I forgot the most important point.
Yes, testing PS-HEAAC at 32/48 kbps against a basic MP3 encoding has some relevance. But it's a dangerous game. Take the atrac3plus test as example. The test was relevant (a basic MP3 coder vs the new Sony's format). Just go on some minidisc boards: people were convinced that atrac3plus at 64 kbps was as good as MP3 at 128 kbps. It's sci-en-ti-fic!
To all people maintaining that MP3 could sound much better => "no, take another look on the scientific test. If you disagree, try to do yourself a better test"... organisation which was of course impossible to do (very expensive).
That's how a flawed listening test setting became a reference and the source of confusion, FUD, etc... exactly the opposite purpose of a listening test (and science in general).

As a consequence: your test has just few relevance? Maybe... But don't cry if several people are going to interpret the result in a distorted manner.
I don't want to fight on french forum against people claiming that PS-HEAAC at 32 kbps is as good as LAME at 128 kbps and basing their opinion on a listening tests posted here, on HA.org. I spent several years to fight against the MP3Pro/ATRAC3/VQF/Vorbis... = same quality as MP3@128 bullshit. And I know that other people did the same in various local forums.
A flawed test (with handicapped MP3, limited persons...) is VERY DANGEROUS for the community (but maybe very helpful for some big companies marketing campaign). It looks scientific and is just parodic.

Sorry to interrupt your professional discussion.
But it's obvios that LAME 128 will be better than HE-AAC2 on music samples.
Let's suppose that Nero HE-AAC2 32 will be good as db HE-AAC2 at 64 kbit/s.
Nero HE-AAC PS 32 kbit/s <= db HE-AAC2 64 kbit/s << LAME 128.
Why to do test if we already know that LAME 128 will be better?

HE-AAC2 will be excelent for movie's soundtrack but for music people will prefer LAME,LC-AAC or VORBIS. I think LAME 128 kbit/s(OGG 85-90 kbit/s, LC-AAC 96-112 kbit/s) is already on limite between transparent and not. Something little worse is already uncceptable for listener even on Low-Fi sound solutions.

Indeed it's wil be better to ask people's opinion if they need such kind of comparison. For me it's pointless.

Lets hypothetically asume that this is true. Then this would mean that Garf either overestimates Nero HE-AACv2 and will loose, or plans to cheat by picking easy samples and incompetent half-deaf testers - therefore just proving that people who dont care much also dont care about quality. In that case, you can just as well throw in WMA-standard @64kbit and it will perform the same.

*If* this test is about the above, then i agree that its pointless. But i disagree that a test with average-joes - if done right - is pointless.

This is just a suggestion to the forum administrators. This thread is becoming a Listening Test one. Why don't you open a new thread in the right forum for this discussion? you can also close this thread or suggest people to post listening test related posts in that thread.

Thanks,
Oki

I would be intersted in seeing the results of Garfs proposed test.

Why there is so much complaining about it I really don't understand.

The scenario was very clear and makes alot of sense in comparative value. (in the context what the new encoder could/would be used for)

Sucks.

Garf deemed his own test impossible to carry out:


He made statements about the target audience but can't realistically do a test with participants of such an audience.

Not at all. guruboolez proposed a public test. I was talking about dragging people into my office. I stated what I wanted to do pretty clearly, I think. (The reason I lost interest in doing it is that there is no point anymore to it, IMNSHO).

One reason why that is needed is that the people that don't hear differences don't want to submit their result "because they can't hear anything anyway". (Actual quotes, actual experience). You end up only with the results of the people that could actually hear differences (and only for the samples where they could). Obviously that makes it impossible to verify the claim if the test is conducted as guruboolez proposed.