Mainly generated noise. The "network speakers" are self-contained beasts; the box can be shielded, wires connecting driver to speaker are short, speaker selected for its inductive characteristics, etc. So, no need for an LC filter (which adds cost -- important if you are thinking about a commodity product).
I'm using a similar *functionality* in how I implement the "doorbell", here. I.e., drive a speaker based on (a) digital input(s) The layout of this house is such that we'd need two speakers (or, one OBNOXIOUSLY LOUD speaker). I suspect this is common for many homes (e.g., growing up, you could never hear the doorbell if you were in the basement, doing laundry, etc.) And, if these were independant units, then they would have to talk to each other -- at the very least to be able to inform each other as to *which* "door" is calling for service.
As this functionality (i.e., doorbell) is a basic function that every home requires (?), I've lumped it in the implementation of the HVAC system -- which has similar DIO's. This also addresses the fact that doorbells tend to have components located *near* the furnace in most homes (e.g., the 24VAC xformer). So, it's a starting point for introducing automation to a home.
It's only logical that I would like to leverage the network speaker design to provide this functionality. Granted, it has more capability than is
*required* for a simple doorbell but easier to leverage that part of the design and treat the speakers as having the same capabilities of the "network speakers" than to have to explain why the doorbell speakers can't be used to play music, announcements, etc. the way the *other* network speakers can perform. ("Why can't I listen to the radio over the doorbell speakers?" "Why can't the doorbell speakers inform me of the identity of the caller on the telephone?" "Why can't my SO use the speakers as a crude 'paging' system when I'm on the other side of the house?" etc.)But, driving two *distant* speakers from a single point means "speaker cables". In our case, two 30 ft lengths in essentially opposite directions. In other homes in which I've lived, that could easily have been 60 ft! Nice long "intentional radiators". I can't rely on the connections to the speaker being short (like in the "integrated" network speaker design). Nor for them to be contained in a shielded enclosure, etc. I probably can't even ensure the user will use a particular *driver* (speaker) with this set up!
So, the question boils down to whether or not I should replicate the network speaker design *in* the HVAC controller (essentially, just add one set of I/O's to the "other" and software, accordingly) *or* just leverage the software and adopt a different hardware configuration. Especially when the hardware changes are largely confined to just the audio amplifier!
I think in the ~400KHz range.