IPB

Welcome Guest ( Log In | Register )

 
Reply to this topicStart new topic
Should I go with CBR or VBR(0) for my mp3's?
DaGrandMastah
post May 3 2012, 03:35
Post #1





Group: Members
Posts: 15
Joined: 3-May 12
Member No.: 99436



Hi all, new to the forums. I had asked this question elsewhere and was told this is the best place to ask.

I'm currently in the process of re-converting my cd library into mp3's on my computer. Ultimately I'd like to get the best (lossy) sound quality I can get...and was originally planning to go with 320 CBR but I'm now reading a lot about VBR v-0 and how it's basically equivalent with an impossible to notice difference between the 2. I'm not concerned about space on my computer hard drive but I do plan on loading up these mp3's to my iPhone (where space can get scarce). I just converted 2 FLAC files and noticed a 1.5 mb difference between the 2, but as far as I can tell, no real degradation in sound quality.

Id just like to get some thoughts on this. I'm certainly no expert on the subject of sound science. To be honest, I'm not an audiophile at all...but at the same time, I want to "future proof" myself so that I never do realize (like I did this time around) that I'd like to upgrade my library (upgraded my headphones and started noticing a difference in quality of some mp3's).

Is there any reason why I may regret going with vbr v-0 over cbr or will I thank myself for saving the space? I figure that if some of the professionals on here cannot notice a difference then there is no chance I ever will. smile.gif

Also, if there are any mac users here, could you please give me your opinion on the best settings for converting FLAC/CD's in XLD? Are the below settings good?

http://i50.tinypic.com/2z9e7u9.jpg


This post has been edited by db1989: May 3 2012, 07:36
Reason for edit: TOS #6: not an MP3 tech issue / TOS #5: post completely different topics to their own threads
Go to the top of the page
+Quote Post
eahm
post May 3 2012, 04:23
Post #2





Group: Members
Posts: 1084
Joined: 11-February 12
Member No.: 97076



1) You will regret it if you don't save them in lossless not because of the sound quality but for the future when you want to switch to another lossy format. Go FLAC.

1.1) The lossy I use for my player (Apple) is AAC CBR256 and the one I use for the random folder (random music with no particular rule or album, music I listen to maybe once a year but I want to keep just in case) is LAME MP3 CBR320.


2) I use CUETools, dBpoweramp and foobar2000 on Windows sorry.

This post has been edited by eahm: May 3 2012, 04:29


--------------------
/lwAsIimz
Go to the top of the page
+Quote Post
Erich_2
post May 3 2012, 11:17
Post #3





Group: Members
Posts: 243
Joined: 10-September 08
Member No.: 58107



cbr versus V0
Go to the top of the page
+Quote Post
twostar
post May 3 2012, 11:22
Post #4





Group: Members
Posts: 487
Joined: 5-August 02
From: Manila
Member No.: 2939



I recommend ripping to ALAC. Itunes could then convert to AAC 128kbps on the fly. If you change your mind and want to fit more into your iPhone, or upgrade to one with bigger capacity, you could just convert to another format/bitrate.
Go to the top of the page
+Quote Post
LosMintos
post May 3 2012, 12:33
Post #5





Group: Members
Posts: 37
Joined: 30-January 08
Member No.: 50912



QUOTE (DaGrandMastah @ May 3 2012, 04:35) *
I want to "future proof" myself so that I never do realize
Go (or stay?) lossless! FLAC, ALAC, whatever. For the rest I agree with twostar (AAC 128 kbps). Regarding your original question I suggest -v2.
Go to the top of the page
+Quote Post
DVDdoug
post May 3 2012, 18:06
Post #6





Group: Members
Posts: 2592
Joined: 24-August 07
From: Silicon Valley
Member No.: 46454



QUOTE
I just converted 2 FLAC files and noticed a 1.5 mb difference between the 2, but as far as I can tell...
What? Two FLAC files were different by 1.5MB? I suppose the original files were different sizes?

FLAC/ALAC will give you a file somewhere in the ballpark of 60% the size of the uncompressed original. So, a FLAC file from a CD (44.1kHz, 16-bt, stereo), will typically end-up a bit more than twice the size of a 320kbps MP3/AAC. If you losslessly compress a file with a higher bitrate, higher bit-depth, or more channels, it will be larger, since it will be compressed to about 60% of its original size.

But since bitrate is related to file size, the 320kbps MP3/AAC file will be the same size (depending on playing-time) no matter what the original format is.

QUOTE
...no real degradation in sound quality.

Since FLAC is lossless, there is NO degredation. When your player decompreses it, it's playing the exact same audio-data as the original uncompressed WAV/AIFF/CD.
Go to the top of the page
+Quote Post
shadowking
post May 4 2012, 01:33
Post #7





Group: Members
Posts: 1527
Joined: 31-January 04
Member No.: 11664



IMO you should take time to find out what is acceptable to you and not follow the current 256 ~ 320k trend. With mp3 there is no full transparency 100% of the time - but public test shows good (near-transparent ) behavior at 135k VBR. The setting is around V5 not V0

Obviously there is a threshold where things fall apart say V8 will likely introduce distortion on casual listening but V5 might not or the difference will be subtle. V4 will offer even more headroom to please nearly everyone. Differences will only be revealed in abx testing and are subtle most of the time. For these reasons many people use V3 ~ V2 or even lower.

So what im saying you can rip your CD to half the rate of 320k and still get very good quality. The current trend is not supported by listening tests. Also many believe higher bitrate will solve rare problems but this is a hit and miss thing.

Consider also the lossless audio suggestions given by others.


--------------------
Wavpack -b450s0.7
Go to the top of the page
+Quote Post
shadowking
post May 4 2012, 04:57
Post #8





Group: Members
Posts: 1527
Joined: 31-January 04
Member No.: 11664



I will add that if you really need to go with high bitrates ,go VBR as 320 is kind of stupid in a way - worse than lower rate CBR that is still portable. V0 recently has been heavily tuned producing a bitrate between 256 ~320.. I think even V1 is very good for such purposes @ 210 .. 240 k

This post has been edited by shadowking: May 4 2012, 04:58


--------------------
Wavpack -b450s0.7
Go to the top of the page
+Quote Post
antman
post May 13 2012, 15:17
Post #9





Group: Members
Posts: 103
Joined: 3-January 07
From: Texas
Member No.: 39241



QUOTE (shadowking @ May 3 2012, 19:33) *
IMO you should take time to find out what is acceptable to you and not follow the current 256 ~ 320k trend. With mp3 there is no full transparency 100% of the time - but public test shows good (near-transparent ) behavior at 135k VBR. The setting is around V5 not V0


Amen. I've noticed this trend too. It always makes me think of this thread:

ABX Just Destroyed My Ego
http://www.hydrogenaudio.org/forums/index....mode=linearplus

And think these tests were done with codecs from 6 years ago!

Go to the top of the page
+Quote Post
timcupery
post Jul 5 2012, 17:43
Post #10





Group: Members
Posts: 780
Joined: 19-December 01
From: Tar Heel country
Member No.: 683



QUOTE (antman @ May 13 2012, 10:17) *
ABX Just Destroyed My Ego
http://www.hydrogenaudio.org/forums/index....mode=linearplus
And think these tests were done with codecs from 6 years ago!

Awesome old thread to mention.
I think these claims would stand up basically unchanged today, as codec quality hasn't made significant strides in the past 6 years. Not that it has gotten worse, but most codecs are running into diminishing marginal returns. Each incremental step or new version is a smaller step forward.


--------------------
God kills a kitten every time you encode with CBR 320
Go to the top of the page
+Quote Post
eahm
post Jul 5 2012, 17:59
Post #11





Group: Members
Posts: 1084
Joined: 11-February 12
Member No.: 97076



Reading my comment up there, I can't believe I used to use AAC CBR 256 Kbps. I now use AAC True VBR ~165 Kbps, I guess ABX destroys everyone's egos. I probably could even go lower but I think ~160 is my sweet spot.

Please test MP3 LAME V2, it sounds transparent to me, always ABX if you have time, you will be amazed.


--------------------
/lwAsIimz
Go to the top of the page
+Quote Post
halb27
post Jul 5 2012, 21:45
Post #12





Group: Members
Posts: 2435
Joined: 9-October 05
From: Dormagen, Germany
Member No.: 25015



It's fine to use moderate bitrate because quality is great.
But it's also fine to use high bitrate like the OP does if you don't have to care about storage space and just feel better about rare events when moderate bitrate quality isn't great.

For the latter case: Apart from using -V0 or CBR320 you can use kind of a combination of VBR and CBR by using the method of my signature. It does essentially what -V0 does but doesn't allow audio data bitrate to be moderate in any situaton other than silence. Moreover available audio data space is maximized a little bit more than with standard 3.99, and accuracy demands for pre-echo situations are increased.

This post has been edited by halb27: Jul 5 2012, 21:54


--------------------
lame3100m -V1 --insane-factor 0.75
Go to the top of the page
+Quote Post
yourlord
post Jul 5 2012, 22:10
Post #13





Group: Members
Posts: 218
Joined: 1-March 11
Member No.: 88621



If one of the reasons for using "needlessly" high bitrates when ripping mp3 is that one doesn't care about how much space it takes up then there is no reason not to rip as lossless.
IMO, the only reason to rip using a lossy codec at high bitrates (beyond 200Kbps) is generally an irrational paranoia about preserving sound quality.

Well, if one is paranoid about preserving sound quality and space consumption isn't an issue, lossless is THE best option.. Once one has their library in a lossless format they can then transcode on the fly to whatever lossy codec they desire and is supported by whatever device without any generational quality issues.

Plus, since one's main library is lossless, and once they do a little ABX testing and realize they can't ABX V5 mp3 or q2 Vorbis (aac, etc) against the lossless versions, they can feel better about transcoding to much lower, rational, bitrates/quality on those devices instead of copying over huge, high bitrate lossy "archival" files. This allows one to store MUCH more music on devices with limited capacity.

Just my opinion..
Go to the top of the page
+Quote Post
halb27
post Jul 5 2012, 22:37
Post #14





Group: Members
Posts: 2435
Joined: 9-October 05
From: Dormagen, Germany
Member No.: 25015



Yes, lossless is the way to go, no doubt after ripping, and in an ideal world also for listening. Unfortunately lossless codecs are not always supported on devices. So we're back to lossy and where we started: moderate bitrate for people who feel fine with the good quality it delivers, and high bitrate for the a bit paranoid who just feel better with it and who don't have to worry about storage space like me.
It's good that everybody's needs can be fulfillled.

This post has been edited by halb27: Jul 5 2012, 22:41


--------------------
lame3100m -V1 --insane-factor 0.75
Go to the top of the page
+Quote Post
AshenTech
post Jul 6 2012, 02:49
Post #15





Group: Members
Posts: 78
Joined: 11-November 08
Member No.: 62144



I agree on the "go lossless" thing, Its silly to rip to mp3 then have to re-rip disks that in the meantime may end up getting damaged, when if you go lossless, you just rip to Flac and convert to whatever format your current device likes the best.
iPhone that would be AAC, android, I would go Ogg, most good audio players today seem to be supporting Ogg and Flac as well.

Another thing to note for everybody is asking the OP what kind of speakers/earphones/headphones/exct he/she plans to use, if its ibuds or the like, theres no point in going high or even medium bitrate in my opinion, you cant really hear the difference with those low quality in ears.(i have a few sets here....i can hear mp3 artifacts only if i really try, unless the encode is stupidly bad or at a really really low rate.

as to ripping with a mac.....Not sure what to tell you, every time I see iTunes I cringe be it on windows or mac....

On windows I use dbpoweramp for most of my ripping/encoding needs anymore.....you may want to try asking for suggested apps on a site like anythingbutipod they have some mac users....you could also check something like http://alternativeto.net/software/itunes/ to find other apps you can use, you will need to look for osx software (thats a basic search not set by license or anything else)

but yeah, I always rip to Flac, then convert to ogg for mobile use, For me, it sounds the best and saves space.....and my sandisk and cowon and samsung players love it wink.gif
Go to the top of the page
+Quote Post
RobertoDomenico
post Jul 6 2012, 03:12
Post #16





Group: Members
Posts: 87
Joined: 16-November 11
Member No.: 95208



If you're planing to use Lame encoded mp3's on iOS device some of them do not work with gapless playback (my experience, iPod Touch and iPhone) over home sharing.

If you're in the Apple ecosystem i would personally rip my Cd's in ALAC and let iTunes encode on the fly. iTunes can encode on the fly to 128, 192 and 256 AAC cbr.
Go to the top of the page
+Quote Post
Canar
post Jul 6 2012, 04:30
Post #17





Group: Super Moderator
Posts: 3361
Joined: 26-July 02
From: princegeorge.ca
Member No.: 2796



320 and V0 are total overkill. I've heard the difference between 320 and lossless on a few samples. If I want to be confident there isn't going to be any huge problem, I use V2. If I want to try fit as much music into as small a space as possible, I use V5. V2 is more a precautionary measure. V5 still sounds great. If I want to eke out every last bit and make things sounds as good as I can make them, I use Vorbis aoTuV q3, but that's besides the point.

The people here are more than willing to help you find that optimal preset, if you care to do the research. 320 vs. V0 is totally a false dichotomy. If you want real quality, use lossless. If you want small files, use around V5. Lossless is a great place to start, because then you can convert to MP3 for your portable. If you aim a bit low, and find that V5 isn't quite enough, you just need to re-encode.


--------------------
You cannot ABX the rustling of jimmies.
No mouse? No problem.
Go to the top of the page
+Quote Post
rhfrjpfdh
post Sep 3 2012, 12:37
Post #18





Group: Members
Posts: 1
Joined: 3-September 12
Member No.: 102860



hello to all
i'm interested in the question (lame 3.99.5): in frames which mp3 file keep more audio information ( -V 0 -b 320 --lowpass 20.5 ) or ( -q 0 -b 320 [default --lowpass 20.5] )?
Go to the top of the page
+Quote Post
Kohlrabi
post Sep 3 2012, 13:20
Post #19





Group: Super Moderator
Posts: 1050
Joined: 12-March 05
From: Kiel, Germany
Member No.: 20561



QUOTE (rhfrjpfdh @ Sep 3 2012, 13:37) *
i'm interested in the question (lame 3.99.5): in frames which mp3 file keep more audio information ( -V 0 -b 320 --lowpass 20.5 ) or ( -q 0 -b 320 [default --lowpass 20.5] )?
If you want to keep as much information as possible you should use lossless compression. Lossy compression is used to produce audibly transparent files (i.e. without artifacts) at small filesizes. Using overkill compression options does little to nothing to diminish the likelihood of artifacts on problem samples.


--------------------
Ceterum censeo Masterdiskem esse delendam.
Go to the top of the page
+Quote Post
halb27
post Sep 3 2012, 16:38
Post #20





Group: Members
Posts: 2435
Joined: 9-October 05
From: Dormagen, Germany
Member No.: 25015



QUOTE (Kohlrabi @ Sep 3 2012, 14:20) *
Using overkill compression options does little to nothing to diminish the likelihood of artifacts on problem samples.

This is not true. Try for instance lead-voice and eig with my variant 3.99.5y -V0+.


--------------------
lame3100m -V1 --insane-factor 0.75
Go to the top of the page
+Quote Post
Dynamic
post Sep 3 2012, 17:22
Post #21





Group: Members
Posts: 819
Joined: 17-September 06
Member No.: 35307



QUOTE (halb27 @ Sep 3 2012, 16:38) *
QUOTE (Kohlrabi @ Sep 3 2012, 14:20) *
Using overkill compression options does little to nothing to diminish the likelihood of artifacts on problem samples.

This is not true. Try for instance lead-voice and eig with my variant 3.99.5y -V0+.


Might it be fairer to state that this is not universally true, but that where the psychoacoustic model has failed at around -V3 or -V2 and we use higher quality options, most of the extra bitrate is spread evenly over the whole sample, so very few of the extra bits will be applied to the features that require more bits to become transparent. If the psymodel was aware of where they were needed, it wouldn't have failed to achieve transparency in the first place and would have provided the extra bits in just the right places.

It's my impression that the lead-voice and eig examples are exceptions, but many other problem samples where there's a defect in MP3's designed capabilities or the psymodel is wrong get only slightly and gradually better as higher quality settings are applied.
Go to the top of the page
+Quote Post
halb27
post Sep 3 2012, 18:51
Post #22





Group: Members
Posts: 2435
Joined: 9-October 05
From: Dormagen, Germany
Member No.: 25015



QUOTE (Dynamic @ Sep 3 2012, 18:22) *
... but many other problem samples where there's a defect in MP3's designed capabilities or the psymodel is wrong get only slightly and gradually better as higher quality settings are applied.

Do you have samples to back up your claims? Please check with 3.99.5y -V0+.


--------------------
lame3100m -V1 --insane-factor 0.75
Go to the top of the page
+Quote Post
halb27
post Sep 6 2012, 07:45
Post #23





Group: Members
Posts: 2435
Joined: 9-October 05
From: Dormagen, Germany
Member No.: 25015



An actual post gives another sample that shows that top quality settings which usually are overkill can save the day for problematic stuff: Nightwish - Angel Falls


--------------------
lame3100m -V1 --insane-factor 0.75
Go to the top of the page
+Quote Post
Dynamic
post Sep 6 2012, 19:48
Post #24





Group: Members
Posts: 819
Joined: 17-September 06
Member No.: 35307



Thanks for the Angel Falls link, halb27. I ABXed that at -V 5 but failed at -V 5+, so I'm very impressed. I'm not very good at hearing this sort of artifact and it doesn't annoy me enormously.

My intention in my previous post was to point out that your -V n+ modes like -V 0+ are a useful exception to a rule-of-thumb that does work fairly well for many codecs with killer samples to describe how it's usually not possible to apply enough extra bits to the right spot, so, as Kohlrabi said, it usually does little to reduce the audibility of artifacts on problem samples.

For example imagine a codec with a -q scale that fails to represent a particular frequency at -q5 by having a masking curve that's 11 dB too high for that feature of that problem sample. If you applies extra bits by changing to -q7 by reducing the quantization noise (increasing the Signal-to-Mask-Ratio) by, say 6dB, it would use a lot of extra bitrate across the spectrum to make a lot of inaudible masked noise quieter still (to no audible benefit) but the one frequency area where the psymodel got it wrong would only get improved by 6 dB too and would still be 5 dB short of the quantization accuracy it requires. Perhaps it would need -q9 or -q8.7 to make the artifact inaudible at the cost of even more bits. Identifying and fixing the psymodel's masking curve inaccuracy could let you apply the required 11 dB improvement at the problem frequency with only a few bits more at -q5, without spreading so many bits around to encode inaudible detail as -q7 would do without fixing the problem.

You seem to have identified that an important class of LAME problem samples that now remain (and seem to get worse with some versions) are associated with certain short blocks receiving insufficient bits, so if I understand it correctly, you've narrowed down where the extra bits are spread, causing less waste with your -V n+ fix than might be the case with most other types of approach short of detecting the specific situations where these artifacts occur and applying the maximum bitrate with high selectivity.

This post has been edited by Dynamic: Sep 6 2012, 19:50
Go to the top of the page
+Quote Post
halb27
post Sep 6 2012, 20:43
Post #25





Group: Members
Posts: 2435
Joined: 9-October 05
From: Dormagen, Germany
Member No.: 25015



You are right in assuming I don't care about the psy model (I am simply not able to do so), but just do anything to encode short blocks with maximim possible accuracy.
But I take care of long blocks as well by not allowing audio data bitrate to fall below a certain (relatively high) threshold. I do so by internally increasing SNR demands the way it was possible up to 3.98 by using the options --ns-bass/alto/treble.
Last not least I keep bit reservoir close to maximum in order to always be able to encode short blocks as well as fulfill Lame 3.99.5 -V0's high quality demands the best possible way.

This post has been edited by halb27: Sep 6 2012, 21:00


--------------------
lame3100m -V1 --insane-factor 0.75
Go to the top of the page
+Quote Post

Reply to this topicStart new topic
1 User(s) are reading this topic (1 Guests and 0 Anonymous Users)
0 Members:

 



RSS Lo-Fi Version Time is now: 19th September 2014 - 16:08