What is the maximum distance that music is transmitted through the air? I would like to hear your advice on which method to use to transmit music or speech in a frequency band that is sensitive to the human ear up to 3km. Everyone knows that the higher the frequency, the sharper the directivity, but instead, the attenuation in the air is severe. The Ganasys company, which made LRAD, dismissed the fact that it used ultrasonic waves as an erroneous view and said that it actually uses the sensitive 2200~3300Hz audible band.
Meanwhile, ZaellixA piqued my curiosity by suggesting that ultrasonic waves can be used to modulate audible sounds to transmit sound with ultrasonic characteristics farther away. Is there really no other way than to modulate with ultrasound?
 A: As you already mention in your question, with increasing frequency two main traits are observed. One is the increase in directivity and the other is the increased attenuation/dissipation from/at the medium (air in your case).
Now, regarding the LRAD device and Ganasys company, I can't really say what the exact reason they decided to deviate from their initial plan is but I assume that relates to the second factor mentioned above. One has to keep in mind that increased directivity only means that the geometrical attenuation is reduced compared to the case, for example, of uniform radiation (or any other case in between). This in no way implies that the radiated power is higher though. Ultrasonic signals tend to be (way) more directive than the majority of the signals in the audible frequency range but that doesn't mean that the radiated power is higher. On top of that, one has to consider the increased attenuation imposed by the medium.
I am no expert on ultrasounds but with my limited knowledge I believe that the emitted energy is lower for ultrasounds and I know for sure that they are attenuated way faster (greater attenuation per distance measure) than, say the mid or low frequencies of the audible spectrum. I believe that those two factors in combination led the aforementioned company to abandon their initial design and move lower in frequency. In the range you mention, one could possibly achieve a good compromise between relatively low dissipation in the medium and directivity. If we also consider using array technology to focus the beam (which is quite easy to achieve for a limited frequency band. You can also see this answer for a quick overview), effectively achieving smaller geometrical attenuation then this frequency range is a quite good candidate for such as system.
Again, to my limited knowledge, ultrasounds are used to transmit focused sound in rather short ranges, especially for partitioning a space into zones or similar applications (excluding medical applications here as they don't seem to be a topic of interest in this question). As far as I know, the two main ways to focus sound towards specific steradian angles are beam-pattern control which is the topic of array processing techniques (among other technologies falling under this field) and the use of horns as used in loudspeaker technology.
