Proceedings of 2010 IEEE International Conference on Ultra-Wideband (ICUWB2010) Abhishek Bit, Dr. Martin Orehek, Waqar Zia Student: Adviser: ID:M9920108 Abstract INTRODUCTION COMPARATIVE ANALYSIS A. Basic Architecture B. Radio Technologies Used C. Broadcast Support D. Number of Devices Supported by a Host E. Extension of the Basic Cell F. Throughputs G. Device Association Mechanisms H. Bluetooth Profiles I. Power Consumption J. Technology Maturity and Market Penetration K. Cost CONCLUSION & SUMMARY REFERENCES This paper presents the results of the comparative analysis of two different protocols that use or could potentially use the Ultra Wideband (UWB) radio technology – Certified Wireless-USB and Bluetooth 3.0 with UWB. Bluetooth 3.0 with UWB is found to have a number of advantages over Certified Wireless-USB. These include lesser overheads, higher throughputs, higher power efficiency, better device association mechanisms, dedicated profiles, and support for data broadcast. The W-USB Specification Revision 1.0 was developed by the USB Implementers Forum (USB-IF). The protocol utilizes the common WiMedia MB-OFDM UWB radio platform for its high data-rate operation. Certified Wireless-USB (CW-USB) is basically that flavour of W-USB that is promoted and standardized by the USB-IF. The protocol can be classified as a full-blown MAC in its own right. Bluetooth is a wireless communication standard intended for exchanging data over short-distances between fixed as well as mobile devices. Bluetooth Core Specification Version 3.0 + High Speed (HS), or Bluetooth 3.0 is the latest iteration of the short-range wireless technology from Bluetooth Special Interest Group (SIG). Instead of UWB, 802.11g was preferred as the alternate high speed radio. However, the Bluetooth 3.0 architecture consists of a ‘generic’ Alternate MAC/PHY (AMP) layer, and the idea of incorporating the UWB radio, as the high speed alternative to 802.11g, is still not completely out of the question. The term ‘generic’ underlines the fact that Bluetooth 3.0 has been architected to use any approved MAC and PHY (including WiMedia UWB MAC and PHY) for its high data rate operation. Furthermore, since Bluetooth 3.0 achieves its 24 Mb/s transfer speeds by pairing devices and then initiating a Wi-Fi connection between them, it could do the same with UWB technology, if it was built in. UWB technology continues to be a potential high speed alternative for the next generation of Bluetooth. Significantly higher data rates, energy efficiency, better tolerance to interference, are some key advantages of UWB over 802.11g. Throughout the discussion in this work, the Bluetooth 3.0 protocol is assumed to use UWB as the high speed radio technology. A. Basic Architecture The CW-USB architecture is based on the hub-spoke model where one device (the Host) is the master and the other devices are the slaves. In such a star formation, it is the Host that controls the entire network of other devices. Different softwares are used for a CW-USB Host and a CW-USB Device i.e. the end-to-end connection architecture is not symmetric. A CW-USB product can either be a limited CW-USB Host, a CW-USB Device or a device capable of assuming both roles (Dual Role Device – DRD). The master-slave topology of CWUSB has another major drawback in a network environment, where one slave can talk to another slave only through the master. This also increases the power consumption of the Hosts as they have to control all slaves. In Bluetooth 3.0, though in any given piconet, one device assumes the role of the master, all devices are born equal. The Bluetooth 3.0 implementation is symmetric. This implies that any device that supports Bluetooth 3.0 can assume the role of either a master or a slave. The Bluetooth 3.0 software on both sides is the same. Bluetooth 3.0 uses two radio technologies – the Bluetooth BR/EDR radio and the high speed UWB radio. This allows switching to UWB in case of high speed communications and reverting back to Bluetooth BR/EDR when the high speed radio link is no longer needed. Protocol complexity is thus increased through an additional AMP manager that controls switching and manages the radios. CW-USB, on the other hand, uses the UWB radio only. This translates to lower protocol complexity compared to Bluetooth 3.0. Connection establishment in Bluetooth takes place out-ofband with respect to the high speed UWB radio. This is in contrast to CW-USB, which uses a Device Notification Time Slot (DNTS) mechanism and all communication takes place over UWB. Connection setup time is thus expected to be higher with Bluetooth 3.0. There is no mechanism that supports data broadcast in WiMedia MAC. The Micro-scheduled Management Control Proceedings of 2010 IEEE International Conference on Ultra-Wideband (ICUWB2010) (MMC) frames in CW-USB are “control frames” that are effectively broadcast. The specification does not allow broadcast for user data. In Bluetooth 3.0 however, the connectionless L2CAP channels allow broadcast transmissions between the master and the slaves in a piconet, over the BR/EDR radio. Thus, a Bluetooth 3.0 implementation with UWB will have data broadcast capability, but over the low bit rate radio. In Bluetooth, there can be at most 7 devices (active slaves) connected to the host in a piconet. The protocol also supports as many as 255 parked devices (inactive slaves) in the same piconet. WiMedia UWB MAC allows support for a maximum of 48 devices through 48 distinct Beacons. However, in Bluetooth 3.0, the BR/EDR link is left in the active state, while the AMP UWB link is used for the high speed communication between the same master-slave pair. Thus, theoretically, the Bluetooth 3.0 with UWB can support 7 active connections and 255 inactive connections. The number of active connections however, can possibly be increased by putting the Bluetooth BR/EDR connection into the parked state while high speed transfer takes place over the UWB radio. CW-USB on the other hand, can support 48 devices (active connections) with a single host. Bluetooth 3.0 has a slight disadvantage when it comes to the maximum number of active connections supported by a single host. However, at any given time there can be a maximum of 262 (255+7) devices (inactive slaves included) in a single piconet. As discussed earlier, Bluetooth 3.0 employs a symmetric end-to-end architecture. Every Bluetooth 3.0 device is capable of assuming the role of either a master or that a slave. Therefore, in Bluetooth, the basic piconet architecture can be easily extended to form a scatternet, with one of the Bluetooth devices assuming the role of a master in one piconet and that of a slave in the other, or functioning as a slave in more than one piconet. The number of active devices attached to a host within one piconet remains the same (maximum of 7). A Bluetooth device can serve as a master in one piconet only. A CW-USB Device cannot assume both roles (unless the complete CW-USB Host functionality is incorporated within the Device as well). CW-USB puts a clear demarcation between a Host and a Device. The only way in which the star network can be expanded is when one CW-USB Device acts as a slave for more than one Host. Thus, the extension of a basic cell in case of CW-USB is an extended star network. Bluetooth 3.0 offers an advantage over CW-USB in terms of achievable data throughputs. This is owing to the following reasons: • Higher signalling overhead in CW-USB owing to MMC frames. Control information in CW-USB is explicitly exchanged through MMC frames (this is in addition to Beacon slots used for signalling such as DRP reservations). • Maximum allowable data payload size in case of Bluetooth 3.0 (with UWB) is 4071 bytes (4075 bytes – 4 bytes for L2CAP header), and for CW-USB it is 3582 bytes (3584 bytes – 2 bytes for WUSB header). This again results in slightly higher data throughput in case of Bluetooth 3.0. This frame size difference is significant when large data transfers are taken into account. With a maximum allowable burst size of 16 packets, this payload difference translates to a (4071 – 3582) bytes x16 = 7824 bytes difference • CW-USB uses dedicated time slots called Device Notification Time Slots (DNTS) for several notifications such as initial connection establishment and communication resumption after flow control. The duration and number of such slots depends on host implementation. CW-USB solutions from CSR use a DNTS slot period of 24 μs with 8 DNTS slots. Bluetooth has its own way of releasing time when it is flow controlling. This is done through special frames called TC or ZA frames. Moreover, connection establishment in Bluetooth takes place over Bluetooth BR/EDR radios. Therefore, in Bluetooth 3.0 with UWB implementation, there is no need for the DNTS mechanism. The DNTS slots thus are a clear overhead in case of CW-USB. With the abovementioned constraints in mind, maximum throughput estimates for both protocols were plotted using Matlab. The following assumptions were made: • Maximum burst size of 16 data packets for both protocols. • Data payload transmission at the rate of 480 Mb/s (Fig. 1) and 53.3 Mb/s (Fig. 2). • Short burst preambles for all but first data packet. • 8 DNTS slots of 24 μs each in CW-USB burst transaction group. • An MMC frame ahead of the data burst in CW-USB. • MIFS of 1.875 μs between burst data packets. • SIFS of 10 μs between control frames and data packets. The throughputs were calculated as the ratio of the user data to the total data transmitted (including protocol overheads and taking into account the abovementioned assumptions). The results in Fig. 1 indicate that theoretically, Bluetooth 3.0 with UWB would give a better throughput performance than CWUSB. With Bluetooth 3.0, theoretical maximum throughputs of 386 Mb/s (with data rates of 480 Mb/s) could be achieved. In case of CW-USB, the maximum attainable throughput is lower at 328 Mb/s (with data rates of 480 Mb/s). At the lowest data rate (53.3 Mb/s); the difference between the throughputs for the two protocols is low, though still in favour of Bluetooth 3.0 with UWB implementation (Fig. 2). Graphs were also plotted to compare maximum attainable throughputs for the two protocols for nonbursty data traffic with the acknowledgment policy set to No-ACK. The following changes were made from the previous assumptions: • Normal PLCP preamble time of 9.375 μs. • SIFS of 10 μs between all data frames (no MIFS). • No acknowledgement frames (No-ACK policy). The plots (shown in Fig. 3) indicate drops in maximum attainable throughputs for both protocols when compared to the burst data traffic scenario. The throughputs for Bluetooth 3.0 however, are still higher than that of CW-USB. The WiMedia PHY supports data rates of 53.3, 80, 106.7, 160, 200, 320, 400 and 480 Mb/s. When data reliability is critical, lower bit-rates are used. The maximum data throughputs that can be achieved with each of these data rates for Bluetooth 3.0 and Certified Wireless-USB were computed using Matlab. They are plotted in Fig. 4 for burst data traffic. The difference between the attainable throughputs for the two protocols increases as the data rates are increased. The difference is minimum (1.1916 Mb/s) for a bit-rate of 53.3 Mb/s and maximum (57.3 Mb/s) for 480 Mb/s data rate. This is because at low data rates, more time is needed for user data transmission. Hence, the overhead ratio (computed as the ratio of the transmission time for signalling overhead to the total transmission time for the complete data transaction) is reduced. From a data communication point of view, even a difference of 1.1916 Mb/s is significant. Note that the highest data rate (one way) supported by Bluetooth BR radio is only 721 Kb/s. Bluetooth 3.0, in contrast to CW-USB, supports a number of device association mechanisms. It uses Secure Simple Pairing (SSP) to simplify the device pairing procedure, and maintain or improve security (protection against passive eavesdropping and MITM attacks). As many as four association models are defined under SSP – Numeric Comparison, Just Works, Out of Band or NFC, and Passkey Entry. One of the device association mechanisms in CW-USB employs the use of a physical USB cable to establish a first time secure connection in effect violating the whole essence of wireless communication. Clearly, Bluetooth has an advantage over CW-USB when it comes to support for device association models. The Bluetooth protocol stack includes dedicated profiles which define the behaviour through which Bluetooth enabled devices “talk” to each other. As many as 22 profiles are supported. Each of these profiles generally includes a user interface, information about dependencies on other profiles and parts of the Bluetooth stack to be used. The specification also allows developers to create their own applications to work with other Bluetooth devices. CW-USB specification does not include any such profiles. In Bluetooth 3.0 device discovery, connection establishment and maintenance, and profile configuration use the Bluetooth (BR/EDR) radio, while bulky data transfer uses the faster alternate MAC and PHY (UWB). In this way, when the system is idle, the proven low power configuration models of Bluetooth are used, and only in case of a large data transfer, the system switches to the UWB radio. When the high speed is not required, the link is turned off or put in idle mode. CW-USB, on the other hand, utilizes the UWB for all radio communication – including discovery, connection setup and radio link management. This leads to higher power consumption (continuous data processing at say 300 Mb/s for UWB radio) in CW-USB as compared to Bluetooth 3.0 where one has the option to utilize the lower bit-rate radios. CW-USB is a published standard. This is probably its main advantage. W-USB solutions in the form of wireless display adapters, wireless docking stations are available in the market. The technology however, has not yet penetrated the mobile device market. Bluetooth 3.0 on the other hand, was released in April in 2009. With Bluetooth and Wi-Fi already present in smart phones and other handheld devices, the newer high speed version should only be a software upgrade over the existing BR/EDR version. However, the use of UWB instead of Wi-Fi (802.11g), as the underlying high speed radio, is an improvement to the standard suggested as one of the outcomes of this study. An end-to-end Bluetooth 3.0 solution would cost more than that of CW-USB mainly due to the following reasons: • There would be two radio controllers in Bluetooth 3.0 (the Bluetooth and the AMP-UWB controller) against the single UWB baseband controller in CW-USB. • The symmetric Bluetooth 3.0 architecture would imply that the complete software stack is used on both sides. CW-USB makes a distinction between a Host and a Device, with the Host managing bulk of the data communication. So, in case of many Devices connected to the Host, the CW-USB solution would have the cost benefit over Bluetooth 3.0. The results of the comparative study for Bluetooth 3.0 with UWB and CW-USB can be summarized using the following table (Table I). The Bluetooth 3.0 protocol has a number of advantages which include a symmetric end-to-end architecture, higher throughputs, power efficiency, support for user-data broadcast, association models, support for dedicated profiles, and scope for better mobile market penetration. The Bluetooth 3.0 protocol with UWB as the alternate high speed radio technology, is therefore preferred over CW-USB, and recommended for integration into mobile devices. [1] E. Ferro, and F. Potorti, “Bluetooth and Wi-Fi protocols: A survey and a comparison,” IEEE Wireless Communications Magazine, June 30, 2004. [2] D. A. Gratton, “Ultra Wideband: I’ll be back,” Incisor, Issue 135, pp. 13- 14, June 2009. [3] Wireless-USB Specification, USB-IF, Revision 1.0, May 12, 2005. [4] Bluetooth 3.0 + HS Specification, Bluetooth SIG, Vol. 0, April 21, 2009. [5] ECMA-368, “High Rate Ultra Wideband PHY and MAC Standard,” 3rd edition, December 2008. [6] J. Walker, “Why Mix Bluetooth with Wi-Fi?”, Wireless Alert section in Network World, May 4, 2009. [7] G. Heidari, WiMedia UWB – Technology of Choice for Wireless-USB and Bluetooth, John Wiley & Sons Ltd., 2008 edition.