Note that each GD(i,j,k)=txgd(i,k)+rxgd(j,k) where txgd(i,k) is the group delay for the transmit chain using the i-th antenna and rxgd(j,k) is the group delay for the receive chain using the j-th antenna. For each k, there are 2N variables txgd(i,k), rxgd(i,k) whereas there are N*(N?1) measurable values of GD(i,j,k) corresponding to the ordered (i,j) pairs where i≠j, with i, j representing respectively the transmit and receive antenna indices. Thus, if N>2, since N*(N?1)>2N, there are more equations than unknowns, allowing the individual delays txgd(i,k) and rxgd(j,k) to be determined if all of the total GD(i,j,k) are measured. For N=2, the individual delays txgd(i,k) and rxgd(j,k) may be estimated under an assumption or approximation that they are same for both antennas, i.e., independent of i,j.
This can be useful for FDD devices, which usually communicate with a receive frequency kR and transmit frequency kT≠kR, for which the total group delay of interest for RTT measurement is thus txgd(i,kT)+rxgd(j,kR). Since measurements GD(i,j,k) made using a receive antenna to receive the transmission from a transmit antenna on the same device are necessarily at the same frequency k, they do not directly yield the above-mentioned quantity of interest txgd(i,kT)+rxgd(j,kR). Nonetheless, they can be used to determine the individual values txgd(i,k) and rxgd(j,k), from which this quantity may then be computed.