Author Topic: Danne's crop_rec_4k experiments for EOS M  (Read 240004 times)

marcelolippi

  • New to the forum
  • *
  • Posts: 42
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #850 on: April 19, 2019, 03:24:36 PM »
Sorry Danne, I was talk about 9 april. I re test with the 11 april version, I think much better. My card is sandisk extreme pro read 170 MBs.

marcelolippi

  • New to the forum
  • *
  • Posts: 42
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #851 on: April 19, 2019, 03:43:57 PM »
the only certain thing is that with 14 lossless, exposing to the right also increases the bit rate by 60%

baladev

  • Freshman
  • **
  • Posts: 81
  • 5Dm3 - 6D - 7D - EOS M
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #852 on: April 19, 2019, 04:37:47 PM »
@marcelolippi Make sure you have no clipped highlights. Picture should be a bit underexposed to get the best result, especially if you have large bright/white objects in the frame. GD causes pink frames in crop mode, so unless you don't mind them, use Kill GD, of course then you won't have LV and no framing, if you need framing, afaik your only option without pink frames is MCM rewired preset. I also reduce audio sampling to 11KHz, this helps to give more room for video. Tonight I was able to film two clips (~12min each) clips of band performance with 1120x1890 14bit anamorphic (used 4K preset). It was dark, in daylight it won't give me continuous.

henricusmaria

  • Freshman
  • **
  • Posts: 70
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #853 on: April 19, 2019, 07:20:13 PM »
@henricusmaria

Normal mv1080 has stretched LiveView, mv1080 with Move Crop mode rewire has correct real-time LiveView.

Consider using the latest build from Danne, some of previous build has Extreme Moire when using MCM rewire, in the newest ones the problem has been fixed.

I have the latest build. Mcm rewire still gives aliasing/moire. Using only Movie crop mode gives me cleaner footage. My question was if there is a difference between using just movie crop mode and (standard) mv1080 with movie crop mode.

theBilalFakhouri

  • Developer
  • Hero Member
  • *****
  • Posts: 1022
  • UHS-I
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #854 on: April 19, 2019, 08:07:16 PM »
@henricusmaria

Movie Crop Mode will give x3 lossless zoom and it uses small area of the sensor 1x1 sampling, in mv1080 you are using the entire sensor 3x3 sampling same as MCM rewire. I don't know why it wasn't clear enough,  have you tested the two mods before?

The aliasing and moire come from the line skipping in 3x3 sampling, to get rid of that and if you want to keep crop factor 1.6 or close (like 3x3), use 1x3 sampling as known as 5K Anamorphic.

henricusmaria

  • Freshman
  • **
  • Posts: 70
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #855 on: April 19, 2019, 11:57:02 PM »
I know.. again, I was responding to what 2blackbar wrote about using mv1080 (no mcm) and moviecrop together. This way you are not using the entire sensor.

2blackbar

  • Hero Member
  • *****
  • Posts: 507
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #856 on: April 20, 2019, 12:05:37 AM »
I think  crop raw records 5x5 not 3x3(had to disable it to record 3x3), so you can force it to record 3x3 (kinda like normal h264 3x crop area) but real noalias mode is 5x5 from what i understand.
I actually tried both 3x3 and 5x5 raw and i prefere 5x5, 3x3 looks a bit artificially upscaled to my eye but that might be my lens resolving power not being good enough.
MAin reason i use mcm rewire is that i can remove focus pixels with it in mlvapp and i cant with regular mv1080, they both looked similar to my eye and mcm can record full sensor which is super amazing.

fanfzero

  • New to the forum
  • *
  • Posts: 11
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #857 on: April 20, 2019, 12:49:48 AM »
So i've just installed today ml on a eos m (after messing with a 100d), I've been testing it and reading the thread. I've been using it in my room, tomorrow I'll take it to the street with daylight. My question is about the resolution of the 4k and 5k modes, for what I'm reading it records in a resolution lower than 1080p and then it gets upscaled when exported to dng, is it then a fake 4k/5k?

2blackbar

  • Hero Member
  • *****
  • Posts: 507
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #858 on: April 20, 2019, 01:58:32 AM »
It records almost two times higher than 1080 -  height is 1900 pixels but width is upscaled later from 1488 pixels to 4464 to match aspect ratio.
dngs are actually 1488x1900, final res is 4464x1900.You can round it to 1920p.
I suspect these are highest resolutions cam bufers can provide at the moment in this combination.Highest width without 5x5 crop is 1736 pixels.
Its the best possible quality that you can get from canon M and its amazing that it even works.

fanfzero

  • New to the forum
  • *
  • Posts: 11
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #859 on: April 20, 2019, 02:22:17 AM »
Well, 4k (uhd) is 2160p, so it gets pretty close, I guess then this is better to use this than using an h264 4k camera (like a lumix) due to the options in post-processing and color, right?

2blackbar

  • Hero Member
  • *****
  • Posts: 507
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #860 on: April 20, 2019, 02:33:34 AM »
Thats up to You , if You prefere RAW file with upscaled width , perceived resolution for me is pretty high , much higher than recording in 1080p modes but i didnt compare it with actual 4k frame.
You also have 4038x2558 in 5x5crop mode if you want highest res no matter if cropped.
I could not get them to work for me tho, highest i could do was 2.5k in 5x5 mode, 3k and 4k have broken preview and they wont record actual image from lens.
Also You need fast card for this, with 45mb writing speed i can barely make anamorphic 5k work for about 10 seconds, i will buy new cards tho cause its totally worth it, get fastest you can actually, who knows what future brings.

Danne

  • Developer
  • Hero Member
  • *****
  • Posts: 7392
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #861 on: April 20, 2019, 08:28:34 AM »
It's not just some simple upscaling. Hardly fake. You can compare resolution and upscale non anamorphic to anamorphic starting point. Start by viewing the absence of aliasing in anamorphic mode.

2blackbar

  • Hero Member
  • *****
  • Posts: 507
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #862 on: April 20, 2019, 05:21:54 PM »
So width of the video in anamorphic mode is limited by hardware binning method or it could be higher in the future ? In 5x5 mode horizontal res is above 2000 pixels, would it be possible someday to have that over 2000 pixels res in mcm rewire ? With 2x2 binning would be 2607 horizontally.
I read this and anamorphic quality is next to full res
https://www.magiclantern.fm/forum/index.php?topic=16516.25
--
Ok i made this test of 5k anamorphic mode, one image is crop from fullscale JPEG photo at 5184x3456(3:2 ratio), and second frame is anamorphic 5k - 4464x1900(cinemascope 2.35:1 ratio), both at 100% resolution.
Its from helios 44 58mm f4 or so, maybe on iso100 i will be able to get better sharpness but this is very impressive, i would not be able to tell the difference if i wouldnt be pixel peeping so close.I have kit lens 18-55 but sharpness looks the same as helios.

Here is full photo with middle part inserted from anamorphic(a bit brighter wall)

https://drive.google.com/open?id=1XxiJt7nWQTKfaTTFlP3AziSkWl86x7Mo
This is amazing, did anyone did some resolution charts with 5k anamorphic mode ?
---
2 more test images on next day, helios 58mm closed to f16, the difference is more visible


on 75degree angle the staircases are most visible, maybe there are some upscalers that handle 75degree angle a bit better and interpolate smoothly.
Also here is interesting experiment using b-spline resize on original dng to 4464:

Looks like using b-spline gets rid of more staircases but perceived sharpness isnt as good.It is interesting option to have IMO.You could then sharpen in post with less aliasing.
fullframe photo,bspline,mlvapp

Jip-Hop

  • Freshman
  • **
  • Posts: 91
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #863 on: April 23, 2019, 07:18:17 PM »
To get started, I'd like to try to record raw continuously with the smallest sensor crop possible in 16x9 ratio at 25fps.

So I loaded crop_rec_4k_mlv_snd_isogain_1x3_presets_2019Apr11.EOSM202 on my camera. Enabled the modules: crop_rec, dual_iso, file_man, lua, mlv_lite, mlv_play, mlv_snd, sd_uhs. Restarted the camera. Ran the cinema 2:35.1 script. It puts me into 4.5K anamorphic mode, with sluggish live view. So I enter ML menu. It shows "Crop mode" set to 5K anamorphic and "RAW video" to ON, 1488x1900 1.88x. I assume this 1.88x is the crop factor in this recording mode. Correct?

I'd like to minimize the crop, so I change "Crop mode" to "mv1080p MCM rewire", and go in and out of the Canon menu. This seems to decrease the crop when checking the FOV in live view, which is what I'm after, but now it shows "RAW video" is "ON, 1736x738 4.83x". What does the 4.83x mean? Is it correct? Haven't seen this discussed before I think.

I enter the "Crop mode" submenu and enable "set 25fps" and change the "ratios" to "16:9". Also in "RAW video" submenu I set "Aspect ratio" to 16:9. The resolution is now at 1736x976. It still says 4.83x. If that's the crop factor I think it's incorrect and would be useful if this could be fixed.

I'd like to set the shutter speed to 1/50 but can 1/45 is the closest I can get (using the Shutter menu in Expo ML menu).

By the way I read all of the posts from page 71 up until here. So I'm trying to put the theory into practice here.

Going to try a recording to see if it's continuous with these settings (I guess it will be, I'm using SanDisk 64GB SD Extreme Pro UHS-I U3 V30 170MB/s)

Danne

  • Developer
  • Hero Member
  • *****
  • Posts: 7392
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #864 on: April 23, 2019, 07:26:45 PM »
Should work. You can ignore the 4x crop stuff. It's all kinds of rewiring and round trips going on but if it looks ok in live view usually things will fly.
There's some fine shutter menu you could use to fine tune to 1/50. Search around.

Edit:
No need to set ratio in RAW video tab.

Jip-Hop

  • Freshman
  • **
  • Posts: 91
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #865 on: April 23, 2019, 08:18:30 PM »
Thanks for the tip and good to know you're aware of the incorrect crop factor value. If possible I'd find it useful if it would show the correct factor. Just made a 27minute test recording that filled up the 64GB card. For me it seems to work continuous without issues in this mode. I greatly overexposed to test, as I understood that increases the datarate.

Danne

  • Developer
  • Hero Member
  • *****
  • Posts: 7392
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #866 on: April 23, 2019, 08:55:49 PM »
I can take a look. Meanwhile. Trust your eyes and test results...

2blackbar

  • Hero Member
  • *****
  • Posts: 507
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #867 on: April 24, 2019, 04:49:07 AM »
Danne I managed to record actual 4000x17000 raw in 10fps (without any upscale later)above 10 seconds and could go longer , im in shock but shutter speed and fps stays at 10 , is this some sort of limitation ? Is it possible to force shutter 50 or 40 and fps to 24 ? Or it wont be possible and is capped by camera not card
 writing speed ?
also where is now 1736x2150 mode ?

Im looking for mode to record 1736 in anamorphic 5k but its limited to 1488, is there any way to get full width of 1736  or there are some technical reasons its cropped a bit ?

Danne

  • Developer
  • Hero Member
  • *****
  • Posts: 7392
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #868 on: April 24, 2019, 08:29:46 AM »
New build:
https://www.magiclantern.fm/forum/index.php?topic=9741.msg208959#msg208959

Commit:
https://bitbucket.org/Dannephoto/magic-lantern/commits/eb45a9540a60188f092a4b45309241d347d4254d

@2blackbar
4000x1700, seems about right for 2.35:1 ratio mode.
Your questions have been answered before so please check other posts. I also suggest you learn how to compile if you want to test older builds like the one you are pointing too. Tutorials for both mac and windows are freely available.

ctfire

  • Freshman
  • **
  • Posts: 53
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #869 on: April 24, 2019, 11:31:39 AM »
New build:
https://www.magiclantern.fm/forum/index.php?topic=9741.msg208959#msg208959

Commit:
https://bitbucket.org/Dannephoto/magic-lantern/commits/eb45a9540a60188f092a4b45309241d347d4254d

@2blackbar
4000x1700, seems about right for 2.35:1 ratio mode.
Your questions have been answered before so please check other posts. I also suggest you learn how to compile if you want to test older builds like the one you are pointing too. Tutorials for both mac and windows are freely available.

Danne can you write what is changed from previous build? To know what to test?

Danne

  • Developer
  • Hero Member
  • *****
  • Posts: 7392
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #870 on: April 24, 2019, 11:36:29 AM »
Come on. Click the commit link, read later posts.

Jip-Hop

  • Freshman
  • **
  • Posts: 91
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #871 on: April 25, 2019, 06:46:01 PM »
Did some recording experiments and noticed my footage got corrupted when I choose the 12-bit lossless and 11...8-bit lossless option.
Then I checked what my settings were in the "Crop mode" and noticed the "Bitdepth" was set to "10 bit".
Just trying to make sense of all the options here, I think it works as follows:

"Bitdepth" in the "Crop mode" submenu should be set to "off" when you want to change the "Data format" in RAW video menu.

In case you want to change "Bitdepth" in the "Crop mode" submenu, then you should set the "Data format" in "RAW video" to 14-bit lossless.

The footage becomes corrupted if you change "Bitdepth" in "Crop mode" submenu and change the "Data format" in "RAW video" menu to something else than 14-bit lossless.

Question: is setting "Bitdepth" in the "Crop mode" submenu (when "Data format" in "RAW video" is set to 14-bit lossless) the same thing as setting "Data format" in "RAW video" to 12-bit lossless when "Bitdepth" in the "Crop mode" submenu is set to "off"?

Just trying to understand where I'm supposed to change what, and what it affects.


Noticed something else wile testing:

In mv1080p 1736x967 46/48fps mode the 11-bits mode gives corrupt footage. Half of one frame combined with part of another, stutter etc. 10-bit and 14-bit work fine.

I get some color artefacts in mv1080p 1736x967 46/48fps and MCM rewired mode that I can't get rid off in MLV App (focus pixels and chroma smoothing didn't help). See picture. Does anyone have an idea how to solve it? When I activate 3x crop mode via the "Crop mode" submenu the same scene is very clean.



Link to MLV: https://www.dropbox.com/s/69jo5hicddv5l0v/M25-1740.MLV?dl=0.

Tested with crop_rec_4k_mlv_snd_isogain_1x3_presets_2019Apr11.EOSM202.

Danne

  • Developer
  • Hero Member
  • *****
  • Posts: 7392
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #872 on: April 25, 2019, 07:00:44 PM »
Thanks for feedback.
Always keep 14bit lossless in RAW viideo menu. If you need reduced bitdepth go through crop mode sub menu. The result is about the same as  going through raw video menu but that way will not work.

Your aliasing issue is what sensor read out brings. Use x5 modes or anamorphic for nonaliased footage.

Jip-Hop

  • Freshman
  • **
  • Posts: 91
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #873 on: April 25, 2019, 07:36:17 PM »
Thanks for clarifying! Will do that from now on :)

Also the mv1080p MCM rewired mode with 3x crop enabled gives very clean (no aliasing) footage and real time color preview (no lag) in 1624x1080 resolution. Continuous recording. Quite nice and easy to use if a big crop is appropriate.

Danne

  • Developer
  • Hero Member
  • *****
  • Posts: 7392
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #874 on: April 25, 2019, 07:42:43 PM »
Yes, digital zoom works like x5 zoom. You can run Movie crop mode without crop rec to get the standard digital zoom. Think width is a little higher then.