It's not so much that the high performance radios are proprietary as it is that building one of these systems is expensive and requires lots of domain specific knowledge. You could somewhat easily build one of these systems for ~$250k using CoT's hardware, in fact Lund University did a few years ago for some of their MASSIVE MIMO research.
If you don't want to use CoT's hardware for some reason then you can roll your own transceivers. You have a choice of Analog Devices chips or Lime Microsystems. Xilinx has a new platform that would also work but it hasn't been widely released yet. Lime's chip is cheap but it's performance is crap and its drivers are even worse. ADI's chips are easy to use and have fantastic performance but you pay for it. You also need to shell out for the massive FPGA's to handle 64*6.4GB/s of streaming data and then you need a system capable of processing it which will still probably be an FPGA or a DSP.
If you don't want to use CoT's hardware for some reason then you can roll your own transceivers. You have a choice of Analog Devices chips or Lime Microsystems. Xilinx has a new platform that would also work but it hasn't been widely released yet. Lime's chip is cheap but it's performance is crap and its drivers are even worse. ADI's chips are easy to use and have fantastic performance but you pay for it. You also need to shell out for the massive FPGA's to handle 64*6.4GB/s of streaming data and then you need a system capable of processing it which will still probably be an FPGA or a DSP.