In signal detection for massive MIMO systems, I have come across mainly three types of research articles.
Type 1: The channel (matrix = H) is taken as Rayleigh fading with mean = 0 and variance = 1. These articles don't employ any path loss models.
Type 2: Just like type 1, but the difference is that the channel matrix is Rayleigh with mean=0 and variance=1/M (M = receiver antennas).
Type 3: The channel matrix is described by the Kronecker model.
I wasn't able to find much about path loss modeling, shadowing, etc., along with other such factors in symbol detection for the uplink scenarios.
As a rough approximate around 80% of the articles are of type 1 (with the assumption that channel is known at the receiver).
What are your comments regarding each type?
Probably you are mistaken by the probablistic model of channel and deterministic model. See the book on wirless communciation by David Tse and Viswanathan. There are well known model for channel. Rayleigh is most widely used for NLOS conditions. Mostly fading channels are modelled probabilistically.
Thank you Rajnish. I received an answer from Dr. Emil Bjornson. Let me share that with you as well. He stated that type 1 and 2 are more or less the same thing since only a scaling factor differs. Type 2 papers are analyzing systems with infinitely many antennas. It is Type 1 that is the most realistic one. Spatial correlation is very important but the Kronecker model isn’t the right model to use in Massive MIMO. It is a model designed for single-user MIMO scenarios. He recommends people to use the spatial correlation based channel models with varying pathloss between users.
Thank you for summarizing my answer!
To be clear, in my vocabulary, "Massive MIMO" is a short form for "Massive multi-user MIMO" so there are many antennas at the base station and many users that are spatially multiplexed. For this scenario, I think the models from Chapter 2 in Massive MIMO networks (https://massivemimobook.com/wp/free-pdf/) is what should be utilized.
For single-user MIMO systems with many antennas at both sides, one can use the Kronecker model. However, I wouldn't call this "Massive MIMO".
Many thanks, Dr. Emil. In my opinion, the performance of existing Massive MIMO detectors needs to be re-evaluated using realistic models, and I agree with your response. And thank you for providing a link to your free e-book.
Thanks Kiran for updating.. You can also see this article.
Millimeter-Wave Communications: Physical Channel Models, Design Considerations, Antenna Constructions, and Link-Budget (IEEE Survey and Tutorials).