I've done ABX testing with the usual suspects - castanets, cymbals, crowd cheering, and decided to use 220+ kbps VBR AAC to archive my CDs onto my Hard disk.

My question is, has anyone tried doing ABX testing and listening for degradation of imaging on loudspeakers? A typical good recording with good imaging on loudspeakers, doesn't image on headphones.

I remember a Stereophile review of the iPod from about 3 years ago. They praised its sonics, the analog and the D-A test results, but the reviewer commented AAC starts to sound really good at IIRC 320 kbps, where below that the image depth and "air" around instruments degrades. Has anyone verified this with ABX?

yes this a common problem with AAC. even lossless sounds less sharp at the top but one have to admit that the general sonic structure in terms of warmness and purple bass figures constistently get's muddled by the yellow airy nerzos in the higher field of the spectrum.
...
no wait a minute, that was all bullshit I heard somewhere...
most AAC encoders sound great at 320kbps. they sound as the original waveform a.k.a. the are transparent. they actually start becoming transparent on most music to most ppl at a much lesser bitrate.
the reviewer and possibly yourself fell for the placebo effect i.e. you're imagining things.
the best way to see for yourself is to do a simple ABX test.

please explain what you understand under 'imaging'

edit: spelling

"air", "breathing room", et cetera, are terms used by audiophiles to try to describe something they think they hear, but is mostly placebo effect, I've done ABX tests with AAC at bitrates of 128, 160 and 192 kbps all of them VBR, and most of the time I have a hard time ABX'ing 160kbps and up on most easy samples. The things I can detect best on lossy encodes are HF smearing on cymbals and the gurgling of low bitrate files, but I have never thought "This compressed file is fine, but it's missing the air and breathing room of the original", because it doesn't make sense, unless you encode to mono.

Yes, Xerophase, with MPC at high bitrate : http://www.hydrogenaudio.org/forums/index....ndpost&p=114968

I took a quick crack at it, using 2 tracks from the 1995 Chesky Records "Ultimate Demonstration CD". I used iTunes AAC encoder at ~256kbps VBR. I played it back on my big speakers, and switched between the CD, and AirTunes (using Apple's little 802.11 box, the AirPort Express, receiving from iTunes on my PC).

I tried to listen for differences in the imaging and soundstaging... and couldn't hear any.

Even some little percussive instruments that sounded like they were a bit back, on one track, and the echo from a wall in the rear in a sax track, were faithfully reproduced.

My big speakers are newish, floorstanding B&WDCM603 S3.

I may try the famous whispering-in-the-left-ear at the end of one song from Sting's Nothing Like the Sun....

On speakers you get the illusion the various instruments are spread out in front of you, and some can sound like they're further in the back. With headphones, they sound like they're spread out between you ears. Instruments that are made to sound like they're far away, vaguely sound like they're far but in no particular direction, and still between your ears.

Pls. note that I said that I have done ABX testing. I used decent Grado headphones and decided to archive a lot of my CDs using ~220 kbps AAC. What I have not done (til today) was do some testing on big speakers, and listening for changes in imaging, because a lot of those aspects don't sound right on headphones.

Has anyone compared ABX results of the same material between loudspeakers and headphones?

Is it easier to ABX some tracks on one or the other?

Obviously, headphones have the advantage of hearing the sound without room interactions etc. but they don't have the stereo crosstalk of speakers, which may or may not affect a codec's performance.

Also given that we have a finite ability to locate sound horizontally (say we can only distinguish differences of 5(?) degrees), do you think loudspeakers would allow us to resolve stereo placement issues better?

I'm just curious that if most codec tunings/listening tests are done on headphones, how well does that relate to reproduction over loudspeakers.

-Iain

I once ABXed a sample on speakers more easily than with headphones : http://www.hydrogenaudio.org/forums/index....ndpost&p=216681

But I did not pay attention to the imaging at all. In fact, the speakers were not in front of me, but rather close to a "headphone" configuration !

BTW a follow up.

I recorded onto CD some short clips of some tracks that have excellent imaging:
- Fairytales of Slavery, Miranda Sex Garden
- forgot song, Ana Caram
- a few others

I tested the original wav file, and a version WAV->AAC (using iTunes, 256kbps avg VBR) -> WAV

I couldn't tell a difference in the depth, etc.

Speakers are B&W DM603-S3 ($1k 2.5 way floorstanders). They sound better than my Grado SR80's, except for the potential for ambient noise, and bass room boom.


Distortion on the recording I find are WAY more audible that AAC codec artifacts even as low as ~160 kbps avg VBR.

I have very crummy headphones, so it's much easier for me to ABX on my speakers (which are decent enough, certainly better than my headphones).

I haven't really tried testing for degradation of imaging, though.

If you have the energy, please try lower bitrates.

I can't reliably ABX 128k AAC VBR when testing on songs (rather than difficult test samples), but differences in mastering quality, etc. are usually very obvious.

I've ABX'ed castanets (the most difficult), I can hear pre-echo on AAC below about ~220 kbps avg. ("extreme") BTW when encoding castanets, the bitrate opens up big time, to about 400+ kbps.

Funny you should bring this up as I do think that imaging is sometimes overlooked in listining test due to the use of headphones and its a good point. If you go to the MP3 tech forum and read my posting in the problem samples thread you'll see that I was able to ABX (LAME -V0) fools.wv (intro to Lemon Tree by Fools Garden) soley by listining to the imaging. This was not detectable on headphones at all. Of course I don't think this is a widespread problem with samples and seems only isolated to special cases.

Audiophile mags have a habit of being able to "see the Emperors clothes" if you know what I mean so I wouldn't put too much faith in one persons subjective view but you could always do an in depth ABX test youself to see if this weakness does exist. I would be very interested in the results.

It's a couple of years later and I did do an imaging test.
Any imaging degradation at higher bitrates is subtle, and is only audible when compared to the original.

At lower bitrates the scattering or "underwater" effect on highs is more immediately obvious, and can be obvious without needing to be compared against the original.

My favorite encoding method, AAC VBR 220+, has no audible degradation including in imaging. Neither does AAC 256 kbps CBR .

For testing I used some Chesky CD's which were recorded using Blumlein mic arrangements.

Imaging degradation is way more obvious on good speakers than headphones.

A quick check for I've used recently is Crystal Method's "Vapor Trail" at the 1 minute mark. The "sigh" while the cymbals are going is a good check, the image is supposed to go sideways.

disclosure: I'm 40 years old and use newish B&W floorstanding speakers. My high frequency hearing is no longer as good as it was, I can now hear only to almost 17 kHz, I used to hear >18.

Maybe younger bucks will hear more degradation than me.