At Mobile World Congress (MWC) in Barcelona last week, there were a number of companies offering improvements to massive MIMO (multiple input multiple output) technology. Massive MIMO systems use a large number of antennas—say, 64—to transmit and receive radio frequency (RF) signals in parallel.
This huge increase in the number of antennas provides a big boost to the number of data streams that such systems can handle compared to today’s base stations. As a result, massive MIMO is now considered a foundational technology for future 5G networks, which explains its strong representation at MWC.
On a massive MIMO base station, those 64 antennas are placed in a planar arrangement (such as 8 x 8 or 4 x 12) termed the “antenna array.” As a result, at any moment, there are 64 transmit signals leaving the antenna array and 64 signals coming into the antenna array.
These massive MIMO systems come with some complications. For example: When you’re transmitting 64 signals at once, how do you ensure the signals are aligned in time and phase with respect to each other?
To address these issues, some of the companies at MWC, such as Blue Danube Systems and Taoglas Antenna Solutions, unveiled a simplified approach to massive MIMO that until recently had been discussed only in academic circles. Dubbed the “hybrid approach,” their technique combines both digital and analog versions of beamforming, which is a signal processing technology used to direct the transmission of radio signals.
This hybrid approach to massive MIMO avoids some of the problems typical of massive MIMO systemsby leveraging something called “coherency.”
To understand coherency, you need to visualize how the 64 signals transmitted from an antenna array may incur different delays and phase-shifts while traveling through different physical paths (or RF chains as these paths are often called). These signals originate at the same time from a common baseband processor and travel through many different digital and analog stages of the system before they arrive at the antenna array to be transmitted.
If the time/phase alignment at the antenna array is perfect or close to perfect (few relative phase errors between the 64 signals), then the system is coherent over the entire array. Sometimes this is referrred to as coherency over the array aperture: the aperture is the actual physical dimension of the array.
A similar effect can be achieved by signals that are being transmitted from a device to a massive MIMO system, rather than being sent from the base station to a device. In this case, if the 64 received signals from the device enter at the same time into the antenna array, and if they all arrive to the baseband processor at the same time or with very little time/phase errors, the system is considered coherent over the entire array.
Mihai Banu, CTO and vice president of R&D at Blue Danube, offered a simple analogy for the concept of coherency. “Think of a large number of people throwing rocks into a lake,” said Banu. “If the rocks hit the water at different times (even close to each other in time) there will be a lot of small ripples and noise like waves in the water. However, if all rocks hit the water at the same time, a large wave or splash will occur. This is the power of coherency.”
Of course, it is possible to build hardware that can transmit or receive coherency over the entire aperture. For instance, traditional military phased arrays are coherent by design. However, they use very expensive components and tune their system in the factory to guarantee coherency in the field. These systems are way too expensive for commercial wireless systems.
When the 3rd Generation Partnership Project (3GPP) 3GPP developed the LTE standard, which relied on multiple antennas, the group recognized the difficulty and cost of using hardware to achieve coherency across an entire array.
So 3GPP came up with ways of using MIMO (not massive MIMO) to deliver LTE from multiple antennas, but without RF coherency over the entire aperture. This worked well when the antenna array contained only two to four antennas, but it is not ideal when there are a large number of antennas in the array, such is in massive MIMO.
To understand why, it’s important to know that 3GPP developed two methods for using MIMO for LTE without coherency. One was based on pilot signals (signals that the system uses to establish a path for a beam to travel) sent from the base station. The second was based on channel reciprocity and pilot signals sent from the user’s device, or smartphone.
The first 3GPP method based on pilot signals sent by the base station is currently used in LTE systems with just a few antennas (up to eight). However, for massive MIMO this is not practical.
The impracticality stems from the fact that the standard assigns a unique pilot signal to each antenna. This means you have to be able to put a separate pilot in every antenna signal that needs to be transmitted.
The smartphone receives all the pilots from all base station antennas and compares them in the digital domain. Then it sees how each pilot changed in phase and magnitude with respect to each other. This change represents the entire “channel” from the base station to the smartphone.
Knowing how the pilots changed, the smartphone assumes that the actual data payload sent next to the pilots changed the same way and it sends a signal back to the base station to correct the phases and magnitudes of the signals it sends.
The reason why this first 3GPP method needs one radio per antenna is so it can produce and process pilots. This works fine for two to four antenna systems because the same radios would be necessary anyway to process the payload, which may be transmitted in two to four independent streams of data (called “layers” in 3GPP).
However, a massive MIMO arragnement would require 58 to 60 additional radios just to generate pilots. This is a lot of unnecessary hardware and leads to a lot of power dissipation.
The second 3GPP method uses pilots sent by the device rather than the base station. So, each smartphone is assigned a unique pilot. The base station receives these pilots with all its 64 antennas and can estimate all channels from the smartphone to the base station via the various radio chains.
In this second method, like in the first method, the only reason for having so many radios is again to be able to process pilots. So, this method, while better for massive MIMO than the first, it is still a brute force approach—throwing radios and power at the problem, all due to lack of coherency.
In Blue Danube’s system, they only build as many analog radio chains as there are layers expected—four to eight. At this point, they can apply the 3GPP methods (first or second) on top of this only for the digital signals they generate to achieve coherency, without the specialized hardware and materials found on military arrays.
“In the case of our system, having coherency over the entire aperture, gives us the capability of shaping the RF energy as we like it and not rely on pilots for this operation,” said Banu. “This is a huge simplification without losing anything in performance.”
A good analogy for this difference between digital and analog beamforming is illustrated by the early days of the digital camera, according to Banu.
“You may remember how with the original digital cameras, everything was digitized massively, and then when you tried to zoom or try motion compensation, they would lose resolution,” said Banu. To fix this problem, camera manufacturers added optical zooming and stabilization features to the device itself.
“So, the hybrid approach of beamforming is exactly like that. It’s like we’re placing an analog lens in front of the digitization part, which kind of gives better signals for the digitizers—exactly like optical zooming does in digital cameras,” he says. “It’s a mixed signal approach.”
Dexter Johnson is a contributing editor at IEEE Spectrum, with a focus on nanotechnology.