why are bitrates >= 112 kbps the only bitrates that support 44.1 khz frequency?

[nukesgoboom]

why are bitrates >= 112 kbps the only bitrates that support 44.1 khz sampling frequency?

i enjoy 128 kbps mp3 files. sounds great and its a low filesize. i figured converting sound files to VBR with 128 kbps as the max and 32 kbps as the minimum would be just as good as 128 CBR but with a smaller filesize.

however, i met with failure, as i could only convert them by downsampling to 22 khz. why is this?

only way i could lower the filesize is setting 112 kbps as the minimum, 128 maximum. now they sound just as good as 128 kbps but with a smaller size.

but why cant i use 32-96 kbps and keep the 44.1 khz frequency?

[Sirber]

but why cant i use 32-96 kbps and keep the 44.1 khz frequency?MP3 ain't good enough for those bitrates.

[setarip_old]

The Fraunhofer Professional .MP3 codec allows for 44,100Hz .MP3s at 96Kbps...

[raquete]

MP3 ain't good enough for those bitrates.agree,you deserve one gold medall.
;)

[Sirber]

ok, find something better then :devil:

[foxyshadis]

Technically 44khz can use any bitrate from 32 to 320, but in practice sirber's right. Some encoders will let you bypass the restrictions - blade will let you encode to anything. (But it's also a 1998 codec, don't bother.) Lots of 96kbps and some 80kbps 44hkz mp3s exist.

Old encoders just sounded so horrible that it became a rule of thumb to disallow any lower bitrates, even though the current version of lame could sound better at 80kbps than blade and xing did at 128. (Which still hurts.)

ps, I still have a 32kbps 44khz "amish paradise" kicking around somewhere. =D

[shon3i]

There is three tehniques which can help in lower rates

1. Inesivity Stereo (aka Joint Stereo IS but now only in Franhofer)
2. Narrowing of Stereo Image (aslo only in Franhofer)
3. Spectral Band Replication (aka SBR aka Mp3PRO aslo only in Franhofer)


First two can't help much but for example file @64kbs sounds little better than without it.

Third tehique SBR can help much, and keep file at 44.1khz, but you need special decoder for Mp3PRO files to play that file with full quality, otherwise that file will sound very terrible.

http://www.mp3prozone.com/assets/basics/mp3PRO_64.gif

[Sirber]

I do not recommend the use of mp3Pro. Great lack of compatibility.

[shon3i]

I do not recommend the use of mp3Pro. Great lack of compatibility.
True, but mp3PRO @ 48-96kbs is not that bad comparing to HE-AAC, Vorbis and others.

Technically 44khz can use any bitrate from 32 to 320, but in practice sirber's right. Some encoders will let you bypass the restrictions - blade will let you encode to anything. (But it's also a 1998 codec, don't bother.) Lots of 96kbps and some 80kbps 44hkz mp3s exist.Not only blade can bypass restrictions, aslo both lame and fraunhofer can do that, lame using --lowpass, --highpass, --resample and other frequency swithes, and fraunhofer insade CoolEdit (Adobe Audition) have lowpass frequency and other useful options to tweak the codec.

Old encoders just sounded so horrible that it became a rule of thumb to disallow any lower bitrates, even though the current version of lame could sound better at 80kbps than blade and xing did at 128. (Which still hurts.)This is because both Xing and Blade use very poor inegration of Joint Stereo MS and IS ( i think that Xnig don't have JS/MS so use only IS which destroy quality at higer bitrates), and this is very destroyable for higer bitrates 96=>, so if turn off JS, sound quality raise up

[nukesgoboom]

i am using a front end for lame called razorlame 1.15.1342 with lame .exe 3.96 is this good/bad?

a long time ago i used to use chun yu shei's mp3 writer plugin. was that a bad/good idea at the time?

i leave the sampling frequency at 'default' and it always comes out to 22 khz when i do VBR min: 32 max: 128. i can definitly tell the difference between a 22 khz 128 mp3 and a 44.1 khz 128 mp3... so its not worth it to do the VBR at such low bittrates i guess.



rockouthippie said:
======================
It is possible to encode mp3s at higher sample rates and low bitrates, but higher sample rates don't produce better quality results. The excess samples actually make the sound worst.

It's like trying to put 50 pounds of mud in a 5 lb sack. I've encoded a lot of low bitrate audio. For an example 32kbps/22050hz mono, sounds superior to 32kbps/44100hz mono.
======================

why would low bittrates not have support for excess samples though? i see your comparison with the mud but i dont understand why it would sound worse and not better. what about the low bittrate stops a high frequency from helping the sound?

[raquete]

MP3 ain't good enough for those bitrates.
agree,you deserve one gold medall.
;)
ok, find something better then :devil:
my post was not critic,was a concordance.sorry if i was unclear.
you know,some people believe that using so low bitrate can be good but later repent.

best regards

[GodofaGap]

i enjoy 128 kbps mp3 files. sounds great and its a low filesize. i figured converting sound files to VBR with 128 kbps as the max and 32 kbps as the minimum would be just as good as 128 CBR but with a smaller filesize.
This is probably false.

A. Because is LAME is not specifically tuned at low bitrates.
B. It lacks an important tool for low bitrates --> intensity stereo.
C. It is very likely that lower resulting bitrate will have a noticable impact on quality anyway. Since bitwise the only way to achieve a lower bitrate would be to sacrifice quality compared to the 128 kbps. (Your maximum frame size is 128 kbps).

why would low bittrates not have support for excess samples though? i see your comparison with the mud but i dont understand why it would sound worse and not better. what about the low bittrate stops a high frequency from helping the sound?
It is trade-off between compression artifacts and loss of frequencies, it is difficult to say which one would be preferable, but with your settings resampling doesn't seem that odd.

[Kurtnoise]

....

edit: sorry. Deleted my post to avoid confusion.

[Slogra]

You can let LAME encode to 44.1khz at 96kbps (--resample 44100) but it's not recommended. A 32khz mp3 at 96kbps will sound better than a 44.1khz mp3 at 96kbps.

To me LAME is best at any bitrate compared to other mp3 encoders. Also the low bitrates even though LAME doesn't have intensity stereo support.

The LAME presets (-V 0 to -V 9) are just awesome so use them and respect them. So if you want a bitrate around 128kbps you should try -V 5 for great quality.

Oh, and never convert an mp3 to another mp3. Cause that way you will have the flaws of the first mp3 encoding and also the flaws of the second mp3 encoding. In other words: it will always sound worse than the original mp3, no matter what bitrate you use.

Finally: Razorlame is a piece of shit imho. The presets suck and it's really too complicated for a frontend. Use something simple like All2lame.

[theReal]

Finally: Razorlame is a piece of shit imho. The presets suck and it's really too complicated for a frontend. Use something simple like All2lame.

???

Razorlame is pretty straightforward. Just use "only use custom settings" and enter your commandline switches. Then you drag&drop files and press encode and you're done. It's very easy to use this way once you've entered the right commandline switches for lame.

[nukesgoboom]

GodofaGap said:
==============
A. Because is LAME is not specifically tuned at low bitrates.
B. It lacks an important tool for low bitrates --> intensity stereo.
C. It is very likely that lower resulting bitrate will have a noticable impact on quality anyway. Since bitwise the only way to achieve a lower bitrate would be to sacrifice quality compared to the 128 kbps. (Your maximum frame size is 128 kbps).
=============

all of your points lead me to believe VBR is only useful for higher bittrates (at least 128 and above). is this what you were trying to tell me? you need higher bittrates to bypass all of those problems.







Slogra said:
===================
The LAME presets (-V 0 to -V 9) are just awesome so use them and respect them. So if you want a bitrate around 128kbps you should try -V 5 for great quality.

Oh, and never convert an mp3 to another mp3. Cause that way you will have the flaws of the first mp3 encoding and also the flaws of the second mp3 encoding. In other words: it will always sound worse than the original mp3, no matter what bitrate you use.

Finally: Razorlame is a piece of shit imho. The presets suck and it's really too complicated for a frontend. Use something simple like All2lame.
======================

i thought the lower the better for -v. why not use -v 0 for maxium quality? or am i confused?

so you say razorlame sucks. how about omniencoder? have you used that before? i will also check out all2lame tonight.

[GodofaGap]

all of your points lead me to believe VBR is only useful for higher bittrates (at least 128 and above). is this what you were trying to tell me? you need higher bittrates to bypass all of those problems.
VBR is only useful if you use it with the default settings and do not add any switches of your own (except perhaps --vbr-new). So no limiting of the bitrates in any kind of way. If you want a lower resulting bitrate use a higher V preset. "The lower the better for V" only makes sense if you don't cripple the presets.

The 2 points I was making is that LAME might not be the best encoder for lowbitrate mp3's (although no large scale test for that has ever been done), but it is certainly lacking a tool. Secondly, 128 CBR will normally be better than a VBR mode that is limited to 128 max.

[foxyshadis]

--vbr-new is default now.

imho mp3 should never be used for < 112 these days anyway, except for hardware compatibility; there are so many codecs designed to fill that space with wide support (aac, aac+, vorbis, wma, or amr) to make up for mp3's shorfalls that it's a bit wasteful to bother. Except for set-top divx players and old cell phones, almost all hardware now supports at least one of those alternate formats.

[nukesgoboom]

GodofaGap said:
=============
VBR is only useful if you use it with the default settings and do not add any switches of your own

128 CBR will normally be better than a VBR mode that is limited to 128 max.
=============

ah, thank you for those answers. im using a program called omniencoder that says it uses the 'recommended' settings for lame by default on VBR. have you heard of this tool? all the stuff ive done so far sounds very good, but the minimum kbps i see in winamps display is always 112.






foxyshadis said:
=============
mp3 should never be used for < 112 these days anyway
=============

good point, but AAC uses trickery to make low kbps music sound better, it does sound 'better' but its not an accurate representation of the source. how do you feel about this??

[foxyshadis]

Every codec uses trickery, especially at low bitrates. Mp3 destroys the stereo image. Wma softens the sound quite a bit. Vorbis raises the volume and lowers the lowpass significantly. And of course aac+ has harmonic extension (sbr) - when you can actually watch a streaming video with 15 kbps sound that sounds like an mp3 4 times that size, it doesn't matter how accurate it is. But even aac-lc sounds less artifacty (and "more accurate") than mp3 at any given sub-128 bitrate.

[gameplaya15143]

And of course aac+ has harmonic extension (sbr) - when you can actually watch a streaming video with 15 kbps sound that sounds like an mp3 4 times that size, it doesn't matter how accurate it is.Comparable audio bandwidth (lowpass) but they don't sound alike. It is possible to stream in NSV with x264 and aac+ ;) by the way.

But even aac-lc sounds less artifacty (and "more accurate") than mp3 at any given sub-128 bitrate.
I don't entirely agree with that statement. My personal tests have shown that lc-aac isn't that much better than mp3, even at bitrates < 96kbps.

@nukesgoboom
Check out the commandline in my sig, perhaps it is what you are looking for (only works on lame 3.93 or earlier). Or even plain old "--alt-preset 104" might work out for you.

If you really insist on using a newer version of lame, try out 3.98a6, it is quite a bit better for problem samples which are most ugly at low bitrates. "-V 9 --resample 44 --lowpass 16" or similar should do what you want with the newer lame versions.

[shon3i]

And of course aac+ has harmonic extension (sbr) - when you can actually watch a streaming video with 15 kbps sound that sounds like an mp3 4 times that size, it doesn't matter how accurate it is.Not only aac+, mp3PRO sounds very good at those bitrates, much better than mp3 or LC-AAC, but HE-AAC @ 15kbps sounds much better than mp3PRO at same rate, aslo mp3PRO @ 64kbs have big chance have same quality like HE-AAC.

don't entirely agree with that statement. My personal tests have shown that lc-aac isn't that much better than mp3, even at bitrates < 96kbps.On 96, maybe if you use --resample switch or some tweaking, on 64 mp3 is horrible sounding 22khz file, where is LC-AAC on 32000(ct)-44100(nero) but sounds very simmilar, for lower rates i don't say anything.

[gameplaya15143]

On 96, maybe if you use --resample switch or some tweaking, on 64 mp3 is horrible sounding 22khz file, where is LC-AAC on 32000(ct)-44100(nero) but sounds very simmilar, for lower rates i don't say anything.
IIRC both CT and Nero lowpass at 12khz.

In my experience lame3.93.1 "-q 5 -V 9 -Z --nspsytune --resample 32 --lowpass 12" will typically end up somewhere in the same bitrate range.

Both lc-aac and mp3 have their own artifacts at such rates though. :sly:

[foxyshadis]

Not only aac+, mp3PRO sounds very good at those bitrates, much better than mp3 or LC-AAC, but HE-AAC @ 15kbps sounds much better than mp3PRO at same rate, aslo mp3PRO @ 64kbs have big chance have same quality like HE-AAC.
Well yes, mp3pro is just a retrofit of some of the then-draft aac standard into mp3, afaik band-limiting and sbr are done exactly the same way. So of course they sound the same. Why would you use mp3pro now that aac is well-supported and common?

I don't entirely agree with that statement. My personal tests have shown that lc-aac isn't that much better than mp3, even at bitrates < 96kbps.
Indeed, when I was making a batch for my portable, some were better and some worse. Dratted things.

[nukesgoboom]

foxyshadis - i see. thanks for informing me of the subtle nuances each codec uses to achieve good sound at such low size. i wonder whats so magical about the 128 kbps bittrate and why at this exact point the codecs sounds similar?





gameplaya15143 - lame 3.93.1 -q 5 -V 9 -k -Z --resample 44 --nspsytune

thank you! i noticed you use lame 3.93.1 ... i thought lame 3.97 was the most recent stable version? i wonder why it would not work with 3.97?

3.98a6! impressive! i wonder where you found it man. perhaps you know a developer, i can only find 3.97, may i ask where you got it?

resample 44 on the new lame will be great to look forward too, but im told having a high sample rate on low bittrates will actually harm the audio more : ( much like having 50 pounds of mud in a bucket that only holds 25 pounds. i heard this analogy on the winamp forums and it seemed helpful.

[DarkZell666]

http://www.rarewares.org/mp3.html ;)

Actually, I would have used :

lame -q 2 -V 5 --resample 32

The --nspsytune isn't listed in the latest beta's cmdline switches in any case. Dunno if it's the reason why it didn't work.

The -q switch is the quality of the algorithm (slow vs. fast) where 0 is very slow (and good) and 9 is very fast (and bad).
You'll probably gain more quality using a lower -q value than resampling to 32khz btw.

Why 32 ? because sampling at 32khz means that frequencies up to 16khz will be encoded (32 / 2), and because after 16khz the loss isn't as audible as the artifacts caused by encoding the higher frequencies (IMHO of course ;))

For the 128kbps stuff, MP3 was designed to stream audio on an ISDN connection (128kbps if dual line, 64kbps if single line). That's why it was the reference spot at the time. Since, everyone uses it because it's near-transparent quality.

[gameplaya15143]

Versions of lame up untill 3.93.1 used gpsycho as the default psychoacoustic model (--nspsytune is an experimental 'tuned' psymodel in those version).

From 3.94 and up, nspsytune became default, as did the -Y switch (which makes vbr ignore noise in sfb21 {noise higher than 16khz}, this is why the commandline I use with 3.93 will not work with newer versions.. and some other reasons as well). Do some reading on sfb21 bloat.

Beware of the -q 0 bug in the older versions of lame. It can make things sound worse. You can find this mentioned in various places on hydrogenaudio.org. I use -q 5 for a reason ;) and this is the reason.

In my opinion, I think it is best to use the lowest samplerate possible required to reproduce the desired bandwidth whenever encoding audio. Some may or may not agree with that.

[nukesgoboom]

DarkZell666
===================
Why 32 ? because sampling at 32khz means that frequencies up to 16khz will be encoded (32 / 2), and because after 16khz the loss isn't as audible as the artifacts caused by encoding the higher frequencies (IMHO of course )

For the 128kbps stuff, MP3 was designed to stream audio on an ISDN connection (128kbps if dual line, 64kbps if single line). That's why it was the reference spot at the time. Since, everyone uses it because it's near-transparent quality.
===================


if the loss isnt as audible after 16khz, then 24khz (44/2) must explain why its so popular. i know 48 isnt seen much because audio cd sampling is 44 khz, but the higher the better correct? or does 128 reach the same problem as 32 does, if you give 128 a high enough sampling rate, the sound is actually worse because of the extra samples for the low bitrate?


ahhh... it comes full circle now. no wonder 128 sounds so great to me. thanks for that interesting piece of info.


thank you for the link also. i definitly want to try lame 3.98!




gameplaya15143
=================
Beware of the -q 0 bug in the older versions of lame. It can make things sound worse. You can find this mentioned in various places on hydrogenaudio.org. I use -q 5 for a reason and this is the reason.

In my opinion, I think it is best to use the lowest samplerate possible required to reproduce the desired bandwidth whenever encoding audio. Some may or may not agree with that.
=================

whoa. i had -q 0 on lame 3.97. am i screwed? you said older versions though. when was this fixed?

desired bandwith? you mean the source sampling rate right? if you have a source of 44 khz, why not keep 44 khz?

[Shinigami-Sama]

heres a food analagy for it

think of the sampling rate as the size of your piece of toast
big number means big piece of toast

and the kbps as how much butter or [insert toast spread here] you have

so if you have a large piece of toast(high khz)
and you want to keep the 44kz, that means you need to spread your butter(kbps) thinner

now for higher rates like 224kbps this isn't to bad
but lower than 128kbps and it gets so thin you can't quite cover the whole peice, so you drop the sample rate(there by making your toast smaller) so that you can have better coverage and a better tasting(sounding) piece of toast

make a bit more sense?
or did I just horribly confuse you?

[gameplaya15143]

DarkZell666
whoa. i had -q 0 on lame 3.97. am i screwed? you said older versions though. when was this fixed?It should be ok with 3.97.

desired bandwith? you mean the source sampling rate right? if you have a source of 44 khz, why not keep 44 khz?
No. I mean audio bandwidth (lowpass frequency). Max audio bandwidth can be half the samplerate. So if a lowpass of 14khz is used, I see no point in using a samplerate higher than 32khz, regardless of what the original samplerate was. When using a low bitrate, higher samplerates can reduce quality. Higher samplerates are more accurate than lower ones but require more bits to encode.

[nukesgoboom]

Shinigami-Sama - i use butter, and i love your analogy. 128 and above is what i use anyway, and i wont worry about higher, since mp3 caps at 320 and this problem at higher bittrates wont even surface. thanks again.




gameplaya15143 - thank god!!! im saved! indeed Higher samplerates are more accurate than lower ones but require more bits to encode, and mp3 caps at 320 so ill never have to worry about it would sound like at 96 khz cause mp3 only allows max 48 khz. im not experienced enough to use AAC or have a need for it, since my devices only support mp3.

thanks for your excellent feedback man.

[theReal]

ill never have to worry about it would sound like at 96 khz cause mp3 only allows max 48 khz.

96kHz won't sound different unless you're a dog or a bat ;)
Seriously, no human being can hear frequencies above 21kHz, most people can only hear frequencies up to 18kHz or even less. Double these 21kHz and you're at 42kHz sampling rate. Give this some extra headroom and you're at 44kHz - that's why 44kHz is totally sufficient for humans :)

[bond]

a good mp3 encoder wil downsample automatically at a specific bitrate when it think its brings better quality. its a feature, not a bug ;)