Shuchi Chawla, Huzur Saran, Mitali Singh

1
QoS based Scheduling for incorporating Variable
Rate Coded Voice in BLUETOOTH

• Shuchi Chawla, Huzur Saran, Mitali Singh

June 13, 2001
2
BLUETOOTH A Wireless Personal Area Network

• At most 1 master 7 slaves in a piconet
• TDMA for medium access
• Slots of length 625µs
• MAC is master-driven

• Two kinds of links
• SCO Synchronous Connection Oriented (Voice)
• 64Kbps for voice
• ACL Asynchronous Connection Less (Data)

3
TDD in Bluetooth

• Master slave transmit alternately

4
Voice BW requirements

• Toll quality voice -- 64Kbps
• Compressed voice -- much lesser
• Empirical studies
• 60 inactivity in phone calls
• Variable Rate Coding (VRC) based on activity
• 23.8 Kbps and 11.4 Kbps
• More than 110 saving!!

5
Scheduling Voice in Bluetooth

• Standard Voice over SCO
• Throughput improvement by Activity based Variable
  Rate Coding of voice.
• Schemes proposed
• Adaptive TSCO
• Voice over ACL

6
Adaptive Tsco

• Tsco period after which SCO connection is
  scheduled
• Increase Tsco when lesser bandwidth for voice is
  required

• Talkspurt Tsco 6 ? 64Kbps
• Silence Tsco 12 ? 32Kbps

7
Voice over ACL

• QoS based scheduling scheme used
• Earliest Due Deadline policy
• Deadline arrival time acceptable latency
• Latency for voice
• normally up to 100ms acceptable
• Avoid queuing of more than 1 packet
• Bandwidth requirement of 23Kbps imposes a max
  latency of 11ms ? 18 time slots.

8
Voice over ACL Issues

• Paging All ACL connections put on hold for 18ms
• We cannot put voice on hold for so long
• During paging ACL voice also scheduled as SCO
• This again limits max voice connections to 3.
• ACL packets fragmented and reassembled at L2CAP
• Send voice down the stack as SCO
• At MAC layer schedule them as ACL
• Assumption Knowledge of slave status to master.
• Separate Reverse Channel for this info
• PiggyBacking requires very few bits

9
Simulation results
1 voice connection
10
Simulation results
2 voice connections
Scheme implemented
Data Throughput (Kbps)
Average delay (msec)
9.98
358.4
Without voice
11
Simulation results
3 voice connections
Scheme implemented
Data Throughput (Kbps)
12
Conclusions

• Scheduling over ACL performs best, but changes to
  standard are required
• Second option Adaptive Tsco, considerable
  improvement in throughput over current scheme

13
Bluetooth protocol stack
14
Paging in Bluetooth