Raw video on 5DMK2

Started by olik, May 15, 2013, 08:19:46 AM

Previous topic - Next topic

0 Members and 2 Guests are viewing this topic.

tihon

Quote from: savale on May 22, 2013, 12:04:27 PM
Does anyone know what is the technical problem of not being able to shoot at 1920 width (but 1880) when not in crop mode? Is it because of the amount of sensor pixels?
I don't really get it because the 5d2 has 5613 pixels width right? divide it with 3 means 1871 pixels, so how is it even possible to shoot at 1880 width? ::)
the same question...
Cinema, cinema, cinema

disuye

Quote from: Africashot on May 22, 2013, 10:39:24 AM
This is already working, just use the zoom button in live view and you can record a 1920 x whatever you choose 1:1 crop

I had no idea! Sorry - I'm still getting up to speed with the raw functionality.

Incidentally, I bought a Toshiba 64GB "Exceria Pro" 1066X card earlier today - specs on paper look totally ballistic and the price was pretty damn good (USD 289 / HK$ 2250 for the 64GB card, I'm based in Hong Kong FYI so this stuff is pretty cheap).

Will reply with performance test tomorrow.

savale

Strange in the magic lantern 5d2 platform header file I see this:

#define SENSOR_RES_X 5792
#define SENSOR_RES_Y 3804

that would mean if you'd divide by 3 you get about 1930 pixels width, so that means 1920 is possible by just skipping 2 pixels. I can't find the code for blocking the max width to 1880. (I expect there is some comment in the code explaining why this is a limit)

JackDaniel412

Real resolution of 5d2!


DAK29

Is the 1880 limitation placed in ML simply because going higher drops too many frames?
If yes, shouldn't it be easy to put a version out that allows us to select all the resolutions, even if it's too slow, just so we can experiment?

I'm still not understanding this anamorphic crop thing - with a true 2.35:1 anamorphic lens, can we record the full pic in 1080 x X?

I keep hearing about cropping. 16:9 cropped to 2.35:1 is NOT anamorphic, it's just less of a 16:9 image.
A true anamorphic picture should take up the entire frame and be vertically stretched, to later be squeezed into the common
letterbox when displayed on 4:3 or with lesser bars when displayed 16:9. Either way, the aspect ratio of true anamorphic
shows way more horizontally than a cropped 16:9 image. I understand I'm stating the obvious, just want clarification.
It seems people are saying you can shoot 1920x(insert various vertical resolutions here which are less than 1080) with anamorph.
I am wondering if this results in true anamorphic shot or a cropped 16:9?

Plan on posting some tests as soon as I order my 1000x, am still unable to do anything without a proper card.

Thanks!

JackDaniel412

Quote from: savale on May 22, 2013, 01:10:55 PM
Strange in the magic lantern 5d2 platform header file I see this:

#define SENSOR_RES_X 5792
#define SENSOR_RES_Y 3804

that would mean if you'd divide by 3 you get about 1930 pixels width, so that means 1920 is possible by just skipping 2 pixels. I can't find the code for blocking the max width to 1880. (I expect there is some comment in the code explaining why this is a limit)

Set different res like 5640x3760 (/3=1880x1253)?

Africashot

Quote from: DAK29 on May 22, 2013, 03:12:09 PM
Is the 1880 limitation placed in ML simply because going higher drops too many frames?
If yes, shouldn't it be easy to put a version out that allows us to select all the resolutions, even if it's too slow, just so we can experiment?

I'm still not understanding this anamorphic crop thing - with a true 2.35:1 anamorphic lens, can we record the full pic in 1080 x X?

I keep hearing about cropping. 16:9 cropped to 2.35:1 is NOT anamorphic, it's just less of a 16:9 image.
A true anamorphic picture should take up the entire frame and be vertically stretched, to later be squeezed into the common
letterbox when displayed on 4:3 or with lesser bars when displayed 16:9. Either way, the aspect ratio of true anamorphic
shows way more horizontally than a cropped 16:9 image. I understand I'm stating the obvious, just want clarification.
It seems people are saying you can shoot 1920x(insert various vertical resolutions here which are less than 1080) with anamorph.
I am wondering if this results in true anamorphic shot or a cropped 16:9?

Plan on posting some tests as soon as I order my 1000x, am still unable to do anything without a proper card.

Thanks!
I guess a lot of people mean the cinemascope aspect ration when they say anamorphic, the custom resolution function however enables you to shoot at a 4:3 aspect ratio (or whatever fits your lens) meaning that you can use anamorphic lenses in order to achieve 2.35, here is a good hands on post on the subject by Andrew Reid: http://www.eoshd.com/content/10450/2-5k-cinemascope-anamorphic-raw-on-the-5d-mark-iii
I personally love this aspect ratio because of the way the image is composed within the wide rectangle, I'd of course love to use true anamorphic lenses to achieve it, for now I do not own any though and I happily accept cropping away some resolution as a trade off to not achieving enough speed with the cards I currently own... 
ML 5D2 & T3i

savale

Quote from: DAK29 on May 22, 2013, 03:12:09 PM
Is the 1880 limitation placed in ML simply because going higher drops too many frames?
If yes, shouldn't it be easy to put a version out that allows us to select all the resolutions, even if it's too slow, just so we can experiment?

Just what I am thinking... It would be nice if we can even select it just to play with it. For some of us it will be useful. It might be possible to use 2.39:1 (1920x817) for the 5d2

still looking at the code to find the 1880 restriction... (so I can remove it hehe :P)

edit: a bit off topic, but does one know how to make sure I an aligned partition on my compact flash card? In windows 7 I can't delete / create partitions on the compact flash disk using my card reader... Is that only possible in linux?

bvogelsang

Quote from: DAK29 on May 22, 2013, 03:12:09 PM
Is the 1880 limitation placed in ML simply because going higher drops too many frames?
If yes, shouldn't it be easy to put a version out that allows us to select all the resolutions, even if it's too slow, just so we can experiment?

As far as I've understood so far, the RAW-function in ML just streams the Liveview-feed of the camera, and this is only 1880 and not 1920 on the 5D2.... ?

1%

Its supposed to be the size it says.... you can't magically get 1920 out of it. Canon upscales this for H264.

disuye

Just curious if there is a limitation in CF card write speeds within the 5Dmk2 other than the card itself?

My 5Dmk2 will not seem to get over 62-63 MB/s despite a card rating of (granted, optimistically) 150 MB/s ... http://www.semicon.toshiba.co.jp/eng/profile/news/newsrelease/memory/topics_130426_e_1.html

DAK29

Cinemascope is 2.66:1 which is not far off from 2.35/2.39:1.

The custom aspect ratio allows 4:3? Okay.... And? Does that allow full frame RAW recording at 4:3?

Remember, anamorphic is full frame.
It's only during playback that bars on top and bottom come into play after the stretched full frame image is squeezed back.
Cropped 16:9 is not full frame and gives even LESS room than 16:9 which imho already isn't enough room.

So in playback, cropped 16:9 to 2.35:1 and true anamorphic 2.35:1 will take up the same area but they are totally different.
The anamorphic source is a full frame worth of data whereas the 16:9 cropped is cutting even more out of a crap ratio.

So I guess next question is - 1920x1080 is not square. 4:3 is more square than 16:9. So... What is the RESOLUTION of the cam's 4:3?

I always assumed the camera just recorded 4:3 and added black crop bars to make the 16:9.

If this is the case than the full uncropped 4:3 would actually be HIGHER resolution than 1920x1080.
It would be like removing the matte, there would be even more picture.

And... If that's the case then the answer must be NO, you can not record true full frame anamorphic raw yet
because if you can't even do 1920x1080 then you wouldn't be able to do higher than 1920x1080 either.

The other case would be that all the custom ratio 4:3 does is crop a 4:3 chunk out of the 1920x1080 image.
If all the custom ratio function does is crop then it is NOT changing the aspect ratio, it's just cropping.
This would be kinda lame and pointless but I guess it would allow for pseudo anamorphic recording.
But it wouldn't really be true at all.

What is the full frame 4:3 resolution?


Yoshiyuki Blade

Full frame video resolution is 1880x1250 for the 5D2. Kinda funky, but it was designed this way. Unless we can get 94+ MB/s of write speed, it's kind of a pipe dream to get full frame resolution continuously, which in turn would've provided even greater resolution at 2.35:1 with an anamorphic lens (effectively 2938x1250 resolution when stretched horizontally). I'd be happy enough with just ~1880x800 which the 5D2 seems just barely able to pull.

DAK29

Yoshiyuki - 1880 is TOPS for horizontal resolution?

So the 5D2 actually can not really do true 1080p? (It "upscales" from 1880 to 1920?) Geeze...
That's false marketing on Canon's part, if they say 1080p it should be 1080p.

So in your opinion, what should I do with my anamorphic?

What would be the optimal settings to retain maximum resolution, proper aspect ratio and remain equally divisible
of course in the numbers so you don't get major distortion when/if upscaled?

I've seen some people say just shoot it 720p and it looks better still RAW upscaled to 1080p than 1080p h.264.

But part of me really hates that of course, I would love to get higher than 720p even if not the full 1080p.

What about 1440x1080i? This is an actual HD resolution it appears. Will HDTVs kick into that mode if something is shot in it?
The high-definition television wiki lists it in the specs:http://en.wikipedia.org/wiki/High-definition_television

I prefer just leaving the image exactly as shot so long as the aspect ratio is proper and letting the tv kick into the proper
mode if possible as opposed to upscaling which is something I consider to be evil. 8)

Yoshiyuki Blade

Just to clarify, full frame is 3:2 ratio and anamorphic is a 4:3 ratio (unstretched), right? You can record at something like 1440x1080 (the 5D2 can barely handle that) and stretch it to 2538x1080. That's well over that full HD can handle (1920x1080) so it even gives you a bit of room to downsample to 1920x817 or some mod2 resolution.

KSphoto

I was recording in 1440x720 and the buffer was handling things fine using a Lexar Pro 600x card, when all of a sudden it quit at about 2900 frames. I wanted to try a full 4gb run using the May 21 build, but it was like the camera went to sleep. What happened?

I got it. When I reset settings the liveview became set on  1 minute sleep.
5DC, 5D2, 5D3, EOS M, too many lenses.

noix222

guys... see if anyone can help.. i have a komputerbay 1000x 64gb but i can record only 1880x840 without stop... anything higher stops max after 500 frames, also if i turn the audio record function the buffer fills up even faster. Using the build 22th. The double buffer is already in the build? Or i have to do something to enable it?? that would allow me to record 1880x960 maybe ?

Hazer

If you can record indefinitely at 1880 x 840 on the Mark II you're doing about as well as can be done.  My Komputerbay 64GB just arrived today and it's definitely faster than my Lexar 32GB 1000x.  But I'm not getting unlimited at 840.  So far I'm seeing max write speeds of about 61/sec, and that's on a good take.  This translates to about 1100-1300 frames at 840.  On the Lexar, sustained 60/sec was the best case, and most often it was high 50s.  Frequently I would not break 1000 frames at 840, so the Komputerbay is an improvement.  This is all with global draw and audio off.

mvejerslev

Perhaps FPS is a factor?

Noix, The cameras internals are the limiting factor on the 5DII as of right now. Not card speed. The camera outputs somewhere about 60-65MB/s max.
5D Mark II, PC

noix222

Yes 1880x840 is pretty good to me, i can work with that! But i've seen some people doing 1920x1080 or higher with the same card i have. So im guessing that is with Mk3, wich probably handle buffer stuff better than Mk2...  :o

hummm i get that mvejerslev... thanks!


Hazer

Alternatively, when you need a more standard 16:9 aspect ratio, you can try 1720x 960.  This looks far better than the stock H264 and only requires minimal scaling to get to 1080.  I'm seeing anywhere from 500 to 800 frames at this size, 61MB/sec.  It sounds like you might do a little better.  Let us know how many frames you get if you test it.

noix222

Yes thanks for the advice Hazer.. it is the best option for a 16:9 for sure! I just did the test you told me... So in 1720x960 im getting almost the 4gb file it stopped with 1300 frames 3.71gb file size... good enough considering the 4gb limit... But tell me you're using a komputerbay card 1000x 64gb as well ?

* so its useless to get cards like a 1066x for the 5d mk2 ? will still not be able to shoot higher res ? or that extra card speed could help to handle at least 1880x840 but with sound ON ?

Hazer

1300 frames is really good!  I'm also on the Komputerbay 64GB 1000x.  It would be perfect if I could do 1300 frames, but the longest clip I have seen at 1720x960 was around 850 frames, 61MB/sec.  A "bad" shot will be more like 500 frames.  What write speed are you seeing at 1720x960?

Re: 1066x cards, you would have to get one and test it.  Even though we are operating well below the maximum rate of these cards on the 5D2, there are still differences between them.  My Lexar 32GB is also 1000x but the Komputerbay is 1-2 MB/sec faster.  This is enough for a few hundred frames at 1720x960.  It is possible that a few MB/sec could make the difference between limited and unlimited recording for a particular resolution.

noix222

Write speed goes around 64MB/sec... and the buffer is filling up really slow... again i got 1288 frames a 3.63GB file size... The strange thing is that we both have same card and camera but different results... One thing i did, i disabled everything in ML menu!

Hazer

Hey, cool, thanks for the upload!  I didn't even realize you could save settings.  I see it in the menu now.  Anyway, no change in my performance.  I can only assume there is great variability in the cards themselves and that is what accounts for the different results people are seeing.  Who knows!  I think there is more to come in terms of performance enhancement though.  Thanks again to the devs...