so i'm encoding some files in low qual vbr for my portable:
--vbr-new -V 9 -h -b 32 -B 160 -m j --lowpass 15.25

however, in many songs with silences in the beginning, it uses a large frame size... 160. in taking off all the switches except --vbr-new -V 9, it uses 224 kbps for all these frames. this occurs all the way down to -V 6 or so. this is taking up valuable space of course.

this is version 3.96.1.

also, the frame size 144 does not appear unless i remove the other flags. I'd think that this would be a valuable frame size for my portable, but i can't use it. -B 144 results in an error, and so does -b 8 or -b 16.

any suggestions?

what sort of bitrates are you looking for ?

maybe this will help ?

The frame size of 144 is only present with MPEG 2 ( or 2.5, which are from 32Khz to lower). If your player supports MPEG2, then you can make a listening test for yourself as to what setting you preffer. ( b 8 and b 16 are MPEG2 too)

About the silence in the beginning using high bitrates, as quoted many times, it might not be digital silence, so there IS actually something there. The only thing i see you could try is reducing the threshold something, but since i am no expert in this regard, i can't give you an advice.

i just didnt think it made sense that something that is almost silent, audibly silent, would take up so much more space... 160 or 224 and then when the audio starts fading in slowly and becomes audible, the frame size jumps down to 56 and so forth...

V9 is tuned to work at 24 Khz sampling rate with lowpassing at 10K, your 15k lowpass is probably bugging it out. V9 is not meant to deal with that bandwidth see, so any number of complications could be happening. I think its safe to lower a presets lowpass, keep samplingrate same to be sure, but raising it is unsupportable.

Idealy lookup the V preset you need for the lowpass you want from the table.

edit: oh yeah, i see you did try without the lowpass
Come to think of it ive seen the odd very high bitrate frame myself with low vbr, but they amount to very little. A small glitch perhaps. Ive used -B 128 to contain it.

~Fair doos.

interesting, what happens wh en you do something like -V 9 --lowpass 20 ??

i would do -B 128 but i like the idea of having 160 and 192 around for very complex moments in songs.

I cant remember specificaly, but in general if you raise the lowpass beyond what the pyschoacoustics are expecting and the track contains steep and persistent, or stronger and briefer but always steep( high frequency ) oscillations, you get alot of artifacting from the encoder trying to include them, resulting in, twingzangs zeeps and stereo yuh yuh.

I think 160 is the max framesize for mono. The graph of bitrates used goes something like a hill, with 2 or 3 framesizes in the middle and sloping edges.

There are mechanisms, to catch and lower too highly demanded framesizes,
which can quietly cope with occasional breakdowns in the normal recommendation, - as long as they occur rarely.
When set into the 'hill curves' top slope, most of the frames are getting as well defined as the vbr-ear likes. why would the vbr-ear want these frames which lie outside its usual preference?

'beats me

That line is bad. Just to point out the obvious, --vbr-old is better than --vbr-new, and there is no such thing as a 15.25kHz lowpass; how about 15 or 15.5?

I suggest you abandon that line in favour of --alt-preset 96 or --alt-preset 128.

It's more outdated than obvious. Check all lame 3.97 alpha threads and blind listening comparison between defaulted vbr mode and --vbr-new: this last one is apparently better on average.

That's just an alpha. We'll see when the final is out.

I'm not sure that 3.96.1 is really different in that aspect. --vbr-new was considered as inferior for 3.90-3.92 (in other words, three years ago), and difference between both VBR modes wasn't AFAIK checked since. If I hadn't tried to compare both VBR mode during 3.97 alpha testing phase, people will still consider the outdated knowledge (--vbr-new inferiority) as a fully valid one.

Many claims and recommandations posted on this board are rather outdated than obvious. It's a simple consequence of the lack of listening tests.