I've done some extended listening tests using various encoders at -q 8, I've used a variety of materials, including my own music which I've played back through exactly the same setup that it was recorded on (RME Hammerfall DSP). I've blind tested the following against the original WAV files...

CVS 1.01
GT3B2
GT3B2+HFR
MTB3
AOTUVB2
AOTUVB2+QK
QK32

One thing is very clear, AOTUVB2 is nowhere near transparent at Q8 - it has a very distinct sound.

The key characteristic (apart from pre-echo which is apparent in all of these encoders), is a separation and lack of integration at high frequencies. It's like the very high frequencies have been put through a bandpass filter.

That means that when Joni Mitchell sings, the breath of her voice is separated from her voice, as if floating in a separate band above it.

This also hold true for accoustic instruments with strong transients - guitars, cymbals etc.

I apologise for not being able to give you any software output - that's not how I judge audio quality - I don't believe that any kind of measurement which is non-human can never hope to to tell the whole truth.

OK, well I'll get to the point with a ranking of each encoder against the original WAV file (out of 5):

QK32 - 4.8 my favourite, it has a beautiful, natural and engaging quality
GT3B2+HFR - 4.5 less noisy and digitally "busy" at high frequencies than GT3B2
GT3B2 - 4.3 - a bit noisier and HF confused than the HFR version
AOTUVB2 - 3.6 very artificial sounding to my ears
CVS 1.01 - 3.5 muddy
AOTUVB2+QK - this seems like an experiment gone wrong
MTB3 - the worst, it sounds very artificial and somehow compressed

I'm a newbie to this forum, so I'm probably in for a barrage of abuse... but the bottome line is I believe Quantum Knot beta 3.2 is the best encoder at this quality by some margin.

I can still ABX it at Q8, which I can't for MPC at Q8. However Ogg Vorbis has a future, and clearly is continuing to improve.

Please follow the forum rules.

That would mean providing test results (I think you're confused about the quality testing bit also -- abc/hr relies upon human perception for results, it just imposes a methodical framework upon that) or test samples, or ideally both.

Where do you download "Quantum Knot beta 3.2"?

Thanks.

if you would have done a search, you would have come up with this thread or someting similar, pointing to:
http://www.rarewares.org/quantumknot/

I didn't check the forum rules, but I've now found this one:

8. Any statement about sound quality must be supported by the author responsible for such statements by a double blind listening test demonstrating that he can hear a difference, together with a test sample. Graphs, non-blind listening tests, subtracting two files and so on are definetely not considered as valid evidences of sound quality

Forget it! Life is too short, and I have music to make. This will be my last post, my apologies for bothering you boys.

have fun.

edit: if your'e intersted WHY the rule exsists, read on here

This is a shame. It's entirely possible that there is a problem with aotuv at high bitrates (it has not been tested rigorously), and you said that you were able to ABX, but because you aren't willing to do a blind ABC/hr comparison, nor post a sample, we wont't be able to learn anything. You claim "life is too short", but had time to make a sighted test and make judgements about quality. A blind test does not take more time to conduct than a sighted test, and the software to do such a test is readily available and easy to use.

Then again, perhaps I'm just taking the bait here. It does seem peculiar that QK32 sounded better than 1.0.1, as I thought its tunings did not take affect above -q 5.

This has all the makings of a troll. In your very first post, mix in some buzz phrases relating to blind testing to draw the regulars in:

"I've blind tested the following against the original WAV files..."
"I can still ABX it at Q8, which I can't for MPC at Q8."

and then contradict the whole idea of blind testing with:

"I apologise for not being able to give you any software output - that's not how I judge audio quality - I don't believe that any kind of measurement which is non-human can never hope to to tell the whole truth."

Throw in some typical audiophile jargon:

"less noisy and digitally "busy" at high frequencies"
"muddy"
"sounds very artificial and somehow compressed"

and then part with an insult when called to task:

"Forget it! Life is too short, and I have music to make. This will be my last post, my apologies for bothering you boys."

I bet I could automate the whole thing with a script.

ff123

Hahaha! I can tell you've been in the game for a long time ff123.

There is no contradiction... a blind listening test is not software output is it? Jeez you guys are so tetchy... please feel free to automate me with a script... I'm sure you have the time and the patience to do it.

You have a good observation ability!

BTW I appreciate what you guys are doing, but perhaps you could try to be a little more friendly to newcomers? You make the Cubase forum look like Sesame Street.

Welcome to HA.org, if you follow this guide closely you will become a very respected member of the commnity in no time at all.

5 Easy Steps to become a respected HA.org member

0. Read the TOS, the FAQ and Pio's introduction "What is a blind ABX test?".

1. Download (and install) ABC/HR or its Java equivalent.

2. Use ABC/HR to set up a test with a sample of your choice and test away (ABX results are always nice too). Especially if you are comparing more than two codecs it's better to do an ABC/HR instead of just an ABX test.

3. Post the results and a lossless sample which causes the problem you noticed.

4. Wait for other members and developers to verify your results and help by providing more information/samples and further testing.

OK... thanks for the practical heads up.

you are the one not behaving with this kind of statements.