video data rate analysis tools

[revsykes]

I'm authoring a DVD and I've have some problematic footage from animated stills (long pans Ken B. style.) I've been having a really hard time getting it to look smooth on playback on computer DVD players, set-top is fine.

I've tried a LOT of different bitrates and settings in Procoder but it doesn't seem to matter, the playback continually seems to get stuttered in certain places on certain computers that just don't seem to be able to keep up. It doesn't seem to happen at the same frames everytime, and the audio stays synced and clean.

Because of the panning animation, I am assuming this is the CPU not being able to keep up with constant changes in the video information? But like I said, I've tried the lowest and highest data rates and literally seen no difference in the outcome.

Is there a utility or tool I can use to analyze these files in depth so I can tell what is going on, if there are spikes in the bitrate, or if there is something I can do to fix it? Would it have anything to do with B-frames or I-frames?

Because of the amount of time spent, I really need to get a solution to this issue so that I can apply it to future productions, where I will be getting a lot of similarly animated stills in the footage.

thanks in advance.

[guns1inger]

Generally, if a settop box ca play it, then all but the slowest PCs should be able too. Standalone players are more likely to choke on bitrate that is too high because of their slower readers and smaller buffers.

There are a number of bitrate viewers in the tools section. Search there, read the comments from users, then try a couple out. Most seem to be inconsistent at best.

[revsykes]

Generally, if a settop box ca play it, then all but the slowest PCs should be able too. Standalone players are more likely to choke on bitrate that is too high because of their slower readers and smaller buffers.

I always understood that to be true as well, but I've always had more trouble with DVDs on computers (besides compatibility/compliance issues.)
Is there a reason this might be happening besides data rate then? Like I said, encoding with a lower bitrate doesn't seem to help the problem at all. What about "keyframes" in relation to motion/transitions? You can see some evidence of it upon playback from a mounted disk image as well, so I've ruled out the media or the burning app.

[mikiem]

Why PCs and not TVs? TVs are probably showing your content at lower rez & interlaced. Most PC players use a couple of methods to deinterlace - show the vid progressive to match display. Best smoothness on a PC would use all progressive frames at higher frame rates, like 60 or 90 or... If it's just playback for yourself, might be simplest to try other players and/or settings for bob, weave etc. (include video card settings). Might also try real interlace if you're using pulldown, or try progressive & pulldown if you're using real interlaced. It is possible to combine the 2, though not sure it can be done at a budget level if that applies.

Purely a guess, might try different frame sizes within DVD spec. My theory is that forcing a resize/redraw might have an effect, but could be totally wrong.

"What about "keyframes" in relation to motion/transitions? "

RE: mpg2... You *might* want to look at the SnellWilcox.com docs if you're interested in the theories of it, but otherwise the nature of the mpg2 video compression is it has problems with pans. Different encoders and a different matrix might help -- might help to look into methods encoding the video in segments, using different setups depending on the scene. And of course this is all assuming there aren't any prob with the original video.

"Because of the panning animation, I am assuming this is the CPU not being able to keep up with constant changes in the video information? But like I said, I've tried the lowest and highest data rates and literally seen no difference in the outcome. "

The cpu is involved only in decoding the video, and in some mainly PC formats like wmv, frames will drop if the PC can't decode them fast enough. But this is talking some *highly* compressed video with large frame sizes or anamorphic. To check it out, open task manager or some similar software that shows CPU utilization while playing video.

Higher bitrates mean less compression so less artifacts etc., but... as you've said not your prob at the moment.

[jagabo]

DVD video will not play perfectly smoothly on a computer monitor unless it's running at the same frame rate -- or an exact multiple.

Even with a monitor refresh that's an exact multiple or the video frame rate, say 75 Hz with a 25 fps video (where in theory each film frame can be displayed for exactly 3 refresh periods), playback will not be perfectly smooth because Windows is not a realtime operating system.

[revsykes]

If it's just playback for yourself, might be simplest to try other players and/or settings for bob, weave etc. (include video card settings). Might also try real interlace if you're using pulldown, or try progressive & pulldown if you're using real interlaced. It is possible to combine the 2, though not sure it can be done at a budget level if that applies.

Thanks for the reply! Can you explain this a little more? I'm not sure I fully understand the settings comment and the suggested interlace ideas.

The source is no fields, animation. As far as deinterlacing, What I am seeing is not a continuous "jitter" or any kind of combing or anything that I would attribute to interlace vs. progressive issues. It definitely seems more like a processing issue, as it is all just progressive and the TV playback looks smooth (Therefore I assume this is not an issue of the set-top de-interlacing or the computer showing the jerky progressive playback.) This is an intermittent "stutter" that doesn't happen consistently or in the same spot, but rather when the frame seems to have a signifigant change while panning. It definitely seems more like a processing problem or a codec issue?

I will ty one of the data rate analysis tools mentioned above and see if there are any clues.

I realize commercial DVDs are compressed by pros with high dollar equipment, but I've made many DVDs in the past that don't have any problem with smooth motion; I'm just finding it hard to believe that you can't animate a pan on a still image and get it to work correctly.

[jagabo]

This is an intermittent "stutter" that doesn't happen consistently or in the same spot, but rather when the frame seems to have a signifigant change while panning. It definitely seems more like a processing problem or a codec issue?

That sounds more like a codec issue. Or a deinterlacing issue. Or mabye an antivirus, virus or spyware issue.

How long are the stutters? A frame or two? 1/4 second? Longer?

Was your DVD encoded as 23.976 fps progressive marked for 3:2 pulldown? Or is the MPEG/VOB file already pulled down to 29.97 fps? If the latter, and the DVD player software is performing a realtime BOB deinterlace, it may be running out of horsepower.

Some combinations of MPEG splitters, MPEG decoders, and video (display) renderers aren't fully compatible. If the computers in question have had different video editing programs and DVD players installed I wouldn't be surprised to see problems like this.

Bitrate is unlikely to be a problem. Any modern processor has the horsepower to decode and display DVD bitrate MPEG2 files. Bitrate is much more likely to be a problem with set-top DVD players.

[revsykes]

How long are the stutters? A frame or two? 1/4 second? Longer?

The video stutters are about a 1/4 second, yes, very quick, and the audio does not stutter.

Was your DVD encoded as 23.976 fps progressive marked for 3:2 pulldown? Or is the MPEG/VOB file already pulled down to 29.97 fps? If the latter, and the DVD player software is performing a realtime BOB deinterlace, it may be running out of horsepower.

Some combinations of MPEG splitters, MPEG decoders, and video (display) renderers aren't fully compatible. If the computers in question have had different video editing programs and DVD players installed I wouldn't be surprised to see problems like this.

Bitrate is unlikely to be a problem. Any modern processor has the horsepower to decode and display DVD bitrate MPEG2 files. Bitrate is much more likely to be a problem with set-top DVD players.

Does deinterlacing really cause this kind of issue? As far as I know, nothing has been de-interlaced, nor is pulldown involved. I am exporting original footage (animation) from After Effects, at 29.97 fps, with "no fields" selected. As far as I know, this means there have been no fields from the first render forward; and so everything should be just frames? Just to clarify, I am not ripping from commercial DVDs or authoring for backup purposes, this is all original production work, the audience is mostly academic, and they will be playing the DVDs on laptops/computers most often.

The computers having problems seem to mostly be older Powerbook laptops, running OSX DVD player. Stutter issue still happens (but rarely) on the Windows machines I've tried. I had no idea that the video editing software installed had anything to do with it, the final encoding is done in Procoder after exporting to 90% PJPEG or HUffYUV from AE.

[guns1inger]

When you encode with procoder, do you output 29.976 interlaced, or 23.976 with pulldown ?

DVD can be 23.976 progressive or interlaced (with pulldown), and 29.976 interlaced. I don't believe it can be 29.975 progressive (only look at NTSC and ignoring PAL for the moment).

I am wondering if you aren't outputing 29.976p from AE, then using ProCoder to produce 23.976p with 2:3 pulldown. This would in effect, throw away 6 frames for each second of footage, and depending on how, could well produce stuttering.

[revsykes]

I am wondering if you aren't outputing 29.976p from AE, then using ProCoder to produce 23.976p with 2:3 pulldown. This would in effect, throw away 6 frames for each second of footage, and depending on how, could well produce stuttering.

Ok, this might be shedding some light. I didn't know this, I thought footage could be just non-interlaced frames and still be 29.97 fps, (just losing half the fields per second.) I am going to do a couple more tests today. Settings from Procoder are non-interlaced on the source settings, but I did notice there is no option available for non-interlaced for the target settings, it is defaulting to BFF. Could this be the problem?

[jagabo]

Frame rate conversion (or improper IVTC) from 29.97 to 23.976 fps would be jerky on any viewing device.

29.97 fps progressive is fine for DVD.

[mikiem]

1st off, apologies to everyone because answering this seems to ignore many great posts. I really don't mean to ignore anything, but felt it *might* be more readable (I don't write well) this way. Most of this has already been mentioned -- just my way of wording it in answer. Thanks.

"Thanks for the reply! Can you explain this a little more? I'm not sure I fully understand the settings comment and the suggested interlace ideas. "

RE: player/card settings... Usually/Normally DVD spec vid will be interlaced, using either pulldown with 24 p or regular interlaced NTSC or PAL video. When you play interlaced video on your PC's progressive scan monitor, it can look bad to many folks, so the video's fields are combined into a progressive frame before it's displayed. In Power DVD 6 you click the advanced button on the video tab of the configuration window. With ATI catalyst there's a choice as well. Changing the way the video's deinterlaced, or if it is deinterlaced, might make a difference for you.

Changing the video's format before it goes on the DVD *Might* have some effect. If the video just has pulldown added to 24 p, rendering say 29.976 video instead might deinterlace better -- or not.

"that I would attribute to interlace vs. progressive issues..... This is an intermittent "stutter" that doesn't happen consistently or in the same spot, but rather when the frame seems to have a signifigant change while panning. It definitely seems more like a processing problem or a codec issue? "

Your TV is getting ~60 fields/second to display, so a given scene has more snapshots or time slices. Deinterlacing on your PC you've got half that number. And, if the content is actually interlaced, some motion is only available in fields, plus combining them (deinterlacing) can muddle things a bit.

That said, it may very well be a processing issue. If you're converting 24 p to 29.976 i, the way those extra frames are derived determines how they look deinterlaced by your PC's player software. During mpg2 encoding, my understanding is that the motion estimation & compensation steps are really critical when you're processing pans. This is something you might alter, though it's a compromise, where what helps your pans may hurt your other video, hence using different processes to encode different scenes & putting it all together.

[edDV]

...
Changing the video's format before it goes on the DVD *Might* have some effect. If the video just has pulldown added to 24 p, rendering say 29.976 video instead might deinterlace better -- or not.

I'm not quite following. My understanding is NTSC DVD is limited to these choices for authoring:
1 29.97 interlaced
2a 23.976 progressive with pulldown flag that telecines to 29.97 in the player.
2b 23.976 progressive with pulldown flag that plays out 59.94 frames per second for progressive playback over YPbPr.

29.97 progressive is not supported.

Now on the standalone DVD player side,
1 29.97 interlaced will play 29.97 interlaced*
2a 23.976 progressive with pulldown flag will telecine to 29.97 in the player.
2b 23.976 progressive with pulldown flag plays out 59.94 frames per second in progressive playback mode over YPbPr. In some players "2a" continues on the composite and S-Video outputs.

On computer deinterlacing DVD players
1 29.97 interlaced will play 29.97 progressive using telecine detection*, bob, weave, blur, blend, or other Hokey-Pokey as chosen in the deinterlace settings menu. Each method has limits and "special" defects.
2a 23.976 progressive with pulldown flag will play progressive at some multiple of 23.976 that best matches screen refresh rate.
2b same as 2a

ATSC DTV and future HD DVD add the following
- 29.97 progressive for 480p, 720p and 1080p
- 59.94 progressive for 720p and 1080p (480p is currently supported for DVD)
- 29.97 interlace for 1080i (480i is currently supported for DVD)

I've probably forgotten some combinations.

* "Cinema" telecine detectors may be present to detect 3:2 sequence. IVTC is optionally performed to convert telecined 29.97 back to the film's 23.976 progressive. The output is then frame repeated to 59.94 progressive output on the component connections.

[revsykes]

I'm not quite following. My understanding is NTSC DVD is limited to these choices for authoring:
1 29.97 interlaced
2a 23.976 progressive with pulldown flag that telecines to 29.97 in the player.
2b 23.976 progressive with pulldown flag that plays out 59.94 frames per second for progressive playback over YPbPr.

29.97 progressive is not supported.[/quote]

If this is the case, I'm really confused now. I've never had anything that was anything other than animation from AE, or regular interalced DV capture. (there is no DV in the sample I am talking about now.) Never any step that would introduce telecine, pulldown or 24p as far as I understand it. When footage is exported from AE, I don't change the frame rate to 24 fps or anything like that, and there are no fields, so it is 29.97 fps progressive as far as I understand it.

[edDV]

I'm not quite following. My understanding is NTSC DVD is limited to these choices for authoring:
1 29.97 interlaced
2a 23.976 progressive with pulldown flag that telecines to 29.97 in the player.
2b 23.976 progressive with pulldown flag that plays out 59.94 frames per second for progressive playback over YPbPr.

29.97 progressive is not supported.

If this is the case, I'm really confused now. I've never had anything that was anything other than animation from AE, or regular interalced DV capture. (there is no DV in the sample I am talking about now.) Never any step that would introduce telecine, pulldown or 24p as far as I understand it. When footage is exported from AE, I don't change the frame rate to 24 fps or anything like that, and there are no fields, so it is 29.97 fps progressive as far as I understand it.

I'm talking about the standards as I understand them. It comes down to 29.97 interlace or 23.976 progressive on the DVD media. It may be possible to author 29.97 progressive for computer playback, but it would be outside of the standard unless I've missed something.

[jagabo]

I'm not quite following. My understanding is NTSC DVD is limited to these choices for authoring:
1 29.97 interlaced
2a 23.976 progressive with pulldown flag that telecines to 29.97 in the player.
2b 23.976 progressive with pulldown flag that plays out 59.94 frames per second for progressive playback over YPbPr.

29.97 progressive is not supported.

If this is the case, I'm really confused now. I've never had anything that was anything other than animation from AE, or regular interalced DV capture. (there is no DV in the sample I am talking about now.) Never any step that would introduce telecine, pulldown or 24p as far as I understand it. When footage is exported from AE, I don't change the frame rate to 24 fps or anything like that, and there are no fields, so it is 29.97 fps progressive as far as I understand it.

There is nothing wrong with burning 29.97 fps progressive source material onto a DVD. I do this all the time with video from my 29.97 fps progressive DV camcorder. In my experience you can encode as either progressive or interlaced MPEG2. Final display on SD TV is the same.

Actually there is a slight difference. Since MPEG internally treats progressive and interlaced data differently (because of the chroma subsampling) there is a difference in colors on a small scale. I can see this looking at enlarged still frames on the computer. Not while watching playback on a TV. I'm sure one could create some test footage where there is an obvious difference on TV.

Aside from the internal MPEG handling, the difference between interlaced and progressive in this context is whether both fields come from the same point in time (or picture). If they do the video is progressive, if they don't the video is interlaced. The DVD player will display them the same way on SD TV -- one field at a time, 59.94 fields per second, the only thing SD TV can display.

Will a progressive DVD player handle the two differently? Does it deinterlace the interlace encoded DVD (which doesn't really need it) and not deinterlace the progressive encoded DVD? I don't know. And what about software players?

If you think progressively encoded MPEG is causing problems then lie to your MPEG encoder and tell it your source is interlaced. It won't know the difference. As far as the encoder is concerned it's just two fields packed into a frame.

[edDV]

I'm not quite following. My understanding is NTSC DVD is limited to these choices for authoring:
1 29.97 interlaced
2a 23.976 progressive with pulldown flag that telecines to 29.97 in the player.
2b 23.976 progressive with pulldown flag that plays out 59.94 frames per second for progressive playback over YPbPr.

29.97 progressive is not supported.

If this is the case, I'm really confused now. I've never had anything that was anything other than animation from AE, or regular interalced DV capture. (there is no DV in the sample I am talking about now.) Never any step that would introduce telecine, pulldown or 24p as far as I understand it. When footage is exported from AE, I don't change the frame rate to 24 fps or anything like that, and there are no fields, so it is 29.97 fps progressive as far as I understand it.

There is nothing wrong with burning 29.97 fps progressive source material onto a DVD. I do this all the time with video from my 29.97 fps progressive DV camcorder. In my experience you can encode as either progressive or interlaced MPEG2. Final display on SD TV is the same.

Actually there is a slight difference. Since MPEG internally treats progressive and interlaced data differently (because of the chroma subsampling) there is a difference in colors on a small scale. I can see this looking at enlarged still frames on the computer. Not while watching playback on a TV. I'm sure one could create some test footage where there is an obvious difference on TV.

Aside from the internal MPEG handling, the difference between interlaced and progressive in this context is whether both fields come from the same point in time (or picture). If they do the video is progressive, if they don't the video is interlaced. The DVD player will display them the same way on SD TV -- one field at a time, 59.94 fields per second, the only thing SD TV can display.

Will a progressive DVD player handle the two differently? Does it deinterlace the interlace encoded DVD (which doesn't really need it) and not deinterlace the progressive encoded DVD? I don't know. And what about software players?

What you are describing is supported in DTV and future HD DVD standards but not in the current DVD standard and standard players as I understand it. That doesn't mean you can't use current DVD technology creatively to get around the standards.

24p DV camcoders (like the AG-DVX-100) record full 24p progressive frames but use standard or "advanced" telecine techniques to build a 29.97 interlaced output. As far as DVD is concerned, this is an interlaced source.

The AG-DVX-100 can use normal 2:3 telecine and be detected as 24p film, or it can use the new 2:3:3:2 "advanced pulldown" sequence that makes 24p editing easy, but would probably screw up player IVTC.

http://www.adamwilt.com/24p/

[jagabo]

The AG-DVX-100 can use normal 2:3 telecine and be detected as 24p film, or it can use the new 2:3:3:2 "advanced pulldown" sequence that makes 24p editing easy, but would probably screw up player IVTC.

Why wouldn't one just edit that as 24p and mark for 3:2 pulldown when encoding to MPEG? Oh nevermind, I know. DV doesn't support 24p so 24p has to be telecined into 30i on the tape.

[jagabo]

revsykes, regarding my earlier speculation about software conflicts, it may help you to know a little about how DirectShow handles video (although it sounds like most of the problems you were seeing were on Macs).

Directshow uses a series of filters (modules) to read and display video files. Data is pipelined between those modules. Every time you start playing a new video the series of modules is pieced together. The linked filters are called a filter graph. A program can build it's own graph, specifying which filters it want's to use, or it can have DirectShow build the graph for it, or anywhere in between.

1) First in the pipeline is a file reader. This simply reads data from the video file. It detects what type of video and audio is in the file and then tells that to directshow. It may perform some read ahead buffering but it doesn't otherwise process the data as it feeds it to the next step in the pipe.

2) Directshow then looks for a filter that knows how to handle multiplexed MPEG video and audio. This is usually an MPEG demultiplexor. The demultiplexor splits the audio and video into separate streams and passes them along.

3) The video stream is piped to an MPEG video decompressor and the audio data is piped to an audio decompressor.

4) The output from the MPEG video decompressor is piped to a display driver filter that puts the data on the screen (it may have to perform a colorspace conversion in the process). The audio is piped to an audio driver filter that plays the audio.

Deinterlacing can take place in steps 3 or 4. That is, either the MPEG decompressor or the
display driver can deinterlace the video. Deinterlacing can take significant CPU resources.

When you have several video applications installed, they may each install their own file readers, demultiplexors and decompressors. When Directshow builds the filter graph it may take a demultiplexor from one program and a decomrpessor from another. This can lead to any type of problem like no playback, jerky playback, messed up video, etc. There are even cases where an audio filters can mess up the video playback.

[manono]

Hi-

29.97 progressive is not supported.

Sure it is. It's usually encoded as interlaced, but may contain 30fps progressive video (2:2 pulldown). I see it quite a lot. A lot of newer anime contains CG 30fps progressive sequences. I've seen a number of extras on DVD containing interviews shot with 30fps progressive cameras. The Criterion logo on their DVDs is 30fps progressive.

The first major film shot in Todd-AO was Oklahoma!. The Oklahoma! DVD features a transfer from the 30fps Todd-AO print. Like all DVD material, it is decoded as interlaced fields, but unlike video, pairs of fields do belong together if you want to recreate the original film frames. Now we've got another type of deinterlacing: 2-2 pulldown where pairs of fields need to be woven together and each displayed twice (2-2-2-2 as oppose to 3-2-3-2).
.
.
.
Before you start to think we are concerned about 2-2 pulldown for the sake of a couple classic films, there is another source of true 30 frames per second material which is more common than you might think: The proverbial sit-com such as NBC's Friends. These shows are shot with film at 30frames/ps and transferred to 60fields/ps video to give them a softer film like look. In consequence, there is a lot of vertical resolution waiting to be reclaimed through 2-2 playback.

You'll find that in the "Deinterlacing Other Sources" section a ways down from the top here:

http://www.hometheaterhifi.com/volume_7_4/dvd-benchmark-part-5-progressive-10-2000.html

Although when reencoding such material, I encode and author it as 30fps progressive, and have no problem with playback via software or standalone (although I do notice periodic short stutters using PowerDVD), it could be that the players used by the OP sometimes have problems with it. If originally encoded as progressive, you might try reencoding it as interlaced. It could be that the OSX player you mentioned can't perform 2:2 pulldown, or chokes on it for other reasons.