Author Topic: Danne's crop_rec_4k experiments for EOS M  (Read 240254 times)

gravitatemediagroup

  • Member
  • ***
  • Posts: 247
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #800 on: April 10, 2019, 04:41:57 PM »
I know the Overlook Hotel when I see it. lmao that's what those screen grabs look like. The Shining.

baladev

  • Freshman
  • **
  • Posts: 81
  • 5Dm3 - 6D - 7D - EOS M
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #801 on: April 10, 2019, 05:08:25 PM »
There's definitely difference, but nothing a quick post processing won't fix. Here's you green frame quickly adjusted to look more or less like the cyan one:




I've been doing digital color correction for 15 years (photos and printers) and pretty much the only thing I care about is details. They can't be brought back if lost. Color can be made almost anything you want nowadays, especially with the power raw presents us.

dfort

  • Guest
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #802 on: April 10, 2019, 05:24:51 PM »
I worked 25 years in editorial and post production and before that work another 20 years as a still photographer. Yeah, I'm an old guy so maybe it is my eyes.

@2blackbar - So you are saying that chroma smoothing (which you referred to as "Chroma blur") is causing a subtle shift in greens? It that the reason you would avoid using it? Seems like that can be easily fixed in color grading.

...pretty much the only thing I care about is details. They can't be brought back if lost. Color can be made almost anything you want nowadays, especially with the power raw presents us.

My point is that chroma smoothing is a useful tool that has gotten an unjustified bad rap. It is worth reading the comments in the source code to see what is going on:

modules/lv_rec/raw2dng.c
Code: [Select]
             * for each red pixel, compute the median value of red minus interpolated green at the same location
             * the median value is then considered the "true" difference between red and green
             * same for blue vs green

By the way, it is very hard to judge what is going on when you are posting 8-bit images. It is best upload DNG or even short MLV files.

...that's what those screen grabs look like. The Shining.

LOL

In any case, we're getting off topic here.

2blackbar

  • Hero Member
  • *****
  • Posts: 507
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #803 on: April 10, 2019, 06:01:50 PM »
Good observations and ideas guys , thanks .
baladev thats quite accurate but its just example, shot is underexposed and was just high iso test of how much i could smooth out noise, is it better to shoot 1600 and brighten up or shoot 64000 and darken shadows.The answer is - bring more light.Sometimes you just dont have them.

baladev

  • Freshman
  • **
  • Posts: 81
  • 5Dm3 - 6D - 7D - EOS M
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #804 on: April 10, 2019, 06:29:15 PM »
Quote
was just high iso test of how much i could smooth out noise, is it better to shoot 1600 and brighten up or shoot 64000 and darken shadows.

I recently tried editing in DaVinci Resolve. One thing that impressed the heck out of me is very powerful noise reduction. It has this tool.. not sure what it does, frame blending or something, but i was able to reduce noise in some quite noisy footage (iso 3200 if remember correctly) to almost nothing without destroying details. It was like magic to me.. used Adobe stuff for years and have never seen anything like this! BTW, there's a free version of it.

masc

  • Contributor
  • Hero Member
  • *****
  • Posts: 2011
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #805 on: April 10, 2019, 07:58:53 PM »
But free version of Resolve has no noise reduction. (as far as I know)
5D3.113 | EOSM.202

2blackbar

  • Hero Member
  • *****
  • Posts: 507
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #806 on: April 10, 2019, 08:44:40 PM »
resolve needs gpu rendering and im on tabletpc used for work, so it doesnt even run after install.

Hek

  • New to the forum
  • *
  • Posts: 2
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #807 on: April 11, 2019, 05:33:07 PM »
When in either of the 2.5k modes, if fps is set to anything other than 23.976, the black bar on the right returns. Tested this on the 4/9 build and the build from today. Tested using 2.35:1 aspect ratio.  To test, set fps override to exact fps, and fps to 24. You will see the black bar on the right in framing preview.

edit- This is on an EOSM

Danne

  • Developer
  • Hero Member
  • *****
  • Posts: 7393
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #808 on: April 11, 2019, 08:14:53 PM »
Not very clear what you are doing or what issue you have.
On a sidenote. Fps override will not be recommended to use if preset includes static fps already set.

Hek

  • New to the forum
  • *
  • Posts: 2
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #809 on: April 12, 2019, 01:48:25 AM »
Here are two stills from an mlv, first one taken at 23.976 fps, the second at 24fps. Notice the black vertical bar on the right.



Also noticed this in the code, not sure if it's related.

Code: [Select]
/* gets rid of the black border to the right */
EngDrvOutLV(0xc0f383d4, 0x4f0010 + reg_83d4);
EngDrvOutLV(0xc0f383dc, 0x42401c6 + reg_83dc);

Just something I noticed when I was editing some clips. I'll stick to 23.976fps from now on.

dfort

  • Guest
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #810 on: April 12, 2019, 02:20:23 AM »
Now that you bring it up--I would prefer if the default fps on the bleeding edge settings could be set to 23.98 instead of 25fps. Yeah, I know that we have users in PAL land and NTSC land but pretty much all film production these days is being done at 23.98 (even if they tell you it is 24). In addition, it does allow for just a tiny bit more max resolution than true 24fps.

Danne

  • Developer
  • Hero Member
  • *****
  • Posts: 7393
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #811 on: April 12, 2019, 03:32:46 AM »
It's been default 24 fps for a while now although 25fps is to prefer often here in pal country. Flicker frequencies and such.
@Hek
You can use shutter fine tuning instead of fps override.

baladev

  • Freshman
  • **
  • Posts: 81
  • 5Dm3 - 6D - 7D - EOS M
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #812 on: April 12, 2019, 04:33:14 PM »
Danne

I tested your last build today. Very nice!
Still get some occasional pink frames. Can't narrow down what's causing them.
The statistics from today is:
5K anamorphic
1136 x 1918 - out of 5 clips 2 had pink frames - 1 and 2 in each
1184 x 1998 - out of 5 clips 1 had 1 pink frame
new 4k anamorphic rewired
1136 x 1900 - out of 3 clips no pink frames

BTW, what's the difference between these two presets?

Will do more tests tomorrow. I film some clips in LV and others in LV with GD off and yet others on non LV mode, maybe this has influence.

Thank you!

Danne

  • Developer
  • Hero Member
  • *****
  • Posts: 7393
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #813 on: April 12, 2019, 04:41:52 PM »
The most trusted way is to run the lua cinema script and record 2:35.1 ratio. If you also enable Kill global draw setting I think youll have a hard time getting pink frames.
Other settings could work and could not work.

baladev

  • Freshman
  • **
  • Posts: 81
  • 5Dm3 - 6D - 7D - EOS M
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #814 on: April 12, 2019, 04:56:33 PM »
I see, I'm a 16:9 guy, so this script won't work for me..
Where is this Kill global draw setting?

Danne

  • Developer
  • Hero Member
  • *****
  • Posts: 7393
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #815 on: April 12, 2019, 05:01:33 PM »
One of the menus.

baladev

  • Freshman
  • **
  • Posts: 81
  • 5Dm3 - 6D - 7D - EOS M
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #816 on: April 12, 2019, 05:28:08 PM »
Ok, Thank you, found it, will test tomorrow.

I've noticed that your build is so smooth and responsive, feels great comparing to my 6D, which feels a little sluggish. Also things don't seem to break, at least I haven't been able to break anything so far, which would require a reboot.. So, congrats, feels very polished.

About pink frames.. I though they were caused by corruption that occurs when a camera runs out of processing power and can't keep up with the data flow or recording it onto a card.. But my observations with today's footage don't seem to support this.. I observed pink frames in places in my footage where I would expect them the least if my theory was correct, like I pan the camera from a brighter spot to a darker spot and a pink frames occurs in the dark spot, where, according to what I've read, the data flow should be smaller.. Strange  ::)

Danne

  • Developer
  • Hero Member
  • *****
  • Posts: 7393
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #817 on: April 12, 2019, 05:44:43 PM »
Could be timer regs, compression code, cpu or a mix of things. Who knows. It's a real pain in the ass, that's for sure...

baladev

  • Freshman
  • **
  • Posts: 81
  • 5Dm3 - 6D - 7D - EOS M
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #818 on: April 12, 2019, 06:17:26 PM »
If it's of any help.. I don't think I ever got any corrupted frames on my 6D. I get 2nd and 3rd frames corrupted all the time, they are practically never good, but apart from that, I can't remember ever getting a pink or mangled frame.

Danne

  • Developer
  • Hero Member
  • *****
  • Posts: 7393
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #819 on: April 12, 2019, 06:31:42 PM »
No help here i'm afraid. I must say I feel lost when tweaking regs. Shooting from hip, testing comparing until I tested all I could think of testing.
At this point I think a1ex could take a look at my anamorphic parts in crop_rec.c and maybe he sees something that can be refined.
What you can help with is tweaking following regs straight in crop mode sub menu:
713c
7150
6014
6008

Maybe you find a better combination for 16:9.

baladev

  • Freshman
  • **
  • Posts: 81
  • 5Dm3 - 6D - 7D - EOS M
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #820 on: April 12, 2019, 06:39:32 PM »
Ok, I will try these.

So, the 6D and EOS M have different values in these regs? Can't take them from one put in the other?

Walter Schulz

  • Contributor
  • Hero Member
  • *****
  • Posts: 8488
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #821 on: April 12, 2019, 06:49:32 PM »
If you ever wondered why it takes so long to get ML ported to new hardware or even to new firmware versions ...
And some cams are worse than that. My 650D won't work with Dual-ISO without tweaking a value which works perfectly well on other 650Ds.
Welcome to the fascinating world of reverse engineering!

Danne

  • Developer
  • Hero Member
  • *****
  • Posts: 7393
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #822 on: April 12, 2019, 06:51:18 PM »
Not the same @baladev. They follow the same relation and they are named the same. A1ex got them all pretty much figured out. Should check his bad ass crop_rec.c for 700D...

Speaking 650D Walter. We should get that one too into anamorphic world.

Walter Schulz

  • Contributor
  • Hero Member
  • *****
  • Posts: 8488
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #823 on: April 12, 2019, 06:58:29 PM »
Speaking 650D Walter. We should get that one too into anamorphic world.

Sorry, unable! I'm pretty sure I'm occupied with other things until April 2020.

Danne

  • Developer
  • Hero Member
  • *****
  • Posts: 7393
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #824 on: April 12, 2019, 07:11:34 PM »
I could prepare a preset for testing. Let me know if any interest here 650D users...