This entertainment broadcast system isn't used on HF, but it's interesting anyway. Parts of it are like things done by utes.
What we see first is the left channel of the receiver's audio output in stereo mode. The 19-kHz pilot tone used by this process is still very much present. I wonder if dogs can hear it on a decent tweeter. Old hi-fis had a "whistle filter," in case this pilot interfered with such things as the bias frequency on a tape recorder. The presence of the pilot is also what turns on that little "Stereo" light.
In this example, the mono-stereo switch is thrown about 3.25 seconds in. The output changes from the sum information (L+R) to the L information only. The receiver begins to actually process the difference information, which is amplitude modulated on a suppressed DSB carrier at 38 kHz. This carrier is restored in the receiver by syncing to the 19-kHz pilot, which is half the frequency. For all our algebra fans out there, the subsequent decoding math is (L+R)+(L-R)=2L. The other channel is derived as (L+R)-(L-R)=2R. In this particular case, a single male announcer, the program does not change significantly when it goes from L+R to 2L. The noise does, however.
Theoretically there's supposed to be very little energy between the program audio cutoff of 15 kHz and the lowest possible difference frequency of (38-15) or 23 kHz. In practice, considerable phase noise immediately forms all around the pilot tone, and effects from distortion also seem to increase.
This plot from 10 hz to 12 kHz shows what happens when the receiver switch is thrown back to mono during a silence. Except for a couple of spiky switching transients, the noise pretty much vanishes instantly. Mono FM, as any old-timer knows, can be extremely quiet, but stereo takes a 10 to 20 dB hit for a number of highly technical reasons having to do with unequal noise density of the sum and difference information. This is seen even more dramatically in the comparison below. The spike at the bottom end is in the computer gear. Everything else drops off fast with frequency, making this increased circuit noise less obtrusive sounding than the numbers would indicate.
All plots made with GRAM.EXE.