Scrutinizing bit-and symbol-errors of IEEE 802.15.4

Communication in Industrial Environments

Filip Barac, Student Member, IEEE, Mikael Gidlund, Member, IEEE, and Tingting Zhang, Member, IEEE

TIM(2013)

Why study error properties• 1.first step in protocol design

• Crucial for designing higher layer protocals• Facilitates the design of FEC

coding,interleaving,retransmission schemes

• 2.bit and symbol-level errors offer more subtle channel-state information• The commonly observed parameters can’t (packet loss,delay)• e.g error pattern,burstiness,ber

Works:1.Study of error properties2.Optimal choice of channel coding

Error Sources1.Physical Environment

Error Sources2.Electromagnetic Interference (e.g WLAN)

Experimental Setup:wlan setup

beacon

predefined content

1.Error distribution

2.Burstiness

3.Channel memory

4.Ber

1. Bit-Error Distribution

1. Bit-Error Distribution

1.randomly placed bit-errors 2.periodic

——Without Wifi Interference

1. Bit-Error Distribution

a ramp like pattern

——With Wifi Interference

2.Bit- and Symbol-Error Burst Length

1.90% bursts are no more than 5 bits

2.single-symbol-error bursts dominate in 802.15.4

3. Channel Memory Length

The burst nature of WLAN-affected errors

4.The Bounds on BER• The performance of FEC

codes depends on how often the number of errors exceeds the correcting capability

• 99.14% of packets corrupted by MFA had a BER≤10% and the mean BER is 1.88%.

• The mean BER averaged over all WLAN experiments is 9.51%

Implication on channel coding selection• Two criteria

• Error correction performance• Computational complexity

• Turbo and LDPC• Good correcting ability• En/decoding slow: hardware-accelerated implementations

result in encoding times in the order of 6–7 ms• Reed-Solomon code

• A tradeoff• require decoding times below 1 ms for certain block lengths

RS(?,?)• 1.Suitability With Respect to Bit-Error Burst

Properties• Burst length is no more than 5 bits• m = 4,n = 2^4 – 1 = 15;RS(15,?)

• 2.Timing Constraints of IEEE 802.15.4-2006-Based Standards• RS(15,7) is the strongest code satisfy the following

constraints:

codeword

symbol

codeword codeword codeword

Interleaving

Implication 1: The Lower Bound of Plain RS(15,7)Performance Under MFA• a measurable named packet salvation ratio (PSR)

• Absoulte improvements of pdr introduced by RS(15,7) in several experiments on links under MFA

• it is not possible to bring quantitative conclusions about RS(15,7) performance under WLAN interference.

Implication 2: BI Versus SI and the Optimal Interleaving Depth

1.SI outperforms BI on both types of link 2.The optimal interleaving depth corresponds to the codeword length

Implication 3: The Gain of SI on Links Affected by MFA• In MFA experiments

• The contribution of SI negligible• BI reduces the PSR in corrupted packets

• So interleaving is not recommendable for default use• should be activated when interference occurs

• FEC and interleaving are preferable to retransmissions• 2,700 times more energy to send one bit than to execute

an instruction

Conclusion• Scrutinizing nature of bit- and symbol-errors of IEEE

802.15.4-2006 transmissions• Bers,symbol burst lengths• Channel memory• Bit-error-bursts : less than five in most cases

• The evaluation on channel coding and interleaving• BI should not be considered for practical

implementations

Thanks

Contributions• 1.a number of conclusions about bit- and symbol-error

behavior • 2.two distinct error patters are identified• 3.the performance of a sufficiently lightweight channel

code is evaluated on the collected error traces• 4.it shows that SI(symbol interleaving) outperforms it

counterpart• 5.the interleaving gain is proven to be negligible on links

affected by MFA • 6.FEC and interleaving are a must on links under

IEEE802.11 interference