Quirk with SVCD to DVD - possible sign of Apocalypse?

[stormcrow]

Hey all, I dont post much, but i spend a couple hours a day here, as I do a lot of production. I don't really have a problem , but I ran across something weird and I was just wanting an explanation if anyone can offer one for me.

Playing around with new dvd writer, sadly it's not supported by Nero yet, so I'm using Ulead DVD Factory. I'm taking great quality svcds i have and converting them to dvd standard to put on disc in Tmpegenc because I dont want to let Ulead do it for me (of course).

Here's the catch - the original svcd mpeg2s are about 386 megs apiece, and bitrate viewer puts them at an average bitrate of a little over 2600(peak 3250).

I converted them all last night, 2 pass vbr at 2600, thinking this would be the same filesize to fit the amount i wanted on the disc. What this gave me though was an increased filesize to about 430 megs apiece -- but bitrate viewer is showing the average at less than 2500(peak 4300)!!

So how is this possible? The increased framesize doesnt effect the filesize, and that (supposedly) is the only thing that changes in the encoding between the two files. How can an mpeg2 with an average 100kbps less produce nearly a 50 meg filesize difference?

Word.

[stormcrow]

Ran another encode with the vbr set to an average of 2400, the filesize still coming out at 395 megs, 10-15 more than the original svcd encoded at 2600 average.

any ideas?

[AceYu]

The horozontal and vert resolution of DVD standard MPG2 is considerably larger and therefore nessicarialy will need more data, that is the simple answer.

The more complex answer basically consists of sampling theroy and frequency domain stuff, but what is happening is that the frames of the SVCD are upsized to DVD resolutions, this not being an exact multiple of the origional frame size (can't remeber the numbers right now) some aliasing effects and other artifacts can be introduced. The amount and type depends on the method used to resize the picture. This will "add" information to the picture, that is MPG compression relies partially on finding the differences between frames, these artifacts increase the differences between the frames and therefore the amount of data required to store them. Re-encoding/transcoding any video with a lossy codex (ie 99.9% of codex) will always add artifacts and noise, so always try to use the origional source where possible, it may make little difference to the final size but the quality will always be better.

Later BRETT

[stormcrow]

Granted, the larger frame size needs more data to carry the same amount of information, but the bitrate is what determines how much is given, correct? My understanding is that a larger frame size at the same data rate should decrease the bits per pixel, and thus the overall quality, which for the sake of what i want is allowable, but in terms of filesize, BPS means filesize per second of video, regardless of what is contained in the video. If two vids are the same length and same bps, they should be the same size even if they are different file formats completely-otherwise bps doesn't actually mean what it stands for.

Or am i just missing something here?

If I'm not mistaken in this, I wonder if its not an issue with the VBR in Tmpgenc - going to try the same comparison with CBR just to see if the same thing happens.

[stormcrow]

Have some more information to add to my mystery now.

Analyzing the mpegs that Ulead DVD Factory creates converting the orginal SVCD mpeg2's to DVD standard, I found some interesting things. Ulead was encoding these at a CBR of 2450 (as bitrate viewer reports). These files came out to be lessthan 350 megs, as would be expected, being less bitrate than the original.

However, I did the same encode in Tmpgenc with VBR set to 2400 avg, and the files were still over 400 megs!

So, it seems that the VBR avg doesnt determine filesize in the same manner that CBR does. Is this expected of VBR and I never have noticed, or is this some glitch? I also can say that when Ulead gives a choice of encoding to DVD, the difference between VBR and CBR changes the Disc Space Needed sunstantially at the same bitrate, so this seems to follow the same pattern.

Anyone know the details of the VBR specification?

[homerpez]

Probably the truest answer is this - Different encoders encode differently.

TMPGEnc, I have never found to be very accurate in file size, even with 2-Pass VBR. It's possible to change a number by only a tiny bit and have it make TONS of difference.

Forget Ulead anything to encode, quality is simply not up to TMPGEnc standards.

If you simply MUST have a file size accurate, you need to A) Use CBR in TMPGEnc, or B) use CCE Multi-Pass encoding instead.

I do the latter, always gets close to the "average" I need, every time. I calculate size with FitCD (assuming I want to fit the movie on one disc), subtract 3 from the AVG bitrate it spits out, and I'm golden.

[Plant_Guy]

CBR guarantees file size. VBR means just that, variable. Depending on how the average is calculated, you can get different results. If you have lots of high-bitrate spikes, then the file size will be larger without affecting the average bitrate (again, depending on the method of calculation), so don't be put off by the apparent discrepancy.

TomG. - aka Plant_Guy