"The 90 dB SNR is squeezed to 50 dB or less by a audio compressor circuit before frequency modulating the RF signal for the Antenna. On the receiver side, once the RF switch has the strongest signal to FM Demodulator, the resulting received audio is expanded back to 90 dB and is ready to by sent onwards to the audio output jacks."
Why is the compression and expansion necessary in transition of signal by wireless medium?
Compression followed by expansion is often called "companding" and is used in various fields to reduce noise. It is effective in reducing the final level of noise that is introduced between the compression and expansion steps. It will do nothing to improve noise introduced after expansion, and could make noise introduced before compression worse, or leave it the same.
Consider the diagram below, which is of a companding noise reduction system created by Dolby for reducing the noise introduced by a tape machine. When looking at the diagram, imagine the "Tape Noise" indicated is instead noise introduced during wireless transmission. Just as the Dolby system reduces the final level of noise introduced by the tape machine and magnetic tape properties, companding a signal reduces the final level of any noise introduced during a wireless transmission.
In this way, companding reduces the bandwidth and/or dynamic range needed in a system that is "inside" the compansion process. That means a smaller bandwidth requirement for wireless transmission, which can save on cost and power consumption.
Note that companding is not necessary, it is only desired.
Because with a 50dB SNR the noise (white noise also known as hiss, and radio interference) would be quite noticable, especially during periods of silence or if the audio signal is not at maximum level.
Companding allows artificially increasing the SNR so the noise is quieter. It can't improve SNR when the signal is at maximum level, you are still limited by the wireless link's SNR - but it is when the signal is quieter that the noise becomes most noticable and companding is beneficial.
For example say your audio signal is at -30dB from maximum. If you have 50dB SNR on the radio link then your actual SNR is now 20dB because your signal is -30dB below maximum. The noise would be very intrusive.
With companding, the same signal at -30dB (with a 2:1 ratio) would be boosted by 15dB so now the signal is at -15dB and your SNR after transmission is 35dB (50dB-15dB). The noise is therefore reduced by 15dB and becomes less noticable. Then at the receiver it is expanded back so the -15dB signal received is adjusted back to the original -30dB level, but the noise is now at -65dB instead of -50dB.
