The reason for the choice of encoding the stereo signal is well documented in other <Wiki> and does not require repeating here. Suffice to say the mathemetics is very simple. It does not matter that the source is either analogue or digital it is encoded in the same manner.
Let the left audio channel level be L and the right R. We combine the channels at the baseband, then modulate the difference on to a 38KHz subcarrier using DSB-SC modulation ( double sideband suppressed carrier). The results is a MPX (multiplex) signal expressed by:
mpx = (L+R + (L-R)cos(2.Π.38KHz.t))
This, of cause is incomplete. There is no reliable means of demodulating this signal at the receiver
It is important to note that the sub-carrier is a difference signal. Under normal circumstance there is little difference between the L,R channels. Thus normally the sub-carrier modulation level will be low. This is a fact that can be used to some usefull end when compression/ expansion is used to enhance the "penetration/presence" of the broadcast signal