Tragic Lantern for 6D

Started by 1%, December 24, 2012, 07:07:02 PM

Previous topic - Next topic

0 Members and 1 Guest are viewing this topic.

gravitatemediagroup

Quote from: 1% on March 06, 2013, 04:57:49 PM


Look at moire on this vid: http://www.youtube.com/watch?v=zpNVUzXiZFY
Defect is one thing but the moire on the fences and grates kinda annoys me. Yes you can shoot around it but you'll never know what will start doing it until you start filming a scene... and then what you tell your whole team to move or buy that filter that jams up the mirror box?


yeah, I knew the 6D was pretty bad news when I seen moire IN DIRT!!! no BS....DIRT

oh, but don't worry, $400+ more for an adapter (that much closer to a MK3) and you can be moire free.

criz_a

Quote from: gravitatemediagroup on March 06, 2013, 09:49:50 PM
yeah, I knew the 6D was pretty bad news when I seen moire IN DIRT!!! no BS....DIRT

oh, but don't worry, $400+ more for an adapter (that much closer to a MK3) and you can be moire free.
;D

but wait, what adapter?


fotojohni

What language do you program arm processors in, I was digging around CHDK and it looks like ubasic?  About the HDMI, basically, from a conceptual standpoint, the sensor is spitting raw data into the digic 5+, which is supposedly 17x faster than the digic 4, though seriously how is this possible?.  And at this point there is probably a compression algorithm.  The digic 5+ is programed to line skip in this camera, where as in the mk iii it is programed to do a 3x3 average or possibly some kind of Gaussian average from a much bigger pixel area and some sharpening algorithm.  Then (maybe, I think) the processor compresses the color into h.264 according to the picture style.  So, it is conceivable that fixing the moire is even tricker than getting uncompressed color. 

It's unfortunate that HDMI doesn't come directly off of that first step for 5.5K video, that would be nice but unfortunately impossible with the current specification (v1.4, max 4K @ 24).


I'm hoping that the HDMI can come off in uncompressed color, that would be very, very cool.  However, unless canon is waiting on the new HDMI 2.0 specs and somehow has some kind of magic built into their HDMI, I wouldn't expect too much in terms of uncompressed resolution.

fotojohni

Quote from: gravitatemediagroup on March 06, 2013, 09:49:50 PM
yeah, I knew the 6D was pretty bad news when I seen moire IN DIRT!!! no BS....DIRT

oh, but don't worry, $400+ more for an adapter (that much closer to a MK3) and you can be moire free.

Please note that the screen is not 1080, it's cropped to something close to 720x1280.  Also, recording in 720 you will have ridiculous moire, I mean, in every sense of the word ridiculous.  It looks like they are taking each pixel color from different lines, so that the color moire is absolutely absurd.

edit: also, the mosaic engineering solution is overpriced.  It made sense in the days of the 5d mk ii domination, but now that the 5d mk iii is out it is just annoying.  For $200 it makes sense.

gravitatemediagroup

I don't shoot 720 and I'm not talking about seeing it on the camera screen.

fotojohni

Quote from: gravitatemediagroup on March 06, 2013, 10:37:48 PM
I don't shoot 720 and I'm not talking about seeing it on the camera screen.

yeah I agree, its real bad.  Mosaic is a very cool solution, I hope they adapt to the lower demand (from people shifting to the 5d mk iii) by lowering the price to $200

xcreativ

Most importantly, this filter - the mechanical implementation of the camera, which prevents its parts. First, it reduces exposure to some level. In their advertising on Vimeo, they refer to the non-permanent cloud of light conditions. However, why the clouds covering the sun is, when there is a filter test, and not without? Secondly, I think the filter to reduce the sharpness.
Cameras:Canon 6D, Canon EOS 550D.
Lens: Sigma AF 70-200mm f/2.8 EX DG OS HSM, Sigma AF 17-50mm f/2.8 EX DC OS HSM, CANON EF 50 f/1.8 II, CANON EF-S 55-250 mm f/4-5.6 IS.
Sound: ZOOM H1, Prof Shotgun Uni-Directional Condenser Mic EM-320E.

criz_a

Quote from: xcreativ on March 06, 2013, 11:04:18 PM
Most importantly, this filter - the mechanical implementation of the camera, which prevents its parts. First, it reduces exposure to some level. In their advertising on Vimeo, they refer to the non-permanent cloud of light conditions. However, why the clouds covering the sun is, when there is a filter test, and not without? Secondly, I think the filter to reduce the sharpness.

Yes, I noticed it, too. The picture got darker with the filter. Didn't like the outcome anyway...

fotojohni

Quote from: xcreativ on March 06, 2013, 11:04:18 PM
Most importantly, this filter - the mechanical implementation of the camera, which prevents its parts. First, it reduces exposure to some level. In their advertising on Vimeo, they refer to the non-permanent cloud of light conditions. However, why the clouds covering the sun is, when there is a filter test, and not without? Secondly, I think the filter to reduce the sharpness.

I believe the filter slightly reduces sharpness, however, this reduction is only very small.  I need to test the optical attenuation, I wouldn't worry about it though, the cameras low light capability is not a problem.

all in all it's a very good solution, albeit an over priced one.

1%

QuoteSo, it is conceivable that fixing the moire is even tricker than getting uncompressed color. 

Not really, just have to sample the sensor differently. They CAN do it... they did it for 600D. Basically a 4:2:2 YUV buffer is created by sampling sensor (skipping some lines) and interpolating. The buffer sizes come out different on each camera because of the math that goes into this and different sensor sizes/timings/etc.

The main problem is you can't fit that 5k image onto <2k sized buffer without skipping a lot of data. They only have a given # of lines they can read fast enough. Then this gets sent to the TI h264 encoder and color sub-sampled to 4:2:0.
Reading the whole thing you get the FPS you get shooting stills.

Without dual digic the best "res" you get is 2k.. that is the size of the zoom window.. we got raw dumps that were 2k as well. Problem is the encoder only takes fixed sizes... so after scaling the whole thing it gets scaled again to 1080P. Effects are applied somewhere in this line... probably onto the original 2k sized window it makes. All this is done to present the "1080p" fixed image to the consumer.. If they got some weird ass 4:3 resolutions in their videos they would be confused. Also the LV not being 100% coverage would be confusing as well.

Some folks want an imaging device, dummies want a packaged product that's ehhhch Deee. There is some wierd sizing stuff that goes with this camera as evidenced by the folding image in FPS override and movie/photo buffer differences of a few pixels. Wish canon would make a 6D pro firmware and leave us with un molested video.

As I figure more low level stuff out and get better I'm going to try to fix some of these shortcomings but its still a long road ahead and only a few people care. I did manage to blow up the photo buffer on 600D and get it to zoom while recording and all that fun so this stuff IS possible just very complicated.

And guess what... 1dc? From BH photo:

4K (4096 x 2160), Full HD (1920 x 1080)

so... still not fixed. They just let you play in the zoom window which btw.. that 2160 is one of its sizes, they just increased vertical resolution by increasing sampling speed (and putting a big ass heatsink). Maybe 1/2 the height is on 1 digic and the other half is on the other?

Sensors are almost the same size but different effective megapixels and orientation... so they cut it down to sample faster. Also probably why we have tall HD buffer, bigger than 1080p

CMOS, 35.8 x 23.9 mm  6D
CMOS, 24 x 36 mm 1DC

sparedog


DTSET123


rktaylor

Am I missing something?

With the latest autoexec.bin-TodayCommits I'm still getting the flicker every 6 to 7 seconds when I use ML in LV?

It looked like that had been resolved under issues.

BTW ... GREAT JOB 1%!!!!
6D | Tamron 24-70 f/2.8 | EF 50 f/1.4 | EF 70-300 f/4-5.6 | Canon EF 100mm f/2.8 L

Armitage

Quote from: rktaylor on March 07, 2013, 03:25:08 AM
Am I missing something?

With the latest autoexec.bin-TodayCommits I'm still getting the flicker every 6 to 7 seconds when I use ML in LV?

It looked like that had been resolved under issues.

BTW ... GREAT JOB 1%!!!!

I just got my 6D in today and this is my first time posting.

I've used Magic Lantern on the 550D and 600D for over a year now and love it.

I just made an account to report this same issue. The menu seems to flicker every 10 seconds or so? And it flickers to an overlay of a standard canon menu. Typically its so quick I cannot tell what menu it is.


also thank you for the development! I'd love to donate to keep your spirits high or just show my appreciation.

Francis

It is probably the picture style menu, as it is a little trick to allow the use of the scroll wheels.

Malex

Quote from: gravitatemediagroup on March 05, 2013, 07:32:32 PM
when you get it, shoot in some low light with various ISO and see if you get some crazy vertical lines in your footage.

I received my 6D yesterday, I don't have crazy vertical lines, without or with ML,

ML is working fine for me, except that one flicker now and then, and the waveforme set as small is smaller than on the 550D.
I didn't play much with bitrate yet.

I'm having a big issue though but not related to ML or the 6D, the 24-105mm Canon that I received with the 6D isn't working properly!
I get Err 01 from time to time, and if I try to turn OFF IS (stabilizer) I hear some noise inside the lens, like if something is stuck!

If the 6D is OFF and set on movie mode and IS is OFF , when I turn back ON the 6D it crash directly! Err 70! WHYYYYYYYY

it's my 1st L serie lens, and it's doesn't work well! :(

1%

New lens, used lens?

The flicker is from the dialog opening in the background so you can use scroll wheels. This was a1ex's fix. Someone complained the wheels were timing out. Its like this on 5d3/m/650D i believe.

Malex

Quote from: 1% on March 07, 2013, 03:34:14 PM
New lens, used lens?

Brand new lens, it came as a kit lens with the 6D, I'll call the shop and see what they tell me.

1%

I don't have any IS lenses to test but get similar error when an iffy lens I have loses contact with the body.

Post the log from the error 70 too. There should be an assert and/or log file.

rktaylor

Quote from: 1% on March 07, 2013, 03:34:14 PM
New lens, used lens?

The flicker is from the dialog opening in the background so you can use scroll wheels. This was a1ex's fix. Someone complained the wheels were timing out. Its like this on 5d3/m/650D i believe.

Are you says that if flickers based on the lens?  I have four lenses of which I bought three new and one used.  I will test it on each lens tonight.

So the solution for now is to delay the flicker is by turning the wheel before it flickers?  When I do it still flickers after 7 seconds.
6D | Tamron 24-70 f/2.8 | EF 50 f/1.4 | EF 70-300 f/4-5.6 | Canon EF 100mm f/2.8 L

Malex

Quote from: 1% on March 07, 2013, 03:43:27 PM
Post the log from the error 70 too. There should be an assert and/or log file.

this is from error 01
https://dl.dropbox.com/u/2478549/LOG000.LOG

this error 70
https://dl.dropbox.com/u/2478549/CRASH00.LOG

do you understand that language...? is it ML related ?

1%

The color calc (internal to canon fw) crashed. Make sure expo override is off but that's about all I can think of.

fotojohni

Quote from: 1% on March 06, 2013, 11:27:04 PM
Not really, just have to sample the sensor differently. They CAN do it... they did it for 600D. Basically a 4:2:2 YUV buffer is created by sampling sensor (skipping some lines) and interpolating. The buffer sizes come out different on each camera because of the math that goes into this and different sensor sizes/timings/etc.


So the problem is the buffer isn't large enough... That seems weird that they would spend all that money making a new digic processor and they wouldn't increase the buffer size/speed.  Maybe it's just me but something must be going on here. 

Edit: I own a 600d as well, are you saying the same moire fix exists in that camera?  Arg... I never imagined this could be such a difficult problem, after all we are 5 years after the 5d mkii, you'd think technology would have caught up by now for 4k imaging to be a breeze.   

Undoubtable they are not skipping lines in the 5d mkiii, so they must be able to processes the entire sensor at some point.  Meaning the buffer is large enough for at least part of the data (maybe just luma channel perhaps) Is this possible, and if so it would technically be possible to create a sort of faux 4k with the 5k luma but 2k chroma.

Maybe this is complete bunk though, I don't really understand image compression that well.  Though I am actively researching.  I appreciate that you are one of the only ones who understands the internal coding structure of the camera.  I am studying computer science, but this is not the sort of coding that I'm used to.  Perhaps I can be to some assistance later on.

Additional Edit (19:41 utc): Wait, 4:2:2 already does this! never mind.  So actually the color channels are considerably less resolution already.  Okay, so maybe, all the 5d mkiii is doing is making an average before it even converts to YUV 4:2:2 color space.  This  makes considerable sense as to why the moire is completely ridiculous on 720, the chroma is only 360!

Additional Edit (20:00 utc) I am very interested in what the moire reduction is on the 5d mk iii on 720p mode.  Is there someone who can look into this?  Also, the fact that they went through all this trouble in 2008 to develop the 5d mkii video and they didn't think it would have an effect on the industry, not even enough to warrant 24p (24p clearly has some kind of magical power considering that magic lantern was originally developed to address that [among other problems] problem)?  Honestly, is the entire global imaging production controlled by some kind of hidden force or are they just completely out of touch with reality? 

1%

600D has a smaller sensor and crop mode so I've never noticed any moire... I can moire out 6D in a heartbeat.

We have 2 different sensor sizes here that have to be sampled with line skipping... even 1dC still skipping. 5d3 has a different sensor and a different resizing algo. Thats probably why they have less moire.

They can make the buffer as big as they want but have to fill it with image data at reasonable speed. The uncompressed we can record shows cuts because the card can't keep up. This is similar to the problem canon faces with these sensors. If they go too fast they get image aberrations... When we increase the timing and sample faster image falls over.

Add data transfer limitations on top. SD is slower than CF and even new CF is not fast enough.

Quoteso they must be able to processes the entire sensor at some point.

Yes, reading line by line when you take a photo.

5d3 sensor is 36 x 24 mm so similar orientation to ours. Weird that it and 1dc have even numbers while we do not. Odd #s of line skips could be doing this.. maybe it doesn't fit perfectly? I certainly have this problem with the high res silent pics.

Only way to fix this is going to be altering what is skipped and what is not. Leaving off crop mode was quite cruel on their part and I know they tested the camera and saw these issues.

QuoteI appreciate that you are one of the only ones who understands the internal coding structure of the camera.

A1ex, aj, indy, and g3gg0 probably have it figured out much more than I do.