[HiFiRE]

There's been a lot of discussion lately about whether or not people can ABX well encoded mp3s or not. There are certainly many people out there that can AB, and this test is simply a chance for you to show off your hearing abilities and put to rest the doubters/haters. I will use Lame 3.97 -V2 as the reference encoder and setting.

There's a great program called Java ABC/HR that will allow us to conduct this test in a fairly reliable manner. Unlike other ABX software, Java ABC/HR makes it possible to encrypt the results so no one can falsify their test. I will set up the test and create an encryption key. After tests are completed you will have to email your test results to me. After some period of time I will process and post the test results.

The purpose of this thread is to discuss what 30 second samples should be used for the test. I suggest we use tracks with a fair amount of dynamics. Maybe one with all acoustic instruments and one with distorted instruments. I'm thinking one track from the Eric Clapton Unplugged album and one track from the Guns and Roses Use Your Illusions album. What do people think about these as choices?

Other feedback is welcome.

 

[Calroth]

I think at least one of the ABX guidelines is to pick a sample that's quiet but segues in to a loud, complicated section. Or the other way around. No suggestions otherwise, except, more well-known music is nice (Eric Clapton and Guns 'n' Roses qualify!).

 

[grawk]

My suggestion is that 30 seconds isn't nearly enough for a valid test.

 

[krmathis]

HiFiRE. I like the idea of a public ABX test.
Java ABC/HR is platform independent, which benefit those of us who not use MS Windows. Great!
Quote:

Originally Posted by grawk /img/forum/go_quote.gif My suggestion is that 30 seconds isn't nearly enough for a valid test.

With sample larger than 30 seconds we might go into copyright problems. If we don't use freely available audio streams that is...

 

[HiFiRE]

Quote:

Originally Posted by grawk /img/forum/go_quote.gif My suggestion is that 30 seconds isn't nearly enough for a valid test.

In addition to copyright problems, longer samples require more bandwidth. As an aside, 30 seconds is more than enough for me to AB.

 

[Calroth]

Quote:

Originally Posted by HiFiRE /img/forum/go_quote.gif In addition to copyright problems, longer samples require more bandwidth. As an aside, 30 seconds is more than enough for me to AB.

Not only that, I don't want to have to be critically listening to some piece for hours on end. With 30 seconds A and B, plus some overhead time, plus some time for replays, that could be around the 20 minute mark for 10 trials. Some people have this time and patience, not me

 

[HiFiRE]

It seems like there isn't much interest in my proposed test. Is anyone interested in participating?

 

[Happy Camper]

I would think the marketing and trials with this forum would crack open a goldmine of information. If we can get some standards established, the industry can better suit us. If loseless is required, more memory. If its not needed, I don't want to buy it.

I would like to participate.

 

[sant430]

I'd like to give it a try. I'll admit right off the bat that I normally can't tell the difference, but what the hell.

Funny, I thought there would be more people who would jump on this opportunity.....

 

[Alleyman]

I'll do it too...

 

[Denim]

I'll give it a go. I've never taken a serious attempt to try and find out if I can tell the difference.

My question would be if you are going to include a lossless reference clip to compare the two unknown clips against? I supose if you did, then some people could use software to figure out what is what. If not, then I would want to compare the clips to my CD's to help me determine what's what. Is that legit?

 

[Joshatdot]

Quote:

Originally Posted by HiFiRE /img/forum/go_quote.gif The purpose of this thread is to discuss what 30 second samples should be used for the test. I suggest we use tracks with a fair amount of dynamics. Maybe one with all acoustic instruments and one with distorted instruments. I'm thinking one track from the Eric Clapton Unplugged album and one track from the Guns and Roses Use Your Illusions album. What do people think about these as choices? Other feedback is welcome.

I propose Quake Soundtrack by Nine Inch Nails/Trent Reznor, I have this CD to make the clips.

 

[ppxnouse]

Quote:

t seems like there isn't much interest in my proposed test. Is anyone interested in participating?

I am in. Just did not want to reply with nothing to say, then rather wait for the test to be set up and just do it.

 

[Denim]

Quote:

Originally Posted by HiFiRE /img/forum/go_quote.gif It seems like there isn't much interest in my proposed test. Is anyone interested in participating?

HiFiRE: How many participants were you shooting for? Start with what you have and more will join in as results become known.

Music selections -- I'd like a couple samples from well engineered albums, such as Steely Dan. I'd also like a couple samples from albums that are known to be way too hot. Almost anything from modern day recordings will fit that bill. I'd be interested in knowing if the quality of the sound plays a major part in hearing the difference between lossy and lossless.

 

[krmathis]

Count me in.