1. doc.: 11-14-0846-00-00ax Submission July 2014 Slide 1 Proposed Calibration For MAC simulator Date: 2014-07-12 Authors:

2. doc.: 11-14-0846-00-00ax Submission Introduction Multiple IEEE presentations have looked at the effects of CCA levels. –See next slide –Many claim significant gain by increasing CCA levels. –Numerous types of scenarios/simulators were used to make these claims –PHY simulator –Real devices –MAC simulator/scenario 3 This presentation focuses on understanding the effect of changing CCA levels in a residential scenario ( scenario 1) –Simulations are done with a MAC simulator Detailed MAC modelled Simple PHY abstraction – Optimal CCA levels for both mean and 5% TPUT are found for a variety of settings. July 2014 Barriac et al (Qualcomm)Slide 2

3. doc.: 11-14-0846-00-00ax Submission Previous Contributions on CCA levels July 2014 Barriac et al (Qualcomm)Slide 3 Document #CompanyTitleNotes 11-14/0628r0SK telecommeasurements on CCA Thresholds in OBSS environments increasing CCA can help target APs overcome deferral due to ACI interference in dense setting. ( experimental results) 11-14/0578r0InterdigitalResidential Scenario CCA/TCPsimulation Discussion Optimal CCA for mean TPUT varies with Txpower, and reuse. 11-14/0523r0OrangeMAC simulations results for Dynamic sensitivity Control ( CCA adaptation) and tranmist power control CCA and TCP are strongly increasing reuse and aggregate throughput 11-14/0635r1DSP GroupDynamic Sensitivity Control implementation CCA level per STA depends on distance to AP. 11-14/0082r0 & 83r0BcomImproved Spatial Reuse Feasibility part I & II 2x-3x gains in both the mean and tail by increasing CCA levels. Densifi: date: 2014- 06-04 Bcomperformacne Gains from CCA Optimiziation

4. doc.: 11-14-0846-00-00ax Submission Simulation Setting I July 2014 Barriac et al (Qualcomm)Slide 4 Residential Scenario –IEEE residential scenario baseline Less apartments modelled for faster run times – Expect comclusions will not change 1 STA per AP –2x5x3 apartments –30 APs ; 30STAs ; –Pathloss as specified in 11ax Simulation Scenarios document –1 80 MHz channel –Results averaged over multiple random drops of APs and STAs locations Floor 1 Floor 2 Example of one drop Floor 3

5. doc.: 11-14-0846-00-00ax Submission Simulation Setting II Simulation Details: –DL only Data traffic (+ UL BAs) –Full buffer UDP –Default Tx Power: 20 dBm per antenna (AP), 15 dBm ( STA) –Antenna Gain: 0 dBi APs, -2dBi STAs –Rate Control: MCS per link chosen to maximize long term goodput – MAC modelling MAC is fully modelled Deferral according to 11ah CCA rules –OBSS packets deferral only if RSSI > ED (assume BSS ID (color) in PPDU ED level controls deferral on a network level Nodes on the same 80 MHz channel have the same primary channel – PHY modelling Pathloss, shadowing, (pathloss as in SS document, shadowing w sd 5dB) no fading. Long term PER curves based on ChD 1x1 (tx antennas x rx antennas) July 2014 Barriac et al (Qualcomm)Slide 5

6. doc.: 11-14-0846-00-00ax Submission Tests Run Using ChD PER curves: –Reuse 1 default power levels –Reuse 1, 6 dB higher tx power levels 6 dB higher power models total power when 4 antennas Using AWGN PER curves: –Reuse 1 default power levels –Reuse 1, 6dB higher tx power levels 6 dB higher power models total power when 4 antennas. July 2014 Barriac et al (Qualcomm)Slide 6

7. doc.: 11-14-0846-00-00ax Submission RESULTS WITH CHD PER CURVES July 2014 Gwen Barriac (Qualcomm)Slide 7

8. doc.: 11-14-0846-00-00ax Submission TPUT as a function of CCA for Reuse 1 July 2014 Barriac et al (Qualcomm)Slide 8 Optimal CCA for mean: -72 Optimal CCA for 5% point: -92 CCA level5 percentile50 percentile90 percentilemean -520.0010.85103.3035.25 -620.0923.44109.2540.35 -721.2730.32107.245.63 -822.8230.97100.8444.65 -924.5024.5474.9833.08

9. doc.: 11-14-0846-00-00ax Submission TPUT as a function of CCA for Reuse 1, with 6 dB higher Tx Power at both AP and STA July 2014 Barriac et al (Qualcomm)Slide 9 Optimal CCA for mean: -72 Optimal CCA for 5% point: -82 CCA level5 percentile50 percentile90 percentilemean -520.018.66105.5534.61 -620.1826.78120.9745.28 -722.3431.27109.5046.43 -822.7627.37102.5043.07 -921.7823.9161.0629.79

10. doc.: 11-14-0846-00-00ax Submission Possible Simultaneous mean & 5% TPUT July 2014 Qualcomm Slide 10 CCA level5 percentile50 percentile90 percentilemean -520.0010.85103.3035.25 -620.0923.44109.2540.35 -721.2730.32107.245.63 -822.8230.97100.8444.65 -924.5024.5474.9833.08 CCA level5% tput Loss compared to optimal Mean tput loss compared to optimal -92026% -8237%2% -7271%0% -6299%11% -52100%22% Reuse 1, default Tx Power, chD PER curves

11. doc.: 11-14-0846-00-00ax Submission Summary of Results so far ( chD PER curves) July 2014 Barriac et al (Qualcomm)Slide 11 Reuse 1 default TX power Reuse 1 6 dB Higher Tx power Optimal CCA for mean-72 Optimal CCA for 5%-92-82 For default power case, the optimal mean TPUT comes with a 70% loss in 5% point. –It can be difficult to optimize for both mean and 5% point simultaneously.

12. doc.: 11-14-0846-00-00ax Submission RESULTS WITH AWGN PER CURVES Note: Chan D curves better reflect reality. AWGN curves run for understanding of how simulator reacts in different settings. Slide 12Qualcomm

13. doc.: 11-14-0846-00-00ax Submission Reuse 1, Default TX power AWGN curves. July 2014 Barriac et al (Qualcomm)Slide 13 Optimal CCA for mean: -62 Optimal CCA for 5% point: -92 CCA level5 percentile50 percentile90 percentilemean -520.0015.03158.7352.21 -620.0030.09170.1662.05 -721.1941.85146.5060.38 -824.6740.63116.7353.21 -924.7925.4680.3635.84

14. doc.: 11-14-0846-00-00ax Submission Reuse 1, 6 dB higher Tx Power, AWGN curves July 2014 Barriac et al (Qualcomm)Slide 14 Optimal CCA for mean: -62 Optimal CCA for 5% point: -82 CCA level5 percentile50 percentile90 percentilemean -520.0010.29177.2155.50 -620.2136.59169.7164.05 -723.6341.80133.9657.99 -824.4933.44107.6147.54 -922.9426.3859.3530.91

15. doc.: 11-14-0846-00-00ax Submission Possible Simultaneous mean & 5% TPUT July 2014 Barriac et al (Qualcomm)Slide 15 Reuse 1, 6 dB higher Tx Power, AWGN curves CCA level5% tput Loss compared to optimal Mean tput loss compared to optimal -9234%51% -820 %26% -7219%11% -6295%0% -52100%14% CCA level5 percentile50 percentile90 percentilemean -520.0010.29177.2155.50 -620.2136.59169.7164.05 -723.6341.80133.9657.99 -824.4933.44107.6147.54 -922.9426.3859.3530.91

16. doc.: 11-14-0846-00-00ax Submission Summary of Results w/ AWGN July 2014 Barriac et al (Qualcomm)Slide 16 Reuse 1 default TX power Reuse 1 6 dB Higher Tx power Optimal CCA for mean-62 Optimal CCA for 5%-92-82 Difficult to optimize for both mean and 5% point simultaneously. –Optimal mean can come with 95% loss in 5% point

17. doc.: 11-14-0846-00-00ax Submission Comparison of Optimal CCA levels for AWGN and Chan D PER curves July 2014 Barriac et al (Qualcomm)Slide 17 Reuse 1 default TX power Channel D PER curves Reuse 1 6 dB Higher Tx power Channel D PER curves Reuse 1 default TX power AWGN curves Reuse 1 6 dB higher TX power AWGN curves Optimal CCA for mean -72 -62 Optimal CCA for 5% -92-82-92-82 Changing PER curves has large effect on optimal CCA level for mean Changing Power level also affects optimal CCA  Optimal CCA level sensitive to parameters.

18. doc.: 11-14-0846-00-00ax Submission Conclusions: Optimal CCA levels are highly dependent on parameter settings – example, Tx Power, PER curves Difficult to get simultaneously optimize both mean and 5% point July 2014 Barriac et al (Qualcomm)Slide 18

19. doc.: 11-14-0846-00-00ax Submission Additional Considerations We would like to propose that the rate control used in these studies be included in the EM document for use with the MAC simulator and integrated simulator. –See document xxxxx We would also like to propose that the deferral method used in these studies be used for the next stage of MAC simulator calibration. July 2014 Barriac et al (Qualcomm)Slide 19