1. Proposal for Statistical Channel Error Model
doc.: IEEE n-proposal-statistical-channel-error-model.ppt
Jan 2004
Proposal for Statistical Channel Error Model
D. Maniezzo, G. Pau, F. Benedetto, F. Cerioli, M. Gerla,
W. Zhu, M. Fitz. UCLA – University of California, Los Angeles
Valerio Filauro
STMicroelectronics, Inc.
UCLA - STMicroelectronics, Inc.

2. doc.: IEEE 802 11-04-0012-00-000n-proposal-statistical-channel-error-model.ppt
Jan 2004
Outline
Introduction
Statistical Channel Model from Testbed Measurements
Tested Configuration
Instant PER vs SNR
Channel Model Assumptions
Results
Conclusions and Future Work
UCLA - STMicroelectronics, Inc.

3. doc.: IEEE 802 11-04-0012-00-000n-proposal-statistical-channel-error-model.ppt
Jan 2004
Introduction
A realistic channel simulation model is mandatory to study a PHY/MAC layer such as n that takes into account the channel conditions.
The channel conditions affect the throughput, the delay and the jitter.
Goal of the simulator: measure network throughput as a function of MIMO parameters such as:
the geometry of the MIMO system,
the model of the channel,
the distance of the stations, etc;
A statistical error model gives us a faster simulator than if we integrated TGn Matlab channel model [1] with the selected network simulator.
UCLA - STMicroelectronics, Inc.

4. Jan 2004
Testbed: 3x4 MIMO System
UCLA - STMicroelectronics, Inc.

5. Statistical Channel Model from Testbed Measurements
Jan 2004
Statistical Channel Model from Testbed Measurements
From real MIMO measurements from the Narrowband Testbed [2], we derive Packet Error Statistic vs. SNR and use it in the network simulator.
We select as a network simulator ns-2 but the approach applies to any simulator.
Test-bed (in UCLA UnWiReD Lab ):
The receiver in fixed location; the transmitter is in different corridor locations for different experiments
Transmission power is -10 dBm for each transmit antenna.
24 different Space-Time coding schemes are tested
Alamouti: 4PSK, 8PSK, 16QAM, 64QAM.
3-TX Orthogonal Block Code: 4PSK, 8PSK, 16QAM, 64QAM.
3-TX Super Orthogonal Block Code with 2 extra bits: 4PSK, 8PSK, 16QAM, 64QAM
3-TX Super Orthogonal Block Code with 3 extra bits: 4PSK, 8PSK, 16QAM, 64QAM.
3-TX Super Orthogonal Block Code with 4 extra bits: 4PSK, 8PSK, 16QAM, 64QAM.
3x4 BLAST: 4PSK, 8PSK, 16QAM, 64QAM.
The measurements are repeated for different receiver positions to have PER vs. SNR.
UCLA - STMicroelectronics, Inc.

6. Testbed Configuration
Jan 2004
Testbed Configuration
Physical layer parameters
Carrier Frequency: MHz
Bandwidth: 4KHz
Symbol Rate: 3.2 kbps
Antenna Configuration: Up to 3x4
No mobility
UCLA - STMicroelectronics, Inc.

7. Testbed Configuration (cont’d)
Jan 2004
Testbed Configuration (cont’d)
Frame Format The transmitter continuously transmits Long frames with the following structure:
Preamble (300 known symbols): used for frame and carrier synchronization.
Control Frame (Alamouti coded frame): used for transmission of the state (seed) of the random number generator, which is used to generate the random information bits to be coded.
Data Block: consists of 24 Data Frames. Each Data Frame
carries random information bits encoded using one of the 24 Space-Time Coding Schemes.
consists of 300 symbols (228 of which are data and 72 are pilot symbols).
Silence Period between Long Frames (70 symbols): the transmitter is silent. The receiver uses this period to estimate the received noise power in the operating frequency band. …
UCLA - STMicroelectronics, Inc.

8. values averaged over a 2 hours experiment
Jan 2004
values averaged over a 2 hours experiment
UCLA - STMicroelectronics, Inc.

9. Channel Model Assumptions
Jan 2004
Channel Model Assumptions
SISO (1x1) instead of MIMO (3x4); single carrier (instead of OFDM)
Propagation model:
Exp. Path loss model
AWGN
Shadowing
No Fading (it is compensated in the real system by: multiple antenna diversity and multiple frequency subchannel spreading).
Error Model: from the measurements we set a look-up table that gives the PER vs. SNR.
Scaling from 4Khz band and 200Mhz carrier to 5Mhz and 2.4Ghz: As a first approximation, for a given encoding scheme, the error rates are dependent only on SNR.
As a first approximation, bit errors are assumed to be randomly and uniformly distributed in a frame; packet error rate is computed based on packet length. Future work will investigate error burstiness
Note: One caveat is that at higher speeds, and using OFDM, better, more robust codes could be used. Anyway, current 4Khz results will provide a conservative estimation
UCLA - STMicroelectronics, Inc.

10. Channel Model Assumptions (cont’d)
Jan 2004
Channel Model Assumptions (cont’d)
Doppler effects:
Doppler effects are present as people move in the environment changing the multipath characteristics and thus causing a drift in carrier frequency (with possible extra errors). This effect is negligible in an indoor slow mobility scenario.
Multipath fade changes and obstruction of direct ray:
In this simplified model we didn’t take into account the relatively slow change in attenuation caused by people moving in the environment.
In the future, careful monitoring of this attenuation in the experiment could be used to set up a proper 2 state Markov Chain to capture these effects.
UCLA - STMicroelectronics, Inc.

11. MAC/PHY Wireless Model in ns2
Jan 2004
MAC/PHY Wireless Model in ns2
MAC
Error Model
Radio Propagation Model
netIF
Received Power
Free Space Shadowing
Wireless Channel
UCLA - STMicroelectronics, Inc.

12. Experimental Scenarios
Jan 2004
Experimental Scenarios
We simulate the following scenarios proposed by TGn Usage Models workgroup:
Residential (#1)
Residential IBSS (#2)
Enterprise (#4)
Hot Spot (#6)
Note: the same scenarios were simulated in [3] by ST-Microelectronics (without error model).
UCLA - STMicroelectronics, Inc.

13. Jan 2004
Simulation Set-up
ns-2.26
MAC: STMicroelectronics extension “MAC b/QoS” [3]
PHY:
Data Rate: 300Mbps
TX power: 17 dbm
Channel:
300Mbps
Exp. Path loss model + AWGN + Shadowing + error model
Simulation time 60 secs
UCLA - STMicroelectronics, Inc.

14. Jan 2004
Results - Scenario #1
UCLA - STMicroelectronics, Inc.

15. Jan 2004
Results - Scenario #2
UCLA - STMicroelectronics, Inc.

16. Jan 2004
Results - Scenario #4
UCLA - STMicroelectronics, Inc.

17. Jan 2004
Results - Scenario #6
UCLA - STMicroelectronics, Inc.

18. Jan 2004
Future Work
More realistic measurements will be available shortly from a new 2.4Ghz testbed.
More experiments will be conducted to evaluate error burstiness and obtain a more accurate dependence of packet error probability on packet size.
A two states Markov chain error model will be developed to take into account the time correlation of the errors (e.g. attenuation caused by people moving in the environment).
The knowledge of time correlation between errors will be exploitedfor optimal parameter setting (eg, fragment aggregation size).
Model parameterization will be used to limit the size of the look-up table (ie PER vs. SNR) as the system becomes more complex
UCLA - STMicroelectronics, Inc.

19. Jan 2004
References
[1] n.doc TGn Channel Model, November 2003.
[2] Narrowband Testbed webpage:
[3] “ TGn Usage Models Simulation Results”, Valerio Filauro, Liwen Chu – STMicroelectronics Inc. IEEE n.doc
UCLA - STMicroelectronics, Inc.