Two pass VBR encoding

[DiscCoasterPro]

Hi all. I'm about to encode and burn to single layer DVD -R a project that is slightly over 2hrs and in order to get the required space to 4.2 GBs I had to drop to 3700.

My question is one I hope can be answered by someone with actual experience testing the 2 pass vs 1 pass VBR encoding.

Is the resulting product actually that noticeably better?


thanks....

[Devil]

101 questions fall from this single question.

Firstly, for a 2 h project (assuming simple authoring and no subtitles or extra audio tracks), you should be able to fit it in with an average 4400 kbit/s video and 192 kbit/s audio (AC-3). This will drop if you opt for LPCM audio, but why would you do that?

So-called CBR is not constant, but does vary somewhat.

A single pass VBR starts out at about average. Then, if it encounters some P or B frames-to-be where there is relatively little data, it slows down the bitrate. Conversely, if the P and B frames contain a lot of data, then it ups the bitrate. As it does these, it recalculates the target average for the rest of the project.

To illustrate this, let's imagine a hypothetical scenario where, from 5 minutes into a video, you visit an F1 racetrack for 30 minutes of video. The first 5 minutes are encoded at a steady 5000 kbit/s. The next few minutes may have peaks of 7000 kbit/s and this may bring the remaining average down from 5000 to 4800, meaning that the last few minutes of the F1 sequence will have crept down to, say, 5500 kbit/s, so that the average can be more or less maintained to fit in the second half, which may be a talking head. This will start out at, say, 3500 kbit/s but will gradually creep up again to avoid excess space. Let's say there is a final 2 minutes of F1 racing. The space available for this may allow it to be encoded at only 5000 kbit/s. Theoretically, the quality of the last 2 minutes will be poorer than the first variable quality sequence of 7000 to 5500, but it woulld require a VERY observant person to see any difference.

With 2-pass encoding, the theory is that the first pass records changes from frame to frame as a code, so the first 5 minutes would say 5000, the next 30 minutes would say 6200, the next 30 minutes would say 3800 and the last two minutes would say 6200 again. The second pass does the actual encoding according to what the first pass dictates.

This is a VERY simplistic explanation. In the example, you may have, say, 800 shots spread over the time, each varying from, say, 1 second to 20 seconds. Some of these will be butted together, while others have transitions. Some may have deliberate or inadvertent camera movements or zooms. These butt joints, transitions, pans, zooms and shake will almost certainly be the major change from frame to frame, not the subject movement, in many, if not most, cases. This means that you will get the highest bitrates for these cases and not for when the subject moves, which tends to beat the purpose of VBR.

Then, you have to consider the quality of the original video and perhaps even the format. If the original is DV, for example, it is already compressed and has artefacts (mostly invisible in normal viewing). These tend to be constant in nature over a number of frames so MPEG encoding has an easier job with fast moving objects and you can get away with a lower bitrate than otherwise with no or little loss in quality. I have had arguments galore about this, as I have stated on many occasions that CBR 6000 kbit/s is difficult to improve upon. I will qualify this as being valid for PAL DV. It may be that NTSC is different as the colour is encoded in a different way on NTSC DV, but this is speculation. If the input video is uncompressed (i.e., no artefacts), then a higher bitrate than 6000 kbit/s will give better results, up to a point, but there are practical limits, not the least of which is the fact that some DVD players don't like a combined A/V bitrate exceeding ~7000 kbit/s on DVD¡ÓR disks.

This is therefore a very complex subject.

Have you ever wondered how a 2½ hour Hollywood blockbuster is recorded onto a 9 Mb disk with such quality, along with ½ an hour of bloopers, whether the director has dandruff, not to mention complex menus, 3 sound tracks, 8 sub-title languages and an interminable copyright notice followed by the logos of 5 companies involved in the production? This is because the 20-odd pass encoding is tweaked so the bitrate is optimal for each scene. Nevertheless, the average bitrate is not high, perhaps only 4000 to 5000 kbit/s or even less. Where there is real action (but not for the transitions), it may peak to 8500 kbit/s or even more, but the sunset at the end is encoded at only 2000 kbit/s. You can get away with this with pressed disks (DVD players are more tolerant of high bitrates because the 1:0 contrast ratio is much higher than with ¡ÓR disks).

[DiscCoasterPro]

Thanks for your reply Devil. I don't honestly know exactly how long this video is. I said slightly over 2 hrs, it may be 2hrs and 20 minutes. I don't remember.

I do remember tweaking to 4.3 required with a 4.4 available and had to do the whole thing over once, so I wanted to be sure to go down to 4.2

I couldn't achieve that without going to 3700 kbits/s using ac3 so, however long that makes it

There isn't really that much movement as it is a pocket billiard match and everything is still except the balls shot.

I was more curious on a generic level if the 2pass really did that much more for the video. There is really no abrupt movement in massive amounts.

thks again.

[Devil]

Cor blimey, a billiards match lasting >2 hours!!! Slow players?

[MrA]

Cor blimey, a billiards match lasting >2 hours!!! Slow players?

Come on now, Devil, you should be more globally politically correct and identify it as "gravity deprived physics".

[DiscCoasterPro]

Ah yes, to explain the wonderful game of pocket billiards to neophytes. All but the real players only know of two games when they think of pocket billiards. 9 ball from its TV exposure, or of course that game everyone plays in their basement called 8 ball. AKA stripes and solids.

Alas, the game is far more complex than the above mentioned, with far more variations than the everyday railbanger is even aware of.

Take a moment, and familiarize yourself somewhat with the ¡§chess¡¨ of pocket billiards here at this link.

http://onepocket.org/

[Devil]

That's not billiards: it's a bastardised form of snooker. Billiards is played with just 3 balls in two versions:
Cannon or French. On a table without pockets and you score in breaks of cannons
Plain billiards. On a table with 6 pockets and you score by cannons, potting and in-off

You Americans pinch and bastardise all sorts of European games:
Basketball: from girl's school netball
American football: from rugby
Baseball: from girl's school rounders
Now "pocket billiards": from snooker.
What are you going to do with cricket?

[DiscCoasterPro]

What are you going to do with cricket?

Billiards is played with no pockets, pocket billiards uses pockets. Snooker is a great game also and I enjoy it.

I step on crickets.