Dan Mills said:
Depends, in something like your run of the mill buck converter, duty cycles at low load can become very small, which raises the upper switching harmonic levels considerably.
You can even get shenanigans like various pulse skipping and DCM modes, good for efficiency, all kinds of bad for possibilities of stuff within the GBP of your opamps.
The synchronization thing is not so much about reducing noise as such, what it does is ensure that all the switchers are putting out the same frequencies so that you do not get intermod beat notes due to product detection either just in some non linear junction or by mixing with the delta sigma sample rate. The variable switch off timing does not matter as it only effects the harmonic content as long as all switchers are running at the same rate (so no possibilities for demodulating a low frequency beat note).
Switching converters for audio:
Layout, Layout Layout and did I mention Layout?
Go with forced CCM if your parts support that.
Sync the things to the sample clock or some harmonic of it if doing the AD-DA thing.
Ferrite is your friend, and ground planes are a good thing.
Most LDOs do nothing at the frequencies modern switchers use (There are some exceptions).
Keep the Q of LC filters sane (A resistor across the L is the usual method if you cannot use something inherently lossy).
Synchronous switchers are sometimes better then allowing a diode commutated buck to go into DCM.
Surface mount beads and MLCCs are very useful, lead inductance is the enemy here so use a couple of separate vias for each bypass cap (Less series L) and practically put the cap in the pad (Only really practical with microvia, but you can get close and use a wide connection).
Keep the loop areas small, tightly twisted wire from an off board switcher is much less offensive then some kind of huge loop, and common mode chokes can be used to send the HF crap around the correct loops rather then all over your board.
Finally, make yourself a set of E and H field probes so you can really see what is going on, RF has its little ways......
73 Dan.
Ooh. I think I know most of these!
Glossary:
DCM : discontinuous conduction mode. Mode of operation of a switching regulator, where the inductor current is zero for part of the switching cycle.
GBP: gain-bandwidth product. Figure of merit related to an operational amplifier. Literally is, as the name suggests, calculated as the product of the -3 dB (aka half power) bandwidth and the gain at some closed loop gain. The idea is that for a simple model of a normal voltage feedback opamp, this product is independent of closed loop gain.
CCM : continuous conduction mode. As DCM, but where the inductor current is non-zero for the full switching cycle.
LDO : low drop-out. A linear pass regulator with a low minimum voltage across the pass element required to stay in regulation. 'Low' in this context means appreciably less than 3 V and is considered low for purely historical reasons (though history is still happening around here, with the heavy use of ICs like LM337, which is not an LDO. I guess cheap and predictable is never obsolete).
Q of LC filter : the 'quality factor' of a filter comprised of inductance (L) and capacitance (C). The Q indicates how pure the resonance phenomenon of the LC interaction is. It is roughly related to the amount of loss (resistance) in the circuit.
MLCCs : multi-layer ceramic capacitors. The most common type of discrete capacitor in use today. The term covers a wide range of components from the high density types used for supply bypassing (with terrifying names like X7R, Y5V ) to the high grade, very linear low capacitance types for linear filtering and such ( with friendly, cuddly names like C0G and NP0 ).
E-field : electric field. A vector field encoding the potential energy of a hypothetical charged particle of unit charge associated with a unit displacement of said particle in relation to a fixed background charge distribution.
H-field : magnetic field. Phwew have to consult a textbook or Wikipedia for this one... Maybe later.
Well you can't escape jargon, but I hope I have provided some google bait for those who want to learn more.
A series LC filter can be better for taming high frequency ripple if the inductor has a high enough impedance at the switch frequency (and a low enough R at DC to prevent unwanted loss of DC regulation). 
