无线通信设计的下一个重要应用

If you are a wireless systems designer, MIMO is in your future. It will change the design paradigm for virtually all wireless technologies from cell phones to broadband. Wherever higher data rates or more robust connections are important --that is, just about everywhere--MIMO will be the technology of choice.

Multiple input, multiple output, as the name implies, broadcasts and receives using multiple radios and antennas. This technology is not simply a paradigm shift; it inverts conventional wireless design principles: that multipath transmission is undesirable and should be eliminated.

This radical departure delivers a radical result: double the physical layer (PHY) data rate of a conventional one-radio, one-receiver system.

Improved performance is not traded for lower robustness. It is accompanied by an 8-dB gain in receive sensitivity as well.

In a two-by-two MIMO configuration, for example, the first step in preparing the source data stream for transmission is to divide the source into two streams. Each stream is broadcast at the same time and over the same frequency.

In a multicast environment, the receiving antennas hear what could be described as an RF tower of Babel. But engineers using high-performance computing hardware have leveraged the unique characteristics of multicast to unscramble the incoming data and reconstruct the source data stream.

Though MIMO is being used in proprietary point-to-point broadband applications by companies like Orthogon Systems, its first widespread deployment began in 2005 in wireless LAN (IEEE 802.11a/b/g) systems fielded by several systems-level OEMs such as Linksys and Belkin -- and powered by Airgo Networks’ MIMO technology. The IEEE 802.11a/g specifications are, of course, mute on MIMO.

But the new system-level products were made to play nice with IEEE 802.11 a/b/g by conforming to 802.11 MAC layer specifications. They don’t get the double-data-rate advantage when talking to legacy a/b/g systems, but at least they understand each other.

MIMO gains its advantage primarily in the physical layer. Once the technology was demonstrated in working products, it was quickly understood that optimizing the 802.11 MAC and PHY specifications for MIMO performance and interoperability between MIMO-based systems was the wave of the future. An IEEE 802.11n was convened to start the ball rolling.

Standards battles
Those tuned in to the 802.11n discussions throughout 2005 witnessed an intense and controversial standards process. Early in 2006, an agreement was reached within the IEEE 802.11n task group. The standard started working its way through the remaining IEEE machinery and is expected to be ratified by the middle of 2007.

The term "spatial diversity" is central to MIMO and means only that multiple transmit or receive antennas are used and that they are separated in space. Data is transmitted over multiple antennas through what is referred to as a "MIMO channel" to multiple receive antennas. If the antennas in the transmit array and the antennas in the receive array are spaced far enough apart, multipath signals between them will fluctuate or fade in an independent manner. It is this characteristic that allows the data to be unscrambled at the receive end. Typically, a training sequence is placed in the preamble at the beginning of each packet to discover how the multipath signals are cross-coupling.

Using a space-time code, the transmitted data is encoded to make use of spatial diversity and to allow digital processing hardware--typically a state machine--at the receiver to extract the underlying data. There are several possible coding schemes. The choice for a particular MIMO system depends on (1) desired performance, (2) the acceptable level of computational complexity in the receiver’s signal processing subsystem and (3) the level of a priori knowledge of the transmission channel.

Some schemes, referred to as space-time diversity codes, optimize for "diversity order"--that is, the performance gain that can be obtained through the number of decorrelated spatial branches are achieved in the MIMO channel. Another popular coding option--spatial multiplexing--optimizes for channel capacity. Coding schemes can be used in combination to obtain the benefits offered by each.

An information source can be scrambled and encoded with forward-error correction redundancy. These bits are in- terleaved to randomize their transmission order. Consecutive coded bits are randomized over the different antennas and orthogonal frequency division multiplexing tones. Each OFDM modulator feeds identical transmit chains and antennas.

The initial version of the IEEE 802.11n draft standard (Version 0.1) has many iterations ahead of it, but its broad outlines can be described. Spatial diversity (the technology described above) is mandatory.

The most exciting advantage of spatial diversity is better data rates. A typical 40-MHz channel as described in 802.11, for example, has a channel capacity of 150 Mbits/second in the physical layer. Using a two-by-two MIMO system, this rises to 300 Mbits/s, enough to broadcast HDTV in a fairly large home--and, because multipath is leveraged--through multiple walls. A four-by-four implementation would deliver 600 Mbits/s. The overhead in the MAC layer and other factors can reduce the theoretical data rate by about 40 percent.

Although MIMO appears to violate Shannon’s Law limiting a channel’s capacity, in reality it does not because the radios use different channels inside the band.

There are other ways to leverage multiple antennas, however, and the draft allows several as options. They can come in handy for specific applications because they basically trade off top-performance data rates for other attributes, like signal robustness.

Using beam forming, the antenna array focuses energy in a particular direction. The narrowness of the beam is in direct proportion to the number of antennas and the gain in the chosen direction. Beam forming does not itself increase the data rate but it does increase range. Complex decorrelation is not required at the receiver, which makes that design fairly straightforward. At the transmit end, however, the trick is to focus energy in the correct direction.

With space-time block coding, one of the problems encountered in a multipath environment where signals can null out each other, can be avoided. Space-time block coding (STBC) transmits identical signals from each antenna. Though that cuts the potential data rate in half, it improves the robustness of the signal and works well with an asymmetric system with different numbers of transmit/ receive radios. Mobile handset designers and others in latency-sensitive domains like voice appreciate STBC’s value because 300- or 450-Mbit/s physical layer data rates are not as important as clear transmissions.

Spatial spreading has beam forming results but unlike beam forming it broadcasts in all directions simultaneously. Because it beam-forms all the OFDM carriers in different directions, it uses more total power. But this creates more diversity in signal paths and insures that most of the carriers are providing good signal strength.

MIMO is clearly not a technology that many system-design engineers are going to create from scratch. MIMO pioneer Airgo is a fabless IC company that provides a complete solution of chip set, reference design and software to systems houses. Atheros Communications, Broadcom, Metalink Broadband, Bandspeed, Ruckus Wireless and Ralink Technologies have also made MIMO announcements.

In addition to mixing and the options afforded in the IEEE 802.11n predraft from the task group, chip companies also have the option of fielding chip sets and reference designs that utilize various numbers of antennas and radios. Most attention so far has been focused on two-by-two systems but some vendors are extolling the virtues of three-by-three systems and combinations of spatial diversity and the three technology options explained above.

MIMO is so new that systems designers will inevitably be at least a bit bewildered as vendors tout their real and imagined advantages. Clarity can be brought to the debate by observing two best practices: (1) Finding the best MIMO system often depends on the target application, and (2) during the prestandard stages of IEEE 802.11n particularly, testing should be conducted in the environment in which the application will be used. Standard tests are not available and MIMO is particularly subject to specmanship claims.

Money issues
For the design engineer, cost is always critical. Multiple radios and signal paths cost more to implement in silicon and, judging from the prestandard implementations, a MIMO chip set is expected to cost between $6 and $15 more than IEEE 802.11a/b/g sets.

Implementations can vary in many ways but MAC efficiency is one area in which it can make a big difference. By aggregating packets assigned to the same destination, MAC efficiency can be increased to 70 percent because it eliminates the overhead associated with each packet when they are sent individually.

Aggregate exchange sequences are made possible with a protocol that acknowledges aggregated MAC protocol data units (A-MPDU) with a single block acknowledgement instead of multiple signals. This protocol effectively eliminates the need to initiate a new transfer for every MPDU and increases the efficient throughput.

A system using a maximum PHY rate of 216 Mbits/s, a 1,000-byte packet length and 32-MPDU aggregation can reach an effective throughput of 178 Mbits/s without bit error rate and 164 Mbits/s with a BER of 1 x 10^-5.

Since MIMO is largely a PHY layer technology, it can be--and will be--adopted across almost the entire range of wireless communications. The added cost of implementing multiple radios and signal paths, of course, makes it unattractive to systems that don’t need more data bandwidth.

The cellular world is already looking at MIMO in anticipation of the convergence of cellular and Wi-Fi in the middle of next year. This will make multimedia on cell phones commonplace and reliable.

Multiantenna handsets are being designed and so are video routers. The evolution of MIMO into cellular base stations is expected to boost capacity by 4x to 7x with fewer dropped calls and better coverage at the cell’s edges.

Similarly, WiBro is expected to add MIMO modes to its evolving specification. Others are sure to follow and MIMO looks like the wireless communications industry’s best bet to make good on its "faster-than-wired" promises.

Suggested reading. The body of MIMO technical literature is expanding rapidly. MIMO articles expressly for design engineers can be found at http://www.wirelessnetdesignline.com/techlibrary/