Extending battery life for dual-mode phones
Back to the future with this Airheads Online article from 2007.
Power consumption and battery life
Power consumption and battery life have long been challenging issues for Wi-Fi device designers. To some degree this can be explained by the origins of the IEEE 802.11 protocol: while the designers of cellular architectures were focused from the outset on small, handheld, battery-powered devices, the fathers of Wi-Fi were subject to a different set of influences; while they all acknowledged the eventual need for handheld devices similar to cellphones, their immediate goal was to provide data protocol interoperability.
As a result, voice engineers might see the history of Wi-Fi protocols as a protracted effort to raise Wi-Fi to the same level of performance as cellular protocols. This is the standard of excellence for battery life, especially for dual-mode cellular-and Wi-Fi phones. Indeed, it was the introduction of the first prototypes of mass-produced dual-mode phones (using the UMA protocol) that sparked the widespread realization of just how poorly Wi-Fi battery life compares; while consumers are accustomed to talk/standby times in the order of 10/100 (10 hours talking or 100 hours standby) on cellphones, the first dual-mode devices late in 2005 were delivering 2/20 in practical tests.
Part of the problem is that a dual-mode phone has two radios rather than one, so one would expect a considerable reduction in battery life. However, performance on the Wi-Fi side (always a concern over the years as the focus moved from laptop PCs to single-mode Wi-Fi phones to dual-mode PDAs) has become much more critical when fitting the technology into a phone form-factor while continuing to strain limited-capacity battery life.
Factors influencing battery life
There are two main reasons for the poor battery life of 802.11.
- Every time an 802.11 station transmits a frame, the cost in energy expended is high compared to cellphones, which use different, less power-hungry modulation techniques. Phone and WLAN architects are now very careful to minimize the number of frames a phone must transmit.
- Every millisecond a phone’s radio is switched on and receiving has a cost in battery life. With the CSMA/CA (Carrier Sense Multiple Access / Collision Avoidance) MAC used by 802.11, it is particularly difficult to predict when a frame will be transmitted by the phone. A frame may be directed to a station at any time, so the station must stay ‘awake’ at all times so as not to lose incoming data. This causes considerable battery drain, and has led to efforts to reduce how often the receiver must be switched on . (The first Wi-Fi sleep-mode, IEEE802.11 power-save is useful but still requires the phone to wake frequently).
Clearly, the amount of energy used to transmit is greater than to receive (perhaps a factor of 5x for phone clients). However, a phone will normally spend much longer in receive than transmit mode, so this cannot be neglected. To ensure maximum battery life, the designer must make sure that the phone does not wake up to receive or transmit any longer than necessary, and that when it transmits it does so with the lowest possible battery draw.
Efforts to improve battery life have concentrated on how to make sure the phone is ‘awake’ for as short a time as possible. For a voice-data device like a phone, this breaks down into two areas of interest, on-call battery life and on-hook (idle mode) battery life. They have fundamentally different characteristics:
- When a phone is on-call, it builds and transmits an 802.11 frame ever 20 or 30 msec. This is a function of the VoIP codec used and network engineering: as the frame interval increases, the overall delay on the voice call lengthens and the perceived call quality degrades, hence the usual figure is one frame every 20 msec. This realization that voice frames are synchronous provides the key to reducing on-call battery usage.
- When the phone is on-hook, one might expect that no frames should be transmitted or received. However, there are a number of important continuous functions:
- The phone must maintain an 802.11 association state with its AP, and with new APs as it moves around. Even if the phone does not move, it must re-key periodically.
- The phone must wake up to monitor the network in case it has an incoming call.
- The phone must receive unicast and broadcast traffic such as ARP requests; often responding to these frames by transmitting as well.
There is work within and outside standards bodies to improve performance in all these areas.
Improving on-call battery life
While a phone will only be on-call for a small part of the day, it must transmit and receive 50 frames per second at these times, consuming considerable battery energy. However, a voice call is essentially synchronous: if we know when the last frame was transmitted or received, we also know that the next frame will be 20msec hence. A technique taking advantage of this has been incorporated in IEEE 802.11e, and is known as U-APSD (Unscheduled Asynchronous Power Save Delivery), and also by its Wi-Fi Alliance certification of WMM-PS (Wireless MultiMedia Power Save). This provides two mechanisms for reducing the amount of time a station’s radio has to receive.
With WMM-PS, the phone can signal the AP that it is going to sleep for a specified period; on a voice call this would be 20 msec. While the phone sleeps, the AP buffers all downlink traffic for it, waiting for an indication that it has reawakened before delivering the pending frames. Hence, the station will sleep for 20 msec, then wake up, send its voice frame on the uplink and receive the frame for the other direction of the call on the downlink and return to sleep mode again for another 20 msec. (The buffering of downlink traffic adds to delay on the voice call, but not to a significant degree.) For voice traffic, U-APSD reduces waking time to the minimum required for frame exchange; further enhancements may include special consideration for missing frames (due to voice activity detection, for instance).
Even with U-APSD, the on-call state involves transmitting at 50 frames/second. In order to further improve a phone’s talk-time, it is necessary to reduce the transmit battery drain as well. There are a number of techniques that can accomplish this:
- There are differences in power consumption between 802.11b, 802.11g and 802.11a. Most phones today still use 802.11b because the chips have been available for a while, are better optimized and tuned for low-power operation, and also use modulation techniques that allow for lower-power components in the transmit chain. When moving to 802.11g or a, OFDM modulation requires very low-distortion transmission, which means transmit power amplifiers are over-specified and take more battery power for a given RF transmit power. Also, 5 GHz components are less power-efficient than 2.4 GHz. This is one reason why phone designers have been slow to move to 802.11g and 802.11a, despite the system-level advantages of operating at higher data rates and in the 5 GHz band. Fortunately the current generation of phones is overcoming some of these challenges. While in 2005 nearly all Wi-Fi phones were 802.11b, 802.11g is now quite common and a few 802.11a phones are already available.
- There is a tradeoff between data rate and time to transmit a frame. Whereas a higher data rate (e.g. 11Mbps vs. 2Mpbs) may draw more power from the battery, the frame is transmitted faster, so the total energy consumed is usually lower. On the other hand, at higher data rates (for a given SNR) there may be more errors and hence more retransmissions. Retransmissions reduce battery life and must be avoided.
- Current phones always transmit at high pre-set RF power levels in order to achieve maximum range. While a high power level may allow high data rates, excessively high power levels result in wasted battery energy. WLAN and phone vendors are now incorporating TPC (Transmit Power Control) features, where the AP can instruct the station to reduce its transmit power when RF conditions are good, using the standard IEEE 802.11h TPC mechanism. Reducing transmit power on the station has other positive effects, including the reduction of co-channel interference across the Wi-Fi network.
Improving on-hook battery life
When a phone is on-hook, the WLAN and LAN treats it as any other data device, and this can result in considerable traffic that consumes battery power. Luckily, once the various sources of traffic are identified they are susceptible to mitigation, to varying degrees.
- Maintaining association. All 802.11 stations must maintain their association with the network even when they are not on a call. Currently, this means a phone must perform a full authentication and 4-way handshake (depending on the authentication/encryption method used) to derive session keys – even if no calls are received during the lifetime of its association with that AP. This will be improved as phones adopt a feature of 802.11i called ‘opportunistic key caching’ that allows a station to derive master keys from one AP, cache them and subsequently use them for associations to any other AP managed by the same Mobility Controller.
- A feature in 802.11 TGk, the ‘neighbour report’ further allows the station to make more intelligent roaming decisions by advertising the Mobility Controller managing each neighbouring AP (‘key scope’). This allows the station to determine the best neighbour AP to roam to, without listening for beacons for extended periods or sending probe requests as would be necessary today.
- Another forthcoming standard, currently 802.11 TGr, will shorten the 4-way handshake, reducing the number of frames to be exchanged each time session keys must be established. All of these features (i. to iii.) are in standards or drafts from IEEE and reduce the number of frames a station must receive and transmit to maintain its association with the WLAN and when roaming between APs.
- While a phone receives a stimulus (e.g. from its keypad) when it must originate an outgoing call, it cannot anticipate an incoming call, and hence has to maintain its registration in the network and monitor periodically to detect incoming calls. A proposal (co-authored by Aruba) in IEEE 802.11 TGv suggests an ‘idle’ mode for devices such as phones, where they can associate once, then establish a relationship with a ‘paging sever’. Subsequently, as the phone moves to new cells (served by different APs) it can use pre-association frames to inform the paging server of its presence. In this way, the phone only needs to associate when it has a call to originate, or when the paging server alerts it to an incoming call via a beacon broadcast. This is a proven technique from cellular architectures, applied to 802.11.
- All data stations on an 802.11 network receive background traffic, the most common traffic being ARP requests from other stations on the LAN. Although these are isolated frames, they make a considerable difference; one recent test showed a 3x improvement in on-hook battery life when ARP frames were proxy-answered by the AP and blocked from downlink delivery to the phone. While the original IEEE 802.11 1999 standard includes sleep modes, and various associated timers (DTIM, LI) can be adjusted to extend on-hook sleep time, preventing the traffic from reaching the phone is a superior solution. Also, depending on protocols used and functions on the phone, it may be possible to identify and filter or proxy other types of traffic, further extending battery life.
- Another technique used in fixed-mobile convergence (FMC) architectures (Wi-Fi and cellular combinations) is to switch off the Wi-Fi radio altogether when the phone’s location, derived from the cellular network, indicates there is no suitable WLAN in range. This can contribute significantly to battery life and is appropriate where, for instance, the FMC design only uses WLAN at the user’s workplace. For general multi-site and hotspot use, however, it is too imprecise to scale well.
As it is often said of security, the quest for improved battery life is a journey, not a destination, and there will always be further ways to improve. This note has discussed the major factors determining on-call and on-hook battery life, showing that while standards and drafts address and solve some of the issues, more progress is necessary to reach the standards of performance established by cellular networks.
Implementing the features described in this note has been shown to double on-call and quadruple on-hook battery life in some of the most advanced dual-mode phones, notably those that have been in use in Japan since early 2005. Most dual-mode phones now advertise on-call battery life in excess of 4 hours, thanks to these battery management changes.