It is a system where the base station antennas are geographically distributed over the cell instead of being co-located in a single array in the cell center.

From a performance perspective distributed massive MIMO is better than co-located massive MIMO, but it might be harder and more expensive to implement. 

This is one approach of massive MIMO where all the large number of antennas are distributed rather than being placed in a single array.

Thank you

Distributed antenna is more hypothetical than practical. Several bunches of antennas are placed across the cell to collect the RF signal. This in turns helps  to generate a strong received signal. How do we connect these bunches ? Wired or wireless ? What are the losses ?

We need to address several problems to realize such a system.

Thank you Sir AR Reddy..

AR Reddy: Sure, it is not straightforward to implement any of these solutions.

I believe that the cloud-RAN architecture is one possible way of implementing distributed Massive MIMO. Each antenna simply quantizes its signal, sends it over fiber to a computer cabinet in the basement which does all the processing.

I am trying to figure out why distributed massive MIMO has better performance. Could you say the reasons?  Is it, among others, because the terminals at the cell edge suffreing less inter cell interference?  It is said in Marzetta paper that the effective SIR in massive MIMO is random variable which  depends on the position of terminals.

Coverage gain: always some BS antenna close to UE, better than centralized M-MIMO

Beamforming gain: coherent transmission and reception, signal from all sources are constructively superposed in both DL and UL, same as M-MIMO

Spatial multiplexing gain: even stronger than centralized M-MIMO