HV20 24p Frame x Frame Breakdown - New Mystery?
OK, here my weak attempt to make a messageboard into a magazine article.
I just spent the last hour putting all this together to put to bed what the HV20 REALLY does when shooting in 24p. Now I'm in TOTAL understanding what the hell is going on here, and even more confused. As we think we all know, the camera performs 3:2 telecine on the 24p footage. By theory, there would be 24 progressive frames tucked into that 30 frame recording. I figured, and other have stated as well, that the camera just sticks the 6 crap interlaced frames to fill it in like any other normal 24p camera does. Guess again. Under close scrutiny, there are actually only 18 progressive frames in the 30 frames. THIS IS NOT TRUE TELECINE. I'll explain in later. Download and watch CLOSELY the below link which is a screen capture of a frame by frame breakdown for 30 frames utilizing a clip that is recorded in 24p on the HV20. (It's in .swf~see link at bottom of thread for link for a free player if needed)
There are 30 frames, I just duped tagging two different frames #8. Notice that there are only 18 progressive frames! The top track in the Vegas timeline is the text overlay pointing out the interlaced frames, the 2nd track being the progressive frames. You see the pattern as well as I do. Does not compute, bunky. The pattern from this HV20 24p recorded timeline looks like this if you aren't familiar with the Vegas format:
p-p-p-i-i-p-p-p-i-i-p-p-p-i-i-p-p-p-i-i-p-p-p-i-i-p-p-p-i-i = 18 progressive frames
Telecining a real progressive source should look like this (3:2 pulldown)
p-p-i-p-p-p-p-i-p-p-p-p-i-p-p-p-p-i-p-p-p-p-i-p-p-p-p-i-p-p = 24 progressive frames
What am I missing here??? There ain't flags that can be integrated with a firmware upgrade with this bastardized pulldown procedure boys and girls.
Below is the same clip above ran through TMPGEnc and broken down & advanced. Again, the marker advancement is done frame by frame, totaling 24 frames in the 1 second timeline.
First off, this does prove that TMPGEnc is doing pull down the correct way if set up right, which is good to know. Second, and more importantly, where the hell is it getting the extra 6 frames from? It must be assembling the 2 consecutive interlaced frames, or blending progressive ones (I think the latter)...I was sick of watching the timeline trying to figure it out (I felt like I was watching the Zapruder Film).
My stupid nature to always try figure everything out and overunderstand it is making me wacko with this camera. I just want to film again...
I think we all have been had as far as thinking we are shooting in true 24p. Am I missing something here? For the record, I think this little HD camera is WAY cooler than anything in it's class, it's just a complex little bugger that has my rusty wheels turning. I like that!!!!
Here's a link to a free swf player if you don't have one already to watch the above clips...
Last edited by acoustiking; 2007 July 26th at 01:01.
Oh yes, trying to understand things drives me wacko. And that, in turns, drives my wife mad. Once I've got my teeth into something, I have to bite, chew, gnaw, rip, and digest. Then I may let go.
Originally Posted by acoustiking
Very interesting observations. In PAL land, this doesn't affect me, but it has nevertheless grabbed my attention. I'll watch the footage...
Great little investigation. Unfortunately I have no answers, but I am most intrigued and look forward to hearing from someone who knows more about this. Couldn't you tell what it was doing to get those extra 6 frames by examining the frames you think it is blending? Then you may be able to confirm it is doing this???
I can confirm that it's always 2 interlaced frames and 3 progressive ones (although my test clip started with interlaced frames while yours with progressive ones). It's a non-standard pulldown removal that must be performed possibly, but as long as you know about it, you should be able to write the right code to do the de-interlacing/pullup.
You just need to blend the fields/frames from the two interlaced frames together and create 1 new progressive frame. The 6 "missing" frames you mentioned are nothing but the blending of the two interlaced ones! That will give you 24 frames.
BTW, many apps seem to get the pulldown right, but some dont, e.g. Premiere LE. However, in all cases the question is: how is the audio syncs back?
Last edited by Eugenia Loli-Queru; 2007 July 26th at 07:20.
So this camera can't record in 24p? That is essentially what this is about isn't it? I wonder if someone with a PAL camera could test their 25p mode to see what they have.
OK, that is really confusing me. But, thanks for the investigation. I hope you are wrong, others can investigate more. Perhaps this explain a little why, in your other post of testing TMPGENC, the after video file size is actually bigger than the before pull down video file size? Theoretically, after pull down, shouldn't the file size be 20% smaller?
Lets not get paranoid here. The 6 frames are being blended together from interlaced frames in a 'legitimate' progressive manner. If you notice, the more expensive progressive Canon HD cameras dont have true Progressive-scan chips. Instead they rely on 24CF, a process whereby a progressive image is derived from interlaced frames.
Now, the important question is, why is Canons method legitimately progressive, while Sony's CF mode not? Its all in the clock, or 'cadence'. Sony cams shoot at 60hz, hence each interlaced frame is slightly out of sync with the other. This gives a great interlaced picture, but trying to derive one progressive image from two consecutive frames, especially if there is motion, is impossible.
The canon, on the other hand, in progressive mode, takes pictures at 24 frames per second. I.e, it has a true 24 FPS cadence. Even if you do have two pictures (interlaced), they are from the same moment in time, and hence they recombine into one progressive picture seamlessly.
I hope that puts some worried minds to rest until some other geek really gets out there and tests the hell out of it.
Originally Posted by sajid
That seems to be the crux of the issue that was confusing me. So it seems that you really are creating a true progressive frame when you combine two of the "i" frames into a "p" frame. Thanks for that info.