To view a PDF version of this article, click here.
When the IEEE 802.11a wireless-LAN standard was defined, it opted for orthogonal frequency-division multiplexing with quadrature amplitude modulation for communication in the 5-GHz band. Although this scheme is inherently very robust, even in the face of multipath interference, the use of antenna diversity can augment the wireless connection. However, antenna diversity, with appropriate signal-combining at the receiver, has traditionally been seen as power-consumptive, thereby restricting its widespread use. Fortunately, advanced techniques such as dynamic diversity can optimize the power/performance trade-off and permit the technology to further penetrate into WLAN applications. But to properly implement dynamic diversity and achieve its full benefits, designers must perform a full and accurate characterization of the receive-side channel.
Through antenna diversity, a receiving station obtains multiple observations of the same signal sent off a transmitting station. The redundancy built into the multiple observations can be used to recover the transmitted signal with a higher degree of accuracy at the receiver.
While antenna diversity has many advantages, it also has the drawback of requiring more extensive signal processing, which in turn leads to increased power dissipation. Hardware overhead also increases, but a fair amount of circuits and building blocks can be shared among multiple signal paths. Also, a high level of chip integration and efficient packaging are important in keeping down real estate and bill-of-materials costs in equipment design.
In this article, we consider a power-efficient wireless system that allows a favorable trade-off between the power consumption and performance of the transceiver/modem. In particular, we are concerned with quadrature-amplitude modulation (QAM) systems in which the receive-antenna connection and the signal-processing method change dynamically and automatically between two diversity modes based on receiver-side link-quality assessments.
The IEEE 802.11a standard, which allows high-speed wireless connection within a local-area network, is based on a multicarrier, quadrature-amplitude-modulated orthogonal frequency-division multiplexing (OFDM) physical layer and a carrier-sense multiple-access media-access control (MAC) protocol. The maximum data rate is 54 Mbits/second, vs. the 11 Mbits/s allowed by the currently popular .11b standard. The latter uses spread-spectrum techniques with complementary-code-keying (CCK) modulation. While OFDM enables an improved multipath tolerance relative to the single-carrier methods such as spread spectrum or CCK, maintaining a robust connection in harsh multipath environments with severe fading is not an easy task. Because each OFDM subcarrier is subject to QAM, a QAM system can be viewed as a building block of an OFDM system.
We shall call the proposed concept “dynamic diversity.” Under this technique, when the link is deemed sufficiently reliable according to the receiver-side quality assessment, a relatively simple form of receive antenna diversity is assumed, which could be switched diversity, selection diversity or no diversity (descriptions of the different diversity modes follow).
As the link condition degrades, however, the receiver assumes a robust but more power-hungry full-diversity mode, which performs signal processing on multiple received antenna paths. When the perceived link quality improves again, the receiver changes its setting back to the simpler diversity mode. The idea is to engage in the full-diversity mode only when the link quality deteriorates to a point that prevents a given data rate.
Dynamic diversity delivers several advantages. It provides a high throughput rate without excessive overall power consumption, and it does not interfere with traditional transmitter-side link-enhancing techniques such as transmitter power control and packet retransmission. Dynamic diversity maintains backward compatibility and interoperability with currently deployed wireless communication equipment, since it is a receiver-side-only enhancement. The receiver-side link-quality assessment techniques play a key role in realizing the dynamic-diversity concept. The dynamic-mode selection can be driven by various physical- (PHY) layer as well as MAC-layer parameters that can indicate a change in the quality of the communication link.
Channel-Fading Model
Let us first consider a representative QAM system (see Figure 1 below) . The received sample is given by:
Eq.1: 
where gk represents the fading in the wireless medium, Ak is the QAM symbol and nk is the additive noise. All variables are complex-valued in general. This is a general description of QAM symbols being transmitted over a fading channel corrupted by additive noise. As such, (1) can be either a time-domain or a frequency-domain model. In an OFDM system that uses a finite number, say N , of some fixed-frequency bins, the model of (1) assumes that the (k+iN )-th symbol transmission occupies the k -th frequency bin, where i is a non-negative integer and k ranges from 1 to N ; that is, the bins are occupied in a successive manner from the first bin to the last, and then back to the first one and so on.
Now consider the channel model that results from the use of one transmit antenna and two receive antennas (see Figure 2 above) . Extension to multiple channels corresponding to multiple receive antennasand, possibly, multiple transmit antennasshould be straightforward. For standards-compliant applications such as the wireless LAN, compatibility with a basic transceiver/modem design must be ensured. For this reason, no transmit antenna diversity is assumed here. The received samples for the two antennas are given by:
Eq. 2(a): 
Eq. 2(b): 
where the double subscript is used for the channel fade parameter and noise to distinguish between the two receive paths. Also shown in Figure 2 is the bit-to-symbol mapping block, which converts a fixed number of bits into a QAM symbol, and the nonlinear processor that produces decisions on the transmitted bits by combining the two received samples in some optimal or suboptimal way. The decisions can be hard or soft, depending on the targeted complexity/performance trade-off. Various techniques exist for generating the decisions. In the context of the particular OFDM format chosen for 802.11a, the above received samples correspond to noisy observations of the complex amplitude of one of 52 subcarriers. The fade parameter g1,k represents the frequency response of the channel at a specific frequency bin and for a particular antenna.
Figure 3 shows the measured frequency response of the channel corresponding to the two antenna paths. The measurements were taken in a location on a fairly spacious floor with a combination of many dry walls and some cinder block walls inside a building with steel exterior walls. If there were no multipath delay spread, a top frequency response that's more or less flat is expected. But as shown in Fig. 3, frequency selectivity of the channel is evident.
Receive Diversity Modes
For clarity, we focus on two receive antennas when we discuss various receive-antenna diversity schemes, but extension to multiple antennas is straightforward.
Link Assessment
In a conventional indoor WLAN, the data rate of a given communications link is typically adjusted at the transmitter side based on some measure of the successful packet transmission rate. As the channel condition worsens (for example, as the receiving station moves away from the transmitting station, or the antenna orientation changes in a mobile station), the link data rate is adjusted downward, since reliable communication at the initial rate is no longer feasible.
Dynamic diversity enables a higher-link rate in more adverse channel conditions than is possible in conventional systems, while keeping the transceiver/ modem components from consuming excessive overall power. In contrast to the conventional WLAN system based on transmitter-side link-quality assessment, dynamic diversity requires a receiver-side link-quality measure.
The state transition diagram in Figure 4 illustrates one particular strategy that allows dynamic selection between full diversity and no diversity. As the link quality deteriorates, the receiver transitions from a single-antenna connection to the full-diversity mode. The transition back to a single-antenna connection is triggered by an indication of significantly improved link quality. When going back to a single-antenna configuration, a comparison of the received-signal strengths in the two antenna paths can easily lead to a preferable antenna connection. In this sense, the scheme of Fig. 4 also incorporates a “slow” form of selection diversity.
A transition between diversity modes and antenna connections is signaled by a change in the perceived quality level of the link. The transmit-side link-quality assessment is typically based on the estimated dropped-packet rate via the observation of the acknowledgement packet and the number of retries attempted. However, the link-quality assessment at the receiver side, as required for dynamic diversity, is based on any combination of three factors: SNR, modem-detection quality measure and MAC-layer link-quality measure.
The SNR can be estimated from the measured received-signal strength indicator (RSSI) and the given noise characteristics of the radio circuits. The average SNR can be combined with a measure of delay spread or multipath interference to provide an accurate link-quality estimate. In OFDM the coherence bandwidth (the reciprocal of the delay spread) of the multipath channel can be measured from the frequency-correlation function that is approximated as:
Eq.3: 
where Hk represents the frequency response of the channel in the k -th bin and N is the total number of subcarriers. The 50 percent coherence bandwidth, denoted as B50 can be defined as the width of this frequency correlation function at 50 percent of the peak. A rule-of-thumb estimate for the rms delay spread is then:
Eq.4: 
Other techniques are possible that can extract the delay-spread information from the measured frequency response of the channel. Depending on the statistical variations that exist in the given network, certain techniques may work better than others. Figure 5 shows the results of a delay-spread estimation technique based on the observation of differences in channel responses associated with neighboring subcarriers. The estimates represent quantities averaged over 100 packets. It is seen that a highly accurate estimate is possible up to the rms delay spread of 200 nanoseconds.
Another useful PHY-level parameter that can lead to a good link-quality measure is a detection-quality measure (DQM) that can be observed within the modem. For example, the detection quality is reflected in an averaged magnitude of the soft decisions captured at the Viterbi detector input. Since the functional relationship between such a DQM and the bit error rate or the packer error rate can be obtained empirically, the DQM can drive the mode selection, the antenna selection or both. No matter which method is adopted, the MAC-layer functions must always verify the Receiver Address field in the MAC header to ensure the packets are intended for the receiving station under consideration.
MAC-Level Assessment
Another class of approaches to assessing link quality is MAC-layer parameters. One method applicable to WLANs is to examine the Retry Subfields in the MAC headers of the received packets and observe the frequency of retried packets. As the frequency reaches a certain threshold, the antenna connection or the diversity mode can be changed in hopes of establishing a better link. The MAC-layer parameters are convenient, since the link quality can be assessed independent of the data rate or multipath effects.
A particularly efficient way of implementing dynamic diversity is to utilize both MAC- and PHY-layer parameters. For example, the receiver can rely on the inspection of the Retry Subfields to sense degradation in the link quality and signal a transition to the full-diversity mode. On the other hand, the reverse transition from the full-diversity mode to the simpler antenna setting can be triggered when the RSSI level increases by some prescribed amount (which could be a data-rate-dependent value).
While detailed implementations of dynamic diversity can vary widely, it's possible to summarize the simulated performance of a conceptually simple scheme relative to that of a single-antenna mode, as well as full diversity and selection diversity (Figure 6) . In the particular dynamic-diversity strategy used here, the RSSI associated with each packet is measured during the preamble period and a decision is made as to whether to keep both receive paths turned on during the reception of the data portion of the packet. When the RSSIs of both packets are below a preset threshold, the full-diversity mode is implemented. Otherwise, the path with a higher RSSI is chosen to assume a single-antenna mode.
The plots are of OFDM packet error rates vs. SNR averaged over a large number of packets. The channel is subject to a 50-ns rms multipath delay spread and the data rate is fixed at 54 Mbits/s with a packet size of 1,000 bytes. The channel is assumed to be static during a packet duration, but to change independently from one packet to the next.
Depending on the threshold level chosen, the performance of dynamic diversity falls between those of full diversity and selection diversity. Also shown is receive-mode power consumption by dynamic diversity at different SNRs. The overall receive-mode power consumption can be estimated by counting the number of packets that were received in full diversity, ignoring the power dissipation during the short preamble period.
It is assumed that the single-antenna mode and the full-diversity mode consume 1 and 1.6 watts, respectively, during packet reception.
It can be seen from the figure that dynamic diversity can achieve performance close to that of full diversity while requiring only a small rise in power consumption relative to the single-antenna mode. In general, adjusting the threshold produces a flexible trade-off between power consumption and performance.
While only the multipath capability associated with the use of multiple receive antennas has been discussed in this article, time and frequency synchronization in the OFDM PHY can also be made considerably more accurate when multiple receive paths are used properly. The front-end signal detect performance also improves with multiple receptions of the signal. Thus, the real performance benefit of employing multiple receive antennas goes well beyond the predicted by classical analyses on diversity gain. To improve bandwidth efficiency even further, multiple transmit antennas can be used. Such a system is known as multiple input, multiple output (MIMO) and with appropriate coding at the transmitter and matching signal processing at the receiver, the data-rate improvement can be large for the same transmission band.
Related Articles
- “Understanding WLAN Trade-offs”; www.commsdesign.com/story/OEG20021101S0015
- “Modeling Multipath in 802.11 Systems”; www.commsdesign.com/story/OEG20021008S0001
- “Compensation for Mixed-Signal Errors in 802.11a ZIF Receivers”; www.commsdesign.com/story/OEG20021030S0017
Acknowledgement
The authors thank members of Bermai Inc.'s PHY, MAC and system integration group for providing valuable input.
About the Authors
Jaekyun Moon () is founder and chief scientist at Bermai Inc. He is also a professor of electrical and computer engineering at the University of Minnesota. He holds a PhD in electrical and computer engineering from Carnegie-Mellon University, in Pittsburgh.
Younggyun Kim () is senior architect at Bermai and has a PhD in electrical and computer engineering from the University of Minnesota.
This looks like an interesting implementation, which employs the benefits of diversity techniques, while taking care of power requirements of low power devices. Since this is an old article, I am keen to know if such a diversity implementation is now available commercially, such as in WiFi specifications (802.11n introduced MIMO), or other implementation