Detailed explanation of wireless LAN VoIP technology application

yxh99

Detailed explanation of wireless LAN VoIP technology application [Copy link]

　As WiFi standards improve and 802.11 chips continue to shrink and expand in functionality, the feasibility of Voice over Wireless Local Area Network (VoWLAN) phone systems is increasing. Dual-band mobile phones can use WLAN connections to provide reliable voice services in the home, while broadband phone services can be connected to laptops via WLAN. On the other hand, WLAN-based VoIP phones can easily support multiple phones with just one WLAN base station, making them comparable to traditional wireless phones with low cost advantages.

　　The 802.11 standards establish the basic mechanisms required to provide reliable, high-performance WiFi VoIP systems. Notable examples are security (802.11i/WPA) and QoS (802.11e/ Wi-Fi Multimedia). In addition, single-button security configuration methods such as Atheros' open source JumpStart for Wireless allow all users to quickly configure WLAN VoIP phones even if the phone cannot display English letters

and numbers. 　　One of the items in WLAN VoIP systems that has not yet been standardized is the polling method. Therefore, this article discusses the different advantages and disadvantages of the two existing polling methods, and specifically focuses on the most critical factor in mobile devices - power consumption.

　　All methods of reducing power consumption must allow user devices to use low-power sleep modes as much as possible, and 802.11 chips must support this practice with the lowest possible power consumption in sleep mode. For example, Atheros' AR6000 radio-on-a-chip mobile (ROCm) device implements extremely low-power sleep mode and Automatic Power-Save Delivery (APSD) technology. ROCm also provides excellent performance, enabling high- speed transmission to shorten send/receive time, and the self-contained driver of the embedded processor on the chip can offload the frequent network maintenance operations from the host processor. Through these and other power-saving strategies, ROCm chips can improve the power efficiency of WLAN operations by up to six times that of traditional WLAN chips, thereby improving battery life. The new generation of 802.11 devices that enable a wide range of VoIP applications now include such chips.

　　Bringing Voice to WLAN

　　802.11 WLANs can take advantage of high-performance components to provide reliable overall performance, however, the characteristics of the medium still present significant challenges when handling voice traffic. Because WLANs use unlicensed spectrum, they must tolerate a lot of interference from various external devices and other WLANs. In addition, like other IP networks, WLANs do not support synchronous operation. Therefore, predictability in the microsecond range is generally unavailable. Since VoIP is a constant bit rate (CBR) application that generates VoIP packets (frames) at fixed time intervals, the WLAN's CSMA contention method clearly lacks centralized synchronous timing.

　　This phenomenon is in even greater contrast to the standard telephony mechanisms implemented in mobile phone systems, which use licensed spectrum and carefully planned base station deployments to minimize radio interference. Mobile phone systems are synchronized from the phone to the backbone, so they know the timing at the microsecond level and never deviate. As a result, they have predictable capacity, and that capacity is designed for a single class of service : voice.

　　These characteristics of mobile phone systems make it easy to comply with the ITU-T Recommendation G.114 standard, which specifies an end-to-end delay budget of no more than 150 microseconds. Because the overall architecture of mobile phone systems uses timed voice packets in a deterministic manner, there is no need to prioritize voice packets with special quality of service (QoS) mechanisms to ensure low latency. Mobile phone systems leverage existing time slots, multiplexing, and voice service management to add data services.

　　WLANs are the opposite, as voice services must leverage features that were originally designed for data. WLANs can only use a portion of the end-to-end delay budget of 150 microseconds, and if both ends of the conversation are using WLAN, the delay budget is further limited. In addition, if voice packets must cross the Internet or a busy corporate network, the packets will inevitably arrive late, or even not arrive at all. Late packets may arrive in groups.

　　Anyone who has used an old transcoder to communicate voice over the Internet or a general WLAN will be familiar with these problems. One approach to building high-quality VoWLAN is to modify the WLAN to meet the needs of the legacy codec. In fact, proprietary implementations have shown that the 802.11 MAC can be modified to use a synchronous, time-slotted TDMA approach, whether full-time or time-sharing; this approach can effectively solve the problem of transmitting voice over WLAN, but such systems are generally incompatible with existing WiFi devices and networks.

　　Although a fully synchronous network is attractive, the lack of strict synchronization is the main strength of 802.11. Over the years, we can see the advantages of such IP networks in the competition between Ethernet and ATM networks. When reliable and adaptive (good enough) channel access is paired with a strictly timed (perfect) sequential approach, the approach that is good enough is usually more popular .

　　Another reason to avoid using synchronization when designing VoWLAN systems is that these systems do not operate in a closed environment. The main selling point of using WLAN for voice is that it allows dual-mode mobile phones and other voice devices to use existing WLAN infrastructure.

One way to improve the existing 802.11 infrastructure is to use new voice decoders developed for Internet applications. These decoders greatly simplify the design of VoWLAN . The inefficient Internet phone environment has led to the development of decoders that can achieve good voice quality at very low bit rates. 　　For example, the iLBC decoder at the heart of the popular Skype Internet phone system can provide features equivalent to the high-end ITU G.729 decoder; the ITU decoder can provide pay phone-like voice quality at only 8kbps; the iLBC decoder from Global IP Sound requires a slightly higher bit rate - 13.3kbps. Global IP Sound claims that their encoder has better voice quality than PSTN and can tolerate up to 30% packet loss. The Internet Engineering Task Force (IETF) has set standards for this decoder. CableLabs ' PacketCable Video and Audio Decoder Specification for Multimedia Terminal Adapters and Media Gatewaysspecifies it as a required decoder. 　　With such codecs, the necessary VoWLAN voice quality is much easier to achieve, and the latency and jitter introduced by the Internet are also addressed, making them particularly suitable for use in asynchronous open systems such as 802.11. If codecs are so flexible, why develop complex timing and synchronization methods? 　　The power consumption challenge　　 　　Despite the flexibility of today's codecs, timing is still very important because it has a significant impact on power consumption. The synchronous nature of mobile phone systems makes it easy and straightforward to implement sleep/wake scheduling for mobile phones. The mobile phone knows between packets that it is safe to go into sleep mode. However, an 802.11 device never knows when it may receive a burst of traffic or must respond to an access point for other reasons. 　　Despite this difference between mobile phones and VoWLAN systems, the latter must still have battery life comparable to that of mobile phone handsets. Dual-mode mobile phones use the same battery for both types of functions, so they will inevitably be compared. 　　At this point, we can't help but want to synchronize WLAN operation. If the access point knows when the handset is in sleep mode and transmits only when it is ready, the handset can go into sleep mode periodically, similar to mobile phones. The access point does not have to transmit to the handset immediately when VoIP frames arrive, but can buffer them if necessary. There 　　are two modes of operation that are synchronized enough to implement good power-saving timing techniques in 802.11 WLANs, so full synchronization is not required. These modes are Hybrid Control Function (HCF) Controlled Channel Access (HCCA) and Enhanced Distributed Channel Access (EDCA). Both modes are part of the Quality of Service (QoS) provisions of the IEEE 802.11e standard, and both can be used in the development of power-saving communication methods that synchronize fixed data rates between access points and stations without synchronizing the entire WLAN.Synchronization with HCCA　　The HCCA mode is like an N-body synchronization mechanism, where the access point sets the CBR polling schedule for N stations. Despite the irregularity of typical 802.11 systems, stations try to synchronize as much as possible to their schedules. It is reasonable to describe such a configuration as an N-body system, because any disruption to the timing of any station on the polling schedule will affect the timing of the other N-1 stations.　　The HCCA mechanism comes into play when the AP receives a CBR request from a station via the traffic specification (TSPEC), and then the AP communicates the CBR schedule with the station. Once the AP accepts a station as a polled user, the station typically goes to sleep until the expected downlink poll or poll plus VoIP frame arrives from the AP (Figure 1). Within the specified time (9μs for OFDM-based 802.11a/g and longer for 802.11b), the station responds with an uplink VoIP data (or QoS-NULL) frame. If the station sends uplink data, the AP responds with an ACK.　　To understand the power efficiency of this mechanism, let's first consider the proportion of time the station needs to stay awake. For the HCCA mechanism to work correctly, the station must wake up from sleep mode before the AP's downlink poll. Depending on the hardware design, the wake-up procedure takes about 0.1 to 1.0 microseconds. The station must then wait for the downlink poll to arrive, which may not arrive when the station expects it to arrive. Delays can be caused by different reasons such as interference, long duration frames on the channel, internal scheduling conflicts in the AP (polling other stations), higher priority operations (the AP must transmit a beacon), the previous frame exceeding the expected exchange time, or relative clock offsets between the AP and the station. However, once the downlink poll arrives, the schedule becomes predictable. Depending on the selected codec and PHY rate, the uplink/downlink frame exchange should occur in less than 1 microsecond.

In the HCCA mechanism, the timing uncertainty mainly comes from the delay of the CBR polling schedule, possible retries after failure, and the change in transmission time when using variable PHY rate. Based on these uncertainties, the station wake-up time is about 2 to 5 microseconds. With a decoder cycle of 20 microseconds, the efficiency ratio achieved by this wake-up to sleep ratio is more than 75%.

　　Fixed bit rate scheduling of HCCA

　　

Wireless LAN Voice Extension VoIP Technology Application" src="http://images.enet.com.cn/iflow/articleimage/200509/1127784195398.gif" border=0>

　　Figure 1: Access stations can implement the HCCA mode of operation specified in the 802.11e standard, providing a predictable VoIP polling schedule to manage when WLAN stations can be in sleep mode to reduce power consumption . 　　Assuming the average call duration is about 100 seconds (average for mobile phone systems) and the AP provides 20 calls at the same time, the WLAN may have to perform call setup/release every 5 seconds. Even if stations frequently enter and leave the polling list, the AP must still maintain the published CBR schedule with each station. Therefore, the AP must also maintain a fixed time slot schedule. 　　A time slot is a channel period specified for a sequence of polling frame exchanges for a specific station. Unless all frames use the same PHY rate so that each exchange occupies the same amount of channel time, the duration of the time slot will vary. In the case of varying time slot duration, it is impossible to achieve efficient power-saving synchronization. 　　One option is to have all stations operate at a fixed PHY speed (6Mbps) to avoid the problem of different durations at one time. Although this option wastes a lot of potential network capacity, coverage is excellent and capacity is good for about 15 stations. 　　If you want to reduce power consumption when using different PHY rates, AP designers can choose two other methods. One method is to change the schedule so that the AP can reliably communicate with each relevant station. This will cause additional overhead and reliability issues. Another method is to let the AP transmit in the unused time of each time slot so that non-polling stations do not fill the empty air with Wi-Fi packets. This method can maintain a synchronous schedule. Polling scheduling faces a greater challenge in that it must support a variety of mobile phones　　using different decoder intervals . In this case, the polling schedule is usually established, and there will be frequent timing conflicts between CBR users . The previous estimate of 75% minimum efficiency does not take into account this scheduling conflict, which will also consume part of the station sleep time budget. 　　The ideal HCCA schedule will also be disturbed by the occasional need to send multiple downstream VoIP frames to the station. Multiple frames are required when packets arrive at the AP in groups due to internet or routing queuing behavior. Multiple downstream frames will delay CBR scheduling unless all conceptual time slots have sufficient extra time budget. Such scheduling delays also occur when upstream retransmissions are required. Reserved extra time slots in every time slot for all stations is a waste of channel time, so scheduling delays are usually solved by extending the power-up time of all affected downstream stations. 　　Engineers usually configure APs to avoid long bursts or other conditions that may increase CBR scheduling delays. This configuration applies to both HCCA and EDCA.

The timing interdependence

of　　HCCA polling contrasts with the timing independence of the power saving method used in EDCA, another 802.11e operating mode. In HCCA, the AP manages all timing and resolves all scheduling conflicts. In EDCA, all stations manage their own timing (so timing management is decentralized), and scheduling conflicts are resolved over the air using the channel access protocol.

　　Therefore, unlike HCCA stations that must humbly cooperate with the AP's polling schedule, EDCA stations can operate in a special power saving mode called Unscheduled APSD (UPSD). This mode begins with a signal handshake between the station and the AP so that the AP knows that the station will go into sleep mode until the station is ready to transmit a VoIP frame (Figure 2). The station wakeup process can occur without scheduling delays, polling waits, or timing effects caused by other stations or conflicting schedules. The station wakes up and transmits the VoIP frame with the highest priority parameters

　　available . The uplink frame is typically initiated with a power-consuming delay of less than 2 microseconds . The AP then responds to the uplink frame with an ACK. If necessary, the station can retransmit and remain awake until the AP sends a VoIP frame or a null indication (indicating that no VoIP packets are available to send). The AP response time is less than 100 microseconds with traditional AP hardware , and improvements may reduce the response time. 　　It is also relatively easy to add EDCA functionality to existing APs because it is much simpler to manage timing for a station at a time than to solve the N-body synchronization problem of HCCA. In addition, EDCA achieves roughly the same 75% power efficiency as HCCA at 20-ms CBR. Using a 30-ms CBR interval improves efficiency by about 83%. 　　Improvements in 802.11 technology will inevitably bring applications such as VoIP, VoIP plus data, video plus data, and VoIP plus video plus data into the mainstream. Although the 802.11 standard does not cover all aspects of VoIP services, it still has important standards. In particular, HCCA and EDCA both provide methods to support voice plus data in the same wireless channel while optimizing the battery life of mobile phones .





　　

　　Figure 2: An alternative to HCCA is the EDCA mode of operation, which is also specified in 802.11e. EDCA can use the Unscheduled Automatic Power-Save Delivery (UPSD) mechanism, where the station manages the power-save polling schedule instead of the access point (AP).

footer