While waiting for more people to vote on poll#4, I'm just going ahead setting up the 5th.

Direct link for samples #5.

There are 5 wav files. 1 is the original, labeled as such. The others are wav decoded from 4 different lossy compression. I'm going to set the poll for 2 weeks, and then disclose which format is which. Anyway, as the title said, this is just an informal poll, for fun. Pick which testfile that sounds the best/worst for you. It is not my intention for this to be a full blown ABX test since obviously every vote will be biased. I do include the original so if you're interested in doing some ABX tests, please do so, and share the result.

This time, the target bitrate is pretty low, and just with this information, some people can probably guess the codecs already. Still, vote away. Track is 紫音-sion-, by Masami Okui

PS: Mods, I really don't see any option to set an expiration for the thread on my end. Where should it be?

Couldn't get your zip using Opera.

I assume comments about the samples are not yet permitted...?

You cannot right click and save-as? Here's another link to the root folder.

Everybody can comment on why they voted what. I just won't disclose the actual codecs used until 2 weeks later.

<rant>
Microsoft apparently hates Opera. How come today even a simple file download can't be acomplished without extremely complex code or graphical bloat...
</rant>

I would rate these samples in the following order:
3 > 1 > 4 > 2

Testfile02 :: The effect where additional component of the crash cymbal on the left channel is put in center is very annoying if you know the sample. Overall stereo very much sucks, but I noticed the defect in direct A/B comparison only. Might get ugly in samples where several persons speak strictly in L and R (we have such an unlistenable HE-AAC "radio" here), or again with known stereo music. I'd prefer mono.
Testfile04 has an unpleasant "MP3"-artifact while the singer pronounces the second word [ro'i] and the crash cymbal noticeably drops out at low level.
Testfile01 is pretty much clean, except that the crash cymbal sounds metallic.
Testfile03 subjectively doesn't appear to have any problems. I probably couldn't ABX reliably. 03 is the only file where the rhythmic HF pulses every 0.440 seconds are clearly audible in the right channel.

All lossy files don't have the clipping (or intentional) distortion at 13.3s. Could be a good thing.

One more thing I gotta ask you. Why do you think this album is so incredibly loud? It comes out 1048 kBit/s compressed and has -11.59 dB ReplayGain. (I don't count the last track "I Wish".) Must be one of the loudest and most unlistenable albums in my collection!! Is Japan so much ahead in the loudness race?

Thank you for your analysis! This is one reason I do this polls, to see detailed feedbacks like yours.

As for loudness, well, TBH, that's the way it is for any pop CDs nowadays, regardless of country. MP3gain to the rescue. It's even worse with Eurobeat CDs.

Seeing as comments are permitted:
The easiest choice for me was selecting sample 04 as the worst. j7n's comment about no clipping or distortoin just doesn't wash with me - sample 04 clearly exhibited clipping in the vocals and some of the drums, and that annoyed me more than the stereo.

Choosing between 01 and 03 was much harder. In the end, it was purely down to the crash. They both had a solid release (I call it ring), but as j7n said, 01 was slightly less natural; too metallic and slightly hollow sounding (and cymbals aren't hollow!).

Thanks for the opportunity to take part. It's been interesting, especially test #5 - I couldn't chose a best one. I look forward to seeing what the codecs were.

This is getting better. Keep the votes and comments coming.

I said that distortion at 13.3s appears gone, not that the whole sample is without clipping. The distortion might be an intention square wave put there, I don't really know. Listening was carried out at 44,100 without further resampling.

I ABX'ed 0 (original) and 3 (my best) based on this fact alone:-

foo_abx 1.3.1 report
foobar2000 v0.9.4.5
2008/07/10 17:32:20

File A: D:\temp\Sample5\Sample5 Original.wav
File B: D:\temp\Sample5\Sample5 Testfile03.wav

17:32:20 : Test started.
17:34:24 : 01/01 50.0%
17:35:36 : 02/02 25.0%
17:36:18 : 03/03 12.5%
17:36:58 : 04/04 6.3%
17:37:35 : 05/05 3.1%
17:38:15 : 06/06 1.6%
17:38:53 : 07/07 0.8%
17:40:57 : Test finished.

----------
Total: 7/7 (0.8%)

2nd week bump

1 was by far the worst for me, and 3 and 4 were pretty close. 1 was let down by the clicking sound starting at :14 (processed electronic closed hi-hat? I dunno) becoming lower in the mix and losing its definition.

C'mon, only a few days left.

Poll #6 is up.

All sounds so bad. I don't even need an ABX for insuring secure result. Detailed ABC/HR log is following:

LOL. Yeah, the target bitrate is pretty low, so ABX with the original is pretty obvious.

C'mon people, time is running out. Only 5 voters out of 900+ views?