In a previous post I referenced a paper of Keysight as a good starting point to understand how Active Antenna Systems (AAS) that have been specified for 3GPP 5G New Radio (NR) could improve overall capacity in a cell and extend the cell range. Today, I have found another interesting paper on the topic that was recently published by Ericsson.
This paper is very complementary to the Keysight paper as it focuses on how active antennas are implemented from a hardware point of view and here’s my summary and takeaways from the paper:
Today, a typical base station is split into three parts: There’s the digital module for signal processing referred to as the ‘baseband’ which forwards digital RF signal information over a fiber cable to the radio unit. The radio unit generates a radio signal, amplifies it and then sends it over a coaxial copper cable to the antenna. Baseband + Radio + Antenna = 3 distinct parts. For reception of data from mobile devices, the system is traversed in the reverse order. For the remainder of the post I’ll only look at the downlink direction, however.
With active antenna systems this changes significantly, There’s a bit of the baseband left which sends information to the Active Antenna System over a fiber link. The active antenna hardware includes part of the baseband, the radio module and the antenna system, all in one box at the top of the mast. There is no separate antenna anymore as before, it is all integrated!
In essence, active antenna systems can be used for two things, potentially interchangeably and at the same time: For far away devices, active antennas can focus the signal in a certain direction instead of equally blanketing its sector with radio energy. This results in a stronger signal for devices far away from the base station and thus in a better throughput. This is what is known as beamforming.
And then there’s MIMO, which is already used for a decade now in LTE to send several data streams simultaneously to a single user by using cross-polarization and refraction at buildings that generate different signal paths. Active Antenna Systems extend the scheme by being able to send streams to several independent devices simultaneously that are located in different places in the coverage area. The paper mentions that 8 simultaneous streams to devices close-by is the optimum. So 4 devices could each receive 2 streams, as is typically the case today in LTE, or 2 devices could receive 3 streams and one device could receive 2 streams, etc. etc. This would result in a significant increase in overall cell capacity but only works when several devices have data waiting at the same time and are close to the base station. They have to be close because the output power of the base station has to be split. If there are 8 simultaneous transmission layers, each layer only receives 1/8 of the transmission power. This is also the case for single user MIMO in LTE today but since only 2 or 4 streams are used, it is not as significant.
How Do The Antennas Look Like?
So how do antennas look like for MU-MIMO and Beamforming? Again, the Ericsson paper has interesting drawings that show that an AAS has 64 antenna elements that are arranged in an 8×8 matrix. Each antenna element has two crossed antennas for cross polarization. In other words there are 128 antennas in the array. The paper then goes on and explains that the 128 antennas are then grouped in 2, 4 or 8 chains which leaves 64, 32 or 16 individually addressable antenna chains in the matrix. A higher number of chains results in more focused radio beams which results in a more fine grained horizontal and vertical beam forming. The downside is that an increasing number of chains requires an increasing amount of hardware and processing power that is not negligible.
Which Array Type For Which Environment
The paper then goes on to say that in practice, a 64 chain antenna array is most suitable for dense urban environments with skyscrapers that result in a 3 dimensional distribution of devices. For urban areas with a high population density but no high-rise buildings, 32 chain antenna arrays seem to be the best compromise as there is little vertical distribution of devices. And for rural areas with a low subscriber density, an 8 to 16 chain antenna array configuration seems to be ideal to extend range. But in any case the antenna array itself always has 64 (x2) antenna elements, its just a question of how many of them are bundled into separate chains.
And finally, the paper also discusses different methods of channel feedback information. Feedback is required as beamforming and multiple simultaneous transmission streams in different directions only work when the base station actually knows where to direct the different parts of the signal energy. For MU-MIMO with close by subscribers, the AAS can use reciprocity based feedback, i.e. it analyzes the device’s uplink transmissions.
For beamforming and far away devices this scheme seems to work less well and in such scenarios the AAS requires feedback information of the UEs via the control channel. In other words, the devices themselves have to analyze the channel based on the signals they receive and then let the AAS know what they have received.
Lots of maths going into AAS obviously and I’m very much looking forward to see how big the benefits are over today’s traditional 2×2 or 4×4 passive antenna systems.