Wireless operators are increasing their focus on data and multimediaservices to drive revenue growth. This is creating demand forsubstantially improved radio equipment performance, given thedifferences in subscriber economics.
Unfortunately, years of wireless innovation have left little newtechnology for performance improvements – with the exception of thedimension of space. Multi-antenna signal processing (MAS) softwareprovides more control over the spatial distribution of radio energy,yielding well-proven order-of-magnitude performance improvements. Thus,MAS is now a key part of next-generation wireless networks.
List of essentials
All MAS architectures start with the same basic principles: theyinvolve two or more antennas on one or both ends of the communicationlink (i.e. the base station and the client device), and they performsome degree of coordinated processing on the signals to and/or fromthese antennas.
From there, the key design choices to be made and variables to takeinto account include:
The number of receive and transmit radio chains in the basestation;
The number of receive and transmit radio chains on the subscriberdevice;
The categories of MAS algorithms used on the traffic channels, andon which directions and ends of the link (uplink and/or downlink,transmit and/or receive);
Approach to control channel processing;
Degree and character of coordination between PHY-layer processing(where MAS is done) and the media-access layer (scheduling);
Approach to transmit calibration;
Basic Architectural Approaches
Systems engineering of MAS-enabled networks involves making the optimumset of these basic architecture choices:
Air interface accommodation of or compatibility with MAS (not allare equally well-suited to MAS implementation);
Functional flexibility – If all users in the cell are restrictedto the same processing modes, algorithms and antenna count, or if thesystem can optimize for heterogeneous conditions with multipleapproaches;
The many attributes of the application itself -i.e. the networkoperation scenarios to be supported such as network scale, loading,spectrum allocation, service definition, subscriber behavior (includingmobility), cost and complexity constraints, especially on siting andclient devices, and greenfield vs. upgrade deployment.
All MAS architectures and processing approaches leverage somecombination of the following four gains and benefits:
Spatialdiversity. This takes advantage of inherent channel differences,where they exist, between antennas separated in space or polarization.Spatial diversity is only useful when not all of the channels are indeep fades at the same time, the likelihood of which is positivelycorrelated to the amount of scattering in the environment. Spatialdiversity link budget gains are on the order of 4dB to 7dB, typically.
Coherent orcombining gain . This uses an understanding of thecharacteristics of the radio channel (channel state information) toweight the signals to or from multiple antennas to create maximumcoherently-combined signal (or sensitivity) in the radio-spacedirection of interest.
This additionally provides some amount of passive interferencemitigation, as the more-focused energy distribution reduces inherentco-channel interference to some degree.
Approaches tapping coherent gains are often referred to as”beamforming,” despite the fact that the most effective approaches, atwork in environments with substantial multipath and scattering, do notcreate “beams” per se.
Examples are maximum ratio combining (MRC) or transmit (MRT).Combining gains are a function of the number of antennas (i.e.10log(n)) and the quality of the channel knowledge used.
Interferencemitigation. This adds calculations to the coherent combininggain baseline to send or receive the absence of energy in theradio-space direction of co-channel interferers. This is also referredto as “nulling” or active interference cancellation (AIC). AIC gainscan be large—on the order of 15dB to 30dB in practice.
Spatialmultiplexing (SM). This is an application of coherent processingto resolve two or more distinct information streams from the same radioresource or channel at the same time in two different places in radioand physical space.
These information streams can be combined at a single receivingendpoint (base station or client device) to enhance the data rate for asingle link—the WiMAX Matrix B flavor ofmultiple input, multiple output (MIMO) is an example of this—orthey can be resolved at different endpoints (e.g. different clientdevices) to enhance system capacity through higher spectrum reuse.
This is also called spatial division multiple access orSDMA. SM does not affect link budget directly, but it can have morethan double the effect on overall spectral efficiency or average clientdata rates or both, depending on how it's used.
|Figure1. MAS algorithms can be grouped according to their leverage of channelknowledge, interference mitigation and spatial multiplexing.|
Figure 1, above , providesan overview of where the most common MAS algorithms fall in thisframework of channel knowledge, AIC and SM. Note that thefirst-generation, simplistic approaches to beamforming based on poorchannel knowledge (such as angle of arrival) are not included in thefield's collective product portfolio today – they have been proven toyield non-viable price/performance results because of poor real-worldresults.
Each of these tools is best applied to particular subsets ofoperator requirements and subscriber behaviors. As examples, for a highdata-rate focus, the bulk of the system's resources should be used inthe spatial multiplexing column, since that will maximize data rates.
For subscriber profiles with very high mobility, the bottom row ismost productive, since it does not rely on channel knowledge (difficultto maintain today with high quality for subscribers moving in excess of100kph) to do its work.
There is a reasonable amount of confusion in the WiMAX communitytoday over the likely performance of MIMO-enabled mobile WiMAXequipment. Since no mobile WiMAX equipment has yet been built andtested in multicell, multi-user environments (i.e. networks underload), performance characterization rests exclusively on link- andnetwork-level simulation.
Building such simulations is an inherently complex business, andtaking shortcuts (such as interference averaging) can yield materiallymisleading results. The Array- Comm research team has been developinglink- and network-level simulations of 802.16 performance for over twoyears now, working closely with our partners at Intel, KT, amongothers.
Mobile WiMAX profiles
The mobile WiMAX profiles include a large number of differentprocessing modes and architectures within the specification:
SISO. For reference, an implementation of mobile WiMAX using a single antennaon both ends of the link (single input/single output);
2×2 Profile. Two antennas each on client device and base station, using the bestapproximation of the baseline MIMO processing included in the WiMAXWave 2 certification profiles;
4×2 Enhanced. Four antenna transmit and receive on the base station, two antennasreceive (one antenna transmit) on the client device. It adds enhancedversions of both MIMO Matrix A and Matrix B algorithms that addressinterference.
|Figure2. Graph shows summary characterization of select mobile WiMAXarchitectures, based on ArrayComm link and network-level simulationwork to date.|
As Figure 2, above, shows,there is substantial difference between the baseline profile MASperformance and the performance of an approach designed to more fullyleverage all four MAS gains.
Many approaches to MAS concepts have been attempted over the past 15years. Some early trials involved large, expensive andprecisely-calibrated arrays that in the end didn't work very well, andsome involved so-called “appliqué” solutions—aftermarket add-onboxes that also yielded generally poor performance because of limitedintegration with the existing radio hardware and necessarilyunsophisticated algorithms.
|Figure3. MAS implementations are under way for WiMAX.|
Here are some general principles for MAS implementations:
Do yourhomework thoroughly. Many tools for network or economicsanalysis and performance simulation from single-antenna domains (e.g.interference averaging) yield misleading results when applied toMAS-enabled gear. Getting MAS analysis right is admittedly morecomplicated, but essential.
Thinkintegrated . The highest performance for the least marginal unitcost is obtained by integrating MAS into client and infrastructuredesigns from the outset, not adding them on after the fact. Thearchitectures outlined in Figure 3, above shows examples of how MASfits into generic WiMAX client and base-station architectures.
Considernetwork performance, not just the link . MAS modes that achieveuseful results for an individual link (e.g. the baseline form of STCMIMO in WiMAX) can fail in a multicell, multi-user context.Network-level analysis of fullyloaded systems is essential.
Use multipleapproaches. Use the right tool for the job. Operatorrequirements and subscriber behaviors vary, from one market to anotherand from one moment to another. Different MAS architectures have uniquestrengths and weaknesses in different applications—there is no single”best” approach. It is much better to include all approaches in thesystem and let environmental conditions dictate which one is used.
Anticipatedynamic, seamless use of all approaches. We have shown in ourPHS implementation, where eight different MAS algorithms are selectedon the fly for optimized performance, on a frame-by-frame anduser-by-user basis, that MAS architectures can be very dynamic systems.There are many levels of radio system control (e.g. beyond individualcells to the network level) that can be incorporated into this sort ofself-organizing optimization process.
Wireless operators face a sizeable technology challenge as theypursue growth through data and multimedia services. The wireless futurethat operators envision is an exciting one – MAS will play a key rolein making it happen.
Steven Glapa is vice president ofmarketing at ArrayCommLLC