The main difference between AM STEREO and FM STEREO is that in AM STEREO, the sum and difference channels take different paths through the transmitter, as opposed to FM, where both go into a composite baseband which goes through the transmitter in a single path.  The WHY of the better sound of AM STEREO lies in the processing.  

In FM, we limit, compress and clip the left and right channels individually in order to create the maximum RMS (i.e., apparent loudness).  Since in FM STEREO, 100% left or 100% right equals 100% modulation, this works just fine.  (This does not take into account the 9% that we must leave for the pilot (SCA modulation can run the baseband over 100% modulation), but it illustrates the point.)  In AM STEREO, however, 50% left plus 50% right equals 100% L+R, and the L+R is the mono sum that most of the listeners hear.  Lose the left or lose the right and there goes half your modulation and half your apparent loudness.  

So in AM STEREO, we don't process left and right individually - we process L+R and L-R instead.  That way when, say, one channel is low or gone, it pumps the L+R in the main channel so that the mono listeners don't notice a decrease in loudness. Since we also process L-R (the difference between the channels or "separation"), we can increase the RMS of the difference and increase the apparent separation by several dB.  

More precisely, in AM STEREO, since we must run the L+R and L-R through two separate paths, if we used conventional left and right processing, we would have to process symmetrically (i.e., at a level where 100% limiting will occur when L=R for the L+R and L=(-R) for the L-R path.) If we modulate with only one channel (left or right), we would take a 6 dB loss in both monaural loudness and coverage.  Instead, we process the L+R and L-R separately, so that when we do have single channel program material, the L+R envelope is "pushed" to 100% (we actually limit this to 90-95% as a result of real-world decoder limitations).  

To a stereo listener, the single channel is actually 6 dB louder than it should be, but this is considered an acceptable trade-off when the alternative is losing 6 dB of the monaural modulation to which almost all the audience is listening.  With FM STEREO, a full left only or right only signal will produce only 50% mono, which can't be increased to 100% with processing, as it can in AM stereo, to recover the 6 dB loss the monophonic listener experiences under that condition.")  So, there you have it.  That's why AM STEREO seems to have more separation than FM STEREO.

AM stereo, of the C-QUAM variety is really quite simple, the transmitter uses a three step process to generate the C-QUAM modulated RF signal from the left and right audio signals. First a QUAM, Quadrature Amplitude Modulated signal is generated. The left and right channels are matrixed to form a L+R monophonic sum signal, and a L-R stereo difference signal. The L+R signal modulates a double sideband full carrier AM signal, and the L-R signal is used to modulate a quadrature carrier, 90 degrees out of phase with the L+R carrier, the quadrature carrier suppressed. These two quadrature signals are then added to form a QUAM signal with a carrier. The QUAM signal could be transmitted in that form, if it weren't for the fact that the FCC mandated that the AM stereo signal had to be fully compatible with the envelope detectors used in the majority of monophonic AM radios. If there is any stereo difference, or L-R component, it will distort the envelope of the QUAM signal, resulting in distortion in monophonic radios using envelope detectors. To get around this problem the QUAM signal is processed by two further steps, to create the C-QUAM, Compatible QUadrature Amplitude Modulation signal. First the QUAM signal is passed through a limiter which removes the envelope information, leaving only the phase information. The second step in generating the C-QUAM signal, from the QUAM signal, is the Amplitude Modulation of the limiter output with the L+R, or monophonic audio signal from the audio matrix. This process gives the C-QUAM signal a normal AM envelope containing the L+R or monophonic audio, for compatibility with existing monophonic AM radios using envelope detectors, while the L-R stereo difference signal is encoded in the carrier phase.

A normal QUAM receiver, would use a phase lock loop, and a VCO to generate a carrier, and a quadrature carrier, to synchronously demodulate the L+R and L-R audio signals. This scheme doesn't work with C-QUAM, because the QUAM envelope was modified in creating the C-QUAM signal. One way to restore the C-QUAM signal to the original QUAM form, so it can be processed by normal synchronous demodulators, is to correct the envelope with a "Cosine Corrector". The Vector math of this is too complicated to get into here, but it is basically an open loop process that depends on highly precise analog functions to generate the cosine of the C-QUAM signal phase, and then divide the C-QUAM signal by the cosine function, to regenerate the original QUAM signal. This process is described in the paper referenced below.

The method actually used in most analog C-QUAM receiver chips makes use of a high gain feedback loop which forces the output of the L+R synchronous demodulator to be equal to the output of a precision envelope detector, this is done as follows. First, since the envelope of the C-QUAM signal is a faithful copy of the L+R sum signal, an envelope detector is used to recover the L+R sum signal, or monophonic audio. The L+R sum signal from the envelope detector is then compared, in an error amplifier, with the L+R audio signal coming out of the QUAM synchronous demodulator. This error signal is then used to control a variable gain amplifier, or analog multiplier in the input to the QUAM synchronous quadrature demodulators, in such a way that the envelope is forced back to the normal QAUM form. The L-R output from the second synchronous quadrature demodulator, is then distortion free, and can be matrixed with the L+R signal from the envelope detector, to recreate the left and right audio signals at the output of the receiver.

To identify C-QUAM AM Stereo Broadcasts, 25 Hz tone is added the L-R stereo difference signal which modulates the quadrature carrier. The 25 Hz tone modulates the quadrature carrier with 5% modulation. Unlike the 19 kHz pilot tone used in FM Stereo, this 25 Hz tone does not play any part in the demodulation of the AM Stereo signal, it's only purpose is to identify AM Stereo Broadcasts, so the receiver can automatically switch into stereo mode, and light an "AM Stereo" light. more

Why broadcast in AM stereo?

Quite simply, we already have the equipment and we would like to appeal to the audiophiles who appreciate the great sound of AM Stereo.