Skip to main content

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: which seems to be higher quality? (Read 2980 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

which seems to be higher quality?

I am trying to determine which files are better.


Media info gave me this data:

Encoded_Library   LAME3.99r
Encoded_Library_Settings   -m s -V 4 -q 3 -lowpass 20.5


the other is

Encoded_Library   LAME3.99
Encoded_Library_Settings   -m j -V 4 -q 0 -lowpass 20.5

Anything in the settings that tells us anything?  Both files are 320.


Here are some images and you can see a firmer cutoff at 16 with maybe more data in the non cut off?

Re: which seems to be higher quality?

Reply #1
You can't always tell by looking at graphs...  You can have a poor-sounding file with a pretty graph...   MP3 throws-away information to make a smaller file and it tries to throw-away details that you can't hear.  You can have a better looking picture by keeping stuff you can see (but can't hear) and throwing away things you can hear...

q0 is supposed to be the best.

And m j is also supposed to be the best.     With joint stereo left & right are converted to mid & side (when it makes sense*) and it makes better use of the bits, basically by saving the information that's common to left & right only once rather than duplicating.   It makes batter use of the bits so you can either get better quality or a smaller file.    The M/S conversion and back is lossless so you don't loose any "stereo".

Ore, since the lower bitrate file is smaller & more compressed we might say the lower bitrate is "better", especially if it sounds identical to a higher bitrate file.


Note that we can only say it's "better" if it sounds better (in a proper blind ABX test).  Of one file has a higher bitrate but sounds the same as a lower bitrate file, or if they both sound identical to the uncompressed original, we can't say the higher bitrate is "better".  




*  Joint Stereo is "smart"...    If left & right are totally different (let's say you have an audiobook with English in the left and French on the right) then Joint Stereo will not use M/S and the result will be the same as the regular Stereo setting.  OR it can switch moment-to-moment depending on the audio content.   About the only time you shouldn't use Joint Stereo is when you have a mono file, because the decoder will falsely report "stereo" to the decoder (and the listener).

Re: which seems to be higher quality?

Reply #2
airtas,
are these files created by you or are you only processing them?
Probably last because you are using 'media info'. Though seems like you have access to similar files where someone is playing around with encoder-switches..
Has the data only be encoded once ?

figuring out 'which files are better'..
Evaluating audio compressed by using a psychoacoustic model and inherent parameters on a technical level is not possible, only very obvious differences can be outlined and even them might not be used to judge about 'being better'.

What is your result after listening several times to both files and determine a personal preference, your 1st choice for 'quality'.
A valid and often used method for properly comparing audio is https://en.wikipedia.org/wiki/ABX_test-ing.

Do you take care for a big project where the choice made should be plausible and valid and generally aiming for the best project outcome, ..if this would really be a matter at all, which might be questionable ?
btw like that music.. :)

Will it make a difference in the universe if you choose one over the other?

Re: which seems to be higher quality?

Reply #3
It tells us that they used too many parameters. They should've used -V4 or -b320, but certainly not both. Those are presets which have been tested many times. Each parameter you add will usually lower the quality.

There is also something strange going on. Are you sure they are 320kbps? -V 4 should give you variable bitrate mp3s around 160kbps, so how can they end up at 320kbps??

If it's really 320kbps then the difference will be extremely small and not hearable. The bitrate of 320bkps is so high that that the impact of these incorrect settings is small.
If they are actually 160kbps then you might hear a difference. In theory i'd say -m j should be better quality at this bitrate. However -q 0 is not the default and not tested much, so it might affect the quality for the worse though.


While i agree that you should listen to the quality and not look at a spectrogram. I have to say that at 320kbps there should be no cut off at 16khz. So i would go with the "No cut off" mp3.

Re: which seems to be higher quality?

Reply #4
-V4 -b320 will produce "-V4" quality files, but pad up with "empty space" to achieve the bitrate of 320.
Try to encode twice with and without -b320 and compare the audio with e.g. foobar2000.

If you don't have the original, you can try to repack it to VBR using https://wiki.hydrogenaud.io/index.php/MP3packer (that is lossless if the file isn't broken). Use the -z option. Then bitcompare.

Files attached.

Re: which seems to be higher quality?

Reply #5
About the only time you shouldn't use Joint Stereo is when you have a mono file, because the decoder will falsely report "stereo" to the decoder (and the listener).
Assuming the encoder can spend almost all the bits on the mid channel and given that (mostly older hardware device) decoders sometimes have massive issues if you stray from 44.1kHz stereo, you should get close to the same quality as mono for the bitrate with increased compatibility.

Re: which seems to be higher quality?

Reply #6
typically someone will only use
Code: [Select]
Encoded_Library_Settings   -V 4
the other shown switches don't really optimize quality, and in all cases the results are VBR files somewhere around 140-185kbps.

see
wiki.hydrogenaud.io - LAME
LAME official homepage - guide to command line options  (scroll down for detailed info)

question is why your files are (CBR) 320kbps, at least spek 0.8.2.3 tells so..
that brings the question if files are encoded more than once to finally 320kbps-CBR-files or pumped up to 320kbps CBR without affecting quality (maybe to better suit older or limited hardware, car radios)
Sure, files are like they are

Re: which seems to be higher quality?

Reply #7
16k cutoff (lack of precision) is by design for CBR / ABR modes. Even for 320.
Bitrate is dictated by a metric rather than noise analysis as in vbr.
 If bands below 16k have too much demand for bitrate due to missing scale factor
band #21 , Then noise in that band is ignored or gets less precision.
Obviously it can be more dramatic if you look at a lower bitrate spectogram say 160k.

In VBR mode it can also happen for quality -V5 to -V3 and higher levels if
-Y switch is used.  If those VBR files have content above 16k, The CBR method
is used for those high freq (sfb21 band).  For other bands its VBR approach.
wavpack -b2.9cs.5
lame --preset cd -f

Re: which seems to be higher quality?

Reply #8
While I am sure others can give you more details, I tend to keep things simple... when it comes to MP3 I would not be too fussy about getting absolute highest quality but aim more for efficient bit rate at a good enough level of sound quality that most people won't notice.

so with that in mind... that's pretty much why I just stick to V5 (130kbps) MP3's as it's efficient and sound quality is high enough as when you just sit back and enjoy your music on any random device, chances are V5(130kbps average) is sufficient for most people quite easily. for lossy formats, MP3 it's always the safest choice since it's pretty much universally supported by anything that can play lossy files.

p.s. I know some might argue storage space is cheap nowadays and has been for quite sometime now. so with that in mind, many might not care about file size efficiency (and then prefer to increase bit rate) but it's more of the principle of it if you ask me as I figure why lose efficiency/storage space if not really needed especially given the sound quality is still high enough as it would be more understandable to increase bitrate/file size if there was more of a obvious drop off in sound quality. but with V5, given the public listening tests around here a while ago, shows it's all-around pretty good.
For music I suggest (using Foobar2000)... MP3 (LAME) @ V5 (130kbps). NOTE: using on AGPTEK-U3 as of Mar 18th 2021. I use 'fatsort' (on Linux) so MP3's are listed in proper order on AGPTEK-U3.

Re: which seems to be higher quality?

Reply #9
hi
may i ask few questions
1)
 it's really needed -lowpass 20.5 ?

2) and do LAME3.99r and LAME3.99 use 2 different engines?
I have always keeped lame updated  , thanks to John 's homepage rarewaves
but here i can't find 3.99 or 3.99r her John 's rarewares

@shadowking
Hi
i have always ripped my cds with eac and lame always updated ,but i can see that with V0 or 320k ( I use mostly) , my songs seems not cutted at
Quote
16k cutoff
at least in foobar2000 spectrum , often are around 18k
thanks

Re: which seems to be higher quality?

Reply #10
I'm not sure about the 'r', but LAME 3.99 began life 17 October 2011 and by 28 February 2012, it had become 3.99.5. If anyone is really interested, I have versions of LAME going back over 20 years. The oldest compiled version I have is 3.10 from 1999, but I have source going back further.

Re: which seems to be higher quality?

Reply #11
I'm not sure about the 'r', but LAME 3.99 began life 17 October 2011 and by 28 February 2012, it had become 3.99.5. If anyone is really interested, I have versions of LAME going back over 20 years. The oldest compiled version I have is 3.10 from 1999, but I have source going back further.
Hi  @john33
maybe you could add v 3.99 on your site
I have understood 3.99r , but several software did not update lame , even they release program updated frequently like reaper , they do still use old version , don't remember which software does use 3.99r
thanks John

@airtas
hi
have you 3.99r compiled?
and may I ask you why do you use -lowpass 20.5  , does make relevant differences in the file size ?
thanks




Re: which seems to be higher quality?

Reply #15
There are known issues with LAME v3.99 and I don't recommend you use it in 2023
https://hydrogenaud.io/index.php/topic,100256.msg840807.html#msg840807
hi @sommio
ok , but many software did not update and do keep using lame 3.99
and it's very weird because i have bought online some album in Flac and i could download in mp3 too ,and they are econded with 3.99r
maybe it's the paid software used to rip the cds
instead i don't want to buy online the mp3 version
some friend of mine do buy music via itunes in aac ,i guess they did sell in mp3 too ,but i'm not sure
i don't like to buy music from Apple , because It's not clear if they remaster the albums
thanks ,appreciate it


Re: which seems to be higher quality?

Reply #17
hi @sPeziFisH
but under construction does it mean it's almost beta and there are some bugs fixed and other not
and need an update?
thanks

Re: which seems to be higher quality?

Reply #18
From 'version.h':
Code: [Select]
# define LAME_MAJOR_VERSION      3 /* Major version number */
# define LAME_MINOR_VERSION    101 /* Minor version number */
# define LAME_TYPE_VERSION       1 /* 0:alpha 1:beta 2:release */
# define LAME_PATCH_VERSION      2 /* Patch level */
# define LAME_ALPHA_VERSION     (LAME_TYPE_VERSION==0)
# define LAME_BETA_VERSION      (LAME_TYPE_VERSION==1)
# define LAME_RELEASE_VERSION   (LAME_TYPE_VERSION==2)
So, technically, yes, it is a beta rather than a full release, but there has been no activity since 19 June 2021.

Re: which seems to be higher quality?

Reply #19
hi @john33
ok  :o
well I will use LAME 3.100 64bit and 32bit
I have ripped lots of cd with  v3.101
I guess I will keep them , they are many , next time I will use 3.100 , i guess it's more stable
and i will replace the 32bit on goldwave 5.70 because it's still 32bit ,and there is a wavelab 32bit very old maybe i could pm for an update
thanks John

Re: which seems to be higher quality?

Reply #20
There is not really any reason not to use 3.101. The biggest change as between 3.100 and 3.101 relates to switching decoder libraries rather than anything to do with encoding quality. But, as always, the choice is yours. ;)

Re: which seems to be higher quality?

Reply #21
This probably wouldn't be needed, but since half of this post talks about the LAME settings, I would warn anyone else that MediaInfo just guesses the settings, as they are not written to the mp3 file.

As such, a CBR file will not change settings related to VBR, so it will show that "-V 4" or the -m j (which is just to indicate that the file uses joint stereo, not that it required that setting). The -q 0 setting might be a real setting, although that could have changed from 3.99 to 3.99r (I don't meant the default, but what was written to the LAME tag). Just like the lowpass. that's also most probably not part of the commandline.   The difference in the spectogram might indicate that one does not use the sbf21 workaround.
The sbf21 workaround is what causes the 16Khz+ band to be less precise in order to benefit the lower bands, because in concrete cases, the precision needed for this band makes all other bands to have excessive precision and in the end have higher bitrate requirements. Given that it is not possible to go above 320kbps, that could actually mean to have less quality, and definitely causes a bitrate bloat in VBR.

Oh, and another thing that could happen on those encodes is that it was encoded as CBR with the VBR engine, making min and max VBR 320. That was not good nor bad, it was just different on a case by case basis.

Re: which seems to be higher quality?

Reply #22
While I am sure others can give you more details, I tend to keep things simple... when it comes to MP3 I would not be too fussy about getting absolute highest quality but aim more for efficient bit rate at a good enough level of sound quality that most people won't notice.

so with that in mind... that's pretty much why I just stick to V5 (130kbps) MP3's as it's efficient and sound quality is high enough as when you just sit back and enjoy your music on any random device, chances are V5(130kbps average) is sufficient for most people quite easily. for lossy formats, MP3 it's always the safest choice since it's pretty much universally supported by anything that can play lossy files.

p.s. I know some might argue storage space is cheap nowadays and has been for quite sometime now. so with that in mind, many might not care about file size efficiency (and then prefer to increase bit rate) but it's more of the principle of it if you ask me as I figure why lose efficiency/storage space if not really needed especially given the sound quality is still high enough as it would be more understandable to increase bitrate/file size if there was more of a obvious drop off in sound quality. but with V5, given the public listening tests around here a while ago, shows it's all-around pretty good.

This is why I get annoyed when people say avoid MP3 without even trying to see if the can tell V2 vs FLAC before mouthing off. My album collection over 385GB much of It being single track albums at 1.1mbit from them being Merzbow style Noise. I seem to fail to tell V2 from flac with some edge cases needing V0/V120(Helix 5.1.1). Basically not even my 512GB microSD enough after I checked so I'm not wasting memory when there no gains for me above 192kbps VBR.

MP3 with Latest encoders hold up shockingly well to newer codecs. The universal compatibility is a great for not having hope if a device uses Vorbis or AAC.