If I have a modulated signal A*cos(2pi*fc*t + phi), I can demodulate it using IQ demodulator and estimate the phase using I and Q channel information. When it comes to MIMO case (specifically SIMO case), I can first accurately measure delay and phase of the individual multipath components (MPCs) which are resolvable, i.e. which are delayed more than the symbol period (BW). Using IQ demodulator helps me to keep sampling rate just 2-5 times higher than the highest frequency of the message signal rather than the carrier frequency. My two questions are: 

1. On what factor does accuracy of phase estimation will depend? Sampling rate or ADC resolution? 

2. How can I measure the phase of the non-resolvable MPCs? Is it even possible with conventional techniques? 

I am asking this question because I am working on tracking user in dense multipath environment and obviously delay information of the MPCs are not sufficient to give high resolution (differentiate two user standing very close (i.e. there path delay to base station is less than symbo period)). So I am thinking of exploiting phase information to use high-resolution beamforming.

Phase estimation accuracy will mostly depend upon received SNR.

To measure the phase in each receiver channel due to each different transmitter channel, you will need some training slots in which the transmitter sends a known signal.