Originally Posted by krabapple
Suuure your did. How long ago was this? What version of what codec? You didn't train the listeners? You used what samples?
What does have anything to do with the point you quoted from me? I said that an untrained listener matched the ability of trained listeners with *no training*. What most people could not hear, they could. You are telling me with time, these people disappear? And what on earth does the version of the codec have anything to do with it? You are making me think I am arguing with dionmaniac all of a sudden
How about this? Take 100 people off the street right now at random, have them take a 16-trial DBT of a LAME 320 mp3 with no training, using whatever music THEY like. I predict you'll find not a single one that could 'pass'. Hell, do it at 192 VBR, I predict the same result.
How about this: it is *your* claim. You go and have that test done and then show it to us. You are the one claiming that LAME does better than its own documentation. My proof point is already made that they don't make the claims you are making. If you disagree, then go and run your objective test. You do believe in them, right? You wouldn't make up the position you have right now in absence of that, right? Right.
I also want to not the bold section where you say with absolute certainly that no matter who I pick from the street, they can’t tell the difference at 320kbps or 192 kbps.
Instead of waiting to run your test, there is a version already done with VBR encoding and all in the maximum PC article I mentioned earlier. People indeed brought their own music and in many cases identified each version: http://www.maximumpc.com/article/do_..._rates_pay_off
Some quotes since you can’t be bothered to read about such tests:
“Each lucky participant was asked to bring in a CD with a track that he or she has listened to for years and knows so intimately that a single missing hi-hat tap would stand out like a sudden blast from a tuba. We ripped each track using iTunes at three quality levels: 160Kb/s, 320Kb/s, and uncompressed WAV. The compressed files were ripped using variable bit rate (VBR) encoding, meaning that a 160Kb/s VBR track allows the bit rate to rise and fall depending on the complexity of the music while maintaining the selected bit rate as the minimum bit rate for the track.”
“In the end, the Hipster confessed that she resorted to guessing in most cases, relying on a hunch based on instinct rather than any identifiable portion of the tracks. Her instincts proved to be accurate when listening to her own track; she correctly identified the bit rates of all three versions…
“Despite this last-minute waffling, she was very confident about identifying the quality levels of her own track, as well she should have been: She correctly identified all three. With songs she was not familiar with, or with tracks that made her burst out laughing (see the last sentence of the Hipster’s report), she identified only two of the 160Kb/s tracks and a single 320Kb/s track.”
“That’s pretty much the story told by his results—he was able to identify the uncompressed version of his own track after considerable A/B comparisons with the compressed versions, but mixed up the 320Kb/s and 160Kb/s versions. He was also able to correctly identify the quality levels of another participant’s track, and the 320Kb/s version of yet another’s track, but had to settle on a tie with the Curmudgeon, at five out of 12….”
So anecdotally in this test, the odds of people detecting the difference is well above zero that you predicted to be.
You think self-selected 'audiophiles' would do better? Fine. Take 100 of those at random. Again, no training. Let them pick their music samples. Wanna bet it'll be any different outcome?
There is an audiophile in the above test (last guy). So you have already lost that bet
And btw, calling yourself an audiophile is no guarantee that you will be 'at the top of the pyramid' of audio discernment. DBTs have shown that, too.
You are confused about that statement. My use of the term Pyramid was to indicate that we are not designing things for the entire population. If we were, then the bar could be very low and even 128kbps would do. I imagine you are not advocating 128kbps as the level of transparency. Yes? If so you yourself are starting to eliminate large portion of the population. Therefore, design rules must be more strict.
The reason they're used is because they illluminate the limits of the technology. Not because they're common.
Solo guitar is not common? A transient is not common? Solo vocals are not common? Do you really think MPEG, ITU, Dolby, DTS, FHG, Sony, etc use these reference clips even though they are not representative of real music?
Why would they evaluate the technology this way and you not? What do you know that they don't? Honestly, how far do you want to dumb down the performance of this wonderful hobby Krab?
Looks like you are already walking down the line some more with that 160kbps. Seems like if CDs went away and all we could download are 256kbps MP3/AAC, you think the music lover needs are met just fine. If I had any doubt of not fitting in with the few of view, this cements it. You should wear the badge of “how low can we go” instead of objectivists. You have no aspirations for quality left.
Again, you want to go on Hydrogenaudio, where LAME codec developers congregate, and spout this Amirian bluster? Be my guest, it will be fun to watch.
You crack me up. You ask me to come outside for a fist fight. When I show up, you say, "wait a minute, why don't you go to the next block and fight the guys there!"
You are the one making these claims. Apparently you don't know how to back them sufficiently on your own and want me to go and fight them in their forum. You didn't just say that, did you?
If you can't own the information, you have no business whatsoever making these claims so strongly and taking on the people in the industry who do this for a living. What you are doing is the most obvious admission of insecurity regarding the topic at hand that I have seen. Remind me of AJ and Arny saying, "let me go get JJ to come and fight you!" Fight your own battles or stop asking people to meet you outside for heaven's sake.
All of this aside, if you think I don't know how to hold my own in my field of expertise, then I don't know what to say. God could show up and you would say you still are an atheist. By that logic, you will never back down from anything. You would be outraged if someone argued with Floyd but boy, be on the other side and no one can know more than your sum total of reading stuff on hydrogen audio?
Do you know the concept of really learning something instead of just reading about it? Do you think for a moment you would know what an instrument sounds without ever hearing it? How on earth do you guys get so emboldened this way? With zero industry experience or broad exposure to testing of such technologies you can be so outspoken? Unbelievable.
I mean really. It is one thing to argue about DACs and amps being the same. Now you lower the bar to 320/160 kbps is good enough and if someone disagrees you are going to send them to hydrogenadio to be taught otherwise? Is this a movie I am watching or is it for real? Educated folks taking such illogical positions?
I don't care what you listen to. (Though if your listening habits are anything like normal, I think you'd fail the 320 LAME mp3 challenge on most of not all of it.)
You have to care if you want to tell me what gear should satisfy my need. Since you can't control my music, the only variable then is the equipment. You don't get a free pass on that. It sure would be nice to tell people to only live where they can have a straight road to work and no curves. That way, handling in a car would not be important. But we can't do that.
As to me hearing the difference, there is that weasel word again. "Most if not all?" If I invert that, does it not indicate that some of the times I can hear a difference? If I can, then you have already lost your case because I am not going to use a suboptimal system that does that when I have a choice of one that doesn’t underperform, period.
I care whether a difference can be expected to matter in a typical listening situation.
You are a full encyclopedia of weasel words. Now it is “typical” listening situations. And that is a precise technical term or a made up one that can mean anything to dig you out of a hole? And “expected to matter?” I am sure that is defined in some audio dictionary.
In your earlier blind test challenge, you didn’t use any of these terms? But toward the end you get worried and we are back to wishy-washing language. Make up your mind and have conviction for heaven’s sake.
So yes, 'random' music -- meaning, you don't get to cherry-pick a 'codec killer' to prove your point.
So I flip a coin and if I happen to get all classical music that is harmonic and doesn’t stress the codec, you think that provides correct conclusion? Why don’t MPEG and ITU do that if that is a valid methodology? They don’t because they don’t like to stick their head in the sand and deal with luck of the draw. They want to *know* of the technology works. We don’t’ drive a sub with a random selection of movies hoping that they are representative of good low frequency extension. You can test 1000 easy cases and you will learn zero about the system capabilities.
Of course, if your point of view is the dumbing down of the standard of audio reproduction, by all means, keep going lower and lower. Let’s get rid of anyone who can hear a difference first. Then let’s randomize the samples and see if that does the job of getting rid of anything that might show a difference. If that doesn’t, let’s just confuse the whole thing by speaking of both sides of our mouths. Did I miss anything?
LIKELIHOODS, Amir. That's what objectivists are talking about, whether they use the word or not. Not worst-case scenarios.
Let’s not confuse you with an objectivist in this topic. An objective person bases his opinions on stated data which you have not provided at all. You are making layman arguments that sound right to you, lacking any data whatsoever to determine said likelihoods. I told you that we ran trials that 90% of the people thought randomly picked content at just 64kbps, 6X lower than the data rate you picked, were transparent to a CD. This was a professionally run test costing $20,000+. You want to say that only 10% of the people can hear the difference between the CD and 64kbps, good luck my friend. Good luck.
Watch MPEG-2 broadcast of sports. See those blocking artifacts during high motion sequences? To a layman, they are a distortion blamed on the display (“it is pixelated”). That is wrong of course. The artifacts show up because we run out of bandwidth to allocate enough bits to the stream and excess quantization rears its ugly head. What are the odds of high motion sports during a week of TV viewing? Very little. Yet, when it occurs, it occurs. Solution was to have better codecs and more bandwidth as needed. No amount of waving your hands about “likelihood” makes that problem on broadcast TV go away.
LOL. And here's what it also says on that same page, about -V (which isn't even 320kps):
"High quality: HiFi, home or quiet listening
-V0 (~245 kbps), -V1 (~225 kbps), -V2 (~190 kbps) or -V3 (~175 kbps) are recommended. These settings will normally produce transparent encoding (transparent = most people can't distinguish the MP3 from the original in an ABX blind test). Audible differences between these presets exist, but are rare."
Are you making my case or yours?
See the bolded section? They are telling that even in blind ABX tests some people can hear the difference (inverse of most cannot). Isn’t that the opposite of your claim? Keep in mind that I already said that we don’t care about the general public. So if we factor them out, the percentage goes higher than the intended audience in their statement. Don’t keep confusing that.
No one said anything about 'magic', Amir. Your clown shoes are showing again. The LAME codec has improved over the last decade and half..
Excuse me. You claimed a lot was changed since I left Microsoft. I left in 2007. So seems like that was another example of hip shooting since you can’t show how the world of compression has changed since. If you want to now claim I didn’t know about LAME while managing the signal processing/compression team at Microsoft, let’s see the proof point. Until then, you are playing Calvinball.
in significant part by addressing the artifacts revealed in different generations of 'killer samples'. Some of which aren't 'killer' any more as a result.
Stated as a layman who can’t really follow the conversations they were having. Maybe this introduction on the LAME web site is helpful for you: http://lame.sourceforge.net/
“LAME development started around mid-1998. Mike Cheng started it as a patch against the 8hz-MP3 encoder sources. After some quality concerns raised by others, he decided to start from scratch based on the dist10 sources. His goal was only to speed up the dist10 sources, and leave its quality untouched. That branch (a patch against the reference sources) became Lame 2.0, and only on Lame 3.81 did we replaced of all dist10 code, making LAME no more only a patch.”
Fact is that the FHG encoder was the best and the one to beat when these guys started (“quality concerns”). Many of these encoder attempts started as a) speeding up the encode as the FHG encoder was so slow and b) to get rid of the prefiltering that FHG used to reduce distortion (remember, high frequency transients are a killer on the codec).
Making things faster often entails shortcuts which reduce quality (anyone remember the Xing encoder?). Ditto for enlarging the bandwidth. Tricks used to get around these are to relax rate control (and hence the strong push to use VBR). And to use processor optimization instead of shortcuts in the algorithm. So yes, LAME has improved over the years but that is because it didn’t sound so good when it started. Even today, it loses to higher efficiency encoder per their own documentation above: “Today, LAME is considered the best MP3 encoder at mid-high bitrates
and at VBR…” You see them excluding lower bit rates? While I appreciate the love of enthusiasts for it, that failing is an unfortunate on their behalf (admittedly, it has no bearing on the current discussion other than to explain the technology).
Anyway, someone said earlier that we get nothing done in these arguments. I must say, I had low expectations for this thread just the same. But I have gotten far more out of it than I could imagine. Folks have now lowered the bar below DACs and Amps making a difference to 320kbps or even 160kbps MP3s being a degradation. I am sure in the next chapter we will be talking about how maybe 8-track tape was also good enough!