Wireless voice over IP is one of the most exciting and promising uses of 802.11 networks. Wireless networks present unique challenges to voice communications and must be carefully designed to provide the level of quality and reliability that people expect from a voice infrastructure. This month, we’ll discuss the unique challenges that 802.11 networks present to voice communications and ways to overcome them.
The most important aspect of a wireless voice network is also the most basic: the underlying RF must be able to provide the required signal strength and signal-to-noise ratio to carry the data without corruption. Data traffic can tolerate a relatively large amount of corruption without degrading performance to unacceptable levels. When data packets are corrupted, they are typically retransmitted. The effect is that it takes longer to get a given block of data across the network, resulting in a net decrease in throughput. Therefore, wireless data networks are often designed with the understanding that some areas may experience sub-optimal signal strength, lower data rates, and corrupted packets. But voice traffic is highly time-sensitive. A delay in the delivery of one or more packets results in an audible drop-out in the conversation, which is unacceptable to most users, who are familiar with the high call quality of wired voice technologies.
Because of the latency-sensitive nature of voice traffic, wireless networks that will carry voice traffic must be designed to the highest standards possible. As a general guideline, they should be designed so that clients in all coverage areas can achieve the maximum data rate that the clients support. VoIP conversations don’t take up much bandwidth—64 Kbps is more than enough for a single conversation, given typical codecs and compression ratios—but designing the network to provide maximum data rates maximizes the network’s ability to cope with interference or other obstacles if they arise. When the network is designed to provide maximum data rate throughout the coverage area, the range of each AP is effectively reduced, since the longer ranges that can be achieved at lower data rates are ignored. Therefore, wireless networks that are designed for voice typically require more access points—perhaps two to three times as many—to cover the same area than if the network were designed for data.
Voice users differ from data users in that data users may be nomadic—that is, they might pick up their computer and move to a different location, but they wouldn’t require connectivity while moving—but data users typically don’t roam—that is, maintain network activity while moving. Voice users, on the other hand, commonly roam—talking on the phone while walking down the hallway, for example. Not only do voice users roam more often than data users, but they are much more sensitive to failures in the roaming architecture. Because data traffic is not highly latency-sensitive, if there was a small drop-out during roaming, any lost data would be retransmitted and the user would probably never notice, but a voice user will instantly notice any drop-out in a conversation. Therefore, a wireless voice network must be designed so that it supports robust roaming.
Roaming has traditionally been one of the less-well-implemented parts of a wireless network, at least in part because it has not been essential to the largely-nomadic data users. The original 802.11 standard didn’t even fully define protocols for roaming, leaving it up to manufacturers, who created proprietary roaming technologies that were not compatible with each other. As a result, it was often the case that clients couldn’t roam between access points unless the client and both of the APs were from the same manufacturer. Roaming standards have since been developed specifically to address the needs of voice clients.
The most important standard related to 802.11 roaming is 802.11r. 802.11r describes “fast BSS transitions”—or roaming methods that allow the station to move between APs fast enough that there is no detectable dropout in a voice conversation. 802.11r is currently targeted to be completed in late 2005 or early 2006. Until then, two other standards are relevant: 802.11f and 802.11i. 802.11f describes a basic roaming procedure that allows stations to move from AP to AP, but only fast enough for uninterrupted data communications and probably not fast enough to avoid a drop-out in a voice call. 802.11i is primarily a security standard, but it describes methods whereby access points can avoid allowing the authentication process to slow down roaming too much. Until 802.11r is completed, 802.11 voice networks should include access points that support 802.11f and clients and access points that support 802.11i, if at all possible. Alternatively, many enterprise manufacturers, such as Cisco and Symbol, have developed proprietary methods of handling roaming. These methods only work between access points and with VoIP handsets from the same vendor. Many designers of 802.11 voice networks avoid the issue of standards-based roaming by buying all of their equipment from the same vendor.
Because of the high performance requirements of a voice network, a network designed for voice should be able to easily handle data. But a converged voice and data network must include QoS methods to ensure that data traffic does not take network capacity away from voice calls. The 802.11e standard (also known as Wi-Fi MultiMedia, or WMM) defines QoS methods specifically for 802.11 networks, and 802.11 access points and client devices that will support voice services should support 802.11e/WMM. As with roaming, vendors have created proprietary QoS solutions that perform the same function as 802.11e/WMM and that require all equipment to be from the same vendor.
In summary, a converged voice/data network must first be designed with a solid RF infrastructure that can provide excellent signal strength and signal-to-noise ratio. Second, it must be able to support fast roaming, which can be accomplished via 802.11f, 802.11i, 802.11r, and/or a proprietary solution. Third, it must be able to prioritize voice and data traffic via some QoS mechanism, such as 802.11e/WMM or a proprietary solution.
Problems With Interior Support Walls and I-Beams (Part 2)
Last month’s “Ask the Expert” column started a discussion of the effects of support walls and I-beams on RF propagation. The discussion of support walls took up enough space that we put off the coverage of I-beams until this month. Because of the complexity of the topic, we’ve moved it from “Ask the Expert” to “Technology and Engineering”. If you missed last month’s topic, we suggest you go back and read it.
The challenge with I-beams is that they are made of metal and can create large shadows in the RF coverage. The shadow is larger the closer the AP is to the beam, simply because the beam takes up more degrees of arc. To make this clear, imagine holding your hand two inches from a flashlight, then holding it fifteen feet from the flashlight. The closer case blocks more of the light. A more subtle effect occurs due to diffraction. The further the AP is from the beam, the fewer degrees of bending due to diffraction are required to allow the AP to fill in behind the beam. For more information on diffraction, see Connect802’s article on diffraction (which deals specifically with RF diffraction) and this article from Wikipedia (which goes in-depth with diffraction of visible light).
The image to the left shows the same sample building from last month. This time, the cement support wall has been removed and only the I-beams are shown.
In the image to the left, an AP has been placed on the left side of one of the support beams. The red contour in the drawing above represents a signal strength roughly sufficient to get 54 Mbps data rates; the yellow contour represents a signal strength roughly sufficient to get 24 Mbps data rates, and the green contour represents a signal strength roughly sufficient to get 6 Mbps data rates. For reference, the AP has an output power of only 5 mW, which is less than a real AP would probably use, but we wanted to have relatively small coverage contours to make the effects of the wall more obvious.
Notice the deep “notch” shadow in the RF coverage that is created to the right.
The notch becomes somewhat less prominent when the AP is moved further from the support beam, but it still has a large effect on the coverage.
The notch is just as prominent when the AP is set to a higher output power (as is seen here), and is cumulative because the beams are set in a line, so each beam adds its absorption and reflection to the ones before it.
Because they create deep shadows in the coverage, it is difficult to cover an area that includes support beams. Placing the AP in the middle of the area spreads coverage more evenly, but doesn’t eliminate the shadows of the beams.
Keep in mind that these shadows are not necessarily areas of zero coverage, but are areas where data rate will drop. Only at more extreme distances will coverage disappear entirely. In indoor environments, reflection of the signal may help to fill in the coverage shadows created by support beams.
The best approach, therefore, involves using multiple access points to fill in each others’ coverage shadows. Mounting the AP to the post itself creates shadows behind the post, but mounting the AP between two posts creates multiple shadows (as seen in the image above). The best approach involves choosing the least detrimental of these two scenarios. Another factor that comes into consideration is that the support beam may be the most convenient and flexible mounting location, somewhat offsetting its negative effect on coverage.
In the example to the left, The AP has been mounted to the right side of a support beam. Coverage to the left of the AP is compromised by the support beam to which the AP is mounted, while the two support beams to the right create a notch in the coverage.
A second AP is mounted diagonally across from the first one. The second AP also has coverage shadows, but the two APs are generally able to fill in where each others coverage is inadequate. Coverage at the exterior of the building is sub-optimal (yellow), but coverage in the center is consistently good.
In summary, environments that include obstructions like support beams are going to have coverage shadows. There’s really no way to prevent them. The best design, then, doesn’t attempt to avoid the shadows, but attempts to place APs so that one AP’s coverage area fills in the other’s shadows, and vice versa. This often results in a diagonal placement of the APs in the area to be covered.
Ask the Expert
The Current state of 802.11 security
What’s the current recommended practice for 802.11 security?
Best practices for 802.11 security are based on finding a balance between available resources and the sensitivity of the data and networks that need to be protected. Home users, with relatively few resources and relatively insensitive data and networks, should use either WPA-PSK or WPA2-PSK, depending on what their access point and client card(s) support. Although WPA2 has somewhat stronger encryption than WPA, WPA is more than sufficient for home users, as long as a strong passphrase is used. A strong passphrase means that it is long and does not contain words that could be found in the dictionary. It is acceptable for home users to use WEP encryption, but only if WPA is not an option.
Corporate users, on the other hand must avoid WEP at all costs. WPA-Enterprise and WPA2-Enterprise are the preferred security mechanisms for enterprise installations with lots of resources and highly sensitive data. These mechanisms require a RADIUS infrastructure, so they’re not practical for most smaller installations, but they offer much greater manageability than the manually-configured passphrases used by WPA-PSK and WPA2-PSK.
In past months, we reported that a draft proposal for 802.11n had been accepted. While this was good news for the relatively hasty completion of the standard, the IEEE standards process is a long and convoluted one, and many more hurdles needed to be cleared. Acceptance of the draft proposal simply meant that there was enough agreement on 802.11n to actually have something to put to a vote. In that vote, taken on May 2, the v1.0 draft not only failed to achieve the 75% supermajority necessary to allow it to move forward towards becoming the final 802.11n standard, but failed to even achieve a 50% simple majority. The media has responded to this development with some furor, but the failure of a v1.0 draft to pass its first vote is hardly a surprise to those who are familiar with the history of 802.11 standards.
After the adoption of the v1.0 draft proposal, vendors were quick to sell products that claimed compliance and some vendors even went so far as to promise firmware-upgradeable compliance with the final standard. Now, analysts say that it is likely that the v1.0 draft will have to be changed, perhaps significantly, before it can move forward in the ratification process. This leaves customers who purchased “pre-n” equipment based on the v1.0 draft in a kind of limbo—not certain what changes will occur in the standard and not certain whether their equipment can be firmware-upgraded to support those changes.
For this very reason, Connect802’s message has always been that customers, especially enterprise customers, should avoid “pre-standard” equipment if at all possible and continue to rely on existing standards that are capable of filling their needs, such as 802.11g and 802.11a. Amid all the hype surrounding an exciting new 802.11 standard, it has sometimes seemed that few voices echoed Connect802’s more conservative advice, but this month, there are a chorus of voices saying the same thing that we’ve been saying all along. We’ll let them speak for themselves.
Connect802’s position is that convergence between wireless technologies such as 802.11, 802.16, and 3G/4G cellular will continue. Although rare, mobile devices already exist that can hand-off a voice call between an 802.11 network and a cellular network—for example, when walking from the street into a building with a Voice-over-Wi-Fi infrastructure. It seems that, as wireless technologies gain more and more functionality, the market for converged devices will only get larger.
But we don’t see the wireless market as a kind of “winner-take-all” proposition in which a single wireless technology must come to dominate. We see 802.11, 802.16, and cellular technologies as each being optimal for a particular niche, and we predict that, although there will be some incursions into each others’ territory (802.11 being used for in-building voice instead of cellular, or 802.16 being used for metro-wireless instead of 3G), each technology’s strengths will allow it to remain supreme in its particular niche.
This month, Wi-Fi Net News has an article supporting Connect802’s position. The argument is that, although 3G cellular may be able to provide data rates that make it a viable alternative to metro Wi-Fi or WiMAX, the cellular infrastructure only has enough bandwidth and RF spectrum to support a limited number of users. For example, a typical cellular tower is connected back to its switching center via a relatively low-speed link like a T1. If a single 3G user can get up to 500 kbps sustained throughput, it’s easy to see that each cell tower can’t support very many active users. The cell companies’ assumption, of course, is based on a typical phone company model of oversubscription: not all users will be active at the same time. But data users, unlike voice users, tend to expand their usage to fill the bandwidth that they’re given. This causes cellular carriers to place restrictive Terms Of Service on their 3G users and to penalize users who violate those terms by using too much bandwidth. The argument, then, is that an entity that wants a high-capacity broadband network would be better off deploying Wi-Fi or WiMAX. At the very least, this would provide the benefit that the entity would control its own bandwidth, and there wouldn’t be a cell carrier waiting to cancel subscriptions or assess penalty fees if usage was too high.
At Connect802 we're your PAGE ONE resource for wireless networking!
Connect802 has the experience, expertise, and resources to help you with your wireless network system. Use your favorite search engine and see what Connect802 is doing. Each month we give you some suggested search terms for you to explore. Here's this month's list. As you look down the search engine results you'll find Connect802 either at the top, or on the first page (true for Google and Excite, unknown for the rest).
encyclopedia data frames encrypted 802.11w
RF CAD modeling for hospitals
Warehouse Wi-Fi design consulting
Mesh routers for RV parks
Rain fade margin calculator
Wi-Fi fade margin calculator
Compare Strix, Tropos, MiLAN
Mesh versus Wireless Distribution
Humorous Request for Comment documents
Why should I have a Wi-Fi wireless LAN installed in my building?