MSP8: MSP rendering DV clips in and out of focus???

[pennstat]

Here's one I've never run into before.

My current project has an interesting problem where the final, rendered video goes in and out of focus in a consistent, rhythmic pattern -- every second or so. It's not a major focus loss, but it is definitely on a consistent pattern and can easily be seen when viewing any details in those clips.

At first I thought that the camera was going in and out of focus and that I never caught it. However, the problem does not exist on the original DV files. I confirmed this by playing back the original, capture files followed by the rendered clips. There are no focus problem on the masters. But after rendering, whether as DV AVI or MPEG-2, they go in and out of focus rhythmically. Most of the master files render without a problem. This is only happening to a few clips.

Also, there is no consistency to the clips that are affected. The clips were all captured the same way; they are not restricted to one particular recording session or camcorder (I use two camcorders simultaneously for my live recordings); they are not from the same capture session; they do not differ in structure or format (all are DV); they all play without any problems directly from their source file. But something about the rendering is making the final video blink in and out of focus. I even thought that being on a particular cue (such as video 3) was causing the problem, but moving a clip to the same cue as an unaffected clip yielded the same results.

I'm baffled by this one. Anyone have any ideas? I really don't want to have to rework the composition for those particular scenes to eliminate this rendering glitch, but I will if I have to.

[tv_news_guy]

pennstat

Where are you seeing this? On the pc monitor? On a tv attached to the pc for viewing source and rendered files? On the viewfinder of your camera after transferring the rendered dv files back to tape?

The only reason I ask is to try to narrow the (pardon me) focus of where your problem is. If they play back fine on the comp.. but not the camera its one thing.. etc etc.

If you haven't exported them to tape yet... try one.

Let us know.. I for one would be glad to help if I had more information.. also are these ntsc or pal and are they dv1 or 2 avis?

Bill tv_news_guy
Non-linear editing is half art/half brain damage

[pennstat]

pennstat

Where are you seeing this? On the pc monitor? On a tv attached to the pc for viewing source and rendered files? On the viewfinder of your camera after transferring the rendered dv files back to tape?

All of the above. Actually, I did not render back to tape but straight to DVD-quality MPEG-2 at variable 8000/bps. I noticed the focus issue when I was watching the DVD for initial testing and writing down problems for further editing. When I found no commonality between the affected clips, I tried viewing all of those scenes' masters through the monitor as well as an external TV that I use in a "dual head" fashion. In all instances, I saw no problems with the original captures but the "focus" issue could be seen on the rendered files. I even tried additional rendering with various options to no avail.

Actually, I shouldn't call it "focus". That's not necessarily accurate, but I don't have another name for it. I did notice, however, that the scanning on the screen (yes, I can see it if I look hard enough) differs when it goes "out of focus", almost like it's an interlacing issue. But that doesn't explain why it happens every second. (Actually it's a bit faster. More like every 2/3 seconds.)

The only reason I ask is to try to narrow the (pardon me) focus of where your problem is. If they play back fine on the comp.. but not the camera its one thing.. etc etc.

Doesn't matter. I can see the focus blips after rendering but not on the masters regardless of the display. I can always let another product, like Pinnacle Studio, render the masters to see what happens, but I don't have a lot of time to play the trial-and-error game. Since I used Studio to do the capture, it might not be an even comparison anyway.

also are these ntsc or pal and are they dv1 or 2 avis?

NTSC. And I'm not quite sure about the format. I capture using Pinnacle Studio over Firewire. (Sorry, Ulead, but I've had tons of problems with MSP's video capture tool in the past.) I was under the impression that it captures DV1, but MSP reports it as DV2. Regardless, I don't think that's an issue here because all of the clips were captured using Studio, but only a few are showing this little issue. In fact, I've been using Studio to capture for over five years and have not run into this until now.

Non-linear editing is half art/half brain damage

With a pinch of masochism.

[pennstat]

Well, I found the cause. Looks like Ulead might have another bug to fix.

It turns out that all of the affected clips were used as overlays to a different video or audio track. In order to keep them in sync with the audio, they were all accelerated to between 102% and 104%. It would seem that MSP's algorithm for accelerating video clips throws the upper/lower field sequence out of whack during the final rendering.

The only thing that I can think of is to change those particular sequences to non-interlaced. There is very little movement in them so combing should not be an issue.

Any other ideas are welcome.

[Devil]

I understand that Speed works by inserting or removing frames at regular intervals. It does not interpolate, which would require some hefty maths operations. For example, if you were running at a slow-down of a setting of, say, 103%, it would add a frame by duplication after 33 frames. I can't see this causing the problem you describe, unless it adds a field after 16 frames, which would upset the field order. If this is the case, I could foresee an image deterioration between frames 17 to 33 of every cycle, but I speculate here, without any certain knowledge.

If I were you, I'd try working the clips you wish to stretch/shrink in deinterlaced mode and see whether that works better.

[Gorf]

With 3 and 4% changes, it would be far easier to adjust the audio than the video, anyway.

With feature films put on PAL video or DVD, the footage is speeded up from 24fps to 25fps and the audio similarly changed. None of this messing about with "pulldown" that NTSC viewers have to endure. The only required adjustment is a slight pitch change in the audio to accommodate the fact that it is running faster.

[Devil]

And, if you wished to correct the pitch afterwards, you can do so in Audacity or even Audio Editor.

[pennstat]

I understand that Speed works by inserting or removing frames at regular intervals. It does not interpolate, which would require some hefty maths operations. For example, if you were running at a slow-down of a setting of, say, 103%, it would add a frame by duplication after 33 frames. I can't see this causing the problem you describe, unless it adds a field after 16 frames, which would upset the field order. If this is the case, I could foresee an image deterioration between frames 17 to 33 of every cycle, but I speculate here, without any certain knowledge.

I'm just speculating as well. I've done this before multiple times with MSP 6 and 7 and never had this problem before. But the only clips that encountered this were ones that were accelerated. Oddly enough, there is one clip that was slightly decelerated (99%) that did not have this problem. So it would appear to be restricted to acceleration based on what I've seen.

I took a closer look at it last night. It looks like for a specific interval the resolution is lost, almost like an off-balance progresive scan, where the scan lines are being replicated one line off where they should be. Again, that's just my observation. But I do know that I never had this problem with any other version of MSP.

Not a major issue, but I'll have to keep it in mind in the future. I got a lot of work done last night, which actually put me ahead of schedule. So I'll see if I can try a few edits with those particular sequences deinterlaced. I'm curious now.

[pennstat]

I tried changing the media source type for all sped-up clips to "frame based" and it got rid of the problem, although if you look really closely you can see where frames are cut out to sped the clip. I'm just about certain that the problem is caused by the inadvertent flipping of the upper and lower frames during the padding or removal of frames to accomodate the speed change.

Unfortunately, I found another bug in MSP 8 because of this, which I will post in another thread.