Mobility Testing for Beamforming Base Stations
by Kang Chen, Randy L. Oltman, and Michael McKernan, Spirent Communications
The continuing goal in today’s data-centric
wireless technology is to improve data rates and fidelity in the
most cost-efficient way. Recent advances in antenna technology
have introduced wide-scale deployment of antenna arrays used in
multiple-in,
multiple-out (MIMO) configurations.
Next-generation wireless technologies such as
WiMAX, Long-Term Evolution (LTE), and Ultra Mobile Broadband
(UMB) all make use of multiple-antenna-array technologies. Now
they also can utilize beamforming, a more advanced technique.
Beamforming is a signal-processing technique
that requires estimating channel information and adaptively
shaping beams to enhance some signals and suppress others. To
implement beamforming, three basic techniques are used: adaptive
beam steering, maximum ratio transmission (MRT), and Eigen
beamforming (EBF).
Adaptive beam steering can be implemented
when the mobile device has a single antenna; MRT and EBF are
MIMO beamforming techniques. All three methods can dynamically
adjust the relative amplitude and phase, the antenna weights, on
each element of the base-station antenna array. Table 1
compares and contrasts the three approaches.
Table 1. Beamforming Techniques
The result is that lobes of constructive
interference or beams can be created and steered on the
transmitted link. On the received link, beams can be created and
aimed toward desired users while nulls can be steered toward
sources of unwanted interference. Figure 1 shows the
basic concept of a beamforming system.
Figure 1. Adaptive Beamforming System
Adaptive Beam Steering
In single-user-antenna beamforming, the
mobile device periodically sends a channel sounding signal. Each
base-station antenna element receives this signal at a slightly
different phase offset and amplitude so that algorithms can
estimate direction-of-arrival (DoA) for each user based on a
unique spatial signature.
The system then adjusts its excitation to the
array by carefully controlling the antenna weights. The
algorithms used to calculate antenna weights are called adaptive
antenna algorithms, and the technique is called adaptive beam
steering.
The accuracy of DoA estimation is critically
important but can be impaired when reflective scatterers near
the user enlarge the signal’s angular spread (AS). Signals
scrambled by scatterers may not add coherently at the user’s
antenna so beamforming gain is significantly weaker. This makes
single-user antenna beamforming most suitable for line-of-sight
(LOS) environments with small AS.
MIMO Beamforming
In addition to existing open-loop MIMO, MIMO
beamforming or closed-loop MIMO is another technique using
multiple antenna arrays. The term MIMO implies that there are
multiple-antenna arrays at both the base station and the mobile
device.
Using MIMO techniques means that base
stations must track a channel characteristic called channel
correlation. This parameter, actually a complex matrix used to
describe the connection, quantifies the relationship between
each path from each transmitting antenna element to each
receiving element. As a very rough definition, high correlation
between two links in a MIMO channel means that the links will
suffer the same effects from environmental conditions.
Nonbeamforming MIMO is well suited to
channels with low correlation because an impairment on one link
may barely affect a different link. MIMO beamforming works best
on channels with high degrees of correlation.
MRT is designed to maximize the
signal-to-noise ratio (SNR) at the receiver. While adaptive beam
steering only uses directional information, MRT requires more
accurate channel data. To calculate MRT antenna weights, the
base station needs to very accurately estimate the channel
correlation coefficients.
MRT maximizes SNR and outperforms adaptive
beam steering at large AS. However, the wideband OFDM-based
signals used in WiMAX and LTE are made up of numerous
subcarriers at slightly different frequencies. Since MRT has to
compute antenna weights for each subcarrier, it is mainly
suitable to stationary or low-speed applications.
In contrast, EBF relies only on the
statistical characteristics of the channel. Here, antenna
weights are calculated for an entire band rather than individual
subcarriers.
While it is suboptimal, EBF has smaller
measurement delay and requires less frequent measurements. It
outperforms MRT in applications with high mobile velocity and
low SNR. As EBF only adapts to significant changes in the
environment, updates take place only rarely, such as when a
mobile user moves from a rural to a suburban area.
As a result, spatial beams can be created and
aimed to more accurately transmit and receive information. Just
as the code and frequency domains can be used to add apparent
gain to desired connections and mitigate the effects of
interferers, beamforming lets systems use the spatial domain for
the same purposes.
Testing Base Stations Under Mobility Conditions
Mobility testing ensures base-station
performance characteristics during operation with moving
wireless terminals. Without mobility testing, base-station
design-verification testing (DVT) inadvertently adds a bias
toward unrealistic static receivers.
Specialized pieces of test equipment called
channel emulators are required since an emulated channel is more
controllable and more repeatable than field-testing conditions.
As mobility testing is applied to beamforming
base stations, the industry has discovered that new test
requirements are needed to maximize coverage potential and
minimize operational costs. Early investigation also points out
critical parameters in both the system under test and the test
equipment being used.
Because beamforming is a function of the
physical radio link, the best approach to developing a test plan
is to use a bottom-up methodology. At this level, there are two
types of tests that serve different purposes but have equal
importance:
• Functionality verification, which ensures
that the beamforming system works as fundamentally designed.
This kind of testing is an effective base line and can be
performed under artificially favorable conditions.
• Performance evaluation, which is intended
to evaluate and quantify performance. It helps engineers
understand how well a system works under realistic or adverse
conditions.
The goal may differ depending on the tester’s
field of interest. For example, network operators may use
performance evaluation tests to plan network deployment and
quantify the variability in performance of a single link in the
system chain. Base-station manufacturers may want to plan
feature roadmaps or identify points of product differentiation
for marketing campaigns. System developers need to evaluate
trade-offs between system complexity and performance. These
separate groups, each with different goals, all require the same
kind of information to be effective.
Location and tracking become much more
difficult under mobility conditions. Accordingly, one key
challenge in the development of beamforming base stations is to
develop effective, efficient algorithms for DoA estimation and
adaptive beamforming. Most systems work well in static
scenarios, but performance in realistically dynamic scenarios
differentiates well-engineered systems from mediocre ones.
These technical differences directly affect
both the base-station and the operator markets. Base-station
manufacturers that deliver performance under extreme conditions
can adopt premium pricing and still provide economic advantages
via increased range to their customers. For network operators,
delivering differentiated performance under dynamic conditions
leads to word-of-mouth marketing which directly translates to
lower churn and increased subscriber adds.
These challenges in development lead to
challenges in testing. Thorough base-station testing must stress
these critical algorithms in the contexts of adaptability to
subscriber motion, change of channel conditions, and multi-user
processing.
Adaptability to Motion
A subscriber motion relative to the antenna
array results in a variation in both DoA and Doppler shift.
After the DoA is estimated, the base station must compute
antenna weights and apply them to the antenna array to steer the
beam. Again, this is something of a challenge in static systems,
but a dynamic system adds the classic feedback-control-system
problem of having to accomplish this within a workable
timeframe.
If DoA is not estimated accurately or the
beam is not able to track the moving subscriber, beamforming
gain cannot be optimized. In the worst case, the communications
link will be lost because the subscriber is moving faster than
the array can track or react.
The impact of DoA error on beamforming gain
is significant. In
Figure2a, a radiation pattern of an eight-antenna
uniform linear array is shown with a DoA error of ε degrees.
Figure 2b shows that an error of 8 degrees results in 10-dB
loss of beamforming gain and that an error of 14 degrees could
cause the loss of the link.
Figure 2. Phase Error and Its Impact on Beamforming Gain
Adaptability to Channel
Mobility implies not only the change of
geometric parameters, but also a change in propagation
environment. The degree of randomness in DoA is a function of
both the subscriber’s motion and the environmental conditions.
Motion causes variability in reflections that is much more
complex than the variability in LOS signals. As
a result, the estimation of DoA at the base station becomes much
more difficult compared with the stationary case.
In an open environment such as a rural area,
there are few scatterers near the user. As a result, the uplink
signal arrives at the antenna array with small AS. In addition,
the channel exhibits very high path-to-path correlation with few
variations so it is relatively easy to deliver coherent signals
to the user. DoA estimation is relatively easy and accurate
using adaptive beam steering.
As the user moves into an environment with
more numerous and significant scatterers, the uplink signal
arrives at the base station with larger AS. If the user is
moving relatively slowly, the channel becomes a candidate to use
MRT. If either condition is not met, then EBF can be used to
deliver even higher system gain than MRT.
To reap the optimal beamforming gain, it is
critical for beamforming
base stations to be aware of any change in channel
characteristics and use the appropriate beamforming scheme.
Testing the beamforming system’s adaptability to the channel is
an integral part of understanding the base-station’s performance
characteristics.
Multi-User Scenarios
One goal of beamforming is to increase system
capacity. As a result, a suitable plan must include testing
system performance in multi-user scenarios, which requires more
complicated signal processing and faster processing speed than
single-user cases.
Two basic multi-user beamforming tests are
depicted in Figure 3. In the first case, both mobiles are
desired users that should see enhanced signals from the base
station. The base station must locate and track both mobiles and
compute antenna weights as before, but now the base station must
form a radiation pattern with two steered beams.
Figure 3. Adaptive Beamforming System Multi-User Scenarios
The second case creates a steered beam for
the user but adds a steered null. This tests the system’s
capability to suppress interferers while continuing to enhance
the signal to the desired user.
Adaptability to both motion and channel must
be accounted for and stressed in test plans for beamforming base
stations. Failure to address these topics will undoubtedly cause
issues upon deployment.
Considerations in Test Systems
Test equipment must never inadvertently
affect testing, and the testing of beamforming base stations
under mobility conditions imposes even more stringent
requirements on the equipment being used. Two critical areas of
concern are the RF performance and the dynamic control of
channel parameters.
RF Performance
Amplitude balance among RF channels, phase
calibration accuracy, and stability are of primary importance to
the validity of beamforming performance tests and even more
critical in beamforming testing than in testing traditional
wireless technologies.
Beamforming base stations use sophisticated
signal processing algorithms to locate users and form radiation
patterns. The effectiveness of these algorithms relies on
channel reciprocity or symmetry between the uplink and downlink
channels for each transmit/receive antenna pair, the phase
difference between antenna elements, and signal amplitude. If
test equipment introduces significant error in any of these
areas, the equipment can seriously and misleadingly degrade the
performance of good algorithms.
Any good channel emulator will have a phase
calibration routine. Before performing calibration, the system
should be powered up and allowed to stabilize for a defined
period of time. This is good practice whenever a channel
emulator is used, but it is absolutely critical when testing
mobility scenarios with beamforming.
A less obvious but more insidious effect is
phase drift in the equipment. Phase stability is a function of
time, temperature, and humidity, but it can be controlled with
proper test equipment design. Certain techniques and choices
made in the mechanical design of a channel emulator will affect
how the emulator reacts to environmental conditions. An RF
emulator can be electrically designed to counteract phase drift.
At a basic level, amplitude imbalance is a
function of the test equipment’s overall output accuracy.
However, the test equipment should be designed so that accuracy
between RF paths is an order of magnitude better than spec since
the slightest error here also can have a significant effect on
results.
Dynamic Channel Control
Mobility testing with beamforming implies
dynamic control of the channel emulation to test the system’s
adaptability to motion and channel. Two aspects of dynamic
channel emulation must be addressed to test the beamforming
base-station’s performance under mobility conditions: dynamic
change of DoA, which simulates the movement of the user, and the
dynamic simulation of channel correlation to mimic the effect of
environmental changes on propagation.
Dynamic DoA Simulation
To simulate DoA changes in a channel
emulator, the phase of each radio link connecting each mobile
antenna element to each base-station antenna element must be
carefully calculated and accurately adjusted. Accuracy of phase
control is critical, but so is the speed at which the angle can
be adjusted. Usually, the transition of the phase change on both
uplink and downlink temporarily breaks channel reciprocity, but
a channel emulator must minimize the time spent applying phase
changes.
Typically, these phase changes are applied
sequentially, but a better practice is to change all of them
simultaneously. For example, the Spirent SR5500 Wireless Channel
Emulator uses a proprietary approach to realize simultaneous
phase changes on all radio links. Not only is transition time
minimized, but it also is independent of the number of antenna
elements.
Dynamic Channel Correlation
Simulation
To test the beamforming system’s adaptability
to the channel, the channel emulator must simulate a change of
channel characteristics. Transition from small AS to large AS
translates into a change in the channel correlation matrix. As
the user moves from a LOS environment to an area with scatterers,
the channel correlation matrix varies from full correlation to
high correlation; the larger the AS, the lower the channel
correlation.
Either MRT or EBF beamforming can be tested
easily under static conditions, and the requirement to test
under changing conditions is an obvious one. It also is
necessary to test under conditions where the environment forces
changes in the technique being used. Test plans should include
forcing the system to change from MRT to EBF and back.
MRT is a better beamforming scheme for a
mobile device moving at low speed and using channels with high
correlation. As a subscriber accelerates, MRT gives way to EBF.
While using EBF, variation in the correlation matrix tests the
EBF adaptability to the channel. Consequently, the channel
emulator being used should control dynamic channel correlation
simulation in conjunction with real-time mobile velocity
control.
Summary
Beamforming has the potential to add great
value to 3.5G and 4G offerings being made by base-station
manufacturers and network operators. Just as MIMO offered a
quantum step forward over existing technologies, beamforming can
add another leap over MIMO in terms of increased data rates,
data fidelity, enhanced range, and more efficient resource
usage.
With this new technology come new
predeployment pitfalls to be avoided. Mobility testing is
important in all wireless technologies, but beamforming adds new
fine points to be considered when creating mobility test plans
and choosing channel emulation equipment.
About the Authors
Kang Chen is a senior applications specialist
at Spirent Communications. Prior to joining the company in 2007,
Mr. Chen held senior engineering positions at Agilent
Technologies and Alcatel. He earned a BEng from Chongqing
University of Posts and Telecom, and an M.S.E.E. from Rutgers
University where he researched MIMO and cooperative
communications. e-mail: Kang.Chen@spirent.com
Randy L. Oltman is the product segment
director for channel emulation products at Spirent
Communications. He joined Spirent in 1997 and has led numerous
wireless product efforts. Mr. Oltman earned a B.S.E.E. from
Rensselaer Polytechnic University and M.S.E.E. and M.B.A.
degrees from Rutgers University. e-mail:
Randy.Oltman@spirent.com
Michael McKernan is a product marketing
manager at Spirent Communications. Before coming to Spirent in
2000, Mr. McKernan spent many years in telecom and
communications engineering. He has a B.S.E.E. from NJIT and an
M.B.A. from Rutgers University. e-mail:
Mike.McKernan@spirent.com
Spirent Communications, Performance Analysis
Wireless Division, 541 Industrial Way West, Eatontown, NJ 07724,
732-544-8700.