Now there is where you're wrong, but I understand why you think you're right. You're thinking what goes in, comes out. Soundstage is a result of delay, phasing, and level between left ear and right ear. If you play a sound in right ear only it sounds like a sound right next to your right ear. Add a little of that sound to the left ear but out of phase and with a slight delay, and your head interprets this as positional information and that sound will sound MORE to the right, even though you are now playing from the left as well.
MP3 at lower bit rates uses a form of stereo image compression which mostly does a good job of not changing the sound stage, however mp3 artifacts are MORE present in the sound field than in any particular sound. MP3 can and does effect the phasing and stereo clarity of certain sounds at lower bitrate, especially percussive type sounds like drums, cymbals etc.
I recently did a ABX test to spot a 128kb mp3 among a 320kb mp3 of the same song. One of the tells I found to distinguish between them was a drum part where the 16th notes alternated between left and right, in the lower bitrate there was muddying of the stereo separation making it harder to hear each individual hit. This was something in the background of the song and nearly inaudible if you weren't looking for it. It wasn't a huge difference but it was enough for me to easily spot the lower bit rate every time.
So I can see where you're coming from but my personal experience diverges from your assessment, stereo information and especially phasing tend to have more information lost in mp3 than other sounds making it have a noticable effect (even though it's very slight) on the sound stage.
Points taken and I concede that low bitrates can indeed affect soundstage. I'd forgotten that delay, phasing, and stereo levels can play a part in the perception of soundstage.
With CBR , and using same encoder , that's actually the case (I guess there might be small differences between encoders, because of headers or whatever) . Also mp3 doesn't remove only information below and above particular frequencies, but use psycho acoustics, to disregard imformations that our brain often miss. For instance a loud sound, immediately followed by a quiet sound, often you don't hear the quiet sound.
Oh and downsampling ... not related at all.
I was assuming VBR, not CBR, but thanks for pointing out that technicality. 1 minute at 192 kb/s VBR can vary wildly in size depending on the musical content, as opposed to CBR which of course will have a fixed size per figure of time. Btw, I was referring to the frequency cut-offs as the user options that LAME provides (lowpass and highpass), not as the MP3 algorithm.
Soundstage is possible on headphones....through two different ways primarily:
- Hardware: Some headphone amps have a crossfeed function that will help simulate soundstage. There are also other devices, like the Smyth Realiser, which will simulate it.
- Software: Binaural recordings, which are designed to be heard on headphones and are mixed appropriately. HDTracks sells a few binaural recordings that you can check out.
Perception of soundstage is also highly dependent on the headphone design. For example, you won't typically find a soundstage with most closed headphones, as the sound waves are trapped within the driver enclosure and aren't allowed to radiate. I've heard a few exceptions though like the Audio-Technica W5000 and Fostex TH900.
Even with open headphones, not all of them will display a "soundstage", or at least the one that the listener will want. For example, Grado headphones in general impose an up-close soundstage. The Senn HD800 imposes a much larger one. Headphones with angled drivers (like the HD800, but also others) will typically have more-convincing soundstages.
Edited by Asr - 3/8/13 at 1:29am