Fig 1. In the ideal home-network implementation, logical connections between devices are located in multiple locations throughout the home.
Fig 2. This 4x4 MIMO system multiplexes four unequally modulated spatial streams. Streams 1 and 3 are modulated with QPSK (per subcarrier-OFDM); streams 2 and 4 are modulated with 64QAM (per subcarrier-OFDM).
Fig 3. Independent test results show the percentage of coverage in terms of packet error rate (PER) versus the overall throughput of the wireless device. It’s clear that beamforming (BF) holds a significant advantage over the other methods.
When deploying home networks, one question always seems to surface first: What wireless technology is “good enough” to support multiple streams of HD video—at the lowest cost? Although several wired options exist for the delivery of compressed video in home networks, including Multimedia over Coax (MoCA), Home Phoneline Networking Alliance (HPNA), and powerline communications (PLC), they don’t offer the same coverage and cost-effectiveness as wireless technology.
Among wireless technologies, emerging 60-GHz solutions lack the necessary reach and are fairly expensive. Thus, designers are left with 802.11n, particularly the 4x4 multiple-input multiple-output (MIMO) version, as the only solution that easily and economically spans every corner of the home.
Potentially, 802.11n can replace wired technologies to significantly extend home-networking coverage for just about every conceivable electronic consumer product, from TVs, residential gateways (RGWs), and set-top boxes (STBs) to game consoles, Internet Protocol (IP) phones, and mobile devices (Fig. 1).
Defining “Good Enough” For HDTV
“Viewing satisfaction” is a fairly complex and involved process to determine, taking into account both objective and subjective components. One specific, objective method that can help quantify and evaluate connection-quality scenarios has to do with measurements of long-term averages for packet loss ratio, also known as packet error rate (PER).
Consumers expect high-quality images at all times (e.g., watching TV for a long period of time without any glitches). This expectation translates into very low PER on the order of 0.01% to 0.001%—hence establishing the base parameter for watching quality HDTV.
Subjective components of viewing satisfaction tend to involve time-dependent, event-driven statistical properties of the system. These errors are fairly hard to simulate or measure directly. However, their combined effects can be seen on a TV screen due to the presence of residual errors that appear for a short time period—only long enough for the eye to see them.
Many parameters could influence subjective testing results. On the wireless physical-layer (PHY) side, however, they’re limited to just two: linear dynamic range of the system and signal-to-noise ratio (SNR) margin. Both are affected when random interfering events occurring in the air bombard the system.
The data rate (also described as total capacity) achievable in a home environment is a product of time and bandwidth per user (i.e., how many megabits per second and for how long). It continues to climb with the move toward higher-quality flat-panel displays and higher-bandwidth and higher-resolution video content. The growth in TV size along with enhanced contrast ratio and color resolutions is driving demand for more bandwidth. Moreover, the transfer of quality HD content requires higher dynamic range.
Consequently, engineers turn to compression technologies such as H.264 to push HD content over the air. A few years ago, most flat panels displayed video content at 1080i-30 resolution, but that’s now shifting toward 1080P-60 resolution. Meanwhile, frame rates of 120/240 frames/s are becoming more commonplace. Couple that with the move to 3D TV, and the cry for more bandwidth becomes louder.
Viewing very high-quality HD content, such as the National Football League’s Super Bowl, on a 72-in. TV screen might require 30 Mbits/s or more of compressed H.264 performance. On that front, Quantenna’s labs demonstrated that using very high-speed H.264 encoders (with an average transfer rate of about 50 Mbits/s) enables compressed video to be displayed on a 60-in. flat-panel TV at a level of quality indistinguishable from Blu-ray uncompressed video.
Video gaming with HD content and low-latency system response creates even greater bandwidth requirements. With the advent of wireless technology using low-latency HD video encoding, all gaming gear can be stowed out of sight, in a central location, far from the flat-panel displays where the games are played. Today’s low-latency (sub-10 ms) video encoding/decoding technology requires that wireless technology deliver even higher bandwidth for low-latency HD video—as much as 60 Mbits/s for a fairly large size screen.
In general, two primary issues drive bandwidth requirements in the home: the amount of compressed video needed per HDTV screen, and the total number of TVs and wireless game consoles. Based on the aforementioned analysis, it’s reasonable to assume a 30-Mbit/s video-encoding rate is needed for each TV and about 60 Mbits/s for wireless gaming.
Of course, some HD compressed sources may have a much higher peak data rate than the average data rate. These rates are likely to be required across as many as three or four HDTVs as well as a variety of gaming gear.
It would be reasonable for service and content providers to allocate a sustained 120 Mbits/s of compressed HD video rates in the downstream direction from various sources, such as STBs, RGWs, and network-attached storage (NAS) boxes. This sustained bandwidth might require an even greater peak data rate when planning ahead for total home network capacity.
MIMO Order For Next-Gen Wi-Fi
Delivering on consumer expectations for wireless-HD video performance and reliability also requires the right MIMO architecture. Moreover, enhancements are needed to optimize connection strength and reliability for consumer entertainment and the viewing experience.
One of the first key considerations is antenna performance, which includes the optimal number of antennas to use for a given application. Since a wireless MIMO channel is a multipath system, multiple reflections create many paths between all of the antennas.
On the transmit side, the transmit signals from any single antenna can be deflected and diffracted into many radio-wave branches as each signal moves forward. On the receive side, any single antenna can receive many of these radio wave branches, with each antenna acting as an independent “observer” that performs independent sampling, improving (SNR).
The greater the number of independent paths, the higher the wireless channel rank. And the higher the channel rank, the more the antennas can leverage independent observation and sampling opportunities, and thus improve SNR.
Each independent path can carry one spatial stream. As the number of spatial streams grows, at least one spatial stream can be assigned per antenna. Each spatial stream in the 802.11n protocol will carry up to 150 Mbits/s in a 40-MHz bandwidth. However, not every spatial stream has equal capacity, or is truly independent, so extra antennas may not always yield additional throughput.
Still, using extra antennas (i.e., two antennas per spatial stream) does benefit the signal reliability by an average factor of two, and even more in some cases. In addition, applying the maximal ratio combining (MRC) algorithm at the receiver can improve reliability because it allows all four antennas to recover the two spatial streams, which optimizes the received SNR.
It’s critical to achieve sufficient SNR to deliver the maximum raw data rate of 150 Mbits/s per spatial stream. Without adequate SNR, the receiver can’t decode an incoming signal, where each spatial stream has a 64-state quadrature-amplitude-modulation (64QAM) index. When the available SNR at the receiver is low, it’s necessary to support a lower modulation index (i.e., 16 QAM or quadrature phase-shift keying, or QPSK).
As discussed earlier, on-time packet delivery is very important for IP television (IPTV) reliability. Consequently, the User Datagram Protocol/Transmission Control Protocol (UDP/TCP) PER must be very low—generally in the range of 0.01% to 0.001%. This contrasts with Web surfing, in which the UDP/TCP is set to about 1% PER. Combining all of these SNR and PER factors ultimately determines the data rate per spatial stream for a given application.
Higher-order MIMO systems hold a major advantage over their lower-order brethren in that they can increase the number of antennas per spatial stream. In addition to committing two antennas to each spatial stream, systems can employ other reliability-enhancing techniques. One of the most advanced techniques is to allow unequal modulation for each spatial stream (Fig. 2).
Reliability also can be enhanced via beamforming, which is a rather straightforward technique. A MIMO receiver attempts to estimate the channel matrix, or the channel between the transmitter and the receiver. This tells the transmitter how to pre-compensate on a tone-by-tone basis for the best possible SNR.
With dynamic beamforming, the MIMO receiver and transmitter can work together to estimate the adverse effects of any objects that would block or deflect the beam. Those effects are then mitigated and/or preempted by redirecting the beams from each of the transmitting antennas.
Generally, correlated MIMO channels possess fewer degrees of freedom relative to ideal, fully scattered channels. For this reason, as the SNR decreases so does the number of spatial streams, which reduces the multiplexing gain of the MIMO system.
In addition, as the distance increases and the SNR decreases, the number of spatial streams also decreases. At sufficiently long distance and with enough attenuation, all versions of 4x4 MIMO systems will operate in one or two spatial streams.
PER and rate/reach curves are generated from long-term averages. Unfortunately, these analyses fall short when analyzing video-distribution quality. In home-networking video distribution, very short-term averages must be analyzed to eliminate various image artifacts. The averages are determined by the HD frame rate (i.e., 60 frames/s) of the video and the compression decoder depth (i.e., 5 to 100 ms).
Basically, this means that channel conditions (channel matrix) have a broad statistical distribution. As a result, based on the characteristics of the wireless channel, any of a wide range of channel situations is possible, with a corresponding impact on throughput.
Simulation of rate/reach curves is another accuracy-challenged analysis method. That’s because the results are upper-bounded due to fixed assumptions about the channel. Therefore, it’s very important to consider the MIMO system’s outage probability, or the percentage of channels that can’t support a specific data rate. Ideally, outage probability is equal to PER, but with non-ideal code, it’s at the lower bound of PER.
Tests performed by Quantenna show a 31.9% outage probability for a 3x3 MIMO system supporting two streams of 64QAM. That means 31.9% of the channels would not provide “reliable” transmission for two streams of 64QAM at the specific received SNR per chain—20 dB in this case. Comparatively, a 4x4 system had an outage probability of 1%.
It’s extremely difficult to achieve full channel capacity (more accurately known as ergodic capacity) without using transmit beamforming. The transmitter must know all necessary channel state information (CSI) for two reasons: the transmitted symbol (codeword) should be spanned over all possible channel conditions (matrices) over all possible locations, and the packet duration should be longer than coherence time. If the channel changes, adaptive beamforming enables the system to adapt quickly and dynamically to avoid capacity (data rate) loss.
The MIMO system’s ability to adapt is also influenced by its beamforming update rate. An adequate beamforming update rate (between 20 and 100 ms) is essential due to environmental factors that impact channel conditions, such as people walking around the home. Parameters like the H.264 (or equivalent) decoder’s memory depth of the H.264 (or equivalent) and ability to conceal certain error conditions can influence the beamforming update rate. Support for high compression ratios requires a memory depth of 100 ms or greater.
Also, because the forward and backward channels both require accurate estimations, only explicit beamforming (which is the most reliable and accurate method) is considered. The alternative—implicit adaptive beamforming—necessitates considerable nonlinear estimation and calibration. This could cause more harm than good if it’s not accurate, even for a short time interval, since it will disrupt image quality.
From a cost perspective, PLC is the least expensive home-networking technology to deploy. It offers a good deal of ubiquity, but is less reliable than alternatives due to challenges related to the availability and uncertain quality of electrical wiring inside homes.
The remaining viable candidates for high-quality video distribution are wireless and MoCA technologies. MoCA has been unable to scale more broadly because many homes still aren’t wired with a coax network, so it’s destined to remain a niche solution.
Moving forward, it has become exceptionally clear that the only option is to do away with wires in the house. The ability to roll services out, ubiquitously, to any home anywhere, without involving field technicians, can bring significant operating-expense savings.
Choosing the right wireless solution must acknowledge that home-networking bandwidth demands in the next three to five years will hover around 100 to 120 Mbits/s. Furthermore, the selected technology must be able to achieve the most coverage in the most homes.
To substantiate these points, statistical measurements were performed by an independent company using an early second-generation Quantenna 4x4 MIMO solution with two spatial streams, over many connections, in nearly 20 homes of different sizes (Fig. 3). The data for alternative, lower-order MIMO solutions is either interpolated or obtained from published results.
If we consider 80% home coverage as a good starting point, a PER of 0.001% for watching HDTV resulted in the 2x2x2 (2x2 MIMO + two spatial streams) solution supporting 10-Mbit/s data rates, the 3x3x2 MIMO solution supporting 55-Mbit/s data rates, and the 4x4x2 MIMO solution supporting 120-Mbit/s data rates.
Clearly, then, neither 2x2 nor 3x3 MIMO offers sufficient performance for a long-term home-networking solution, and 4x4 MIMO solutions are the superior alternative. The only remaining question is the cost-effectiveness of 4x4 systems versus 3x3 systems.
Generally speaking, 3x3 systems are about 15% to 20% smaller than 4x4 systems. Over time, economies of scale are expected to bring the cost of goods (COG) delta between a 3x3 system module and 4x4 system module to below $1.
The new generation of 4x4 MIMO technology (with four spatial streams capable of unequal modulation, plus low-density parity check (LDPC) and dynamic beamforming) will provide even better coverage than alternative solutions, with expectations of exceeding 120 to 150 Mbits/s throughout 95% of homes worldwide. This will reduce overall deployment costs for carriers and consumers. The remaining, very large homes of up to 7000 square feet can use a single mesh node with frequency reuse capability to achieve coverage in nearly 100% of homes.