First of all, I have an intermediate understanding of audio formats. I know the difference between the more popular formats, CBR vs. VBR. I know about LAME and EAC and I think I know how to configure them correctly.

That said, I'm under the impression that the --alt-preset standard/extreme/insane settings are for the best quality at the lowest size - but can they offer better quality than 320CBR? That's mainly what I want to know.

For most music, I just throw it at Windows Media Player and have it do 192CBR. I know it could sound better, but for "most music" it doesn't matter.

But I have something I need to have the best possible quality. I would use a lossless codec (FLAC) and call it a day. In fact I have the CDs as FLAC, via EAC's settings as configured by Chris Myden (which, I think is for .mp3, but I just changed it to FLAC), on my computer. But my 40GB Zen Touch only supports Mp3, WMA, and DRM'd WMA. I don't think it supports WMA Lossless, but I haven't tried. (That would be acceptable, I'm using less than a third of that 40GB.)

Basically, I don't care about size, I just want the best quality my Zen will play in the car. Would that be 320CBR or a VBR format?.

PS... The music is Nightwish, a Finnish symphonic metal act with opera vocals. On the computer it sounds great (in FLAC) but in the car, I feel like I'm missing something, and with something that complex, with such a range, I feel I need the best quality I can get. And they're worth it, they're my favorite band.

PPS - I saw the name Sebastian Mares on the site coming in. Does he run this board or just post here? I knew him on another board (which is where I remember the name HydrogenAudio from, hence why I came here for audio advice). (IPS board, I posted as Dark Reality and then //Nathan.)

You should do some ABX testing to convince your mind that -Vx preset are really good enough - Especially for use in a car where the noise floor is quite high!

Hi NathanJ, 320 cbr is the best mp3 setting. If you are wanting the best for your player and your music taste. 40GB Zen Touch supports wave files. Not recommending either but just answering two of your questions.

cheers,
herefornow

320CBR is the best, and the alt-preset settings have long been replaced by a new system. See the wiki.

I'm inclined to add the LAME -k and -q0 switch to CBR 320 for highest possible quality (in theory), but I'm more inclined to use ABR 320 -k -q0 for smallest 320 mp3 file, with all the frequencies intact and most efficient compression.

--alt-preset insane IS 320 kbps CBR so it is already as good as you can get with mp3.

Sebastian is a regular contributor and highly valued member of HA, but does not run the board.

I love (just discovered) this band too. But 320 kbps should be more than enough. Try ABXing it against lossless if you are not convinced.

In addition to 320kbit CBR, you could use mp3repacker to remove the empty frames, making it a slightly smaller VBR file "losslessy".

Still I think you should try to convince yourself that it's a waste of space. I was of the same oppinion just a few years ago that I wanted the best possible mp3 quality, but still at this bitrate lowpass and psychoacoustics filters are applied to the source. Also keep in mind that even 320 kbit would suffer from some problematic samples.

I was convinced that bitrate is not everything when 32kbit aac fooled me in ABX test... To keep my mind clear I use lossless nowadays and transcode to something like mp3 -V5 when I need good quality for portable/car/etc.

What you do for yourself is up to you, but on these boards, refrain from giving advices that contradict the common knowledge and reasoning.

-k does not help having more quality. It helps on having *less* bits to store information.
-q 0 should be ok to use nowadays (in CBR it is still valid, as opposed to VBR with --vbr-new ), but it has been said long and loud that doesn't offer much over -q 2.

Ok, maybe what I do I should keep to myself, but I *didn't* write in this thread something like "you should try to do this or that..." (maybe on some threads I may have done it). I didn't offer advice. Thanks for the gentle reminder.

Since I got criticized on this, I would just like to defend my point (even if it is way too uncommon):

- It has been agreed that CBR 320 is the highest quality available, even breaking the barrier of the common knowladge knowledge and reasoning of "ABR is better than CBR". At the default settings for LAME CBR 320 (lowpass of around 20.5 IIRC and q value of 3), I believe the quality 320 kbps can still be stretched further via switches.

- It is also a common knowledge that CBR 320 is way too overkill (wasted bits), so to optimize these bits, why not include the full frequency spectrum using the -k switch (no lowpass). Many people use losssless lossless like FLAC because it is identical to the source (and that includes all frequencies at least up to 22khz for CD audio). One might say it is immaterial to include all frequencies due to the limitiations limitations of the human ear, but then again why lossless too if you can't hear them all. Conversely, why it would not be acceptable to include them all in MP3 if many swear by lossless. I haven't really pushed the OP to try the -k switch at 320, but maybe that is what would fill what he says "is missing". LAME documentation says -k switch may cause ringing or twinkling (in low bitrates like 128 I'm sure).

- q value of 0 does not offer additional quality, but it optimizes the compression of ABR 320 kbps by around 20kbps less. If two files would sound the same/identical (hey, don't throw in "have you ABX'ed?") but one is smaller than the other, then that is efficient, and indirectly saying that it is of higher quality given its smaller size. (I firmly believe this is a common reasoning). Compacting the size of the MP3 file would allocate the saved bits for the higher frequencies included by the -k switch.

- ABR 320, to save bits as well. AFAIK (correct me if I'm wrong or please point me to the evidence), there has been no direct conclusion that CBR 320 is better than ABR 320. So far, the general rule is that ABR is equal or better than CBR at any given bitrate (the common knowledge is that 320 is an exception). If ABR 320 and CBR 320 would sound equal (that is, neither sounds better), then I would go for ABR for its smaller size again. Same quality but smaller size = better quality for size.

That's why I rest at ABR 320 -k -q0. Criticisms are welcome.

Cheers!

EDIT: typos

Your defence of the -k switch shows that you just don't get the sfb21 issue with mp3.

I don't think it is an issue in ABR/CBR...

http://www.hydrogenaudio.org/forums/index....ost&p=46392

EDIT: At any rate, it wouldn't still change a thing between ABR 320 and CBR 320, ABR is just smaller

-k is no longer from 3.98

Deepest apolgies, Rio, both for the tone of my post as well as the inaccuracy. I was sure that I had seen a recent post by Robert that indicated otherwise, but I can't seem to find it, and other posts clearly support your position. I still don't go along with the use of -k though.

So far I'm using -k at 3.97. At any rate, at 3.98b6 ABR 320 would still be at least equal (if not superior) to CBR 320, consequently making it better value for bitrate with or without the -k switch.

ABR is underrated here. I'm inclined to believe that ABR 320 is better than CBR 320, given its smaller bitrate. ABX? No one still dared to do so to prove otherwise. Again, equality yet smaller size = better quality.

From the LAME HA Knowledgebase:

"The rule of thumb when considering encoding options: at a given bitrate, VBR is higher quality than ABR, which is higher quality than CBR (VBR > ABR > CBR in terms of quality). The exception to this is when you choose the highest possible CBR bitrate, which is 320 kbps (-b 320 = --alt-preset insane), but this produces very large filesizes for very little audible benefit."

It only mentioned an exception, but to what extent it is excluded was not mentioned. The only exception I can infer is that ABR 320 cannot be considered of higher quality than CBR 320, only equal.

Going back to the OP's question: My answer is ABR 320 -q0 (with or without -k)

Thnaks pdq. You got me thinking about the -k switch (even shadowking got me thinking twice). Maybe it's about the sfb21 bloating issue, maybe again it's not. Having the -k switch does not fully include all high frequencies at the same amplitude (but attenuated), only it does not apply lowpass altogether. At any rate, in my later posts you would find that I would have the -k switch optional too.

The point I'm driving at is that 320 MP3 can still be squeezed using ABR. Thanks for all your replies.

Cheers!

EDIT: Just tried 3.98b6. The -k switch doesn't apply at all. However, it appears to have a regression in ABR encoding (or inefficiency).



Encoding the same track using ABR 320 -q0, 3.97 yielded 298 kbps, while 3.98b6 yielded 317 kbps.

83.7% LR stereo.

Either your track has different songs in each channel or you have the bitrate way too high.

Pardon me, but I think I got you confused Mike and, yes I have the bitrate way too high. I'm encoding the said track using two LAME versions 3.97 and 3.98b6 set at ABR 320 -q0. 3.97 yielded 298 kbps, while 3.98b6 yielded 317 kbps. I also tested 3.98b5 and it has an identical 317 kbps bitrate.

Much as LR stereo is at 83.7% on all three versions tested, I'm just wondering why the sudden increase in bitrate, from 298 to 317. The most I can infer is that the -q switch is negligible in the newer version, which on LAME 3.97 has a great impact on shaving bits.

I also posted the same finding on the Lame 3.98b6 thread.

the reason why you get 317kbps is plain and simple

You are telling it to be around (ABR) 320kbps, and 3.98 is more correctly tuned than 3.97.

You should really change your setting to ABR 270, which halb27 (spelling?) recommends. That setting will give you what you're asking it to do. (288 would be the exact middle of the two highest bitrates, 256 and 320, if you want it higher)


About the usage of -k, the missing scalefactor band affects all encoding modes, not only VBR. That thread (aside of being from 5 years ago), just talks that it affected the most to VBR, because of the nature of VBR (note that it was vbr-old at the time, "fast" wasn't yet recommended)

JAZ,

Much as ABR really intends to hit the target bitrate (in this case 320), could somebody please explain why LAME 3.97 would play around 280-305 and the newer one almost hits 320.

What would be the offending version here? Would somebody say it's a bug in 3.97 and now corrected in 3.98?

Again the most likely culprit here is the -q switch (particularly at 0 value), which is operational in 3.97 and now defunct in 3.98. As with past experiences, encoding ABR 320 using 3.97 without the -q0 switch would tend to reach ~310 and above kbps, while using the -q0 switch shaves at least ~20kbps or more.

As with the -k switch, as posted earlier, I would drop it altogether though.

EDIT: It is a sad reality that an open source encoder would be rendered uncustomizable on newer versions. Would this be the direction of LAME? Sorry got it mixed up

Only ABX testing will tell you. I would suggest using LAME with -V 1 (second highest quality VBR mode, 19KHz lowpass) and start testing from there.

Nightwish is awsome! (so is the game in the vid)
..didn't know they were Finnish..

Why would you EVER use -k for lossy encoding? Waste bits to encode frequencies you don't hear - losing the bits of the spectrum that you CAN hear. That would definately defeat the claim that 320kbit should be suprerior to any other bitrate.

Also I don't believe that ABR320 just throws away the unused frames (similar to mp3repacker), but that could at least be tested.

This is LOSSY - WHY would you want to do this instead of using lossless, or maybe just lossyFLAC?

Why use anything much higher than 16k ? Who can provide good abx results with several MUSIC samples using -V2 vs V2 --lowpass 16.. No one I suspect.

An ABX test will certainly help you determine which format is best for your current set of equipment, and many would say (based on their own tests with whatever equipment they were using at the time) that a more highly compressed format (192-256 kbps) is more than enough. Many fail to consider the future, however (the next set of speakers or headphones they may purchase, or some revolutionary new type of DAC that hasn't been released yet) when recommending lossy formats. Since storage is no longer an issue in many systems, you might consider a lossless format for archiving if you are ripping a large collection of music that you really would rather not rip again. My own tests show that 245kbps sounds as good as lossless, by the way, but I still rip to a lossless format for these reasons (who knows what my next system might sound like, and my 500GB hard drive can easily hold my entire collection in FLAC format).

Hmm, there's a NIN song on Year Zero with a 19khz tone during a guitar solo, I ABX'd a really low -v preset (like 9 or 7) because of the low lowpass (also, those settings are bad quality all over).

BUT...I guess that's only one sample, not several

First, a 19kHz tone is not a normal component of music.

Second, you have to be very careful, because high amplitude ultrasonics that are really beyond the threshold of hearing, can easily result in intermodulation distortion that IS audible. A higher quality audio system could eliminate this distortion and make the ultrasonics inaudible again.