Canon 100D / SL1

Started by nikfreak, October 19, 2015, 10:41:29 PM

Previous topic - Next topic

0 Members and 3 Guests are viewing this topic.

Danne

I checked focus pixel behaviour in certain modes on my 100D and had a wtf moment. Needs to be verified. First I checked 3x zoom mode since I´ve seen reports on focus pixel free output in this mode. I then checked regular mv1080p mode. This is my conclusion.
Lossless 12 and 9-11 bit settings are free from focus pixels
14bit, both lossless and regular 14bit are filled with focus pixels
10 and 12 bit (non lossless) are also filled with focus pixels

The command in the test was this in terminal:
mlv_dump --relaxed -f 2 --dng --no-fixcp INPUT.MLV

All MLV testfiles here(100MB):
https://bitbucket.org/Dannephoto/magic-lantern/downloads/100D_Focus_pixel%20test.zip

I would like to verify this behaviour so if someone with a 100D could do this test as well. Even maybe test an eosm, or a 700D the same way. Record 9-11bit lossless a 12bit lossless and a 14bit lossless MLV file and run the above mlv_dump command. It can be tested with mlv_dump on steroids for example:
http://www.magiclantern.fm/forum/index.php?topic=18975.0

Only the 14bit lossless file should contain focus pixels.


Danne

Yes, remember the preferred raw type testing and the other issues around this changing the stream. A quick check on these files without focus pixels seems cleaner and without artifacts which could indicate something else going on. Didn´t check but isn´t the raw stream the same for all lossless bits? Or is it related to digital gain changes in liveview(static 14bit brightness for all bits in recent crop_rec_4k code?). Havn´t had time to test with older builds...

Danne

haha, the wtf moment is gone. You were right.



a1ex

Quote from: Danne on October 10, 2017, 10:57:20 AM
isn´t the raw stream the same for all lossless bits?

Nope - see the above links.

If you can find a raw type affected by digital ISO and with predictable focus pixel patterns, we should probably use that one instead.

OlRivrRat

           I see "9-11 bit settings" mentioned often lately, where are these "9-11 bit settings" found?
ORR~DeanB  ~~  80D-ML  &  SL1+ML  &  5D2+ML  &  5DC+ML  &  70D+ML(AliveAgain)

DeafEyeJedi

Quote from: Danne on October 10, 2017, 10:46:03 AM
I checked focus pixel behaviour in certain modes on my 100D and had a wtf moment. Needs to be verified. First I checked 3x zoom mode since I´ve seen reports on focus pixel free output in this mode. I then checked regular mv1080p mode. This is my conclusion.
Lossless 12 and 9-11 bit settings are free from focus pixels
14bit, both lossless and regular 14bit are filled with focus pixels
10 and 12 bit (non lossless) are also filled with focus pixels

Only the 14bit lossless file should contain focus pixels.

Tested and confirmed.





100D Lossless Samples:

9-11-bit (8.8 MB) - https://bitbucket.org/DeafEyeJedi/magic-lantern/downloads/9-11_bit_Lossless_M10-1015_samples.zip

12-bit (9.5 MB) - https://bitbucket.org/DeafEyeJedi/magic-lantern/downloads/12-bit_Lossless_M10-1016_samples.zip

14-bit (12.6 MB) - https://bitbucket.org/DeafEyeJedi/magic-lantern/downloads/14-bit_Lossless_M10-1017_samples.zip
5D3.113 | 5D3.123 | EOSM.203 | 7D.203 | 70D.112 | 100D.101 | EOSM2.* | 50D.109

Danne

Had the feeling you would pop up here Deafeyejedi. However, you missed this one ;)
http://www.magiclantern.fm/forum/index.php?topic=16040.msg191424#msg191424

DeafEyeJedi

Ah, yeah I sure missed that one. Thanks, Danne!

However, in case anyone hasn't noticed yet but when viewing MLV's directly (in this case MLV App) it seems 14-bit lossless looks a tad brighter (not much but noticeable) than 12-bit or 9-11 bit lossless?

Watch the vimeo link above and you will see that M10-1017 (14-bit lossless) is brighter than both M10-1015 & M10-1016 (9-11 bit/ 12-bit lossless) or is it just me?

All were shot under the same settings from 100D.

Quote from: OlRivrRat on October 10, 2017, 07:14:36 PM
           I see "9-11 bit settings" mentioned often lately, where are these "9-11 bit settings" found?

If you go under ML Video and prompt inside raw_rec.mo and where you will normally see options for 14-bit lossless or 12-bit lossless (9-11 bit is sitting underneath these options below) within the enlist.
5D3.113 | 5D3.123 | EOSM.203 | 7D.203 | 70D.112 | 100D.101 | EOSM2.* | 50D.109

OlRivrRat

                     @DeafEyeJedi

       Not finding it ~ Have searched Both Experimental Builds & Latest Nightly. Can You put up a ScreenShot

or send direct to Me @   [email protected]
ORR~DeanB  ~~  80D-ML  &  SL1+ML  &  5D2+ML  &  5DC+ML  &  70D+ML(AliveAgain)

Danne

I´m having an issue and wonder if it can be confirmed by another user:
http://www.magiclantern.fm/forum/index.php?topic=19300.msg191397#msg191397

The problem I´m having is that 5xzoom mode is producing corrupt MLV files. About every two seconds I´m getting a doubled, identical frames causing jump cuts all over the dng sequence. It´s reproduced every time. Could someone upload a file filmed with following settings with latest crop_rec_4k(experimental) version for the 100D:
https://builds.magiclantern.fm/jenkins/job/crop_rec_4k/56/artifact/platform/100D.101/magiclantern-crop_rec_4k.2017Oct10.100D101.zip

mlv_lite
FPS override 24
5xzoom, 1920x1080 will do
lossless 12bit


It´s enough with a 5-8 second recording.

Nex

Ok, that is strange... With latest build (Oct02) - camera is booting and working fine, whats interesting... Memory card used to install broken build - doesnt boot. With a card that was used for older builds - all is fine.

Danne

I don´t like to report vague issues but I have something similar. Sometimes I come back to my camera and it won´t boot. I need to take out the sd-card, start the camera without the sd card and then put the card back in. Then it usually let ML kick in, camera wakin up again etc. Many times it needs a couple of try like this. Lately it seems to work fine. Will try to reproduce the issue.

nikfreak

I can confirm the behaviour that builds prior 02 Oct could break the boot. I experienced exactly the same when switching through Af methods / modes prior to this fix. Only solution was to format the card. I confirmed this earlier:

Quote from: nikfreak on October 01, 2017, 09:50:25 PM
Ok, cam is alive again. After the crash I experienced described above the card seems to have somehow got an error (bootflag? cause unpacking ML zip to it didn't help)...

So I suggest formatting the card used and start from scratch if you used any build between July and 02nd Oct.
[size=8pt]70D.112 & 100D.101[/size]

GianlucaM83

I have a similar problem with build: 2 oct, 4K raw video recording; lossless compression.
As usual before installing a new ML build I uninstalled ML With in cam firmware updater and reset all camera setting, then formatting the card in my card reader in exfat, copy all build files to memorycard and update to new build using  firmware updater.
That build work nice but sometimes My 100D boot up after few second and I see an error like this: Memory card not found or damaged. At this point I have to turn off the camera, remove and insert the memory card several times, until the red LED start to blink, then turn on camera again and all work fine. No problem without ML installed  or last 10/12-bit RAW video build.

IDA_ML

Hi all,

I would like to confirm Danne's observations on the quite annoying frame jumps.  Here is what I did:


1) I used the 27-th Sept. Experimental build for the 100D to shoot a few clips from my window on the 18-th floor.  I experimented with:

      A. MLV_rec module at the 2496x1080 resolution, 10 bit and sound at 5x magnification  (about 83 frames possible in that mode) and
      B. RAW_rec module at the 2520x1080 resolution, 8 ... 11 bit lossless at 5x magnification  (about 150 to 250 frames possible in that mode, dependent on the fine detail in the scene).

I used MLV_dump to extract the cDNGs.   Frame grab image quality in both modes was fantastic. I never noticed focus pixels on the videos shot in those two modes. If there are any, MLV_dump seems to kill them all, so does MLV Producer too.  But I might be wrong.  I am not so good at pixel peeping.

2) Unfortunately, despite the high image quality, the videos did not come out 100% OK.  I also get the frame jumps that Danne observed  but I noticed them much later when I tried to play the clips.  I call these jumps "hick-ups".  They occur at random locations in the clip, so I wouldn't say that they appear at some periodic time intervals of 2s or similar.  Unfortunately, these hick-ups occur also in earlier builds, I remember observing them last winter at high resolutions, and also in the later builds from Oct. 6 and 7.  I have not tested the latest build yet.  Pretty busy at the moment.

For those who wish to try to reproduce this issue, may be on the latest Experimental build, just dial in the settings from 1) B above and shoot a clip until recording stops. Then open it with the latest version of MLV Producer and convert it into H.264.  Then play it with VLC Player or whatever player you use and you will notice the hick-up jumps appearing at random locations in the clip.

OlRivrRat

                     @IDA_ML

       The 27Sep Build zip that I have does not have a raw_rec module, in fact none of the crop_rec_4k builds

that I have downloaded have a raw_rec mod', did You import it from elsewhere?
ORR~DeanB  ~~  80D-ML  &  SL1+ML  &  5D2+ML  &  5DC+ML  &  70D+ML(AliveAgain)

IDA_ML

No, I didn't.  As far as I remember it was there.  But you can try some of the later modules, e.g. the Oct. 6 or 7 build.  Those behaved in the same way as I described in previous post. 

Did anyone try the Oct. 10-th build?

Danne

Thanks IDA_ML for confirming the "frame skip" issue on 5x zoom mode. So, to me it seems we have to issues that could need some extra focus on solving.

- Frame skip issue in 5xzoom mode
http://www.magiclantern.fm/forum/index.php?topic=16040.msg191480#msg191480

- The seemingly focus pixel free raw stream with lossless 12 and 10bit footage.
http://www.magiclantern.fm/forum/index.php?topic=16040.msg191426#msg191426

For the second issue it seems we need to find the correct preferred raw type.
QuoteIf you can find a raw type affected by digital ISO and with predictable focus pixel patterns, we should probably use that one instead.

The first issue I really don´t know where to look for a solution. Is this raw type related? According to dfort his eosm is filming 5xzoom lossless just fine. It doesn´t seem related to speed either. I lowered resoution to continuous and still the frame skip jumps are there just the same. Any pointers here are very welcome.

Regarding the preferred raw type. Is this way to scroll trough raw types still valid?
in raw.c
#define RAW_DEBUG_TYPE   /* this lets you select the raw type (for PREFERRED_RAW_TYPE) from menu */
It sure shows in the menu and I am able to change the register number but it seems it has no effect to the live view window. Any help on this too is very appreciated. 100D is an awesome cam, but two of the best features asking me, 12bit lossless(continuous) and 2520x1080 9-11bit are a little crippled right now :)

DeafEyeJedi

100D_crop_rec_4k.2017Oct10.100D101_Test run through.



Video Samples (ProRes Proxy in Rec.709 via Switch -- Thanks Danne!):

RAW (mlv_lite) FPS override 23.976p while in 5x zoom, 1920x1080p 12bit lossless @ 15mm 2.8


RAW (mlv_lite) FPS override 24p while in 5x zoom, 1920x1080p 12bit lossless @ 15mm 2.8
5D3.113 | 5D3.123 | EOSM.203 | 7D.203 | 70D.112 | 100D.101 | EOSM2.* | 50D.109

Danne

Unclear what the first clip is, seems a little bit of crop_mode and then something else. BUT.
The last clip seems to be 5xzoom and 10 seconds in is the famous jump cut.

DeafEyeJedi

First clip was shot in 23.976p and second in 24p within FPS override -- could that be the reason why?

However, in the second clip (24p) I do notice the famous jump cut (turns out it's usually at dng #265) but will need to do more tests to be sure if it's reproductive on the same certain numbered dng file.
5D3.113 | 5D3.123 | EOSM.203 | 7D.203 | 70D.112 | 100D.101 | EOSM2.* | 50D.109

Danne

You can check your dng sequence in Mlrawviewer. Faster than exporting to mov files.

a1ex

Quote from: Danne on October 11, 2017, 07:18:28 PM
Regarding the preferred raw type. Is this way to scroll trough raw types still valid?
in raw.c
#define RAW_DEBUG_TYPE   /* this lets you select the raw type (for PREFERRED_RAW_TYPE) from menu */
It sure shows in the menu and I am able to change the register number but it seems it has no effect to the live view window.

That's right - it only changes the raw stream that gets recorded as MLV. It doesn't change Canon's image pipeline (so the end result is the same); it only picks some Bayer stream from a different location in their pipeline (at least that's my understanding). The raw type "CCD" (default) is probably the earliest in the pipeline, and the "DEFCORRE" one (used with 8...12-bit lossless) is after some preprocessing (such as digital gain with the SHAD module*), or defect correction, or whatever else Canon does).

*) I use SHAD_GAIN to darken the image (reducing the number of useful levels and therefore the bit depth).

There are many raw types, but where exactly they are in the pipeline, it's not known. What I know is summarized here.

The easiest way to test is to enable the Framing preview in mlv_lite (or raw_diag, if you prefer).

Danne

Thanks a lot for explaining/verifying.