There have been some great posts. I would like to take it one step further. Much of what I'm about to say will make no sense. Sorry.
If you have time, do a quick internet search. Look for 'record groove under a microscope'.
Now keep things simple: the stylus moves to the left (1st axis), the stylus moves right (2nd axis), and there is a slight up and down movement (as can be seen by the varying depth of the grooves) (3rd axis). A vinyl record groove can potentially store, and therefore output a 3D signal.
Don't think of a band, with the drummer behind the singer, this is simply the location of the 'instruments'. Every stereo, and every source media, can portray this.
Instead, think of a song with a lone singer. There is a 3 dimensional aspect to their voice.
This 3D picture is captured on the master tape. The master tape is fed it's information via a balanced cable. Pin 1 is ground, pin 2 is the positive portion of the waveform, pin 3 is the negative portion of the wave form.
The positive portion of the waveform pushes your speaker driver out, the negative portion of the waveform pulls the driver in. Full, fluid uniform motion. People that own crossoverless full range driver speakers know exactly what I'm talking about. There is a 'rightness' to the sound. It sounds 'real'.
If you have a fully balanced stereo (Atma-Sphere, BAT, Lamm, etc.) this waveform is preserved.
Vinyl joins master tapes as a 3D source.
CD's do NOT preserve this 'balanced' nature of the waveform. It is a single ended source. If you have RCA connectors in your stereo chain, it is also no longer balanced.
To my ears, DSD is balanced. I will therefore conclude, with no technical information to support my claim, that the digitization of the original waveform to CD destroys the 3D nature of the event.
Here's the problem with my entire post. Until someone builds a fully balanced crossover speaker, you won't be able to 'hear' how huge the difference is. This between CD and vinyl. Unless of course you have a full range planar (no crossover, not a single capacitor or inductor in the signal path), a full range electrostatic (no crossover), or as mentioned, a full range driver (Lowther/Fostex etc.).
Only time will tell if I have a clue?