IEEE 802.19 Wireless Coexistence TAG

1
IEEE 802.19Wireless Coexistence TAG
An Analytic Coexistence Assurance Model

• Steve Shellhammer
• shellhammer_at_ieee.org

2
An Analytic CA Model

• Make reasonable approximations of PHY and MAC
  layers.
• Provide a method of predicting the impact of
  interference in a timely manner.
• Not a detailed model intended to predict absolute
  performance of either system.
• Is intended to predict relative impact of
  interference.
• Only considering non-hoppers at this point
• Intended as a first-order approximation.

3
Model of Interferer

• Interferer sends pulses
• When transmitting a pulse the interferer is
  models in the frequency domain as band-limited
  white noise of power PT

PT
B
fc
fc B/2
fc - B/2
4
Model of Interferer

• Based on our knowledge of the interferer traffic
  the temporal model of the interferer is a
  stochastic process of pulses. Need to consider
  various models.
• Distribution of pulse durations
• Distribution of spacing between pulses

5
Model of Interferer
TP
TS
TP
TP
TS

• Pulse TP duration is a random variable
• Space TS between pulses is a random variable.

6
Example of Pulse Model

• The interferer is sending TCP IP packets.
• There is an AP far away sending ACK packets. So
  we dont consider this an interferer.
• Throughput is about half the data rate.
• TP 1.0 ms
• TS is a uniform RV
• TS U(30, 1300) us

7
Path Loss Model

• Some standard path loss model will be
  recommended, like the one used in 802.15.2.
• Other path loss models could be used.
• Give a topology of devices you can determine the
  interference power at the receiver based on path
  loss model.
• pl(d) path loss in dB, with d in meters.

8
Topology of Wireless Devices

• One possible topology

Transmitter
Is not interfered with due to distance from
interferers
Does not interfere due to distance from NUT
System A Network Under Test
Receiver
d
System B Interferer
Primary Interferer
9
Receiver Model

• Model receiver filter as an ideal brick wall
  filter, as far as interference goes.
• The portion of the interfering signal that is
  within the passband of the receiver filter is
  pass though undisturbed
• Any portion of the interfering signal outside the
  filter passband is eliminated entirely.

10
Receiver Model
NI
Interferer PSD at Receiver
1
Receiver Filter
NI

• Noise after the receiver filter is the same
  height as before the filter, but possibly a
  smaller bandwidth

11
Bit Error Rate

• It is assumed that there is formula for BER for
  the receiver in AWGN.
• ber(?) BER versus SNR for AWGN.
• There are two periods of stationarity when we
  want to calculate the BER (which will help us get
  PER)
• During a portion of the received packet when
  there is no interference during the packet
• During a portion of the received packet when
  there is no interference during the packet

12
Bit Error Rate

• BER when there is no interference is based on
  thermal noise.
• Since this is not very high we can
• Assume it is very low
• Or set up realistic topology and calculate BER
• Since absolute performance is not a primary
  concern method one is recommended.

13
Bit Error Rate

• BER when interference is present is based on
  equivalent AWGN.
• Pick AWGN level that would give equivalent power
  after the receiver filter.

14
Bit Error Rate
1
Receiver Filter
BF
NI
BAF
BAF
)
(
NI
BF
BAF
15
Effective AWGN

• Power after receiver is NI BAF
• To get the same power after filter we have to
  have,
• Neff BF NI BAF
• The issue is that the interfere may not be as
  wide as filter. So we are dropping the PSD and
  widening the bandwidth
• This is another approximation

16
Bit Error Rate Summary

• We now have a method to calculate the BER when
  there is no interference and when there is
  interference.
• Calculate Eb from path-loss
• With no interference use N0
• With interference use Neff
• Can also add N0 to Neff

17
Packet Error Rate

• A packet in the network under test (NUT) is sent
  from transmitter to the receiver. There is a
  (possible) overlap between that packet and an
  interfering pulse.

TD
T
18
Probability Analysis

• Calculate probability density function for the
  random variable T. (Work still to be done).
• T is a mixed random variable. There will be a
  finite probability that T is zero, and some
  density function over the interval (0,TD)

19
Probability Density of T

• An example of a PDF for T

½
fT(t)
1/(2T)
0
TD
20
Packet Error Rate

• Step 1
• Calculate PER for a fixed value of T
• Step 2
• Average over all values of T using the previously
  calculated PDF for T
• Step 3
• If necessary, average over packet duration, TD,
  assuming it is variable

21
Other Metrics

• Calculate other metrics based on PER and
  necessary approximations (e.g. independence)
• Throughput
• Latency
• Packet Loss Rate (assuming a fixed time to
  complete transmission)
• Other

22
Conclusions

• Outlined an approach to analytic solution.
• Next steps
• Work out technique for determining PDF of
  collision time.
• Write up document giving details.
• Apply to an example and use for comparison with
  other techniques.