MC's MPeg Encoder: underflow?

[Tsplat]

It appears the Mainconcept team either doesn't know or will not talk about this problem. I'm using their MPeg Encoder v1.05.01.00 and am now getting a lot of underflow error messages using settings I've used for a long time now. I've asked on their forums but get no response. What's the cure for underflow? Use a lower minimum, average, or max bitrate.... or lower all three? My current settings are 2500/min, 6500/avg, 8250/max. Thanks.
---
Splat

[Amnon82]

Make a template and upload it to megaupload.com. I'll take a look into it. 'Options' > 'Save settings as Profile'. You'll find a new mef-file in

'C:\Dokumente und Einstellungen\<YOUR USERNAME>\Anwendungsdaten\MCMPEGEnc'

[Tsplat]

I've u/l the profile to my server. Here it is: http://home.comcast.net/~splatdrums/myhigh.mef

BTW, I'm also using Pegasys's TMPGEnc DVDAuthor to encode everything for my DVDs, always encoding the audio to AC3. I thank you for looking into this.
---
Splat

[dragongodz]

why are you using a min bitrate so high by the way ?

[Tsplat]

I usually use either 2000 or 2500 for min. I guess for converting VHS to DVD it's overkill... just trying to get the best resulting pic as possible.

PS, diggin your sig line. One of the funniest flix ever. "It's good to be the king!"

[Mr_Odwin]

I would've thought that having a higher minimum bitrate would be counter-productive. The bitrate you could've saved on an 'easy' scene needs to be taken from somewhere else to meet the specified average. If that makes sense ...

[dragongodz]

would've thought that having a higher minimum bitrate would be counter-productive. The bitrate you could've saved on an 'easy' scene needs to be taken from somewhere else to meet the specified average. If that makes sense ...
i would have to agree. having some min bitrate can be good to make sure the encoder doesnt fubar what it thinks is an easy scene but really 2000+ seems a bit extreme. try lower such as 500 or even 1000 and then compare a clip that has both slow scenes and action scenes.

PS, diggin your sig line. One of the funniest flix ever. "It's good to be the king!"
hehe i just wish someone would release it here on dvd. my video is falling apart with age. it was release on dvd in the U.S. back in 1999 for frells sake. ;)

[Tsplat]

I'll see about lowering the minimum. Still though, I don't know if these underflow messages are because of something Mainconcept changed with this latest release, or it was happening even with earlier versions but wasn't output to the status window or logs.

[dragongodz]

Still though, I don't know if these underflow messages are because of something Mainconcept changed with this latest release, or it was happening even with earlier versions but wasn't output to the status window or logs.
either way MC should know about it and fix it. they have in the past fixed problems with rate control, i know because they did it for me.
when i get the chance i will look at that mef and see if i can reproduce the problem and if so then report it to those that need to know.

[Amnon82]

I can't reproduce the bufferunderflow.
K, I used a PAL 720x567 Progressive source but the source don't matters (so think I).

Mainconcept 1.05.01.00 Log:

multi pass encoding, pass 1
MPEG video encoding
Video source:

AMD 3DNOWEXT

MMX:y SSE:y SSE2:y 28:0 3DNowExt:y 3DNow:y CMOV:y
Using library: MainConcept optimized Software Encoder (C:\Programme\MainConcept\MPEG Encoder\mcmpgvout.004)
Initializing video encoder
Number of CPU: 1, available: 1
SIMD: SSE2
Video encoder initialized(SMP_MC+)
Frame width/height = 352x480
Frame size avg = 140486 bits, min = 36176 bits, max = 302096 bits
Bitrate avg = 4210379 bps, min = 2768239 bps, max = 5973738 bps
Mquant avg = 2.94, min = 1, max = 4
Total size = 18052493 bytes
I:mquant= 2.94, X=192.51, d= 30.141, size= 32452, frames: 69
P:mquant= 2.94, X= 68.04, d= 30.431, size= 22200, frames: 275
B:mquant= 2.94, X= 29.37, d= 30.732, size= 14164, frames: 684
::mquant= 2.94, X= 50.66, d= 30.612, size= 17541, frames: 1028
Total time: 22 seconds (00:00:22), time per frame: 0.021 seconds.


multi pass encoding, pass 2
MPEG video encoding
MPEG output file: H:\testmc.mpv
Video source:

AMD 3DNOWEXT

MMX:y SSE:y SSE2:y 28:0 3DNowExt:y 3DNow:y CMOV:y
Using library: MainConcept optimized Software Encoder (C:\Programme\MainConcept\MPEG Encoder\mcmpgvout.004)
Initializing video encoder
Number of CPU: 1, available: 1
SIMD: SSE2
Video encoder initialized(SMP_MC+)
Frame width/height = 352x480
Frame size avg = 140205 bits, min = 34280 bits, max = 302096 bits
Bitrate avg = 4201965 bps, min = 2769310 bps, max = 5973738 bps
Mquant avg = 2.94, min = 1, max = 8
Total size = 18016414 bytes
I:mquant= 2.94, X=192.51, d= 30.221, size= 32452, frames: 69
P:mquant= 2.94, X= 68.04, d= 30.239, size= 22200, frames: 275
B:mquant= 2.95, X= 29.37, d= 32.976, size= 14111, frames: 684
::mquant= 2.94, X= 50.66, d= 32.059, size= 17506, frames: 1028
Total time: 23 seconds (00:00:23), time per frame: 0.022 seconds.


Muxman Log:

MuxMan version 0.14g
Accepted video H:\testmc.mpv size = 18016414
Accepted audio H:\testmc T01 3_2ch 448Kbps DELAY -208ms.ac3
18:10:46 Begin multiplex.
Maximum audio duration 2056 fields.
Starting scene Segment_1_scn1 at 00:00:00:00
SeqEnd at 112E89A.
Bytes remaining in buffer = 0.
18:10:46 End multiplex.
Bitrate - avg: 4755529, min: 3142585 (lba 7786), max: 6612523 (lba 3437).
Shortest GOP has 20 fields, longest GOP has 30 fields.
Fields: 2056, VOBU: 68, Sectors: 9956.

[Mug Funky]

hmm. so the same settings have been working fine without a hitch. does that mean it's just this sample that's making MC underflow?

try run a bitrate scan on your encoded output, and look for spikes that go above the max bitrate (by a lot - bitrate scanners all give different results... i use BitrateView mostly, and that can be quite inaccurate). if it also shows the quantizer raising sharply, then there's your spot. now note the time it happens and see what caused it.

there's certain problem samples that just annihilate the ratecontrol in most encoders... uncorellated noise is an obvious one (like the closeups of the TV in "Ring") or very complex particle motion (the "star child" logo, or the gateway scenes in "cowboy bebop"). i wouldn't be surprised if there's something similar in your source.

[Tsplat]

The source was a live concert VHS vid I bought and am now trying to archive it to DVD. I'm still waiting a comment from the MC crew, btw.

[Mug Funky]

have you tried a bitrate scan to find the actual part that caused the underflow?

[dragongodz]

i ran some tests on some high action footage and even tried doing things a 2000 CBR encode and could not get it to report any underflows. so i dont know what you are doing but i can not reproduce your problem.

is it MC thats reporting the underflows or Tmpgenc author ?

[Tsplat]

I logged the encoding process and it lists ever frame where underflow was reported. I cannot see any problems with all the frames I checked...which were about 20-30, but there was about 100-200 underflow warnings in the log. Since it looked OK I'm forgetting about it, for now, and the resulting DVD came out fine. BTW, I've never had DVDAuthor report underflow problems, if it even can. For my next conversion if I get all these underflows again I'll worry.

[dragongodz]

I've never had DVDAuthor report underflow problems, if it even can.
yes i have read other posts where it has reported underflows. if its not then there may not be any.

I logged the encoding process and it lists ever frame where underflow was reported.
sorry as i said i am unable to reproduce it. all i can suggest is uninstalling, rebooting and doing a fresh install and see what happens. if that still fails then i dont know why you are getting this.

EDIT:
actually can you make a copy of the log file for me to look at ?
also a small piece of footage that does it every time would be useful to see if we can reproduce it that way aswell would be good if possible.