.35 setting?

[masc]

but it is already stretched to 16:9. it just uses a lower resolution. Can we not stretch it to 16:9 but using the resolution in the 2.35 setting?

There is nothing stretched. It is cropped!

[lightspeed]

So it is not possible to turn 1664 x 2126 into 3763 x 2126

[masc]

Have you read and understood this post?

This is the sensor:
RGRGRGRGR...
GBGBGBGBG...
RGRGRGRGR...
....

Now read each column (=1x1=no stretching needed):
RGRGRGRGR...
GBGBGBGBG...
RGRGRGRGR...

Now read each 3rd column (=1x3=needs 3x horizontally stretching in post)
RGR...
GBG...
RGR...

Now read you suggestion, each 2nd column (=1x2=needs 2x horizontally stretching in post)
RRRR...
GGGG...
RRRR...
But where is the blue gone?! ---> impossible.

On the other side, 1664x2126 is already to large to get continuous recording. How do you want to save the double data at the same time?

[lightspeed]

I see. thanks

[baladev]

- Possible fix for corrupted frames in anamorphic modes

For all "anamorphers" out there you may wanna test this version since it seems we can get corruptionless footage. Not that there was a lot of it but still.

@Danne

Tested your latest build - crop_rec_4k_mlv_snd_isogain_1x3_presets_2019May20.EOSM202
Filmed two clips - 10 and 5 minutes long at 1072x1810 25fps 12bit with 4k anam preset.
In the 10 min clip I found 23 corrupted frames in the first 4m20s of it, didn't check after that.
In the 5 min clip I found 0 corrupted frames, the difference is that I was filming during sunset and the second clip was much darker.

[Danne]

Nice with test.
I wonder if the original preset 1072x1900 works with 10bit and 24fps? The 16:9 is really only experimental. I rely on 2:35.1 to see if they can be considered reliable.

[baladev]

Do you mean the 1488x1900 preset you get by running your cinema 2:35 script? I will test it, but it seems to me current anamorphic presets get corruption when they are close to maximum writing speed when data pipeline is saturated and something starts to break. So, one needs to make sure not to go above 60-62mbs or maybe even lower and things should be ok?

[Danne]

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

@Baladev.
Untested:
I did following(ananorphic rewired mode). If selecting 16:9 10 bit you´ll get 1120x1900 but if you select 12bit it will instead do 1064x1800. Hit menu button if changing bitdepth. Test if 12bit is corruption free with 25fps.

[masc]

I did following(ananorphic rewired mode). If selecting 16:9 10 bit you´ll get 1120x1900 but if you select 12bit it will instead do 1164x1800. Hit menu button if changing bitdepth. Test if 12bit is corruption free with 25fps.

Really? But 1164x1800 is 17.5x9 ... is that correct?

[Danne]

Yes, the magic ratio @masc 
Na, sorry typo, should be 1064x1800.

Added also 12bit reduction resolution in 2.35:1 mode so this will jump into 1328x1700 when selecting 12bit(use menu button to activate regs).
Reuploaded a build in above link.

[baladev]

Ok, did a tests with these settings:
crop_rec_4k_mlv_snd_isogain_1x3_presets_2019May23.EOSM202
4k anamorph preset, 1064x1800 12bit 25fps iso100
Bright sunny day, lots of trees and some white roof tops in the frame (can't post pictures for some reason).
Scene underexposed by ~1 stop (i believe vertical lines in the histogram are stops?) to avoid highlights clipping and to reduce data rate.
Two clips - 10 and 5 minutes.
For the first 6 minutes the time indicator was green, then became orange when the sun came out from behind a cloud, making rooftops very bright.
Result - 0 corrupted frames.

[Danne]

Wow. That sounds great. Thanks for the test.

[Jip-Hop]

Bought the EF-M 22mm f/2 STM today and it came with a free EOS M. Now I've got 2! Can't wait to start using ML extensively on both of them ^^

[baladev]

Try turning off all the modules, start camera in photo mode and shoot a still with that kit lens attached. Chances are the shutter won't fire.

I'm now on the latest Danne's build - crop_rec_4k_mlv_snd_isogain_1x3_presets_2019May23.EOSM202
Just repeated the test with all 3 of my 18-55mm kit lenses.

With two I couldn't reproduce the shutter bug by doing this:
1 disable all modules
2 start camera in photo mode
3 take pictures
4 switch to video mode
5 take pictures
6 switch back to photo mode
7 take pictures

The camera focused and took photos fine.

With the third lens I was getting one misfire (when I press the shutter button and nothing happened) at 2nd and 7th steps. First attempt of taking photo didn't work, consecutive attempts worked fine.

A few days ago when I was trying to purposefully reproduce the shutter bug with the original lens, I managed to succeed by messing around in ML menu and doing crazy things like disabling modules and re-enabling them again right away:) Then I would get one misfire and then it worked fine again. One time I managed to get it permanent - all attempts of taking photos let to shutter not firing - don't remember what I did ( I think I started in video mode and filmed a short raw clip and then switched to photo and attempted to take a shot), but a camera reboot fixed it and it went back to normal.

[berechiah]

Bought the EF-M 22mm f/2 STM today and it came with a free EOS M. Now I've got 2! Can't wait to start using ML extensively on both of them ^^

Can u give me a like I'd actually buy one right now if I could

I'm in Ireland