Hi ..

If i'm correct --alt-preset standard uses some --scale 0.xx value. I rather use mp3gain to control this. If i use --alt-preset standard --scale 1, what will lame do ?

- Lame will disable scaling
- Lame really multiplies every sample with 1

What if it is option 2, will it affect sound quality ?

Just something i was wondering about.

Thanx.

This is incorrect. --alt-preset standard/extreme/insane, in fact all of the --alt-preset vbr modes, do [b]NOT use any scaling.

The only things that use --scale are the abr/cbr presets.



Well again, --alt-preset standard does NOT scale, but if it did..



They would both do the same thing.



No.

And just to be absolutely sure, you don't need to do this anyway since this preset doesnt' use scaling.

Hope that clears it up some

Ah ok, now i know that.



I like short answers !



It does ! Thank you ..

does this also include --alt-preset <bitrate>
or only --abr <bitrate> etc...?

No, no.
Only --alt-preset [cbr] <bitrate> has the automatic scaling
Normal --abr doesn't

ahh, ok.

but, if --alt-preset (abr) does scale, is it recommendable for a divx audio track anyway?

i do dynamic range compression with azid to push up the volume before encoding with lame, so i dont want my audiotrack's volume to be downscaled.

My prefered method is to use azid with normal dynamic compression, but without normalizing.

I then encode with --alt-preset xxx --resample 44 --scale 1

Then use mp3gain to normalize to max noclip gain.


Using mp3gain is much better than doing any --scaling. The default --scale is put in there for the 99% of people who hasn't heard of mp3gain

why are you resampling.. it give worse qiality and almsot 95% of audio cards out "there" can handle 48khz today

anyway i use --alt-preset 128 --lowpass 18.5

Sorry i just doesn't like/want any lowpass below 18.5


however i woudle wis ther was some program the did run the encoding multiply tiems asdjusting the bitrates in the encoding syntax so that then endign file woudle fit around the given bitrates


E.g many times you need to specify hgigher/lower bitrates to get 128kbits

LAME is tuned for 44.1kHz and not well tuned for 48kHz. There are some known audible quality problems with 48kHz.

--alt-preset 128 uses a lowpass of 17.5

If you want to raise it to 18.5, you have to bear in mind that low frequencies will be affected. ff123's recommendation of --lowpass 16 --ns-bass -8 was made to ensure that low frequencies are done correctly (--nspsytune is known to have some problems with bass at low-mid bitrates). Dibrom tuned it to --lowpass 17.5 --ns-bass -6 to get a good balance between preserving the high frequencies without over affecting the low frequencies. If you are going to raise it further, you are going to upset that balance.

i guess alt-presets <abr> default scaling decreases the volume, right? by what value does it multiply the pcm data?

since i do dynamic compression w/ find-max-gain in azid before lame enconing, i think there is no need to modify (--scaling) the volume again. am i right?

TIA

The --scale used by ABR and CBR --alt-preset depends on the bitrate used. Generally the lower the bitrate, the more likely it is to clip, and the lower the --scale value.

If you use 100% max gain in Azid, then you will have to scale down again in LAME to prevent clipping. I think that's kind of silly to have a two step lossy process. You should either cut it down to one step by modifying Azid command lines to scale only to 93% or whatever is recommended for the ABR bitrate you are using, or better still, don't normalize and don't scale at all, then use mp3gain after encoding.

But, from the other side,
resampling from 48 to 44 KHz doesn't it introduce per-sample (anti)aliasing problem?
Must admit that I don't know exatly the matter, but time-space compression "at digital stage" can't generate (lately by FFT) false rounding high-freqs??

Not if properly done. Don't know about LAME resampling, but SSRC does it properly.

i know that dibrom have doen alot of work in this preset but

1: he optimize mostly for music and not movies track (which benefits more from vbr/abr then music)

2: he have different ears/better artifact detections that i do
i can hardly heard the differense between normal movie tracks and --alt-preset 128

note that i use AAC for my music encoding and not mp3's
note2 i use vorbis -4.99 sound now in my movies


about the 48khz lame bug
i read a test somwhre on or linked from HA or doom9.
it showed that the so called tunde for 44khz only wasn't giving bad quality in 48khz
or to puts it this way
there is no bug/quality drops in keeping the files at 48khz

also consider this

frist all you loose (sligtly)quality be resampling down.
on playbakc creative card (mothe msotly used i think) upsamles to 48khz in a very bad way causin quality drops again.

these to resamplingstep (qualty drops) woudd be gone if going for a clear 48khz file

btw
sorry for lack of english skills in the morning :sleeping:

There are methods to prevent the aliasing problem, simplest is to do a simple low-pass filter just below 22kHz before changing the sample-rate.

This is new for me. If anything, I thought it did better at 48khz because it reduces some preecho problems.

--
GCP

Theoretically, 48khz should reduce the pre-echo by around 10%. The problem though is that I think some part of the psymodel is not tuned properly for this sampling rate.. because sometimes the pre-echo (which is supposed to be better) get's smeared into one of the side channels, quite blatantly.

There are some discussions about this somewhere in the archives here... and there were some similar observations made by David over on r3mix.net.

If whatever problem exists in the psymodel were fixed, then 48khz would be better almost all the time. Due to this inconsistency though, I don't recommend using it over 44.1khz.

assuming you were to use --alt-preset 128 --scale 1 is there still no quality influence? (of course this still in the mp3gain case)

If you use mp3gain instead then yes, using --scale 1 will result in no quality degredation. But if you turn off scaling and you don't use mp3gain, then you will get quite a bit of audible clipping in most music which will defintely lead to lower quality.