Originally Posted by DonoMan
The load of the speaker is still going to mostly-short-out any RFI. The closest point of induction does not simply send a voltage signal both ways down the cable where it will enter the amplifier's output at its usual voltage. It will split according to Kirchoff's Laws. Almost all the current the RFI produces will travel to the speaker and its voltage will be pulled down, which also affects the voltage that makes it back to the amp. Dimmers aren't exactly switching at super high frequency either. Probably 100-250kHz or so. Take a tweeter I like, the Vifa/Peerless XT25, rated 0.01mH. It has a reactance of 6.3 ohms at 100kHz and about 63 ohms at 1MHz (assuming no lowpass filters in your speaker crossover for the tweeter, anyway). Still going to pull most normal sources of RFI way down in voltage. Will some make it back to the amplifier? Yes. Is it possible for it to cause a problem? Yes. How likely is it? In most homes, extremely unlikely.
But hey, I won't say don't do it. I'll just say I wouldn't (and never have). Hell, my speaker wires, currently not in-wall (but I've done in-wall in multiple setups), are parallel to power cables, with very little separation. No problem at all. Edit: I have a dimmer about 10 feet away and another about 20 feet away. And there is a difference between 10 feet and 4 inches in that regard, I do get that.
I do like conduit, though!
Signals induced into speaker wire will likely be common mode since they will affect both conductors more or less the same. The speaker is unlikely to have much effect on this common mode signal since it will be on both inputs, and the speaker doesn't have a separate ground connection or differential amplifier to absorb much of this common mode signal.
At the amplifier end the common mode signal will be mostly converted to differential mode since each side of the signal will like encounter different conditions such as impedance. One side of the signal will see the low impedance outputs of the emitters of the power transistors, the impedance of an output inductor, if used, perhaps some capacitance to ground, and anything else attached to this conductor. This signal will also travel back through the feedback network that is connected to the output. The feedback network is likely connected to a ground and one of the inputs of the input differential amplifier in a typical design, through resistors and capacitors. There may other connections as well. It's hard to predict the result, which would be very frequency and signal level dependent, as well as dependent on the circuit design and physical implementation.
The signal on the other speaker wire conductor would first enter the amplifier ground. For reference, AVR's don't have a physical safety ground, they use two prong plugs. Most of the grounds in an AVR/AVP are ultimately connected together, although not all at the same point. Most of the grounds (various signal, digital, etc.) are also connected to the case at various points thorough small capacitors, something like 1 uF capacitors. There must be a lot of art in how all of these ground connections and signal flows are designed. The are also typically a connections from the case to each of the AC power inputs through small capacitors, around 1 uF, that are rated for this sort of safety critical location. One power input is essentially at earth ground, at least in the U.S., and other as 120 VAC. Since the outlet might be miswired, the capacitors are connected to both AC input lines. The idea is to dump collected high frequency noise, especially from the switch-mode power supply, into an earth ground.
The ground connections described above mean that noise introduced on the ground side of the speaker connector can go about anywhere and do what it will do. It's a tribute to the people that design AVR's and AVP's that there are few reported problems with noise that comes back through speaker wires, in instances when this happens, or most other entrances for that matter.