For live recording, what DVD "speed"/length to use?

Factual Questions
1

I work for our local PEG Channel (Public/Education/Government), or Community Access. We are starting to use a standalone DVD recorder (Panasonic DMR-T3030) instead of the VHS units to make our government-meeting recordings.

I noticed that this recorder, using DVD R, can be set in advance for 1,2,4 and 6 hours for a single-sided, single layer DVD blank. The question is, which to use? We certainly would like to avoid the time it takes and the loss of information during a media change under live conditions, but we also want to get the best quality of recording possible.

I don’t believe DVDs could be correctly described as having “speeds,” like a tape recorder, where the lower speed results in a poorer signal. So what is the diff between the 1,2,4,6-hour settings?

My guess is the diff is in the compression scheme and/or level. I am familiar with how compression works in computer data (LZW, JPG, GIF, ZIP) and audio (MP3), but the nature of video provides other options. For ex, video can compare adjacent scenes and only store the differences, not the entire scene, giving a high-compression ratio with a potentially high-quality image. But is this how DVD works?

If so, then I imagine a sporting event, where adjacent scenes might be significantly different, and therefore suffer from data loss at high compression, should use the 1 hour setting. But a government meeting, where almost nothing changes from frame to frame, would work jes’ fine at the 6-hour setting.

I also wonder how this setting would affect the audio.

Any experts out there that can advise?

2

Bitrate. DVD video is MPEG-2 and can vary from 0 to 9800 kbps. If the divisions are 1, 2, 4 and 6 hours that corresponds to averages bitrates of 9600, 6200, 3000 and 1900 kbps, respectively (assuming a soundtrack of AC3 stereo at 192 kbps). The visual quality achievable at these bitrates varies according to factors such as frame format and net motion within the visual field. A picture generated by a handheld camera will, due to random movement, require a higher average bitrate to record than the picture generated by a tripod mounted camera. Also, picture quality tends to degrade rapidly below 2 mbps (though this might not be an issue if you’re using a stationary camera to record talking heads).

Generally speaking, yes. An overview of the process can be read here.

Absolutely correct, with the proviso that at the 6 hour setting you’ll want to use a tripod mount to prevent random camera motion from draining detail from the picture.

Audio can be coded as linear PCM (similar to audio CD) or AC3 (similar to MP3) at 48 KHz. A stereo PCM recording consumes about 1500 kbps, or about 700 Mb per hour, while a stereo AC3 recording consumes about 192 kbps, or about 85 Mb per hour.

Very probably, your DVD recorder uses AC3 by default with an option for PCM at the higher quality settings.

3

Wow, thanks for that informative reply, KoalaBear. I could hug you. :slight_smile:

Since our recordings are mostly Talking Heads, our multiple cameras are all wall or ceiling-mounted, and we try to avoid panning and zooming when the camera is active, maybe the 4 or 6 hour settings would work fine.

I’ll have to check the recorder’s manual to see what options are available for audio. But would the stereo AC3 process be the same sound quality regardless of the DVD length setting?

I wonder about the video adjacent-frame encoding, however. When we switch cameras, the entire frame is likely to require storage as a starting point for following frames. Would this situation produce any kind of visible “artifacts” in the image analagous to what happens in low-res JPG still images?

4

For the video, it would depend on whether the multiple cameras were fed to an in-line switcher or edited together afterwards.

MPEG introduces an I-frame roughly twice a second. Most encoders are smart enough to insert an I-frame whenever the difference between two successive frames is above some threshhold – it’s ‘cheaper’ from a compression standpoint to use a fresh I-frame than to describe the change as a set of differences. But in any case the strategy is invisible: if your videotape equipment doesn’t show a glitch at a camera change, neither will your DVD.

All of this changes when your objective is to record video for later editing, however. In that case you need to record I-frames only at the highest quality level you can afford so that the video handles like motion picture film when it’s loaded into an editor. You want to defer the inter-frame portion of the compression process (where data starts getting discarded en masse to achieve the target bitrate) until the disc is mastered.

The audio bitrate is fixed at some (selectable) value such as 192, 224, 384 kbps etc., which delivers constant audio quality throughout the length of the recording. 384 kbps is typical for a 5.1 channel surround track on a commercial DVD, so 192 kbps should be plenty for stereo.

5

So for the audio, 192 kpbs is what I would expect for a 6-hour setting? Do I understand that right – the audio, if AC3 is used, would be the same quality regardless of the DVD length setting?

As far as the camera/image consideration, we do all editing in-line, real time, for two reasons. One – it’s good enough, and our budget would never pay for the time and equipment to edit later, and two – if we edited anything, the political accusation could theoretically be made that we were altering an historical record of what actually happened at a Council meeting!

It sure sounds like the 6-hour setting makes the best choice for us. If so, that would be great – while most meetings do not run that long, it would be good to know that we’re covered if they do.

6

Considering how much more space-consuming video is than audio, I can’t imagine your DVD recorder would actually alter the audio-compression just to save space. I think it would only amount to a few more minutes of video. But that’s just my guess…

7

hobbes730, I suggest you read the technical details as in the link Koala Bear gave above.

But I’ll admit the tech talk can be overwhelming, so let me see if I can sum up what happens in MPEG2 compression.

Let’s say you have a minute’s worth of video; the image consists of a fruit basket sitting on a table. Nothing moves, nothing changes. A minute of images, at 30/second, is a shitload of data. BUT, imagine if the recorder were smart enough to know that, once the first image is snapped, it could be stored only once, then played back 30 times a second for a whole minute. Lessee, that would save 30 x 60 x 60, or 108,000-1 images from being recorded, with the only overhead being the instruction to “repeat this one 107,799 times.” That would be pretty compact, eh – a compression ratio of nearly 108000:1? And no loss of image quality whatsoever.

Of course, that is an oversimplification of real life. The MPEG2 encoding scheme stores full images once in a while, then stores only what changes until the next full image. To use the fruit basket analogy, if someone instantaneously removed the apple, the next image would be the same as the first except in the small area where the apple once was.

So you can see that sports events, where successive frames may differ significantly during action sequences, cannot be compressed as much as talking heads on a news show, where very little changes. The sports action requires much more unique data storage.

Does that explain things a little? :slight_smile:

Koala Bear, although our cameras are rigidly mounted, they are not very new or of super-high quality. Do you think low-level video noise would be a factor in the frame-to-frame compression scheme at the 6 hour setting? That is, would video noise fool the compression scheme into thinking it had to store more I-frames or toss out more data?

8

Well, slap me silly with a wet video camera. Hobbes730, you were asking about audio, not video compression. Mea culpa.

[Emily Litella]
Never mind.
[/Emily Litella]

9

Musicat: Noise can seriously affect the encoded quality and bitrate used. Noise will be interpreted as detail, and the encoder will have to spend bits saying when noise appears, where it moves to, and when it disappears. I would suggest that you just try recording once at the 6-hour setting, and if the results are not satisfactory, move back up to the 4-hour setting, etc. However, I’d imagine that the 6-hour setting would be fine for your purposes.

10

Hehe, no problem Musicat. That was one of the best descriptions of MPEG-encoding I’ve ever read :slight_smile:

11

That’s kinda what I was worried about. But what would be the symptoms that the results are not satisfactory? What kind of “artifacts” should I look for? Jerky video? color changes? Other?

12

I’m most familiar with MPEG4, but I’d guess artifacts produced by MPEG2 would be similar. Look for what are called “macroblocks,” especially in areas that are darker or in flat backgrounds. These appear as square areas of different brightness or colors. Pay special attention to any movement, to see if it smears or gets blocky. You can also look for unacceptable noise (“ringing”) around sharp edges or bright objects. Realistically, I’d think that 1900kbps would be more than adequate for your needs.