What causes undersizing?

[raddygast]

Well, 0.84.1 Pro resolved the output file issue and greatly aided my undersizing cause, so I can't expect to get any closer to ideal. I think this title just creates prediction problems -- there are a buttload of VOBs that I've blanked out, and only the main VTS is being encoded as a consequence. However, it is already not too big, only a 90.9% reduction is required. I've selected MPEG Standard matrix. Could the qmat be fiddling with prediction?

I remember reading movies with LOTS of VTSes can cause slight prediction errors, so I was wondering if that was it. I increased targetsectors to make it hit about 4462MB (as it says in RBOpt, so about 2255xxxxx).
Last time (with 0.83Pro) it came to about 4260MB (200MB undersize), this time it came to about 4412MB (only about 50 MB off).

But it's just surprising because usually when I tell it to increase targetsectors like that, sometimes it even overshoots.

I encoded another film in this same batch, also with many blanked VTSes (The Incredibles) and that one came out BANG ON in terms of size. So it must be the title, but what I wonder? I guess I can run another re-encode with even huger target sectors, but for 50MB I really doubt it's worth it.

[jdobbs]

I highly advise anyone that if you are within 100MB -- leave it alone and don't change the target size. You're just asking for trouble. This isn't transcoding we're talking about -- it's encoding, and it can sometimes be off a little.

Why in the world would someone want to waste 3 hours encoding just to be over by a little and have to reencode... or even worse, butcher it through a transcoder after all that work making it high quality.

Sometimes I regret even putting the TargetSectors variable in...

[raddygast]

Yeah, I agree with ya man. I know that altering targetsectors is a real risk, but since I run these batches overnight I'm not in a huge hurry.

The thing is, the undersize occurs even with my targetsectors increase. So it's true that for this title the prediction or "target size" isn't working that great with CCE. I bet it's just something weird about the source material.

I won't bother re-encoding it, I just wanted to get the hang of when RB and CCE are a bit off and when they aren't. 3 out of 4 of the last encodes were pretty bang-on, and 1 was off by only 50MB. Not a huge deal -- I'm just interested, that's all.

But I have noticed that RB's default targetsectors is quite conservative -- much more than DVD Shrink's, for example. I was wondering, why is thaht? I think you may have answered that already though -- is it because re-encoding can cause a lot of variability, and is thus greater risk? Shrink, as a transcoder, probably doesn't have to worry nearly as much so can set the target size to something much more aggressive and near the barrier for DVD+R.

The last thing is, maybe I should back off a little, because recently halfway through my spindle it seems I'm getting more PIF errors on my media scans, but only toward the end of the disc. Maybe being 100MB off or so is a good thing. :)

[jdobbs]

This isn't transcoding we're talking about -- it's encoding, and it can sometimes be off a little. DVD Shrink is a transcoder.

[raddygast]

Yup, that's what I said. So I guess predictions with a transcoder are not a thing of variability?

Anyway I'll be content with +/- 75MB or so. Don't want to spend too much time agonizing about this sort of thing.

[jdobbs]

75MB will only give about 83Kbs additional bandwidth on a two hour movie. In video that's like peeing in the ocean and trying and measure how much it rose.