Error reporting bit rate?

[Anole]

I've recently started forcing VRD to output an mpg file with a different VOB size in an attempt to clean up some unknown video problem that QSF isn't handling.
It has worked in a few instances, and I will continue to use it as a work-around 'till I can come up with a proper solution.

Yesteday, I re-saved (re-encoded?) a 1.1gb mpg video, forcing GOP=11.
(don't ask why, it was just an arbitrary size chioce)
The resultant file went from about 1.1 gb to about 1.2gb, or around 100mb larger.
I dont have the exact numbers, but the bit rate reported by VRD after completing a save went from 2.2mb (?) to 3.3mb (?), or 2.5 to 3.2 or possibly 2.2 to 3.5 ... whatever.
Anyway, the file size grew by maybe 8..9%, but the reported bit rate grew by 30..50%.
My Tools | Options | Stream Parameters | Encoder Quality was set to High.

I wonder how the bit rate is calculated, and how GOP size affects it?


Though a feature request instead of a bug report, could we get a GOP size report somewhere in VRD?
I suppose it would have to be after a save, since a static report when the file loads wouldn't be right.
Any chance that number could be reported after the user has spent some time editing?? :)

[BaysideBas]

It stands to reason that making the GOP size smaller will increase the file size (more I frames). Now you have more bits to push through the pipe in the same amount of time, therefore the bit rate has to also go up.

[DanR]

Its highly unlikely there's a bug in the bit rate calculation as that is the simplest calculation in the whole program: 8 * # of video bytes output / program duration.

FYI, I would really recommend using the auto quality on the encoder quality especially if you're planning on moving to DVDs.

Regarding the GOP analysis. If you set the max gop to anything > 0, say 100 you'll see a GOP analysis in the log file on output.

[Anole]

In the first post of this thread, I was shooting from the hip, using the best info at my disposal.
However, I did not run extensive tests to confirm my theory.
It looks like I made a mistake, which could have been due to the late hour of some of my adventures. ;)

Here is actual test data taken from a number of saves of the same file based on a single .Vprj, and written with different GOP settings.



Qualifying at Silverstone, GB, for Formula One race.
Dish Network satellite broadcast of the Speed Channel.
VRD 475 @ 12 June 2006
All writes set the bit rate to 10mb/s
source file is: 2447484 in size, and 2h 10m 45s in length.
There are 10 cuts to extract just the show, no header, no trailer, no commercials.
The original recording contained a 30 minute copy of MotorWeek, which was discarded.
output = 1h 7m 21s


write with NO GOP change
========================
Output Complete

Video output frames: 121104
Audio output frames: 168378
Processing time (secs): 125
Processed frames/sec: 962.69
Actual Video Bitrate: 2.45 Mbps

file size: 1308711



Write with GoP = 60
===================
Output Complete

Video output frames: 121104
Audio output frames: 168378
Processing time (secs): 128
Processed frames/sec: 943.35
Actual Video Bitrate: 2.45 Mbps

file size: 1308727



Write with GOP = 48
===================
Output Complete

Video output frames: 121104
Audio output frames: 168378
Processing time (secs): 141
Processed frames/sec: 857.46
Actual Video Bitrate: 2.46 Mbps

file size: 1310614



Write with GOP = 36
===================
Output Complete

Video output frames: 121104
Audio output frames: 168378
Processing time (secs): 533
Processed frames/sec: 226.86
Actual Video Bitrate: 2.53 Mbps

file written: 1346467



Write with GOP = 24
===================
Output Complete

Video output frames: 121104
Audio output frames: 168378
Processing time (secs): 705
Processed frames/sec: 171.64
Actual Video Bitrate: 2.59 Mbps

File written: 1377605 kb



Write with GOP = 18
===================
Output Complete

Input Sequence Errors: 6
Video output frames: 121104
Audio output frames: 168378
Processing time (secs): 831
Processed frames/sec: 145.57
Actual Video Bitrate: 2.62 Mbps

file written: 1395081

You'll notice that even with GOP=60, there was at least one fixup.
Setting the GOP=48 did not increase output time drastically.
When choosing a GOP of 36, VRD had a lot of work to do.

Note also, at GOP=18, there were input sequence errors.
How can that be, considering all the previous times the same file was processed??

Anyway, what this all shows is how the satellite company increases the GOP size, way past all wisdom, and gains very little in the way of compression.

.(search keywords: GOPs long short twentyfour thirtysix fortyeight eighteen time )

[litz]

Cable feeds are the same way .... easily 60 frames+ on GOPs.

It's like they simply go until the frame counter overflows, then pop a GOP. Great fun when there's a frame error and it takes 4-5 seconds for a GOP to roll around and your picture stabilizes ... :-(

As for the encoding quality, you should use AUTO not HIGH ... high will encode your recode frames @ the header bitrate, auto will approximate a bitrate based on the surrounding frames.

If, for instance, you have an average bitrate of 2.5mb/s, and a header bitrate of 15mb/s, HIGH can result in unpleasant surprises in the bitrate @ your edit points ...

I've also seen GOP weirdness without the "Convert I-frames to GOP" box checked.

(DRD - do I have that right?)

- litz

[Anole]

I usually try to pay attention to what Phd and DanR say.



FYI, I would really recommend using the auto quality on the encoder quality especially if you're planning on moving to DVDs.

And as a result of the above, I did change over and start using auto mode for Encoder Quality.



As for the encoding quality, you should use AUTO not HIGH ... high will encode your recode frames @ the header bitrate, auto will approximate a bitrate based on the surrounding frames.

If, for instance, you have an average bitrate of 2.5mb/s, and a header bitrate of 15mb/s, HIGH can result in unpleasant surprises in the bitrate @ your edit points ...

Well, if this is what Dan said elsewhere, I must've been asleep at the time.
This could very well be what gave me such a surprise before.
When I get a chance, I'll try rerunning a few of the above tests at Quality = High just to see what happens.

Since my goal is to make DVDs, I now set the bit rate in the header to the recommended 10mb, as noted in my previous post.
.