1. copyright 2002 Custom Coding, Adaptive Rate Control, and Distributed Detection for Bluetooth Matthew C. Valenti Assistant Professor Lane Dept. of Comp. Sci. & Elect. Eng. West Virginia University Morgantown, WV mvalenti@wvu.edu Max Robert Mobile & Portable Radio Research Group Virginia Tech Blacksburgh, VA This work was supported by the Office of Naval Research under grant N00014-00-0655, AOL, and the MPRG Affiliates Program

2. © 2002 Motivation & Goals Motivation  Bluetooth enables low cost/power wireless connectivity.  However, range is restricted to ~10 m due to limited power, inefficient modulation, and modest error control capabilities. Goal of this study  Develop strategies for improving the performance of Bluetooth in low SNR environments.  Benefits: Range extension. Operate in noisy industrial environments. Tolerate more interference.  However, all proposed strategies comply with the standard. We are not suggesting changes to the standard. 2/16

3. © 2002 Features of Bluetooth Radio layer  Gaussian frequency shift keying (BT=0.5). Nonorthogonal: 0.28  h  0.35 1 Megabaud over 1 MHz occupied bandwidth. Baseband layer  Transmissions are broken into 625  sec slots. A packet may be 1, 3, or 5 slots long.  Time-division duplexing (TDD). Master/slave take turns transmitting.  Packet-by-packet frequency hopping. 79 or 23 channels spaced 1 MHz apart. Piconet synchronized to master’s clock.  ACL Packets for data. DHx (Data high rate): No FEC. DMx (Data medium rate): (15,10) Hamming FEC code. ARQ used by both DMx & DHx (assisted by CRC). 3/16

4. © 2002 ACL Packet Structure Causes of frame error:  Failure to synchronize with access code. Sufficient for T>65 bits of the 72 to be correct.  Failure to decode the packet header. Protected by triple redundancy code.  Failure to decode the payload. Access Code Payload Packet Header 72 bits54 bits Payload Header Payload Data CRC 0-2744 bits 8 or 16 bits 16 bits 0-2712 bits 4/16

5. Throughput over BSC Channel 10 -5 10 -4 10 -3 10 -2 10 -1 0 100 200 300 400 500 600 700 800  Data Rate in kbps DH5 DH3 DH1 DM5 DM3 DM1 Data bits per frame Average number of ARQ transmissions Slots per frame 5/16

6. 5101520 0 100 200 300 400 500 600 700 800  =E s /N o in dB Data Rate in kbps DH5 DH3 DH1 DM5 DM3 DM1 Throughput in AWGN Performance of noncoherent & nonorthogonal FSK: 6/16 We assume h=0.32

7. 800 051015202530 0 100 200 300 400 500 600 700  =E s /N o in dB Data Rate in kbps DH5 DH3 DH1 DM5 DM3 DM1 Throughput in Quasi-Static Rayleigh Fading Quasi-static Rayleigh fading: SNR constant for entire frame. Varies from frame to frame. SNR is exponentially distributed. Average throughput. 7/16

8. © 2002 Custom Error Control The AUX1 packet  A seventh ACL packet type.  Occupies one slot.  CRC & ARQ are turned off. Operates as a “noisy bit pipe”. Whatever is received is passed up to application.  29 bytes of payload data. Can use AUX1 to transport a custom code  Implement FEC & ARQ on host computers. Sender: First CRC encode, then FEC encode. Any FEC code can be used: BCH, Reed Solomon, turbo. Some FEC codes can also perform error detection. Receiver: Decode FEC code, then CRC code. If errors, must manually ask for retransmission.  No modification of Bluetooth standard is needed. 8/16

9. Data Rate in kbps 55.566.577.588.599.510 0 50 100 150 E s /N o in dB BCH coding bound 1  t  43 BCH t=10 DM1 DM 3 Example: BCH Coding in AWGN 9/16 Notes: Used 16 bit CRC plus (232,k) shortened BCH code t is the error correction capability of the code up to 2 dB gain by using custom coding

10. © 2002 Adaptive Rate Control Optimal packet type depends on instantaneous SNR. Can select the packet to match the current SNR.  If custom coding is used, then can also pick the code parameters (e.g. t). Most of the benefit comes from selecting from among a small set of packets.  Set {DH5, DM5, and DM1} gives most of the gain.  CQDDR is a protocol from CSR (David McCall) which operates under same principle. Problem is that the channel SNR must be known a priori (predicted).  An alternative approach is to use hybrid-ARQ with incremental redundancy (which is “blind”). 10/16

11. Adaptive Coding for Quasi-Static Fading Adaptive BCH: Use AUX1 to transport custom code. Adapt t to match instantaneous SNR Fully Adaptive: Choose from among 6 standard packets. Can also choose a custom coded AUX1 packet. Gain is 1.5 dB. The set {DH5,DM5,DM1} yields almost same performance (within 0.1 dB). 0510152025 0 100 200 300 400 500 600 700 800 E s /N o in dB Data Rate in kbps DM1 DM3 DM5 DH5 BCH10 Adaptive BCH “Fully” Adaptive 11/16

12. © 2002 Antenna Diversity Performance in fading can be improved by using multiple (receive) antenna elements. Best performance improvement is achieved using maximal ratio combining.  However, this is too complex and requires coherent detection. Instead, we perform post-detection combining on a packet level.  Use CRC to determine if packet is correct or not.  If a packet is correct at any antenna, then it will be accepted by the system.  Packet is only needs to be retransmitted if it is incorrect at all antennas.  Note that this requires a separate receiver for each antenna. 12/16

13. 5101520253035 0 100 200 300 400 500 600 700 800 Average E s /N o in dB Average Throughput (kbps) M=1 M=2 M=6 Gain @500 kbps 3.2 dB for M=2 6 dB for M=6 Performance of Bluetooth With M-antenna elements Using packet-level combining Of DH5 packets In quasi-static Rayleigh fading 13/16

14. Distributed Detection Packet-level combining required the M antennas to be attached to M transceivers.  No reason why they must be colocated.  The transceivers could be connected through a backbone as in an infrastructure-based WLAN.  Detection is distributed over space.  When the mobile is equidistant to the M transceivers, performance is as if they are connected to the same device.  However, the diversity advantage diminishes if mobile not in center. MS location A AP #1 AP #2AP #3 AP #4 AP #5 AP #6 MS location B 14/16

15. 5101520253035 0 100 200 300 400 500 600 700 800 Average E s /N o in dB Average Throughput (kbps) M=1 M=6 Gain @500 kbps 0.4 dB for M=2 1 dB for M=6 Performance of Bluetooth With M-antenna elements Using packet-level combining Of DH5 packets In quasi-static Rayleigh fading When mobile is at location B 15/16 MS location A AP #1 AP #2 AP #3 AP #4 AP #5AP #6 MS location B

16. © 2002 Conclusion Several strategies can be used to improve performance of Bluetooth.  Each strategies complies with standard.  Custom coding: Use AUX1 to transport custom BCH code.  Adaptive rate control: Match the frame type to prevailing channel condition.  Antenna diversity: Use M antennas, but combine at packet level. Antennas don’t need to be co-located. Multiple Bluetooth devices can mimic antenna array. Future work:  Channel tracking and prediction. Hybrid ARQ with incremental redundancy.  Actual implementation of these strategies. Validation of channel models.  Application of similar concepts to 802.11 16/16