Soft-in/ Soft-out Noncoherent Sequence Detection for Bluetooth: Capacity, Error Rate and Throughput Analysis

1
Soft-in/ Soft-out Noncoherent SequenceDetection
for BluetoothCapacity, Error Rate and
Throughput Analysis

• Rohit Iyer Seshadri and Matthew C. Valenti
• Lane Dept. of Computer Science and Electrical
  Engineering
• West Virginia University
• iyerr, mvalenti _at_csee.wvu.edu

2
Objectives

• Achieve dramatic improvements in energy
  efficiency and throughput for Bluetooth with a
  minimal increase in complexity by using
• Sequence based, noncoherent demodulator
• Bit-interleaving
• Soft-decision decoding
• Feedback from channel decoder to demodulator
• Obtain an information theoretic bound on the
  minimum signal to noise ratio required for
  reliable signaling
• Bit-wise log-likelihood ratios used to compute
  Shannon capacity under modulation, channel and
  receiver design constraints
• Demonstrate performance improvements over popular
  receivers using an extensive simulation campaign
• Evaluate packet error rate (PER) and throughput
  performance for data medium (DM) - rate packet
  types

3
Bluetooth

• Low cost/ low power connectivity for wireless
  personal area networks
• Operates in the license free 2.4 GHz ISM band
• Band divided into 79 channels, each 1 MHz wide.
  Channels changed up to 1600 times per second
• Channel symbol rate of 1 Mbps
• Uses Gaussian frequency shift keying (GFSK)
• M 2
• BgT 0.5
• 0.28 h 0.35

4
Benchmark Bluetooth System
Encoder (15, 10) Shortened Hamming Code (SHC),
single error correction code
Baseband GFSK signal during kT t (k1)T
GFSK phase
Detector Limiter discriminator integrator (LDI)
5
Bluetooth System with Sequence Detection
GFSK pulse shape causes adjacent symbol
interference
Detector Soft-Decision differential phase
detector with Viterbi decoding (SDDPD-VD),
Fonseka, 2001
Viterbi decoding can produce burst errors, which
could be mitigated by bit-interleaving
6
Bluetooth System with SISO-SDDPD
SDDPD-VD forms hard estimates on code bits
SISO-SDDPD generates bit-wise LLRs for the code
bits
LLRs from detector passed to decoder, which
performs soft-decision decoding
Bit-interleaved coded modulation (BICM)
Additionally, soft-information can be also be fed
from decoder to detector BICM with iterative
decoding (BICM-ID)
No gains over BICM Behavior explained using EXIT
curves
7
SISO-Soft-Decision Differential Phase Detection

• Received signal at the output of a frequency
  nonselective, Rician channel, before filtering
• r(t, a) c(t) x(t, a) n(t)
• Received signal after filtering
• r(t, a) c(t) x(t, a) n(t)
• Received signal phase

(t, a)
(t, a)
8
SISO-Soft-Decision Differential Phase Detection

• Detector finds the phase difference between
  successive symbol intervals
• The GFSK pulse shape causes adjacent symbol
  interference
• The phase difference space from 0 to 2? is
  divided into R sub-regions
• Detector selects the sub-region Dk in which
  lies
• The sequence of phase regions (D0, DI, ) is
  sent to a branch metric calculator

9
SISO-Soft-Decision Differential Phase Detection

• Let be the phase
  differences corresponding to any transmitted
  sequence
• A branch metric calculator finds the conditional
  probabilities
• Branch metrics sent to a 4-state MAP decoder
  whose state transition is from
• to
• The SISO-SDDPD estimates the LLR zk for ak as

10
FEC for Bluetooth

• Bluetooth specifies 7 types of ACL packets for
  data transfer
• 6 out of the 7 packet types use cyclic redundancy
  check (CRC) and ARQ
• 3 out of these 6, i.e. data medium (DM1, DM3,
  DM5) also use a (15, 10) shortened Hamming code
  (SHC) for forward error correction (FEC)
• The (15, 10) SHC is cyclic and described by the
  generator polynomial
• The cyclic code can hence be expressed using a 25
  32 state trellis and decoded by running either
  a Viterbi or MAP algorithm over the trellis

11
Capacity Under Modulation, Channel And Receiver
Design Constraints

• Channel capacity denotes maximum allowable data
  rate for reliable communication over noisy
  channels
• In any practical system, the input distribution
  is constrained by the choice of modulation
• Capacity is mutual information between the bit at
  modulator input and LLR at detector output
• Constrained capacity in nats is Caire, 1998

12
Capacity Under Modulation, Channel And Receiver
Design Constraints

• Constrained capacity for the proposed system is
  now
• In bits per channel use
• Constrained capacity hence influenced by
• Modulation parameters (M, h and BgT)
• Channel

13
Performance Evaluation and Comparisons

• Performance of proposed SISO-SDDPD with BICM
  compared against
• Limiter discriminator integrator detector with
  hard decision channel decoding, with and without
  bit-interleaving LDI-HDD
• SDDPD-VD, followed by hard decision channel
  decoding, with and without bit-interleaving
  SDDPD-HDD
• SISO-SDDPD followed by soft decision channel
  decoding, without bit-interleaving
  SISO-SDDPD-SDD
• Comparisons made on the basis of
• Bit error rate
• Packet error rate
• Throughput

14
Bit Error Rate Comparison
Scenario Minimum Eb/No to achieve BER 10-4. Six
simulated points from top to bottom are 1)
LDI-HDD 2) LDI-HDD with interleaving 3)
SDDPD-HDD 4) SDDPD-HDD with interleaving 5)
SISO-SDDPD-SDD 6) SISO-SDDPD with
BICM Information theoretic bound for SISO-SDDPD
based BICM SDDPD specifications
R24 uniform sub-regions Channel
parameters Nonselective, Rician fading, K 2
dB Bluetooth specifications h 0.315, DM1
packet types SISO-SDDPD with BICM gives the best
BER performance
15
Packet Error Rate Comparison
Scenario Packet error rate for DM1 packet types.
SDDPD specifications R24 uniform
sub-regions Channel parameters Nonselective,
Rician fading, K 2 dB Bluetooth specifications
h 0.315 DM1 packet types SISO-SDDPD with
BICM gives the best packet error rate
performance. Gain over LDI based systems 9
dB Gain over SDDPD-HDD based systems 4 dB
16
Throughput Comparison
Scenario Throughput for DM1, DM3 and DM5 packet
types Solid curve Systems without
interleaving Dotted curve Systems with
interleaving SDDPD specifications R24 uniform
sub-regions Channel parameters Nonselective,
Rician fading, K 2 dB Bluetooth specifications
h 0.315 SISO-SDDPD with BICM gives the best
throughput performance For maximal throughput,
packet type should be adaptively selected to
match SNR
17
Conclusions

• An energy efficient, noncoherent receiver design
  investigated for Bluetooth
• Soft-in/ soft-out, soft decision differential
  phase detector developed
• BICM paradigm applied to Bluetooth
• Error rate and throughput compared against LDI
  detector and Fonsekas SDDPD with Viterbi
  decoding
• SISO-SDDPD-SDD shown to outperform LDI-HDD and
  SDDPD-HDD
• Additional gains possible with interleaving
• Constrained capacity found using Monte Carlo
  simulations

18
Future Work

• An algorithm that designs nonuniform phase
  regions using received phase differences and
  adapts itself to varying channel conditions and
  GFSK parameters
• Nonunifrom regions can perform better than
  uniformly phase regions Fonseka, 1999
• Results in a smaller look-up table
• Estimating the Rician K factor and Eb/No at the
  receiver using the Expectation-Maximization
  algorithm

19
Complexity

• Branch metric calculations in SISO-SDDPD
• Metric calculations involve nonlinear functions
• Pre-calculated and stored in a
  look-up table
• Table needs to be updated once at each Eb/No
• Number of states in the detector
• SISO-SDDPD operates on a M2- state trellis
• Number for states in the channel decoder, with
  soft-decision decoding
• ML/ MAP decoding performed on a 32- state trellis

20
Sensitivity to h estimation errors
Scenario Effect of incorrect estimates of h on
SISO-SDDPD and LDI detectors SDDPD
specifications R24 uniform sub-regions Channel
parameters Nonselective, Rician fading, K 2
dB Bluetooth specifications Correct value of h
0.315 Values assumed at detector 0.28, 0.35
DM1 packet types SISO-SDDPD more robust to
incorrect estimates of h
21
EXIT Chart
Scenario EXIT chart for the SISO-SDDPD based
BICM receiver SD-DPD specifications R24
uniform sub-regions Channel parameters Nonselect
ive, Rician fading, K 2 dB Bluetooth
specifications h 0.315, BgT 0.5 Detector
EXIT curve predicts no improvement with BICM-ID
22
Throughput Calculations

• Throughput Maximum achievable, one way data rate
  Valenti, 2002
• Nt Total number of times a given packet
  must be transmitted (on an average) until it is
  successfully decoded
• Ns Number of slots occupied per round
  trip, including one return slot
• Duration of each slot 625 µsec
• Ku Number of data bits in the packet type