Any GH2 news on the HDMI output front?

Ian-T said:
That's exactly how Cineform does it on all telecined clips. It takes about a second before it starts to remove the puldown...so it's not unusual to see some interlacing in the first second or so.
Click to expand...

So I have high hopes that they'll figure out whatever is unique to the GH2 and update the app to handle it.
 
I don't think I am confident enough with this to go through the trouble during real production. I may do a few quick tests though. I am surprised more people are not weighing in on this.
 
I found the 17 second cripple/bug/thing!

Out of curiosity I decided to do these tests in 720p mode. Of course, the capture is still 1080i (that is the only thing that the GH2 send out live. Interestingly enough, it has progressive frames for 17 seconds, then interlaced for 17 seconds, then progressive for 17 seconds. Actually, it is more like 16.75 seconds... but this is certainly the magic (evil) number we have heard mentioned in the past.

Here is a sample: http://bit.ly/hKbhHO

Another interesting note: I have run the test multiple times. When I increased the shutter speed to 60+ the hdmi signal becomes ALL INTERLACED. Anything under 1/60 has the 50/50 mix of interlaced and progressive. This is obviously tied to how the gh2 changes to 30fps when in 720p mode if you have a low shutter speed that makes 60p impossible. I will now test what happens when you change the shutter speed one the fly!
 
Last edited:
Even some test shots tomorrow would be awesome - to demonstrate the difference in keying from AVCHD to something a little less aggressive (in the compression department). So if you can spare the time to run a 1 minute test involving flying hair or even a dancing mop head (:)), that would be cool. If the difference can be demonstrated clearly enough, more people might actually start to get interested in this thread. For now, a lot of people may just be thinking, "Well, the footage LOOKS good enough to me, so why the extra headaches?" Call it a marketing effort :)

I tried to remove pulldown on your 1_min test (1080 stuff), and it's impossible. My guess is that you exported that piece from a timeline as opposed to just making a shorter version of the 5_min test using Quicktime's copy-paste method. Is that what happenend, or are we seeing a brand new bug?
Interesting about the 720p thing and the 17 seconds popping up again! Have you tried higher shutter speeds in 1080 24P mode? See if we lose p-frames there too?
(Also, your link for the 720p stuff isn't working yet.)

We may be looking at a semi-crippled workflow here, where the interlacing gets messed up every 8 seconds or so. I don't know if Cineform can solve this - never used them. But if it takes their exports a second or two to figure out the cadence, then you still lose those seconds every 8 seconds - doesn't seem like a great solution. I think I would just do my editing in 29.97 timeline, and once I know what my clips are, send them to shake to have the pull-down removed. And if there's an interruption in the middle of a clip, I'd break the clips up for each solution. Then just bring everything back onto a 24P timeline. I haven't tried that workflow, other than for effects work, but it's where I'd start if I was shooting for perfection.
 
Can you confirm how the 1_min test file was made? Was it exported from a timeline or created by copying a piece of the original file using QT Pro?
Thanks.
 
It should be fine. Have you looked at it? It's an irreparable mess. Interlacing all over the place. QTP should have been perfect. Now I'm a little scared. Fixing a repeating blip is one thing. This is not. Guess I better go for the big file and look at it.
Can you think of anything that might have changed? Have you tried to remove pull-down with it using Cineform (I hope that's not cheating and just doing interlace interpolation).
Hmmm . . .
 
I meant for that particular clip . . . the 1_min one. It's the one I can't get pulldown removed from.

I haven't downloaded the 5_min one yet - it's on my ToDo list :).
 
Good question. The main advantage to shooting 24 fps is that the camera can do it in a progressive mode, which gives you your best resolution. I don't think any of the PAL options are progressive (yet?). So you're limited to interlaced footage that can only be interpolated. It's the biggest curse of video - the interlace.
I'm not sure how the 24P streams on PAL cameras for the 1080i over HDMI, but we haven't seen a PAL camera record over HDMI with pull-down. Don't know why, but wonder if there's a 50i cadence being applied, which would really mess up a pull-down being added to it. I'm not saying the PAL cameras are doing it, but aemonk's been trying to get 24 p through HDMI using AJA product, and everything stays interlaced, no matter what modes one might try to remove the pull down. At least with my tests. Of course, you're welcome to try and see if you can figure out what's happening there.
 
Just took a look at this vimeo post http://vimeo.com/17727380
y. kimura has kindly been doing his own testing. I'm downloading his direct link but it's hours from being done.
I have downloaded the vimeo file, and found a section where he had recorded 24p over HDMI for over 8 seconds. It begins around :55 and continues on to 1:04 or so.
Sure enough, as you approach the 8 second mark, the interlacing starts changing cadence all over the place. Up to that point, it's consistent.
So, it's nice to see that this characteristic isn't limited to just one camera. Well, maybe it's not nice, but . . . knowing what you're up against is better than being blind.

I guess the next step is to come up with a working strategy to recover each frame as an original progressive frame.
 
OK. Figured out what was wrong with 1_min clip. It got scaled down! At least the one I downloaded did. HA! That'll get me spinning.
Finally went through the 5_min file. Everything is fine. Predictable trouble every 180 frames or so, and easily solved. All progressive frames are still there. Just a little crochet work, and then advance to next setting, and off we go. Feeling pretty good about a sub-$1000 camera that can offer this capability. Is it worth another $4,000 to have all the other features of the AF-100 and NOT have to mess with this?
Depends on your budget and time requirements. Either way, an off-camera recorder is needed. So who's not going to spring for this little baby just to play around with, knowing that if the pull-down doesn't work on a certain point in time, it can be quickly solved with a little massaging?
Now, beggars can't be choosers, but if someone shoots some chromakey footage, and posts both file types somewhere so we can do some proof-of-concept stuff, that would be pretty sweet.
Just sayin' :)
 
Hmmm, so you were even able to get the rough 8-second-ish parts to work? It's just a change in cadence? Were you able to get the whole thing to come out progressive with "massaging?"

If this is all true, it is reasonable to assume that a tool could be made as a one-click solution to this since it is predictable and consistent. I am not willing to babysit a clip modify pulldown settings by hand every 8 seconds, because I am working with long-form content.
 
Yes, but the key is "massaging." In other words, I haven't found a predictable way to find when the trouble starts. Maybe it is predictable - I just haven't charted it. But, I now know where to look. And it doesn't take long to fix. But, yes, a complete progressive clip can be made with your 5_minute piece. Automatically? Not yet.

For long form work, it would be tedious to conform first -using this method - and then edit. Especially if you're shooting like a 10:1 or more ratio!
This may be best left for doing effects work and shorts.

I'll do some more poking around later tonight (it's morning now). Got another schedule I have to keep . . .

What I am seeing is that the clip starts with a cadence of AA, for example. When the interlacing starts to stumble, I select BB for a frame, then go back to AA for another 4 or 5 frames or so. Then, when the interlacing problem hits again, I go to BB for another 180 or so frames. When interlacing skips a beat, I go to BC for that frame, then back to BB for another 4 or 5 frames, then the interlacing stumbles, and I set it to BC for another 180 or so frames. It seems easy to find, and the solution is predictable, and all the data is there. Can someone write a script to predict it? I don't know. I'd love to try a script in Shake, but I have to see if the trouble spots are that predictable or not. I guess, if it's being driven by technical issues, it should be predictable.

I'll let you know what I find out once I get back into it again.
 
Thanks for continuing to investigate this. If it turn out this issue can be precisely defined, then maybe an automated software fix is possible in post. Perhaps the Cineform folks will want to take it on? I also wonder if this is something that Panasonic may eventually address in firmware. In any event, thanks for scoping it out.
 
I won't hold my breath on Panasonic helping us get an unintended feature from the camera. But true, I think the Cineform guys would be able to figure out the pattern quite easily. And considering that they added GH1 specific features to the software already, it's not unlikely that they have their eye on this camera already.
 
You must log in or register to reply here.
Back
Top