24fps HDV seen on other site....stinks

[Obin Olson]

dvcpro-hd ROCKS..been shooting it..great stuff indeed ;)

[Frank Aalbers]

<<<-- Originally posted by David Warrilow : Hi,

I'm thinking this is going to improve the look of CF24 footage over what we're seeing now - as the shutter is locked at 1/60th in CF24 on the FX-1. Producing the freaky stuttering effects that we've seen to this point.

DW. -->>>

The stutering will not go away. The problem is that the time base that those cameras use to do the 2:3 pulldown on is not real 24p.

I took footage of FX1 and took out the clean 24p footage. Outside of the 1/60 shutter the frame times between frames don't look equally the same. They are off.

The biker that passes in front of the camera is a constant move. But the motion difference per frame is not equal. That makes me believe it's not really shot with a rea 24 fr/sec time base.

Here is the clip:
http://home.comcast.net/~chalbers/fx1_24ptest.mov

Frank

[Sam Edwards]

I don't think Frank's posted 24p test is valid. I believe the shutter was on auto and much faster than 1/60th because the motion blur is almost nonexistant. If the shutter was open 1/2 of the frame length you would see motion blur that would fill half of the distance the wheel travels. In this example the motion blur is much less. Look at the spokes - they're sharp!
Frank, could you cue up the tape to this part and turn on the data display? If this is what you get at 1/60th of a second then the camera is broken.
Cheers,
Sam

[Frank Aalbers]

<<<-- Originally posted by Sam Edwards : I don't think Frank's posted 24p test is valid. I believe the shutter was on auto and much faster than 1/60th because the motion blur is almost nonexistant. If the shutter was open 1/2 of the frame length you would see motion blur that would fill half of the distance the wheel travels. In this example the motion blur is much less. Look at the spokes - they're sharp!
Frank, could you cue up the tape to this part and turn on the data display? If this is what you get at 1/60th of a second then the camera is broken.
Cheers,
Sam -->>>

You are maybe correct that the shutter was maybe faster, but it definately shows acurately that the frames are not constant motion. You don't need moblur to see that !

Frank

[Sam Edwards]

Hi Frank,
I don't think you can accurately judge 'strobiness' or 'studder' from footage that was shot with a fast shutter...
Fast shutters give that effect. I'm trying to find a clock with a sweep second hand that I can shoot some tests with to give an idea. I don't feel I've seen a decent test of a 24p extraction from this camera.
Thanks for taking the time to post something. I think it has a kind of cool look.
cheers,
Sam

[Barry Green]

The easiest way to see a true 24P extraction would be to shoot in DV mode in CineFrame 24, and then use Vegas' "remove 2-3 pulldown" feature from the DV file. It won't be full-res HDV, but it will show you the raw 24P frames.

I think I have some CineFrame 24 DV footage here, but I don't know if it has panning or something that will easily show the motion between frames... I wonder if I can get AfterEffects to remove the pulldown... I have some side-by-side shots on the XL2/24P and FX1/CF24 at a skating rink that would make an excellent comparison to see what CF24 is actually doing with motion and how it compares to 24Hz progressive-scan...

[Frank Aalbers]

<<<-- Originally posted by Sam Edwards : Hi Frank,
I don't think you can accurately judge 'strobiness' or 'studder' from footage that was shot with a fast shutter...
Fast shutters give that effect. I'm trying to find a clock with a sweep second hand that I can shoot some tests with to give an idea. I don't feel I've seen a decent test of a 24p extraction from this camera.
Thanks for taking the time to post something. I think it has a kind of cool look.
cheers,
Sam -->>>

I am NOT basing myself on the stutter ... I'm basic myself on flipping through the frames one by one and looking at the distance the subject travels PER FRAME. It's not constant between the frames while it schould be ... THAT is what I mean.

And to check that you don't need moblur.

Frank

[Frank Aalbers]

<<<-- Originally posted by Barry Green : The easiest way to see a true 24P extraction would be to shoot in DV mode in CineFrame 24, and then use Vegas' "remove 2-3 pulldown" feature from the DV file. It won't be full-res HDV, but it will show you the raw 24P frames. -->>>

That is EXACTLY what that footage shows. I did not use blending or anything. It's pure extraction of the 2:3 pulldown to get the raw 24p footage.

I took it frame the HDV footage though. I used Shake to do this.

1. I exported each individual HDV res frame of the footage to single frames.
2. Looking through the frames it clearly had 2:3 pulldown.
3. In HDV the field priority is reversed though. upper field first. So I moved the whole frame one scanline up to make the lower field first.
4. I seperated both fields per frame. and resized them to 720*240 each.
5. Then I took the resized results and interlaced them back together in the same order to get back full frames with interlaced fields.
6. Now had a DV sized interlaced footage of the original HDV footage keeping the 2:3 pulldown correctly.
7. I then imported the frames to Vegas as a 29.96 fps lower field first DV footage.
8. There I selected in the file properties to remove 2:3 pulldown by selecting the correct frame for the removal sequence.
9. This resulted me to clean raw 24p DV footage from the Cineframe 24 HDV footage.

So ... It's not just downressed and resampled.

That's why I'm stating that it's not a true 24p time base.

Of course that pipeline is not possible to constantly use it. It was just to test out the so called 24p 2:3 pulldown they talked about in DVExpo. It's a FLUKE !


Frank

[Barry Green]

Turns out I do have an FX1 CF24 DV clip, and after removing the pulldown automatically in Vegas, I can step through it frame-by-frame. This looks to be the same basic technique you used, but it's much simpler to get at, and I think there are fewer potential variables involved. It's footage straight from the camera, vs. having to do any post-processing on it.

The footage features some cars driving by on the street. On the DVX footage you can step through frame-by-frame, and the cars advance by equal distances each frame.

On the FX1 footage, there's a very different motion pattern. It's like on even frames the car advances a little, then on odd frames it leaps forward much further, then on the even frames it goes a little, but it seems to be consistent on an even/odd basis.

Okay, so now we're getting to the bottom of the CF24 mystery. I exported 8 frames of the DVX footage, and found that yes, the car moves almost exactly 30 pixels each frame. For a consistent-speed object, the object spacing is exactly the same, showing that the CCD was sampling at the same rate as the footage created.

Then I exported the same frames from the CF24 footage (they were shot at identical times and sync'd up to exactly the same frame). In the CF24 footage, the car moves 27 pixels on one frame, then 39 pixels on the next. 27, 39, 27, 39, 27, 39... no wonder it's herky-jerky and stuttery!

Well, so, there you have it, the mystery of CF24 seems to be closer to being solved. However it's creating its 24P simulation, it's not doing it on anything approaching a consistent 24Hz time basis.

And this should forever put to rest the notion that "if you just remove the pulldown it'll be fine", because it isn't. Frank, your observations are exactly right.

[Ignacio Rodriguez]

There are by now variyng reports on the web about Cineframe 24. Most say it is terrible. Some say it is great. Granted some people may not "see" how bad it is but... could it be that the Cineframe firmware on some cams is better? Perhaps it would be a good idea for people testing Cineframe 24 to inform their cam's serial numbers. Just an idea.

[Charles Papert]

We "experimented" with shooting the Cinealta in 24 frame mode with the shutter off, and by experimented I mean we left the shutter off by accident. Switching it on and off, I could clearly see the difference with fast motion through the frame.

[Frank Aalbers]

<<<-- Originally posted by Barry Green : Frank, your observations are exactly right. -->>>

Woohoo ! ...

... actually ... too bad really ... :-(

Frank

[Sam Edwards]

Hi Guys,
I would say, based on my, Frank and Barry's tests that cineframe 24 is just the 2nd then 3rd field extracted and interpolated to create pseudo 24fps footage. To me the only advantage to shooting in cineframe mode is that in the GOP compression there's going to be less difference between frames because they're reapeated, so less stress on the compressor. This might give you sharper 24 fps footage than if you would have if you shot 1080i and converted in post. It would be a hard thing to prove. For cinema work a PAL fx1 in cineframe mode is probably the best choice if you need to do your project on the fx1.
Oh well. I still love the way the 1080i looks.
Thanks again guys!
Sam