|
|
|
| IBM home | Products & services | Support & downloads | My account | ||
 |
|
|
|
|
|
|
| ||||
| Roaming charges: Pumping the piconet: A first peek at the new WPANs | |   | |
| ||||
| ||||
| IEEE 802.15.3 spells good news -- but is it worth the wait?
Larry Loeb (larryloeb@prodigy.net) The two specifications that comprise the "WiMedia" brand of WPAN (the p is for personal) networks have generated no small amount of buzz -- particularly given that one was only recently standardized and the other enters its first draft proposal in November of this year. But then, it's been a long time since a new wireless technology had so much crowd-pleasing potential. This month, Larry give us a first look at the big ideas behind WiMedia. Most wireless networks are built with the single goal of extending device mobility beyond where the connecting wire ends. As such, industry has allowed professional groups like the IEEE to set the specification for the lowest two layers of the OSI (open systems interconnection) model; namely the physical (PHY) layer and the media access control (MAC) layer. This works well for manufacturers and consumers alike, giving the former an interoperability standard they can rely on and ensuring the latter that they won't be locked into some proprietary networking scheme. The past few years have seen broad consumer acceptance of several IEEE standards, including 802.11 for Wireless LANs and 802.15.1 for Bluetooth networking. But the profusion of wireless multimedia devices such as digital cameras, digital widescreen TVs, laptops, and MP3 players has begun to push the performance boundaries of the existing standards. Consumers don't just want these devices sitting side-by-side in their living rooms anymore; they want wireless multimedia networking -- and they want it to work right out of the box. Bluetooth was the first wire-replacement technology to try to address the need for serious wireless interoperability. Unfortunately, while Bluetooth isn't very complicated to use, it has a low effective data transfer rate, maxing out at 700 Kbps. This makes even routine data transfers (such as downloading pictures from your digital camera onto your laptop) too slow for widespread consumer acceptance. So, the IEEE has begun to expand on the Bluetooth concept. The proposed 802.15.3 WPAN (wireless personal area network) would operate on the 2.4 GHz band, delivering up to 55-Mbps throughput at ranges of about 30 feet. IEEE 802.15.3a would operate on the ultra-wideband system (UWS), where data rates could easily soar into the 110 Mbps to 480 Mbps range. Neither of these two new specs (dubbed "WiMedia" by the eponymous marketing/user organization) is yet a reality. The 802.15.3 spec has recently been standardized but implementation is a ways off, and the initial draft proposal for 802.15.3a will be issued in November of this year. But, even with the wait, these two new specs spell good news for wireless multimedia developers. Don't take my word for it, though; read on and I think you too will soon be preaching like a believer. Life in the superframe Figure 1. Schema of time-slot-to-device allocation In this schema, the beacon header transmits control information to the piconet, which is the assemblage of controller (PNC) and device clients (DEV) that make up the 802.15.3 network. Figure 2 shows the interrelationship of elements in the piconet. Figure 2. Interrelationship of elements in the piconet As you can see, one device typically assumes the role of controller. The primary function of the controller is to establish the basic timing for the piconet, using beacon headers to designate time allocations for different devices. If necessary, the controller can also implement announce commands to extend the beacon's control mechanisms. The controller is also responsible for security on the piconet, which involves all the usual public-key techniques. Figure 3 is representative of a typical superframe. Figure 3. A typical superframe CTAP, which stands for channel time allocation period, is a mechanism that uses TDMA (Time Division Multiple Access) to manage devices operating under a specified time window. CTAP is composed of several elements, starting with the contention access period, or CAP. CAP utilizes the CSMA/CA protocol for medium access to communicate commands or small amounts of asynchronous data from the controller to client devices. CAP is followed by the channel time allocations, or CTAs. CTAs come in various flavors, including management CTAs (MCTAs), commands, and isochronous or asynchronous data connections. The mechanics of QOS If a device needs a certain amount of time to send its data, it can request an asynchronous allocation. The controller then schedules this time when it becomes available. In this kind of transfer, only the source device or the controller can terminate the allocation. The controller can place the MCTAs wherever it needs to in the superframe, not only at the head of the stream. This gives the controller the flexibility to make changes on the fly, based on incoming requests for CTAs. This flexible structure allows QOS requests to be met in the overall system architecture. A device can use the channel status request (CSR) command to determine the link quality between itself and another device. Based on the result, the device might change the transmit power, data rate, or the channel time to improve the connection. CSR also enables the controller to poll the piconet for trouble spots. If the controller discovers a weak link, it can change the radio channel, hopefully improving the quality of the piconet. The decision to override a transmission channel is one that only the controller can make. The internal alarm clock
DSPS puts devices to sleep at intervals defined by the controller. Basically, the device alerts the PNC when it wants to go to sleep (that is, listen but not transmit), the PNC informs the device of which beacon will be its wakeup call, and the device stops transmitting. PSPS is used for groups of devices that will sleep together for a certain number of superframes and then wake in the same superframe, such as a group of devices in a sensor network. Such devices form a PSPS set with its own specified interval between sleeping and waking periods. With these periods defined, other devices in the piconet set always know when the set will next be awake and handling traffic. APS is used for devices that need to power down for extended periods of time. Under the APS mechanism, each device sleeps and wakes up on its own. This kind of behavior is different from the routine kind of power management that allows each device to power down when it isn't transmitting or receiving data. PHY layer characteristics 802.15.3 supports five data rates. The base data rate of 22 Mbps is uncoded. The remaining data rates (11-Mbps, 33-Mbps, 44-Mbps, and 55-Mbps) use trellis-coded modulation. The header for all frames is sent at the 22-Mbps rate, which allows devices to detect traffic on the piconet. The bandwidth is limited to 15 MHz in a standard 802.15.3 network, which simultaneously allows more channels and decreases the piconet's susceptibility to interference from other systems. Receivers report both the signal level and whether the high-order modulations are in use, which would indicate good signal quality. Receiver reports are used by devices to determine whether channel errors are occurring due to a poor signal or due to interference. 802.15.3's designers have done their best to ensure it will be a good neighbor to all the other entities riding the 2.4-GHz spectrum. The protocol uses passive scanning, lower transmit power (coupled with transmit power control), and dynamic channel selection to proactively minimize interference to other systems. It also has mechanisms to minimize the effects of interference on its own transmissions. Ultra-wideband velocity The answer to that question lies with the FCC's February 2002 decision to create an unlicensed ultra-wideband system (UWS). Figure 4 compares the frequency usage of the ultra-wideband system to that of 802.11a. The graph also illustrates the characteristic differences between a UWS transmission and more conventional kinds of PHY transmissions. Figure 4. Frequency usage of UWS vs. 802.11a The difference is obvious. While 802.11a concentrates its transmission in 100-MHz of the 5.7 GHz range, UWS operates at a much lower amplitude level of signal, extending from 3.1 GHz to 10.6 GHz. This kind of transmission can sound like noise unless it's properly received, but it really doesn't interfere with other transmissions in the same frequencies. The 802.11a spike of power into a limited-RF spectrum versus the even (but low) distribution of power over the UWS spectrum tells the story here. With UWS you get the wide spectrum, but the signal is way down there, decibel-wise. The IEEE 802.15.3a task force has been attempting to draft a set of standards that would let 802.15.3 harness the transmission power of UWS. Like the original 802.15.3 spec, the proposed 802.15.3a spec is full of big ideas. Among other things, 802.15.3a would require that the interfering average power of the network always be at least 6 decibels below the minimum sensitivity level of a non-802.15.3a device. As a result, 802.15.3a transmissions would be easily rejected by other receivers operating in the same frequency bands. In sum, 802.15.3a devices operating on the UWS could obtain much higher data rates (110 Mbps to 480 Mbps) than those achievable on any currently existing or proposed wireless network. While a draft is expected to emerge for voting in November 2003, the transmission approach is still under debate. Impulse Radio (IR) was the original proposal, but Multiband Approach is the popular favorite (that is, Intel likes it) and might have the inside track to ratification. IR uses short-duration baseband pulses with a bandwidth of a few GHz and supports multiple accesses using a time-hopping scheme. MA divides the spectrum into separate bands greater than 500 MHz. MA bands can be used either statically or dynamically, and data is modulated by concatenation of the bands (using a scheme similar to the OFDM of 802.11g). Figure 5 shows how these groups can be organized. Figure 5. Organization of bands and signals
on the PHY frequencies The MA approach offers greater flexibility and scalability than the IR approach, and its pulse repetition frequency can actually be lower than that of IR at the same peak power. MA's timing acquisition is also faster, and -- if we can believe Intel -- it's also more feasible to implement. In conclusion Taken together these two specs signal an important shift for wireless multimedia, albeit one that will take some time to fully materialize. While standardization isn't far off, routine implementation might not occur for two or three years -- which is about how long it will take for some manufacturers to crank out the jellybeans and others to incorporate them into devices. It's worth keeping an eye on these specs now, though, especially if you're developing applications for wireless multimedia. Like Bluetooth, WiMedia networks will allow chip-level fabrication of devices that link disparate consumer electronics. Unlike Bluetooth, it will allow consumers to send and receive data at speeds they can live with. Knowing what's coming down the pike tomorrow can influence your development decisions today, and also help you steer clear of protocol decisions you could later regret. And ... Larry's tip of the month
| | ||||||||||||||
|
| About IBM | Privacy | Terms of use | Contact |
Larry Loeb has written for many of the last century's major "dead tree" computer magazines, having been -- among other things -- a consulting editor for BYTE magazine and senior editor for the launch of WebWeek. Larry's newest book has the contractually obligatory title of Hackproofing XML and is published by Syngress (Rockland, MA). He's been online since