Anyone else use 320 kbps mp3's?, n00b here. Options

[JAKE196]

Well I've done some more tests and I was surprised, high-ish bitrate mp3's are harder to tell from .wav files than I thought. Thats a very good thing. I was wrong again though. doh.

Them ABX tests are hard.

I don't have any important conclusive tests worth posting, sorry to disappoint. So I didn't document anything.

There was this one test were I got the first 10 trials right right when comparing a really high VBR file and a 320 kbps file. But after that I failed and passed 50/50 so in the end the results were no good. Maybe I just have no patience. I was comparing lossy to lossy anyway so I guess it wouldn't count even if I did get good results.

One day I'll probably switch to lower bitrate VBR mp3's after I do some more extensive tests, but for now I'm kind of sick of tests. I'm still half convinced that if you are looking for the best quality possible and you have no access to lossless formats, 320 kbps might be reasonable overkill.

I can't be bothered to do any more tests right now because summer break is almost over and I have books to read, the Olympics to watch and beaches to go to and junk. I don't have access to this computer every day either.

If and when I do some more tests, if I get any interesting results I'll be sure to post them on here and document everything. I won't claim anything, I'll just say here is some results from an ABX test.

This post has been edited by JAKE196: Aug 8 2008, 21:26

[kornchild2002]

One day I'll probably switch to lower bitrate VBR mp3's after I do some more extensive tests, but for now I'm kind of sick of tests. I'm still half convinced that if you are looking for the best quality possible and you have no access to lossless formats, 320 kbps might be reasonable overkill

Thank you, thank you, thank you. That is all I wanted to hear from you. I wanted to hear that you would test lower bitrate files. You don't have to use lower bitrate files but I think that you should at least test them given the quality of the Lame mp3 encoder. I am sorry but that one comment of "You guys aren't going to convince me to use a lot lower bitrates..." just rubbed me the wrong way. We weren't trying to convince you to use lower bitrate files. We were only trying to get you to look at the possibility of using them. In other words, we wanted you to test the lower bitrate files.

You might want to re-think that "I'm still half convinced" statement. It is true that 320kbps is the highest bitrate that the Lame mp3 encoder can offer without going into the free format range. Technically speaking, the mp3 format can encode at bitrates of up to 600kbps but those files won't be compatible with any portable players (at least the ones that I know of) and software playback will be slim to none as well. Having the highest bitrate does not always mean the best quality. It is often thought that -V 0 can produce the same perceptual results as 320kbps. That means that one can have the same quality as 320kbps with file sizes that are much lower. Of course your mileage will vary. Still, I think you should get that "320kbps is the best" mentality out of your head as it is only going to cloud your judgment. When it comes to lossy encoding, it is always better to have an open mind and not be influenced by any of these age old arguments.

[digital_music]

Hi, am interested in the general discussion you are having on high bitrate mp3s.
I have discovered through experimentation since 2002/3 that a higher bitrate does improve the sound quality, using LAME (all versions) and outboard Hi-fi amp/headphones.
The actual audio is 'closer' to the original CD.
In fact, taking advantage of the '--athlower' setting you can increase the amount of audio you can hear, by moving the absolute threshold of hearing below that of which has been set by the algorithms.
By using V0 and Q1 (and for an ultimate quality - Q0) the file quality is very good, with no audible artifacts and using Metal tracks with complex passages and other - pop dance punk newage synth etc.
Being middle age, I cannot hear above 16khz myself and give an HF rolloff at 16k5Hz (it may affect cymbals and other transients)
I do find it important, however, to actually enjoy the music once encoded!
So the settings I use are :- (LAME 3.97)
-b 128 -m j -V 0 -B 320 -F --lowpass 18.8 --lowpass-width 3 -q 0 --athlower 12
and have never had any problems.

This post has been edited by digital_music: Aug 9 2008, 00:37

[Slipstreem]

Another person who thinks he knows better than the devs and the guys who've helped to road-test LAME over many years! I'm getting very close to giving up now.

Cheers, Slipstreem.

This post has been edited by Slipstreem: Aug 9 2008, 03:00

[lesswire]

The actual audio is 'closer' to the original CD.

By closer do you mean you can absolutely distinguish a high bitrate mp3 from the lossless source? Have you done any ABX tests?

So the settings I use are :- (LAME 3.97)
-b 128 -m j -V 0 -B 320 -F --lowpass 18.8 --lowpass-width 3 -q 0 --athlower 12

Instead of this complicated command line why don't you just use the V settings? I'm not by any means a Lame expert but the experts here will tell you to simply stick to the V settings instead of this crazy line you have come up with.

Cheers

This post has been edited by lesswire: Aug 9 2008, 04:09

[carpman]

-b 128 -m j -V 0 -B 320 -F --lowpass 18.8 --lowpass-width 3 -q 0 --athlower 12

Yeah, but wow! Doesn't it look impressive!

C.

"Often, mud gives the illusion of depth"

[slimserver]

From following this thread I went ahead and did an ABX test myself with 4 tracks from my favorite artists. I started with a FLAC and a transcoded LAME 3.98 V5. Using my cherry picked tracks, I thought I would be able to easily pick the V5 but I could not. I actually got pretty upset when I could not tell the difference between A and B. Now I wonder if LAME V5 is that impressive or is my hearing nowhere near as good as I thought it was. I used to use V2 for my portables now I have switched to V4.
Thanks to those who've contributed to this thread. My opinion and use of lower bit rates has positively changed.

[Synthetic Soul]

Hi, am interested in the general discussion you are having on high bitrate mp3s.

Yet you appear to have ignored all the warnings that were made to JAKE196.

I have discovered through experimentation since 2002/3 that a higher bitrate does improve the sound quality, using LAME (all versions) and outboard Hi-fi amp/headphones.
The actual audio is 'closer' to the original CD.

There is obviously some logic in saying that higher bitrate means better quality; however this is not necessarily the case, and certainly becomes untrue the higher the bitrate.

In fact, taking advantage of the '--athlower' setting you can increase the amount of audio you can hear, by moving the absolute threshold of hearing below that of which has been set by the algorithms.
By using V0 and Q1 (and for an ultimate quality - Q0) the file quality is very good, with no audible artifacts and using Metal tracks with complex passages and other - pop dance punk newage synth etc.
Being middle age, I cannot hear above 16khz myself and give an HF rolloff at 16k5Hz (it may affect cymbals and other transients)
I do find it important, however, to actually enjoy the music once encoded!
So the settings I use are :- (LAME 3.97)
-b 128 -m j -V 0 -B 320 -F --lowpass 18.8 --lowpass-width 3 -q 0 --athlower 12
and have never had any problems.

We've already had a report of trolling, and I can't help but think it may be right. Posting a command line like this in such a thread is like sneaking up behind a calmed bull and flicking his testes.

To conclude: Where is the proof to your claims? ABX?

Thanks to those who've contributed to this thread. My opinion and use of lower bit rates has positively changed.

At least some use has come from this thread. Have you purchased a mu'umu'u yet?

[mr budzone]

From following this thread I went ahead and did an ABX test myself with 4 tracks from my favorite artists. I started with a FLAC and a transcoded LAME 3.98 V5. Using my cherry picked tracks, I thought I would be able to easily pick the V5 but I could not. I actually got pretty upset when I could not tell the difference between A and B. Now I wonder if LAME V5 is that impressive or is my hearing nowhere near as good as I thought it was. I used to use V2 for my portables now I have switched to V4.
Thanks to those who've contributed to this thread. My opinion and use of lower bit rates has positively changed.

Welcome to the club man! =)
When I tried out the latest 64kbps test on HA I got kinda blown away since the high anchor (Im kinda certain it was itunes aac 96kbps) could not be distinguished for me on any track from the original. I mostly use V 5 for my portable nowadays and V 4 for the stereo @ home.

digital_music is either a troll or unable to read the previous text in this thread.

[DonP]

I think i've just about read enough of this messy cymbals, HF lowpass, i am young stuff on the net over the years. Even if its true I think its nothing to brag about . If normal masking which works on most people doesn't for you its NOT a good sign of ear health.

Being able to hear high frequencies is a different issue than masking.

[xSerpentx]

I always stay away from 320, you never know what it's encoded with.

[Synthetic Soul]

I always stay away from 320, you never know what it's encoded with.

Unless you encode it yourself.

If you are concerned about legal downloads from stores I am sure that you could easily find out.

[bwat47]

320 kb/s mp3s are a total waste of space.

[kwanbis]

This post needs more votes, so it is more representative

[B7k]

I have only used Mpeg-1 layer2 (Mp2) @320kbps or AC3 @ 384kbps for my collection of music as for mp3 iv'e used lame @ 192-160kbps vbr it is not bad in my own opinion

This post has been edited by B7k: Apr 14 2009, 07:22

[Rei Murasame]

I have discovered through experimentation since 2002/3 that a higher bitrate does improve the sound quality, using LAME (all versions) and outboard Hi-fi amp/headphones.
The actual audio is 'closer' to the original CD.
[...]

Being middle age, I cannot hear above 16khz myself and give an HF rolloff at 16k5Hz (it may affect cymbals and other transients)
[...]

So the settings I use are :- (LAME 3.97)
-B 320 -F --lowpass 18.8 --lowpass-width 3 -q 0 --athlower 12

Sorry for reviving this thread, but I just couldn't resist, in light of that muddled-looking commandline. Once upon a time I felt the need to start submitting to ABX tests to see if I could actually find any difference between lossless source and 320kbps LAME.

My hearing is pretty decent, and my listening as well, but over the years I've failed to pass an ABX with LAME 320kbps about 100% of the time (which is why I personally never encode at 320). I'm now 23yrs old. I'd be astonished if you were able to tell the difference between the files you're generating with that 'custom' commandline, versus just using just -b 230 by itself - and that's before we even get into asking if 320kbps is overkill or not.


An ABX I took last month:

CODE

foo_abx 1.3.4 report
foobar2000 v0.9.6.6
2009/06/12 08:57:43

File A: J:\[LonE]_Various_Artists_-_CODE_GEASS_Lelouch_of_the_Rebellion_O.S.T._[w_scans]_(FLAC)\[LonE]_Various_Artists_-_CODE_GEASS_Lelouch_of_the_Rebellion_O.S.T._[w_scans]_(FLAC)\02 Stories.flac
File B: J:\[Nipponsei]_Code_Geass_Hangyaku_no_Lelouch_Original_Soundtrack\02 - Stories.mp3

08:57:43 : Test started.
09:01:59 : 00/01 100.0%
09:02:39 : 00/02 100.0%
09:04:27 : 01/03 87.5%
09:06:57 : 01/04 93.8%
09:09:49 : 02/05 81.3%
09:11:55 : 02/06 89.1%
09:14:23 : 02/07 93.8%
09:15:04 : 03/08 85.5%
09:15:43 : 04/09 74.6%
09:16:27 : 05/10 62.3%
09:17:22 : 06/11 50.0%
09:18:18 : 06/12 61.3%
09:21:59 : 06/13 70.9%
09:22:19 : 07/14 60.5%
09:23:00 : 07/15 69.6%
09:23:31 : 08/16 59.8%
09:24:30 : 09/17 50.0%
13:37:01 : 09/18 59.3%
13:37:57 : 10/19 50.0%
13:39:59 : 11/20 41.2%
13:41:35 : 11/21 50.0%
13:42:36 : 11/22 58.4%
13:44:19 : 12/23 50.0%
13:45:34 : 12/24 58.1%
13:48:39 : 13/25 50.0%
13:51:42 : 14/26 42.3%
13:53:30 : 14/27 50.0%
13:54:19 : 15/28 42.5%
13:55:49 : 15/29 50.0%
13:57:43 : 16/30 42.8%
13:58:29 : 16/31 50.0%
13:59:18 : 16/32 57.0%
14:00:10 : 17/33 50.0%
14:00:50 : 17/34 56.8%
14:02:24 : 17/35 63.2%
14:05:34 : 17/36 69.1%
14:08:54 : 18/37 62.9%
14:09:57 : 19/38 56.4%
14:11:58 : 20/39 50.0%
14:13:09 : 21/40 43.7%
14:14:05 : 22/41 37.8%
14:15:17 : 22/42 43.9%
14:16:24 : 23/43 38.0%
14:17:20 : 23/44 44.0%
14:21:05 : 23/45 50.0%
14:23:18 : 24/46 44.1%
14:24:00 : 24/47 50.0%
14:26:05 : 25/48 44.3%
14:26:26 : 26/49 38.8%
14:26:49 : 26/50 44.4%
14:27:41 : 26/51 50.0%
14:29:36 : Test finished.

----------
Total: 26/51 (50.0%)

A 7 hour ABX test of FLAC versus LAME CBR320, which as you can see, I failed spectacularly. You claimed that a high bitrate like that is 'closer' to the original, but I think that it's actually transparent.

Can you show me any ABX test you've taken where you've actually detected a difference?

This post has been edited by Rei Murasame: Jul 3 2009, 21:01

[rpp3po]

Can you show me any ABX test you've taken where you've actually detected a difference?

The Castanets sample should bring that 'custom' command line to its knees without too much effort...

[/mnt]

Can you show me any ABX test you've taken where you've actually detected a difference?

Here's a list of samples that i've successfully ABXed at 320kbps CBR:

PTP - Show Me Your Spine [sample] (dance club track from Robocop)
Welcome To Drexciya [sample]
Skinny Puppy - Human Disease [sample]
eig sample (no ABX log, but almost anyone can ABX this at 320)

This post has been edited by /mnt: Jul 3 2009, 23:00

[Eli]

I use lossless (FLAC) for archiving and ABR ~128 mp3 for my lossy files. This is more then good enough for me in most situations in terms of quality, the files are small, and mp3 is compatible with all players. I use ABR because I had some problems with ipods and VBR LAME files.

[gorgekko]

I use VBR. Back in ye olden dayes when MP3s were new and annual deficits measured in hundreds of billions were considered enormous I used 320 but ABXing later proved that years of Slayer turned my ears into concrete.

I haven't ABXed lately but I wouldn't be surprised if these days I didn't have problems differentiating between 160 and 192.

I wonder whatever happened with Jake and his ABXing.

[Rei Murasame]

Can you show me any ABX test you've taken where you've actually detected a difference?

Here's a list of samples that i've successfully ABXed at 320kbps CBR:

PTP - Show Me Your Spine [sample] (dance club track from Robocop)
Welcome To Drexciya [sample]
Skinny Puppy - Human Disease [sample]
eig sample (no ABX log, but almost anyone can ABX this at 320)

Oh my goodness. Well, let me just say, you've inadvertently put me through quite a lot there! I chose one of those samples at random (the "Show Me Your Spine" sample), and tried to ABX it. I used LAME 3.98 @ 320kbps and went for 50 turns just to be sure.

CODE

foo_abx 1.3.4 report
foobar2000 v0.9.6.8
2009/07/04 01:28:24

File A: D:\Show_Me_Your_Spine__Sample_.flac
File B: D:\_ Show_Me_Your_Spine__Sample_.mp3

01:28:24 : Test started.
01:30:39 : 01/01 50.0%
01:35:43 : 02/02 25.0%
01:37:57 : 02/03 50.0%
01:41:25 : 03/04 31.3%
01:42:41 : 04/05 18.8%
01:44:22 : 05/06 10.9%
01:47:03 : 06/07 6.3%
01:54:24 : 07/08 3.5%
01:54:58 : 08/09 2.0%
01:56:19 : 09/10 1.1%
01:58:49 : 09/11 3.3%
01:59:16 : 10/12 1.9%
02:03:22 : 11/13 1.1%
02:04:57 : 12/14 0.6%
02:05:28 : 12/15 1.8%
02:07:20 : 13/16 1.1%
02:11:03 : 13/17 2.5%
02:11:53 : 14/18 1.5%
02:12:30 : 15/19 1.0%
02:12:56 : 16/20 0.6%
02:13:26 : 17/21 0.4%
02:13:49 : 17/22 0.8%
02:14:20 : 18/23 0.5%
02:15:16 : 19/24 0.3%
02:16:11 : 19/25 0.7%
02:16:31 : 20/26 0.5%
02:18:32 : 21/27 0.3%
02:18:57 : 22/28 0.2%
02:19:56 : 22/29 0.4%
02:20:44 : 22/30 0.8%
02:22:41 : 23/31 0.5%
02:24:36 : 23/32 1.0%
02:26:08 : 24/33 0.7%
02:26:40 : 25/34 0.5%
02:27:47 : 25/35 0.8%
02:28:50 : 25/36 1.4%
02:29:38 : 25/37 2.4%
02:35:24 : 26/38 1.7%
02:35:47 : 27/39 1.2%
02:37:36 : 28/40 0.8%
02:39:01 : 29/41 0.6%
02:39:46 : 30/42 0.4%
02:40:25 : 31/43 0.3%
02:43:58 : 31/44 0.5%
02:44:14 : 31/45 0.8%
02:44:46 : 32/46 0.6%
02:52:42 : 33/47 0.4%
02:54:04 : 34/48 0.3%
02:55:21 : 35/49 0.2%
02:55:53 : 36/50 0.1%
02:56:14 : Test finished.

----------
Total: 36/50 (0.1%)

It's one of those "what? how is this happening?", kind of moments. This evidently is not transparent to me either. I'll have to retract my previous claim after all.

As an aside, I have to say, /mnt, that the speed at which you arrive at the finish in your tests is seriously, really impressive, it's like you're able to latch onto an anomaly immediately and just find it again and again in rapid succession.

[/mnt]

It's one of those "what? how is this happening?", kind of moments. This evidently is not transparent to me either. I'll have to retract my previous claim after all.

As an aside, I have to say, /mnt, that the speed at which you arrive at the finish in your tests is seriously, really impressive, it's like you're able to latch onto an anomaly immediately and just find it again and again in rapid succession.

The Show Me Your Spine sample has a very oblivious precho artifact on first few secends with the third synth drum. Also theres precho all over the place on this track at 320.

Most of those samples do have oblivious atifacts at a certain time i.e 0:05 - 0:08, which i can detect and hear and then ABX to see if there is a real problem. Sadly after passing a lot of ABX tests you can start to hear or just sense artifacts all over the place, at lower bitrates or your prefered bitrate i.e V 2.

Anyway i managed to ABX yet another track from my music collection at 320 CBR:

CODE

foo_abx 1.3.4 report
foobar2000 v0.9.6.8
2009/07/04 11:58:34

File A: E:\Music\Albums\Skinny Puppy - Too Dark Park\01. Convulsion.flac
File B: C:\Temp\Convulsion b 320.mp3

11:58:34 : Test started.
11:59:24 : 01/01 50.0%
11:59:35 : 02/02 25.0%
11:59:44 : 03/03 12.5%
11:59:55 : 04/04 6.3%
12:00:09 : 05/05 3.1%
12:00:31 : 06/06 1.6%
12:00:54 : 07/07 0.8%
12:01:11 : 08/08 0.4%
12:01:24 : 09/09 0.2%
12:01:33 : 10/10 0.1%
12:01:42 : 10/11 0.6%
12:01:47 : 10/12 1.9%
12:01:56 : 11/13 1.1%
12:02:05 : 12/14 0.6%
12:02:16 : 13/15 0.4%
12:02:25 : 14/16 0.2%
12:02:35 : 15/17 0.1%
12:02:53 : 16/18 0.1%
12:03:01 : 17/19 0.0%
12:03:20 : 18/20 0.0%
12:03:24 : Test finished.

----------
Total: 18/20 (0.0%)

Precho on the distortion clipping at 0:04, making it sound louder.

Convulsion sample

This post has been edited by /mnt: Jul 4 2009, 12:34

[shadowking]

320k for transform codecs is too high because wavpack lossy can be transparent with even lower bitrate - Why does a super tuned psymodel need 320k?

I say no more than 250 k for the quality headroom / transcoding stuff . AAC transcodes great @ 220k and even 190k. It typicaly won't need 320k to fight preecho either.

Sometime in the near future AAC 200k is the new 320k mp3. A universaly compatible high quality solution for everything.

This post has been edited by shadowking: Jul 4 2009, 16:48

[/mnt]

It typicaly won't need 320k to fight preecho either.

Yeah 320kbps is still useless against major precho artifacts, due to the limitions on the Mp3 codec spec. Which is why i don't bother with higher bitrates and stick to a bitrate or setting that sounds not too bad or transparent most of the time e.g -V 2 or 192 CBR (with Joint Stereo).

Sometime in the near future AAC 200k is the new 320k mp3. A universaly compatible high quality solution for everything.

IMO Nero AAC at q 0.55 (~200kbps VBR), seems to almost come close or better with LAME V0 on some personal tests i did. Also all those killer samples linked above, that i've ABXed at 320 Mp3 sound alot better on Nero at 0.50 - 0.50, but still ABXable though.

This post has been edited by /mnt: Jul 4 2009, 19:18

[Rei Murasame]

Precho on the distortion clipping at 0:04, making it sound louder.

Convulsion sample

Well, this one I seemed to be totally unable to distinguish even though you told me it was there. The clipping sound seemed to be about the same loudness to me on the encode, so I had a lot of difficulty telling it apart.

CODE

foo_abx 1.3.4 report
foobar2000 v0.9.6.8
2009/07/04 18:58:57

File A: D:\Convulsion__Edit_.flac
File B: D:\_ Convulsion__Edit_.mp3

18:58:57 : Test started.
19:02:19 : 01/01 50.0%
19:02:49 : 02/02 25.0%
19:03:18 : 03/03 12.5%
19:04:14 : 03/04 31.3%
19:04:37 : 03/05 50.0%
19:05:04 : 03/06 65.6%
19:05:35 : 04/07 50.0%
19:05:56 : 05/08 36.3%
19:06:14 : 05/09 50.0%
19:06:28 : 06/10 37.7%
19:06:57 : 07/11 27.4%
19:07:15 : 08/12 19.4%
19:07:38 : 08/13 29.1%
19:07:55 : 08/14 39.5%
19:08:49 : 08/15 50.0%
19:10:11 : 09/16 40.2%
19:10:43 : 10/17 31.5%
19:11:07 : 10/18 40.7%
19:12:21 : 10/19 50.0%
19:12:58 : 10/20 58.8%
19:13:24 : 10/21 66.8%
19:13:35 : 10/22 73.8%
19:13:51 : 11/23 66.1%
19:14:21 : 12/24 58.1%
19:15:09 : 12/25 65.5%
19:15:52 : 13/26 57.7%
19:16:25 : 14/27 50.0%
19:17:40 : 15/28 42.5%
19:18:04 : 16/29 35.6%
19:18:30 : 16/30 42.8%
19:19:04 : 17/31 36.0%
19:19:29 : 18/32 29.8%
19:19:52 : 19/33 24.3%
19:20:47 : 19/34 30.4%
19:21:10 : 20/35 25.0%
19:21:37 : 21/36 20.3%
19:22:28 : 22/37 16.2%
19:23:08 : 22/38 20.9%
19:23:33 : 23/39 16.8%
19:23:57 : 23/40 21.5%
19:24:22 : 24/41 17.4%
19:24:46 : 25/42 14.0%
19:26:09 : 25/43 18.0%
19:26:32 : 26/44 14.6%
19:27:59 : 26/45 18.6%
19:28:53 : 26/46 23.1%
19:29:17 : 26/47 28.0%
19:29:53 : 26/48 33.3%
19:30:23 : 26/49 38.8%
19:30:50 : 26/50 44.4%
19:30:57 : Test finished.

----------
Total: 26/50 (44.4%)

Any suggestions on how it could be so elusive? I'm assuming it's just a lack of experience on my part?