This section of our series on deploying 802.11n wireless LANs explains the impacts (good and bad) that 802.11n can have on voice applications and describes techniques to deliver toll-quality voice over wireless LAN,
Don't miss any of the articles in this tutorial on deploying 802.11n for improved wireless LAN performance:
- 802.11n for improved WLAN performance
- 802.11n WLAN design best practices
- Supporting legacy devices
- Optimizing 802.11n for voice over WLAN
- Delivering video over your 802.11n WLAN
Mobile VoIP handsets and Wi-Fi enabled smartphones still tend to speak 802.11a/g, but this is changing. In-Stat expects more than 750 million phones to ship with embedded 11n by 2013. One reason: 802.11n clients can now implement one spatial stream, bypassing MIMO cost, size and power impacts.
Smartphones can use 802.11 for high-speed data applications like streaming video, but how can VoIP benefit from 802.11n? After all, VoIP does not need high throughput and may be degraded by contention with clients that do. But VoIP can still benefit from 11n's range and reliability, given measures that ensure quality of service (QoS).
During VoIP calls, clients send and receive short fixed-length frames carrying digitized voice, compressed and encoded by a codec (e.g., G.711). VoIP likes consistency, tolerates little loss (depending on codec), and hates long or variable delay (jitter). VoIP over Wi-Fi challenges thus include contending for airtime on heavily used channels, scanning for new APs before roaming, and re-authenticating during handoff.
11n features that most benefit VoIP include dual-band support (increasing WLAN capacity, thus reducing contention), MRC and STBC (reducing error rates), and WMM Power Save (extending battery life). But just deploying 11n APs does not create a voice-ready wireless LAN. Rather, you must design your WLAN to meet specifications – including those defined by each VoIP device manufacturer. For example, the following settings are recommended by Vocera for voice over wireless LANs.
This excerpt applies to any kind of WLAN, but Vocera also advises specific settings for Cisco and Aruba. Spectralink recommends slightly different settings for Meru, Motorola and Ruckus WLANs, to name just a few. And most enterprise WLAN vendors publish their own design guides as well. Numbers can be confusing, so let's consider the rationale behind them.
- Transmit power: APs can use more powerful antennas, but reciprocity means that it is not useful to crank an AP's transmit power higher than a client's. (While you're at it, be sure to orient APs and their antennas according to vendor specs.)
- Coverage and signal: Survey the edges of each cell to ensure that clients can hear the AP at -65 dBm or better and vice versa. For consistency, signal-to-noise ratio should be no less than 25 dB throughout the entire cell.
- Beacons, DTIM and SSIDs: These affect how long clients can sleep and thus affect battery life. A Delivery Traffic Indication Message (DTIM) interval of 1 or 2 ms is often recommended. Keeping the number of SSIDs down can also reduce beacon overhead.
- Priority: WMM specifies access classes that clients can use to indicate they are sending voice, video, best-effort or background frames. 11n APs use these to implement prioritized transmit queues (per SSID or per client). Voice traffic must be marked using WMM to ensure frequent regular airtime when other apps share the channel. At the AP, WMM can be mapped to 802.1p (Layer 2) or DSCP (Layer 3) priority markings. (Note: Mapping is critical but very difficult – and more than we can cover here.)
- PSPF: Public Secure Packet Forwarding refers to filters often used to block client-to-client communication in public hotspots. To permit VoIP flows between 802.11 clients, PSPF must be disabled.
- Rates: Many manufacturers recommend disabling slow data rates (typically <11 Mbps) to avoid contention between 11a/g handsets and very slow/distant clients. Disabling 1-2 Mbps basic rates can also reduce WLAN overhead.
- Channels: Advice here varies and depends on type of handset, as well as overall RF design. The Vocera example assumes an 11g handset (2.4 GHz), but 11a and 11n dual-band handsets should use (non-DFS) 5 GHz channels. Priority helps, but if you want to avoid non-voice contention entirely, put VoIP on its own channels.
- Density and Call Admission Control: Although not in these tables, a max of somewhere between seven and 12 simultaneous VoIP clients per AP is recommended by most manufacturers. More generally, as utilization exceeds 50%, contention causes problems for VoIP. To better manage resource use, use WLAN Call Admission Control features to enforce limits, reject or redirect calls or handoff requests, and perhaps even reserve capacity for voice.
- Timeouts: To minimize delay incurred when clients roam between APs in a micro-cell WLAN, start by sizing cells correctly and keeping client channel scan lists short. In secure WLANs, minimize full 802.1X re-authentications and make roams finish faster with opportunistic key caching, or use PSKs instead.
In the end, actual WLAN performance should be verified against voice design goals. Common goals include frame loss <2%, jitter <10 ms, and WLAN-induced latency and AP handoff delay <50 ms. These metrics contribute to overall call quality, typically measured by a mean opinion score (MOS), where 4+ is toll-quality.
This was first published in December 2010