7D slow-motion at 60 FPS and 50 FPS over ride. Now working.

[ted ramasola]

Testing the 7D for slo-mo with MLV+SND.
Buffer-1 or 2

Some of the resolutions that go continuous are:
1728 x 516  2.39:1 at 48 fps
1664 x 506  2.35:1 at 50 fps
1408 x 566  16:9 at 50fps
1344 x 540 at 16:9 at 50 fps

Notes:
1. Continuous shooting is possible at this resolution with sound.
Just like the 5DmkII, audio is advanced 2 frames and has extra frames of 7-8. After a nudge audio is perfectly synced with no drifting.

2. To achieve high frame rate, canon menu must be set to 1280x720 60fps.

3. Use FPS over ride in ML menu for 50FPS and lower.  48 or 96 shutter speed depending on amount of blur you want.

4. The 50/60fps video captured is squeezed and framing is not correct.
   ML menu tip will show you have to resize 1.4 to achieve correct 16:9 ratio of, for example, 1344 x 756 in post, the aspect ratio options will show 1344 x 540.
   However, using a chart and some guide lines, resizing to 1.4 is still not enough.
   You have to stretch height further and cropping the top and bottom to achieve proper proportion. Resize should be approx, 1.65. (edit 1.66, thanks to larrycafe)

Maybe a dev can edit the tool tip to reflect these.
Either correct the cropping of the video or change the info/tip to use 1.66 instead of 1.4 stretch.

Conforming the 50/60 FPS video to 23.976 in post gives a smooth slow motion.

[ted ramasola]

Added images to show the resizing of the 50 fps footage to fit a 16:9 timeline properly.

[OSCA LEE]

CAN YOU POST SOME EXAMPLES...I REALLY NEED TO LOOK INTO THIS I AM ALSO WILLING HELP YOU ON IT

[arrinkiiii]

Hi Ted,

Can i assume that we should use buffer 1 or 2 for slow motion in the 7D? And for 24fps also the best?

[gonstock]

Hello, up to now I've been streching Height 1.5x  and It looks good.

Just my experience!
Cheers!

[larrycafe]

with 7D sensor (I believe it is the same as my 60D)

the whole sensor is 5184 X 3456
in 720P mode, the raw output is reading 1 out of 5 lines, 3456 divided by 5 = 692

and the sensor's 3:2 aspect will be converting to 1728 X 692. if recording at 16:9 is needed, it be calculated by the required sensor height and then divided by 5. it is 5184 / 16 * 9 / 5 = 584.

with 1728 X 584 stretching to 1728 X 972, it is actually 1.66X

[ted ramasola]

Hi Ted,

Can i assume that we should use buffer 1 or 2 for slow motion in the 7D? And for 24fps also the best?

in my tests for that build, buffer 2 is best for the 7D.

[ted ramasola]

with 7D sensor (I believe it is the same as my 60D)

the whole sensor is 5184 X 3456
in 720P mode, the raw output is reading 1 out of 5 lines, 3456 divided by 5 = 692

and the sensor's 3:2 aspect will be converting to 1728 X 692. if recording at 16:9 is needed, it be calculated by the required sensor height and then divided by 5. it is 5184 / 16 * 9 / 5 = 584.

with 1728 X 584 stretching to 1728 X 972, it is actually 1.66X

Thanks! I have poor math skills so my 1.65 scale was based by eye. I will test it again using 1.66, but if users can avoid using 50P by "twixtoring" 30P then 50P should be avoided as the quality drop is noticeable.

[arrinkiiii]

in my tests for that build, buffer 2 is best for the 7D.

Thanks Ted =))

[OSCA LEE]

ok used the TL feb13 build...followed all the instructions....

very choppy playback but it is in slow motion....

and tips on how not to get the lag lines out of the footage...

[OSCA LEE]

I have a casio ex f1....I used it mainly for slow motion only problem it requires a lot of light to get great quality footage...

if we can crack this it will be awesome

[ted ramasola]

ok used the TL feb13 build...followed all the instructions....

very choppy playback but it is in slow motion....

and tips on how not to get the lag lines out of the footage...

what is "lag lines" ? Do you mean the aliasing? or are frames corrupted?

[OSCA LEE]

no frame corruption...I will have to play it back on the pc..I used the MLv player to test it...that may have been the issue

[OSCA LEE]

ok...unless we can ramp up the fps to 300+ the footage isn't smooth and it has ghost images as well

[arrinkiiii]

Well,  trying today the last nightly ML for the 7D and didn't yet have the fps override. So making some tests in 2.35:1 and trying  achieve higher frames rate Im using NTSC but changing the fps in canon menu just have 60fps with this the MLV files stay corrupted/pink... So without fps override i change from NTSC to PAL and change the fps in canon menu to 50fps and recorded a clip.

I did the MLV2DNG, open After Effects and the  dng's, AE assume 30fps i just need new interpretation and change to 23,976fps and everything went good. Ok slow motion without corrupted frames or with fps override ML tool.

[OSCA LEE]

Well,  trying today the last nightly ML for the 7D and didn't yet have the fps override. So making some tests in 2.35:1 and trying  achieve higher frames rate Im using NTSC but changing the fps in canon menu just have 60fps with this the MLV files stay corrupted/pink... So without fps override i change from NTSC to PAL and change the fps in canon menu to 50fps and recorded a clip.

I did the MLV2DNG, open After Effects and the  dng's, AE assume 30fps i just need new interpretation and change to 23,976fps and everything went good. Ok slow motion without corrupted frames or with fps override ML tool.

what settings did you have your shutter speed at...

[arrinkiiii]

I think was 96

[ted ramasola]

High Framerates raw+ sound recording Now Possible with no errors and corrupt frames. Renaming this thread to reflect this development.

60FPS via canon menu (turn OFF FPS Override)
   2.35:1 AR
   1536x466 --1'53"
   1472x448 --continuous recording

   2.20:1 AR
   1472x478 --1'57"
   1408x458 --continuous recording

   2:1 AR
   1408x504 --1'31"
   1344x480 --continuous recording

   16:9 AR
   1344x540 --1'
   1280x514 --continuous recording

50fps FPS Override

   2.35:1 AR
   1664x506 --continuous recording

   2.20:1 AR
   1600x520 --Continuous recording

   2:1 AR
   1536x548 --Continuous recording

   16:9 AR
   1408x566 --continuous recording


Note that the proper scale ratio is 1.66 and not 1.4,

1.4 scale will Fill up the screen BUT the image is still vertically distorted. (see first post regarding this)

So, a 2.35:1 ratio recording can actually fit a 2:1 timeline project.

[arrinkiiii]

Hi Ted, This last test that you post is with what date?  and version, TL or ML ?


 

[ted ramasola]

the 60fps used the nightly feb 24-25

the 50fps used TL feb 25-26

[arrinkiiii]

-Thanks Ted 

[arrinkiiii]

I try with 60fps from canon menu and have pink frames, a lot... and camera with the orange temperature on 50 degrees. With 50 from canon menu it's ok.

[ted ramasola]

use the feb 24 nightly or TL.

make sure fps override is off and just canon menu is 1280x720-60fps. FPS override 60 is pinkish/magenta colored.

I think even using feb 25 or 26 is ok.

[arrinkiiii]

Im using the ML, that don't have fps override. But i will change the buffer method for see what i can get =))

[ted ramasola]

Buffer -1 is what I used for all tests and when shooting on 7D.

[arrinkiiii]

ok, thanks... i test with buffer 2  =))

[arrinkiiii]

With ML i still get pink frames in 60fps but with TL everything is good.

[Flyingsheep]

I hope eventually the stretching issue would be solved.  Whenever you resize it to a proper 16:9 ratio there are definitely some aliasing problems. 

[ted ramasola]

I hope eventually the stretching issue would be solved.  Whenever you resize it to a proper 16:9 ratio there are definitely some aliasing problems.

I don't think ML can do much about the aliasing as even when using h264 the 720P is already heavily aliased. Its a hardware limitation IIRC.

What ML has done is giving you a slightly better solution since you can now record an image bigger than 1280x720 at 60/50/48P

[Flyingsheep]

Right but it looks like stretching the image on the top and bottom makes it much more pronounced.  I shoot 720p 60p in h.264 all the time and it's not nearly as bad.  I think fixing the stretching problem would be a better solution than increasing resolution a little bit.

[crwn_schlr]

Right but it looks like stretching the image on the top and bottom makes it much more pronounced.  I shoot 720p 60p in h.264 all the time and it's not nearly as bad.  I think fixing the stretching problem would be a better solution than increasing resolution a little bit.

I just recently used 48fps Raw at the full 16:9 res on a concert tour, in conjunction with 60fps h264, and I actually have some examples where the canon 720/60 shows pretty significant aliasing, and the raw/48 does not.

Not trying to disprove your claim, different scenario, different results. But I'll note I do remember one specific 720/60 clip that shows pretty heavy aliasing in a hat, and on most of the raw/48fps clips it wasn't there, or i didn't notice it.

I'll try to find some examples, as I'm pulling still frames now anyway.


On another note, I've been testing different aspect ratios with the 1.66 stretch factor. I haven't done any math but I think the 2.35 stretched is nearly 1.85. So I shot some stuff recently at 1.85 to see if it hits 16:9. That way you can get the maximum write speed with out throwing away any of the image.

Any one else played with this?