20 www.rfdesign.

com November 2005

Next-Generation Wireless
W
e are currently witnessing an unprecedented increase in the
demand for higher speed and better coverage of wireless
networks. To meet this enormous demand, one approach is to increase
the channel bandwidth over which radio signals are transmitted.
However, this approach is not practical because frequency spectrums
are expensive and transmitter and channel bandwidths are determined
by regulatory standards. Other ways to improve the throughput is to
use more complex modulation schemes. This, however, increases
the complexity of the radio system and, thus, the cost. This problem
requires a better solution.
In the past two years, an emerging technology known as MIMO
has been one of the most promising technologies to improve the
performance of a wireless link. MIMO refers to a radio link with
multiple transmitter antennas and receiver antennas. In wireless links,
radio signals from the transmitter travel in space, reect off objects and
reach the receiver over multiple paths. Multipaths can cause interference
and signal fading in conventional radios. MIMO takes advantage of
multipaths by multiplexing those signals with advanced DSP algorithms
to boost wireless bandwidth efciency and range. Wireless systems using
MIMO can signicantly improve the spectral efciency of a system.
For example, a wireless local area network (WLAN) system using two
transmit antennas to two receive antennas (2 x 2 MIMO) can boost
the maximum raw data rate for 802.11a and 802.11g networks from
54 Mbps to more than 100 Mbps.
MIMO orthogonal frequency-division multiplexing (MIMO-OFDM)
technology has been adopted by the IEEE 802.11 standards group
as the foundation for a high throughput amendment to multimedia
wireless delity (WiFi) applications. In addition, a consortium of chipset
developers has formed task groups such as the TGnSync, WWiSE and,
most recently, the Enhanced Wireless Consortium (EWC) are working
together to create an IEEE 802.11n specication.
Recently, a number of new products based on MIMO technology were
introduced in the market and have delivered signicant improvements
in data transfer speed and coverage area over products using standard
802.11a/b/g technology. Although currently more expensive, the cost
of these MIMO-based devices is expected to drop to levels similar to
802.11 a/b/g devices as the technology gets widely deployed, increasing
bandwidth and meeting more user expectations.
MIMO, as a new technology, poses great challenges for silicon
chipset vendors, contract manufacturers and brand owners with respect
to research and development and production test methods. This article
focuses on the physical layer issues and challenges involved with
testing MIMO devices. It aims to demystify these challenges as well
as offer readers fast, accurate, scalable, and low-cost ways to identify
impairments and help improve MIMO system performance.
How a MIMO system works
A standard 802.11a/b/g system uses one transmit antenna and one
receive antenna in a radio link as shown in Figure 1. Radio signals
from a transmitter traveling in space may reect off multiple objects
and arrive at the receiver through multiple paths. The receiver sees
the vector combination of radio signals from these paths. Due to the
phase delay difference over these paths, these signals sometimes add
up in phase and, sometimes, when they are out of phase, they cancel
each other out at the receiver. This causes the received signal strength
to uctuate constantly or fade and can signicantly degrade the data
throughput of the wireless system.
In wireless systems, radio signals from different users are typically
separated by frequency, time or code. With beam-forming technol-
ogy, also referred as smart antenna technology, each user can also be
distinguished by their physical location in space.
Wireless systems use smart antenna technology to reduce the effect
of multipath fading and to improve radio link quality and coverage.
As shown in Figure 2, smart antenna technology uses adaptive antenna
The challenges of testing MIMO
To meet the demand for higher data rates and better coverage of wireless
networks without increasing bandwidth or acquiring expensive frequency bands,
an emerging technology called multiple input, multiple output (MIMO) has
appeared. MIMO is capable of signicantly increasing wireless data throughput.
Because this technology presents technical hurdles to chipset vendors, this
article will focus on demystifying physical layer issues with MIMO and present
ways of improving MIMO performance.
By Fan Liang
Figure 1. Standard WLAN system.
Figure 2. Wireless system using beam-forming (smart antenna)
technology.
22 www.rfdesign.com November 2005
arrays that provide spatial diversity from the propagation channel and
signal-processing algorithms in order to detect the direction of the
client. A smart antenna is able to steer a transmitted beam by accurately
controlling the phase of the signal over each element of the antenna
array to the client. Another way to improve the range is to use maximum
ratio combining on the receiver side. Here, two independent receiv-
ers are used to receive the same signal, and the two received signals
are then combined using signal processing to get the desired signal.
Antenna arrays are designed using traditional metrics from antenna
theory. With beam-forming technology, a single datastream is trans-
mitted over the communication link. Smart antenna technology can
be used with existing 802.11a/b/g systems to improve performance.
The data packet is compatible with the 802.11a/b/g standard that has
the same spectral efciency.
Real MIMO systems use multiple transmitter streams and multiple
receivers. As shown in Figure 3, each transmitter sends independent
data [Tx
1
; Tx
2
; Tx
n
] from different transmit antennas simultaneously
and using the same radio channel. At the receiver end, each antenna
receives the composite signal from all transmitters represented by
[Rx
1
; Rx
2
; Rx
m
]

where m and n represent the number of receivers
and transmitters respectively. In a practical application, m and n are
typically less than 4.
The different paths may be represented mathematically as:
(1)
Rx
1
= h
11
Tx
1
+h
12
Tx
2
+.... h
1n
Tx
n
Rx
2
= h
21
Tx
1
+h
22
Tx
2
+.... h
2n
Tx
n
:
:
Rx
m
= h
m1
Tx
1
+h
m2
Tx
2
+.... h
mn
Tx
n

or, in matrix form as
(2)
[Rx] = [H][Tx].
The [H] in equation (2) represents the transfer matrix of
a MIMO channel.
In a traditional radio system, multipath signals decrease throughput
as they cause co-channel interference. On the other hand, a MIMO
system relies on this interference suppression to implement multi-
datastream detection and then separate the individual transmitted
streams. By carefully designing a MIMO packet and by using advanced
digital signal processing (DSP) techniques in the MIMO decoder, we
can recover the variously independent transmitted datastreams.
To recover the transmitted datastream [Tx] at the [Rx], the MIMO
system decoder must rst estimate the individual channel transfer
coefcient h
ij
to determine the channel transfer matrix [H] during
the MIMO preamble of the packet. Once the estimated [H] has been
produced, the transmitted datastream [Tx] can be reconstructed by
multiplying the vector [Rx] with the inverse of transfer matrix [H]
-1
.
This is represented by
(3)
[Tx] = [H]
-1
[Rx].
The process, in principle, is equivalent to solving a set of N
unknowns with N linear equations. To ensure that the channel
matrix is invertible, MIMO systems require an environment rich in
multipath.
It is important to note that unlike traditional methods of increasing
throughput by increasing bandwidth, MIMO systems can increase
throughput without increasing bandwidth. This is accomplished in a
MIMO system by exploiting the spatial dimensions and increasing the
number of signal paths between the transmitters and the receivers.
Because each independent datastream is transmitted in parallel
from separate antennas, the data throughput increases linearly with
every pair of antennas added to the MIMO system. This means that by
using a MIMO system, wireless network operators can increase their
broadband services within the currently allocated spectrum without
having to expand to more spectrums.
Challenges and solutions to testing MIMO devices
MIMO-OFDM technology brings signicant performance improve-
ments to wireless systems. However, it also brings many challenges
to product development and testing due to the OFDM modulation and
additional complexity of multiple radio architectures involved.
For the benet of lower cost and higher-efciency digital modu-
lation, a zero intermediate frequency (ZIF) radio architecture with
in-phase and quadrature (I/Q) modulation is often used in a MIMO
system. In this type of architecture, the baseband signal is split into
I, the in-phase component of the waveform, and Q, the quadrature-phase
component of the waveform. On the transmitter side, the baseband
I and Q components directly convert into radio frequency and feed
the power amplier and antenna for transmission. On the receiver
side, the RF signal goes through the I/Q demodulator and converts
directly to baseband I and Q components. Imbalances (amplitude,
phase and group delay) between I and Q signal paths will directly
affect modulation accuracy. In addition other impairments such as
carrier frequency accuracy, phase noise, local oscillator (LO) leak-
age, spurious interference, and amplier compression can affect
the performance of a MIMO system. These are the imbalances and
impairments we are most concerned with uncovering.
A MIMO packet consists of a legacy 802.11 a/g preamble eld
and a MIMO preamble eld followed by the payload data. For each
composite data packet received, the MIMO decoder must be able to
estimate the channel transfer matrix [H] during the MIMO preamble
of the packet. We assume that within the length of each MIMO
signal packet, the characteristics of the communication link between
the MIMO transmitters and receivers remain constant. The channel
transfer matrix [H] determines the link characteristic. The quality of
Figure 4. Measurement setup with LitePoint IQnxn MIMO
test system.
Figure 3. Wireless system using real MIMO technology.
24 www.rfdesign.com November 2005
the transmitted signal during the preamble has a signicant impact
on the accuracy of the channel transfer matrix [H] estimation, and thus
the quality of the MIMO system.
For the transmitters, the traditional impairments like I/Q mismatch,
group delay and group delay variation, compression and phase noise
will affect the system performance. There are many impairments that
can affect transmitter performance such as variations in the baseband
and radio-frequency integrated circuits (RFICs), component tolerance,
impedance mismatch-
ing in the transmis-
sion line, components
along the signal path,
differences in the para-
sitic capacitances and
inductances along the
printed circuit board
(PCB) traces for the I and Q signal path variations, spurious interfer-
ence, non-linear effects of the amplier, etc.
Error vector magnitude (EVM) dened as the vector difference
between the ideal error-free decision points in a signal constellation
and that of a measured signal, is a direct measure of the modulation
accuracy and overall signal quality of the transmitter. EVM captures
both amplitude error and phase error and reduces the many param-
eters that characterize distortions of a transmitted RF signal into a
single one.
Parameters such as transmit EVM, symbol constellation diagram,
transmitter power, transmitter spectral mask, and crosstalk between
transmitters are all important measurements of the transmitted signal
quality. They provide valuable information about the root cause
of impairments during the phases of MIMO chip design, product devel-
opment and design validation. Accurately measuring these parameters
becomes even more important to designers of MIMO-OFDM-based
systems due the complexity involved when multiple radios operate
in the same frequency.
To measure and analyze the above parameters, designers and test
engineers require a test instrument that combines a vector signal
analyzer (VSA), spectrum analyzer, and a power meter. Figure 4 shows
the block diagram of the IQnxn MIMO test system from LitePoint
Corporation. The system, as shown, consists of two VSAs and two
vector signal generators (VSGs). This test system is scalable and can
be expanded to include more VSAs and VSGs to enable concurrent
measurement of MIMO devices.
Figure 5 shows the measurement results of a device under test (DUT)
with MIMO radio. The data rate transmitted from the DUT measures
108 Mbps total with a 54 Mbps datastream coming from each transmit-
ter. The RF signal was sent through coax cables with 20 dB attenuators.
The IQnxn MIMO system from LitePoint measured the following key
parameters simultaneously in a single capture. Those measurements
This test system is scalable and can be expanded to
include more VSAs and VSGs to enable concurrent
measurement of MIMO devices.
26 www.rfdesign.com November 2005
provide detailed information about each individual transmitter.
The transmitter spectral mask for both channels is shown at the
top left and center window of the display. The blue trace is the power
spectrum density. The red trace denes the spectrum mask as dened
by the 802.11 standard. For each Tx channel in the MIMO system, the
power spectrum measurement needs to stay below the spectrum mask
limit in order to reduce adjacent-channel interference.
The time domain waveforms for both transmitters are shown at the
top right window.
The channel estimation results are shown at the middle right
window of the display. Channel 1 [h
11
] and channel 2 [h
22
] display
the channel atness and transmitter power balance of the OFDM
signals. Channel 3 [h
21
] and channel 4 [h
12
] display isolation or leakage
between channels.
Average EVM over each subcarrier for both channels is shown in
the middle left window of the display. Notice that both EVM measure-
ments are reasonably good over the 52 subcarriers except at the edge
of channel 2 where EVM degrades due to group delay. Group delay
imbalance between I and Q signal paths can adversely affect modula-
tion accuracy and cause constellation distortion. Such an imbalance
usually relates to the different trace lengths and dielectric constant of
the PCB layer for the baseband I and Q signals. The difference in the
I and Q signal paths inside an RFIC can also contribute to the imbal-
ance. Group delay imbalance is frequency dependent, thus, affects
each OFDM subcarrier differently. Typically, the end subcarrier will
be affected the most.
The symbol constellation diagram combines with the correspond-
ing system EVM measurement to give a good indication of signal
quality. Sharp, well-dened points in the symbol constellation
diagram represent good signal quality while distorted or smeared symbol
constellation points represent poor signal quality. Designers can
use a symbol constellation
diagram as an easy way
of qualitatively assessing
and diagnosing adverse
impacts such as I/Q im-
balance, phase noise, and
amplitude compression.
A symbol constella-
tion diagram for both
datastreams is shown in
the lower part of the dis-
play (Figure 5). Each
stream is transmitting
54 Mbps with OFDM
and with a 64-quadrature
amplitude modulation
(QAM) constellation. The
two green constellation
points representing the
four subcarriers carry
binary phase shift key-
ing (BPSK) modulated
pilot tones. These pi-
lots are used to create
a continuous series of
amplitude and phase ref-
erences throughout the
data frame. Demodulation
is then performed relative
to these pilot carriers and
this allows for signal im-
pairments to be corrected
continuously throughout
the burst. The 64 red constellation points represent 64-QAM symbol
constellation measurements taken from each of the 48 OFDM data
subcarriers and over many symbols. The constellation points of pilot
tones are relatively well dened compared to constellation points of
data subcarriers. I/Q amplitude mismatch results in the pilot tones
separating mostly along the I axis, while I/Q phase mismatch results in
the pilot tones separating mostly along the Q axis.
Phase noise affects both modulation accuracy and EVM. Phase
noise is usually introduced during frequency conversion when a
baseband signal is mixed with a local oscillator (LO) to translate to
RF frequency. The LO phase noise consists of contributions from
three main sources in a frequency synthesizer: 1) the frequency sta-
bility of the reference crystal oscillator, 2) the frequency stability of
the free-running voltage controlled oscillator (VCO) used by the
phase locked loop (PLL), and 3) the loop bandwidth and the noise
from the PLL used in the frequency synthesizer. The impact of phase
noise can be seen as a circular distortion of the signal points around
the center of the symbol constellation diagram.
At the bottom right of the display as shown in Figure 5, there is
a window with numerical results of transmitter power, EVM, carrier
frequency error, and phase noise.
Figure 6. Composite MIMO measurement setup.
Figure 5. Measurement results from a LitePoint IQnxn test system.
28 www.rfdesign.com November 2005
One of the important elements in a radio system is the RF power
amplier. To achieve maximum efciency, the RF power ampli-
er is ideally operated close to its saturation point, but not above
it. This fact directly affects product cost and quality since it requires
a trade off between power consumption and amplier bias. The gain
of the power amplier is compressed when maximum, or peak power,
exceeds the ampliers saturation point. When the amplier operates
into its saturation region, non-linearity of the amplier can lead to
many undesired effects such as harmonic distortion, intermodulation,
spectral re-growth, cross modulation, and modulation inaccuracy.
A MIMO-OFDM radio generally requires a greater degree of power
back-off from the power amplier saturation point because of its high
peak-to-average power ratio. The effect of amplier compression can
be seen in the symbol constellation diagram as constellation points are
spread to the point where decision errors are likely.
On the production line, fast, accurate, and low-cost test meth-
odologies are critical to lowering the overall cost of devices and to
enable MIMO technology to gain widespread acceptance. Transmitter
EVM testing along with transmitter power and transmitter spectrum
mask testing of each of the transmitters can be used as a pass-fail metric
for the system. Figure 6 is the block diagram of a MIMO test solution
for mass production from LitePoint Corporation. The radio signals
from two independent transmitters are combined through an RF
combiner and fed into a single VSA and analyzed by proprietary
DSP software.
Figure 7 shows the measurement results from a composite signal
on a MIMO DUT. The data rate transmitted from the DUT measures
108 Mbps total with a 54 Mbps datastream from each transmitter. Notice
that spectral mask, transmitter waveform, average EVM, frequency
error, phase noise and symbol constellation are all shown as a composite
measurement. Any defect or failure on any of the individual MIMO
channels will result in a
failure report in the com-
posite measurement.
Channel estimation
results for the individu-
al MIMO channels are
shown in the middle
right window of the
display. The graph pro-
vides information on the
channel atness of each
channel and transmitted
power balance between
channels.
These proven, fast
and accurate test meth-
odologies greatly sim-
plify MIMO system test
in a production line, guard
quality and increase test
throughput.
Conclusion
MIMO-OFDM tech-
nology offers a promis-
ing way for next-genera-
tion wireless systems to
enhance their channel
capacity and robustness
of the link. Accurate and
fast measurements in the
frequency, time and mod-
ulation domain can help greatly shorten the design cycle times and
improve overall product quality and protability. However, traditional
test systems are not designed to handle the multiple simultaneous
transmitters and receivers in a real MIMO system. Two test methods
were investigated here. The rst method proposed consisted of multiple
synchronized VSAs and VSGs capable of simultaneously measuring
all key parameters for system measurements involving multiple radios,
simplifying a complex task to a straightforward one. This rst test
method proposed provided a fast, accurate, scalable, and affordable
test solution for MIMO product development environments.
The second test method proposed involved a one-box MIMO
test solution, which used the composite measurement approach to
greatly simplify MIMO test procedures and improve test throughput
on the manufacturing oor allowing for a fast and low-cost solution
for mass production environments. The advantage of this method
is that it can be used with existing test instruments such as the
LitePoint IQex. RFD
ABOUT THE AUTHOR
Fan Liang is a senior member of the technical staff at LitePoint
Corp. Liang has extensive experience in wireless networks,
RF/microwave circuits and system design. He has worked with
numerous wireless standards and has designed multiple 802.11
test systems for WLAN applications. He earned an M.S. degree
from Xian Jiaotong University in Electrical Engineering with
a focus on wireless communications. He may be reached at
[email protected].
Figure 7. Test results from composite measurement.