I've been comparing codecs and came down to these two setups as the best:
I used LAME 3.92 with EAC and --alt-preset standard (uberstandard)
and then used the Radium Fraunhofer at a mere 128kb

...and I can't tell the difference when I listen to them. I have $150 Sony DJ headphones. Am I crazy?

Both MP3s produced the same exact audio as far as I can tell.
The LAME encoded one is almost twice as large due to it's higher bitrate usage.
Any other encode with another codec or a lower LAME bitrate produces flawed audio.

I just recently started my little research into this, the new Uberstandard sounded like a great idea to me. But then there's this Fraunhofer build thats been around since '99?!

Here's my samples (on my slow ftp, it's always up though).
fraunhofer
lame

Not crazy, but probably some kind deaf...

I don't know a single song Radium manages to be transparent on...

dev0

I'm not sure what you mean by this statement. I'm directly comparing these two codecs...
Did you listen to my samples? And if you did can you describe the difference between the two?

I might be able to imagine why you couldn't tell the difference between the two MP3s, i.e. your hearing isn't very good (no offence, but I think 99% of people find it extremely easy to pick out a 128kbps MP3 anywhere).

But "LAME at a lower bitrate produces flawed audio"? So, to you, FhG CBR 128 and LAME --aps sound identical; but FhG CBR 128 sounds better than, say, LAME CBR 192? That's insane.

yeah, 128 is pretty easy to spot... i've stopped downloading them... 192 is "acceptable" but i've been trying to get everything in 256/320 lately... especially knowing not everyone encodes with lame (god only knows, why not)... personally i encode my stuff with ap insane, and you can spot those files from 128bit fhg files a mile away with earmuffs on...

Well I did listen to your samples and I could hear the difference easy. The Radium file does for this particular file a good job, the differences weren't really irritating. But the overall sound is less 'full'. For your type of music I guess it ain't so bad to encode in 128k/bit, Mp3 ain't that bad at music which is dominated by bass.

But for instrumental music (with higher frequencies) the result between the two will be very obvious. You should see for yourself and encode a regular popsong. You can surely hear the difference!

where did you get that info from? (r3mix?)

Maybe try doing your listening test earlier in the day...

I often can't tell the difference between vorbis q6 and the same file transcoded down to vorbis q2 when I compare using my pc's speakers (apple soundsticks) in the evening.
First thing in the morning I can hear the difference very clearly even when it's just playing as background music.

I presume listening to loudish music all day at work is temporarily damaging my hearing .

If your "I have $150 Sony DJ headphones" are Sony MDR-V700DJ's or similar V-series cans, then this is surely part of your problem since those phones aren't even close to the quality one would need to perform a real listening test.

I own these exact headphones, and I purchased them for their intended use (DJ'ing using tables) and I found them way too punchy on the bass and the highs feel "overly extended". These cans have so much bass that if you turn them up to a volume even resembling "loud" you can feel them vibrate and hear obvious distortions in the mid-range. Even for DJ use, these headphones are crap as you would expect DJ monitors to be more neutral. For the price, these headphones are total crap and have been mentioned as so in many reviews:

http://www.audioreview.com/PRD_124031_2750crx.aspx review by Eagle3195
http://www.hifichoice.co.uk/archive/perl/9...printreview.htm

However, what is entirely scary is how often these headphones are praised as being the "best cans ever" by those who are obviously audio idiots. However, these cans probably are the "best ever" if you buy all your gear from low-end high-markup-mass-market-generic-conumser-megastores, like Brandsmart, Best Buy, Sears or Circuit City.

Generally, these cans are decent for mobile mixing work where you have tables going and you have to hear the BEAT rather than the music, but for listening for critical detail in music, it's crap. However, a plus for MP3 listeners might be the ability of these cans to make Xing files sound better.

So, in all, I'm suprised that you would be able to ABX anything using them, except maybe VQF and the original CD.

Edit: Also, the Sony MDR-V900/900DJ cans aren't much better, the only major differences being the shape of the earpads ("Circum-Aural Earcup Design") an improved OFC cord, and, according to Sony's specs, "Amorphous Diamond Evaporated Diaphragm" to "stiffen the drivers", which are the exact same drivers present in the V700's. Doesn't sound to me like these can be much of an improvement since they are otherwise identical, down the the driver type and construction.

Part of the problem with this particular passage is that the artifacts I expect from too low a bitrate
sound kinda like what you started with. Lots of looped samples that have been through all kinds of
processing that may munge them up more than the mp3 compression.

I did find a section I could distinguish consistantly with winabx, so the 2 files for sure aren't the same,
but I couldn't tell you which is closer to your original CD.

edit: as to the nature of the differences.. hard to describe as this isn't the sort of thing I normally listen
to, but in one section I focussed on what sounded like a brushed cymbal rising note which sounded
smoother on the fraunhoffer file. On another part there were some transients that didn't happen at quite the
same time relative to each other. The second one I was able to identify in a blind test 12/12 times.
But when it all is very synthetic, how can you say which is more real? You might as well stick
with the lower bit rate.

frodoontop: I didn't know that, thanks!

TrNSZ: whoa, you are right on about my headphones. I had no idea... I did indeed get them from "The Good Guys" and they were the best that I could find at my local low-end high-markup-mass-market-generic-conumser-megastores
I know better now, thank you

DonP: "But when it all is very synthetic, how can you say which is more real?" Good point! Kind of a funny quote too heh

Thank you all again, I've learned a bit today...

lame is great... but i have always liked that radium fraunhofer codec... i know lame is supposedly better, but i still use the fhg codec quite a bit more than lame B)

i think that lame is better than fhg codec from 128 to 320 kbs
i only use fhg codec for my videos encoding when a need files at 96 kbs because lame can't encode at cd sampling rate at this bit rate (44,1 khz)

I'm pretty sure he's trolling... Saying that Fraunhofer @ 128 kbps CBR is indistinguishable (or sounds as good) from an APS encoded file is a wee bit too hard for me to swallow... But if he's serious about it... no offense!!

Rotten

Lame APS is definitely better than any 128 Kbps encoding, no matter the encoder. At very high bitrates, Lame is noticeabily better at pre-echo, whilst FhG is better with trumpet-like sounds, the only ones where I have heard differences with APS and APE (with exception of pre-echo killer samples).

No, I was just generalising - I think that most people can pick out the obvious artefacts that a 128kbps MP3 produces, i.e. the horrible "swishiness" and nasty smearing of "s" and "shhh" sounds on a lot of vocals.

Hearing more complex artefacts such as pre-echo takes some "training", but I think that when trying, there are very few people who would honestly declare a 128 MP3 transparent.

I feel obliged to say "obvoiusly there could be samples which just suit an encoder down to the ground at 128, and I dare say that some test tones might emerge transparent from the likes of Radium. There's also the fact that if you've bought a pair of £5 plastic speakers from a supermarket's special offer, then they'll pollute the sound so much that most artefacts could become masked; but in the majority of real-world situations, I think I'll stand by my statement

Whoa, where the hell did you get that?
I guess, after some training, most people *could* tell a diffrence, but as for now 99% of them think 128kbps mp3s are 'cd-quality'.


Not true. Once you've learned how to spot artifacts, the equipment is not really that important.

If a 128 kbit mp3 has been properly made then for me its the 16khz low pass that makes it stand out as a low quality encode. And this clearly is equipment dependent- there's no way that i could hear that low pass on my old pc speakers because they don't go that high anyhow. I used to find 128 kbit transparent on those

Spotting artifacts can be sometimes hard on low quality equipment.
But anyway, 128kbps MP3 is far too easy to distinguish from original.
I can mostly tell if something was MP3 encoded (poorly - no aps) without original.

On some songs I've encoded, I can easily tell LAME 3.90.2 @ 128kbps or 160kbps from the CD. The reason actually isn't artifacts. I find it performs well enough.

The problem is too much masking. For instance, in the song Gold from Megaherz's recent album Herzwerk II, some sort of soft cymbal sound is completely fine in one situation, but very abruptly cut in another.

Ask and I shall provide a sample (FLAC vs LAME@128?)

I'm serious; do you have allergies or somethin' like that?

I have very bad hearing, due to fluid in my ears. The fluid makes it harder to hear, and causes ringing. Tons-a people have that. I take some normal stuff to make it better, but it's hard to. I have some songs that I can tell at --alt preset extreme 13/16 a lot of the time, which on a laptop with $20 headphones, is saying something(there's static, even when nothing's playing, with everything but WAVE and VOLUME muted). Now, with all this stuff, it's hard to ABX a 96k MP3 from the original on an average song. I'm a perfectionist with sound, but if I can't hear, I can't complain.

Then, I just tried ff123's training site again after I had gotten the bad hearing, I'm just saying to myself, "um, where are the artifacts?". It doesn't matter how obvious they are if you can't hear them.

Ignorance is bliss though...

I can ignore artifacts or even normally don't notice them (if MP3 is high quality).
But on some songs they are quite annoying. (esp. ringing)

Hehe that's the reason I posted again, what I meant was that I thought "99% of people would be able to hear the difference should they try to hear it". I think the reason most people accept 128 as CD quality is that they're not aware that it's a bit rubbish, i.e. there's a massive placebo effect there - they want it to be CD quality, and if they don't know better then as far as they're concerned it should be CD quality... and so if they're not listening with an open mind (i.e. they've got the attitude "this is CD quality" as opposed to the attitude "hmmm, is this CD quality?"). So, because of all that, they don't hear any artefacts.

I've not got particularly good hearing, I can ABX --alt-preset 192 from the WAV but not --aps. I can, however, ABX an --ap 160 10/10, and I find it easy - the most obvious artefact, the high frequency smearing and swishiness, is still very audible when I listen for it.

That's why I made the comments I did in my original post, because the poster was saying that he couldn't tell the difference between --aps and FhG 128 when trying to.

Well, my equipment is so so hi-fi (SB Live 5.1, Technics AX720 5x100W und Senheiser HD600), but I can say that with all due respect to the LAME (mostly I encode with LAME), even though FHG corrupts samples, it has very 'light' sound which is much easier to listen and I very much like songs with many guitar samples encoded with FHG. If I listen to LAME MP3s for all day long, hearing a FHG encoded MP3 can be a relief I can tell you that!

Guys, we are talking about perceptual audio here and that term goes beyond all these graphs - the best perceptual encoder is encoder which can eliminate most of what can't be heard and at the same time preserve most of what can be heard. Spoken in english language, the one that can fool human ear at the given bitrate more that the others.

We do not discuss preferences here.

Anyway, either these MP3s are made with 'wrong' commandline
or you don't like the lowpass.

:x

This is the kind of thread that leaves a sort of "bitter aftertaste". Lots of myths and unverified claims here, no kind of hard evidence - that's not exactly what HA is about.


Isuldor:



Could you please verify by means of an ABX test that you are actually hearing this, and that you have not fallen victim to the placebo effect? Thanks.


frodoontop:



This assumption is based on what, this single sample (without even an uncompressed, original version for comparison)? That's ridiculous.

Also, your assumption that "MP3 isn't bad for music dominated by bass" is, in many cases, wrong - especially "pulse bass" (as found often in Techno-type music, I suppose) gives MP3 trouble even at higher bitrates.


TrNSZ:



No offence, but have you ever tried it? With headphones, even if they're extremely crappy (I was using a pair of half-broken 10 EUR Vivanco headphones for more than a year, and performed many ABX tests with them), ABXing a 128kbps MP3 still is - at least in my experience - a fairly easy task. 150$ headphones, no matter what brand, should definitely be enough. Many will agree that the equipment used is probably the least important factor when spotting artifacts (once you know what they sound like).


Timothyw:



Evidence please...


Med0:



This is absolutely ridiculous - I'm 200% sure you are imagining all that. Again, please provide ABX results to support your claims, or don't post about your listening experience at all.


I'm not very happy to see HA become increasingly polluted by these kinds of posts, and I don't think I'm alone...

CU

Dominic

Dominic has been a bit harsh, but he's right.
Any heard difference must come with an ABX result, and if possible with the listening setup.

Please respect the rules.

WTF... all the tags started showing for no reason after editing, including the "[Quote]" tags. Hell with it!

I've noticed a kind of "lite" or thin sound from FhG as well (in particular, Producer Pro) as compared to Lame. No, I'm not providing any ABX results because I don't care enough to bother, but if you imagine that MP3 doesn't color the sound of the original (at *any* bitrate) then it's you with the overactive imagination. IMHO, some differences perceived only over time may not be ABX'able (yeah, I know that's a controversial opinion).

I guess, volcano saw my number of posts and haven't even bother to understand what I am saying here and said something meaning that i do not belong here and that "I am polluting this gruop". volcano, I think that you have an attitude problem here.

i'm glad that fewtch undrestanded entirely what i wanted to say. the topic of this thread is: "LAME vs. FHG" and that's what I was writing about, wasn't I??

seems to me that the goal of the LAME developers is to make it as transparent as possible and to make it sound exactly like the original as much as possible.

FHG encoder (Opticom & AudioActive 2.x) doesn't sound exactly like original. you can clearly hear flange, "swishing" efects and I am sure that there are threads about FHG encoders in HA database.

however, I found that FHG MP3s are more pleasant to human ear than LAME ones. And yes, like fewtch nicely confirmed, you can only perceive that after quite some time and I can swear on it! I hear it!

take for example comparison between transistor and tube amps, compared at the same price - transistor amps have a lot better characteristics and their graphs ans specs from the lab tests are sometimes (far) superior to the tube amps. however, people like 'warm' sound of tube amps and they are realy buying it and realyenjoying it and they know why!

same way I compare LAME and FHG.

I read somewhere that FHG encoder was mostly tweaked by A.Capella version of song "Tom's Dinner" performed by Suzane Vega. I like vocals and guitars encoded with FHG - they are realy beautifull!

I think you just don't like the lowpass... try this (or similar) commandline:
--alt-preset 128 --lowpass 18
(If you still think, that FhG is brighter, try --lowpass 20)

Is there a middle ground between "audio mysticism" and the reductionist viewpoint that music is only a series of frequencies and amplitudes? News at 11.

BTW, I prefer Lame --alt-preset standard, and that's all I use. The "thinner" quality with FhG I don't care for much (nor the artifacts either).

Edit -- I understand the need to maintain technical clarity on HA (especially in the Tech forums... apologies to the moderators, I'll try & stay out of these threads from now on).

Some people like tube amps "warm" sound, some don't. If you are going for transparency, "warm" (=addition of soft even-harmonic distortion) tubes are a "no-no". In psychoacoustic compression, transparency is the final goal, isn't it?

Well, at last, do whatever you want to enjoy most the music.

that's my point. someone likes sound "as is" and someother likes "warm" sound. someone likes lame, someother fhg... and someothers like both of them, up to situation.

I tried to explain main difference between LAME and FHG.

Well, that is your personal subjective preference, and nobody can argue against that. But no doubt at high bitrates LAME usually does a better job than FHG, that is, adds less artifacting. So, what you want to say is that you like FHG artifacting as opposed to transparent sound?

Well said Kike, that's the bottom line... the original poster didn't say "I prefer the artefacts which FhG produces", he said "LAME produces flawed audio". That's why not many of us have taken too kindly to his post.

Even med0, whilst being a little more honest about things, has said some pretty nonsensical things:



So music with swishing added is more pleasant to the human ear than the original? Please...

This is laughable, FhG encoders don't add the swishing intentionally, if they could get rid of the above mentioned artefacts then they would. They add the swishing because they aren't very good encoders (or the bitrate that has been selected is too low to produce quality audio with the MP3 format).

Med0:



No, that's not at all what I was thinking. I didn't even mean you in particular. I was ranting about the general increase of posts claiming this and that, but then not providing any evidence supportive of these claims (another typical example is "foobar2000 sounds so much better than Winamp"). It just pisses me off, because if everybody provided ABX results in the first place, lots of stupid discussion and flamewars could be avoided. I could just choose to turn away and ignore these kinds of posts, but I'd like to contribute to keeping HA as a board generally free of misinformation and unverified, totally subjective claims which both a. harm HA's reputation and b. make finding relevant information unnecessarily hard for newbies.

You might want to think about the basic goal of psychoacoustic audio compression - the goal of a good encoder is not to sound "warm" or whatever you like to call it, but to reproduce the sound as closely to the original as possible.

Actually, as far as I'm concerned, there is no such thing as "warm" or "cold" or whatever sound in psychoacoustic audio compression - these are terms typically used by (pseudo-)audiophiles who "know what they're hearing" or musicians who "hear things differently from 'normal' people" and hence don't believe in blind testing, but don't really know a thing about audio compression. Compressed music is either transparent or not transparent, i.e. artifacted (by ringing, flanging, pre-echo, stereo collapse, or whatever). As simple as that. It doesn't take much technical knowledge to work that out, just a little experience in listening to compressed music.

You can't compare audio encoders like you would speakers or different types of amplifiers (which is where words like "warm sound" could really come into play, although I bet there's a lot of placebo going on there as well), they are two totally different things.

Sorry for the rant. Med0, please realise that this is not about you being "the bad guy" or something like that, it's more or less based on general observations.

</rant>

That's what people do anyway... what I really meant is, is there a middle ground as far as discussion is concerned -- and specifically, discussion on HA.

Bedeox's statement "we do not discuss preferences here" was what prompted me to think about this. Perhaps he meant a "tech" forum specifically, but if he meant HA in general... I know the importance of clearing up ignorance concerning audio, but if someone doesn't take a purely "engineering" viewpoint of audio, will they be lumped into the "audiophool" camp without further consideration?

Just some thoughts running through my brain.

The easiest way for me to tell the difference between 128 FhG and any higher bitrate or VBR is to use crowd appluase, like the kind heard in a "live" music performance. Then listen for what I call "garblies". It's some weird time-shift sounding noise like certain claps have been smeared in time or mangled or echoed improperly. The worst kind of "garble" is the kind that has a very "metallic sheen" on the garble. (Xing is especially notorious for that sound). FhG cannot escape the garbled applause either. You would have to be deaf to not hear it. It's the most clear and irritating artifact I've found at 128 kb/sec and even people without great frequency extention on their hearing can pick it out easily. The artifacts also show up very lightly at 160 kb/sec, but IMO on most encoders, you cannot hear the artifacts at 192 kb/sec. The "metallic" sound of Xing however never goes away at any bitrate.

mp3

Sorry, this is a blanket statement and was made out of ignorance. I apologize to anyone I may have offended. That said I've learned alot from this thread and my own little investigation into the performance of audio codecs. It obviously all comes down to personal prefrence. My own conclusion has been that my encodes should be as small in filesize as possible. All MP3 is lossy audio, right? If I care about the quality of a certain track, then I should have the original digitally mastered CD with it.

I've started using ogg q4, it uses vbr so it's even smaller than fhg@128kb (too bad there's no vbr for fhg)

Actually there is a vbr for fhg. I'd recommend Cool Edit Pro's version if you really want to go with fhg vbr, since it doesn't automatically cut everything off at 16 kHz, like MMJB does (even at 100% vbr quality).

ff123

Hmm.. my take is that there will be artifacts. These are detectable or not (transparency)
and (slightly different) annoying or not.

There are some differences that I can
reliably detect using ABX, but they don't especially bother me. I guess this runs counter
to those faults that golden ears hear which disappear under ABX.

I'm not about to dispute what you say, rather I'm just stating my peculiar situation... I have Rammstein - Live Aus Berlin and I'm pretty sure I hear warblies even at 192k CBR. Oddly enough, Nirvana - Unplugged In NY isn't nearly as bad, sounding half-decent (with egards to distortion on applause) even at 128kbps.

Why doesn't anybody make a head-to-head comparison of 128kbps FhG and LAME here?

I compressed a clip using FhG and LAME 3.93.1, using --alt-preset 128 (I presume it's not the worst setting?)

Since I'm a newbie, it was my first time using WinABX. Therefore it's not completely clear to me how one should proceed with the tests - i.e. how many times one should be allowed to compare the clips during one trial, or if it doesn't matter. Anyhow, it was much more difficult for me to distinguish between the FhG and the original than between LAME and the original. I could ABX LAME 100% of times after listening once to each clip. I could ABX FhG about 90% of times after listening probably up to ten times to each clip/a special part of it.

As I'm not a very experienced audiophile or mp3 user, I don't have any preferences, I'm just trying to find the better codec, and I have been as neutral as I could, so don't flame me

Feel free to download my clips and compare for yourself.

Original WAV (2.71MB)
Fraunhofer 128 CBR (252KB)
LAME 128 --alt-preset 128 (249KB)

Cheers,
Immo

The most important thing is to try out lots of different samples before you make up your mind which one you prefer.

I also recommend abchr: http://ff123.net/abchr/abchr.html

if you're going to pit one codec against another (vs. against the original).
The thing is that you could be unconsciously biasing yourself when you separately ABX a codec against the original, since you know which codec is being tested at that particular time.

Example: say I separately ABX wma, and then I ABX mpc, both against the original. Maybe I want the results to come out better for mpc, so I kind of let my attention wander when I'm ABX'ing the mpc (not intentionally, of course).

ff123

ff123: I downloaded your prog and will probably try it as soon as I've prepared some other samples.

I agree that one could be biased while wanting the outcome for a codec be better, but then again it's sometimes hard to be biased even intentionally if you use ABX, because if you're not sure which is the original and which not, then you simply can't be sure which button to press to be biased, so to say

As for 128kbps FhG vs. LAME, could anybody provide us with an example where he/she thinks LAME sounds nearer to the original, since I have posted an oppisite opinion?

Immo

Try this one:

http://ff123.net/samples/ma11short.flac

ff123: I tried out your program by testing the sample which you gave me, compressing it with FhG (in CoolEdit 2.0) and LAME 3.93.1 --alt-preset 128.

I must say that here the (what seems to me) advantage of FhG has shrunk. In case of FhG I was able to tell the difference from the original 20/20 times. Unfortunately, in case of LAME it was 20/20 as well.

It seems that in this case LAME did a better job preserving the higher frequencies of the hi-hat, and if it wasn't for the distortions of this same hi-hat, it would be much harder to distinguish it from the original.

Similar "whistling" sounds could be observed in the FhG, only that they seemed to be of lower frequency and less juggled, but more permanent.

I don't know which codec I would choose to compress this track. Probably I would tend to FhG, because despite producing some whistling, it seems not to distort the audio as much as LAME (which, anyway, isn't much). At least I could try to imagine that the instrument actually sounds like that
Surely I would use a higher bitrate, if size wouldn't be too important. And last, but not least, I would hardly ever need to encode this music

Cheers,
Immo