PHY Abstraction for HEW System Level Simulation

Mar. 2014 doc.: IEEE 802.11-14/0335r0 Instantaneous SINR Calibration for System Simulation Date: 2014-03-17 Authors: Name Affiliations Address Phone email Yakun Sun Marvell Semiconductor 5488 Marvell Ln, Santa Clara, CA 95054 1-408-222-3847 [email protected] Jinjing Jiang Marvell Semiconductor Yan Zhang Marvell Semiconductor Hongyuan Zhang Marvell Semiconductor Submission Slide 1 Yakun Sun, et. al. (Marvell) Mar. 2014 doc.: IEEE 802.11-14/0335r0 Overview A step-by-step calibration was proposed in [1,2] with

high level descriptions. The first step of static radio statistics (long-term SINR) calibration has been presented in [3]. More companies have been worked together on the step-1 calibration [4]. We follow up on the next step of SLS calibration. PHY statistics (Freq-domain SINR distribution) Simulation Scenario Static Radio statistics (S/I distribution) Submission MAC calibration PHY Tput calibration Slide 2 Yakun Sun, et. al. (Marvell) Mar. 2014 doc.: IEEE 802.11-14/0335r0 Instantaneous SINR calibration The objective is to align physical layer receiver characteristics in a dynamic environment. Dynamic physical layer receiver characteristics reflect the frequency domain SINR calculation, small-scale fading channel generation, and equalization. Option 1: Instantaneous receiver-output SINR per tone Includes fading channels from both the desired transmitter and interferers Includes the MIMO receiver algorithms such as MMSE for MIMO cases Includes Doppler effects of channel generations Includes antenna correlation for MIMO cases Option 2: Effective SINR per frame Also include all the physical layer factors as in option 1

Essential value for later PER decision Less number of values to save Aligning effective SINR implies aligning PER/throughput (to some extent) Submission Slide 3 Yakun Sun, et. al. (Marvell) Mar. 2014 doc.: IEEE 802.11-14/0335r0 Instantaneous SINR calibration (2) Option 2a: alternative to option 2, use (SNReff) Given the convergence to an upper bound (RBIR, MMIB), effective SNR is sensitive to mapping offsets (in different implementation) at high SNR region. Avoid the ambiguity at high SNR by using (SNReff) as a bounded value, 256QAM 8 256QAM 64QAM 7 Mutual Information, (SNR) 6 5 4 3 2 1 0 -10 Submission -5 0 5 10 SNR (dB) Slide 4

15 20 25 30 Yakun Sun, et. al. (Marvell) Mar. 2014 doc.: IEEE 802.11-14/0335r0 Comparison of Option 1 and 2 Option 1: Pro: to avoid using the same PHY abstraction method, easier to agree and implement Con: less strong physical meaning Option 2: Pro: strong physical meaning (effective SNR per frames can be easily translated to PER, and infer throughput). Con: Need a unified PHY abstraction method (lack of consensus at this moment) Need to watch out the mapping offsets at high SNR (avoided by option2a) Submission Slide 5 Yakun Sun, et. al. (Marvell) Mar. 2014 doc.: IEEE 802.11-14/0335r0 Procedure of Statistics Collection Detailed PHY is assumed Fading channel models, Doppler spectrum, and antenna correlation (if MIMO) are defined by the scenarios Receiver algorithm is reflected (MMSE for MIMO, or MRC for single stream) Effective SNR per frame (mapping can be done for an agreed modulation level other than the MCS of the frame) PER decision is not required at this step (always successfully decoding the packet) Some simplest MAC is assumed. CCA-only, basic CSMA, or EDCA with the same AC for all STAs/APs. Full buffer traffic

Each AP and STA transmits a packet of a fixed (and equal) size at a fixed MCS. Multiple drops of AP/STAs are simulated for a scenario In each drop, collect the physical layer receiver characteristics observed at each STA/AP for each packet. Only collect the data frame (exclude beacons, etc.) Generate the distribution (CDF) of dynamic physical layer receiver characteristics at STAs (downlink) and APs (uplink) over multiple drops. Submission Slide 6 Yakun Sun, et. al. (Marvell) Mar. 2014 doc.: IEEE 802.11-14/0335r0 Simulation Setup Simulation is based on scenario 1 to 4 in [5]. Distribution of uplink instantaneous SINR are plotted as an example. We can select only one scenario for calibration. Detailed/optional simulation assumptions: 2.4GHz Channel with 20MHz Bandwidth No antenna gain, no cable loss 1 Tx and 1 Rx are assumed (other than defined in [5]) EDCA with AC2 for all STAs/APs (using default parameters) MCS 7, each packet of 1584 bytes STAs and APs are dropped and associated based on scenario [5] Submission Slide 7 Yakun Sun, et. al. (Marvell) Mar. 2014 doc.: IEEE 802.11-14/0335r0 Simulation Assumptions (Scenario 1) Parameter Value Number of STAs 4 STAs per apartment Channel Model

TGn B (AP-AP, STA-STA, AP-STA) Penetration Loss Wall 12dB, Floor 17dB, linear for multiple walls/floors BW 20MHz at 2.4GHz. Each BSS randomly selects one channel out of 3. TX Power AP: 23dBm, STA: 17dBm Association 100% STA in an apartment associated with the AP in the room. Submission Slide 8 Yakun Sun, et. al. (Marvell) Mar. 2014 doc.: IEEE 802.11-14/0335r0 Instantaneous UL SINR Per Tone A large portion of STAs frames come with high received SINR a high probability of successful packet. Also a long tail of low SINR 1 0.9 SINR per tone for unicast frames 0.8 0.7 CDF 0.6 0.5 0.4 0.3 0.2 0.1 0 -10 10

30 50 70 90 SINR Per Tone (dB) Submission Slide 9 Yakun Sun, et. al. (Marvell) Mar. 2014 doc.: IEEE 802.11-14/0335r0 Effective SINR Per Frame RBIR is used for effective SNR mapping. We truncate the SNR vs. RBIR mapping at 27dB for 64QAM and 30dB for 256QAM. 1 1 0.9 0.9 0.8 0.8 0.7 0.6 64QAM RBIR 0.6 EffSINR per unicast frames (256QAM) CDF CDF 0.7 EffSINR per unicast frames (64QAM) 0.5

256QAM RBIR 0.5 0.4 0.4 0.3 0.3 0.2 0.2 0.1 0.1 0 0 -10 -5 0 5 10 15 20 25 0 30 Effective SINR Per Frame (dB) Submission 1 2 3 4

5 6 7 8 RBIR Per Frame (bit) Slide 10 Yakun Sun, et. al. (Marvell) Mar. 2014 doc.: IEEE 802.11-14/0335r0 Simulation Assumptions (Scenario 2) Parameter Value Number of STAs 4 STAs per cubicle, 4 AP per BSS Channel Model TGn D (AP-AP, STA-STA, AP-STA) Penetration Loss Wall 7dB, linear for multiple walls BW 20MHz at 2.4GHz. Each AP selects one channel out of 4 in a BSS. (BSS4k+1,BSS4k+2,BSS4k+3,BSS4k+4)= (ch1,ch2,ch3,ch4) TX Power AP: 24dBm, STA: 21dBm Association 100% STA in a BSS associated with an AP in the BSS by RSSI, no P2P STA Based on [2] before the document was updated at the meeting. Submission Slide 11

Yakun Sun, et. al. (Marvell) Mar. 2014 doc.: IEEE 802.11-14/0335r0 Instantaneous UL SINR Per Tone 1 0.9 0.8 0.7 CDF 0.6 0.5 0.4 0.3 0.2 SINR per tone for unicast frames 0.1 0 -10 0 10 20 30 40 50 60 70 80 SINR per tone for Unicast frames (dB)

Submission Slide 12 Yakun Sun, et. al. (Marvell) Mar. 2014 doc.: IEEE 802.11-14/0335r0 Effective SINR Per Frame 1 1 0.9 0.9 EffSINR per unicast frames (64QAM) 0.8 EffSINR per unicast frames (256QAM) 0.7 0.7 0.6 0.6 CDF CDF 0.8 0.5 0.5 0.4 0.4 0.3 0.3 0.2 0.2

0.1 0.1 RBIR per unicast frame (64QAM) RBIR per unicast frame (256QAM) 0 0 -10 -5 0 5 10 15 20 25 0 30 Submission 1 2 3 4 5 6 7 RBIR per Unicast frames (bit) Effective SINR per Unicast frames (dB) Slide 13 Yakun Sun, et. al. (Marvell) 8

Mar. 2014 doc.: IEEE 802.11-14/0335r0 Simulation Assumptions (Scenario 3) Parameter Value Environment BSSs in Hexagon (figure 5), simulated BSS in 1 channel (figure 6) BSS radius: R=7m Number of STAs 30 STAs per BSS Channel Model TGn D (AP-AP, AP-STA), TGn B (STA-STA) Penetration Loss None BW 20MHz at 2.4GHz. Each simulated BSS selects the same channel. TX Power AP: 17dBm, STA: 15dBm Association 100% STA associated with the strongest AP Submission Slide 14 Yakun Sun, et. al. (Marvell) Mar. 2014 doc.: IEEE 802.11-14/0335r0 Instantaneous UL SINR Per Tone 100

90 80 70 CDF (%) 60 50 40 30 20 10 0 -20 -10 0 10 20 30 40 50 60 70 SINR per Tone for unicast frames (dB) Submission Slide 15 Yakun Sun, et. al. (Marvell) Mar. 2014 doc.: IEEE 802.11-14/0335r0 Effective SINR Per Frame

100 100 90 90 80 80 64QAM 64QAM 256QAM 70 60 CDF (%) 60 CDF (%) 256QAM 70 50 50 40 40 30 30 20 20 10 10 0 -5 0

5 10 15 20 25 30 0 35 0 Effective SINR per unicast frames (dB) Submission 1 2 3 4 5 6 7 RBIR per unicast frame (bits) Slide 16 Yakun Sun, et. al. (Marvell) 8 Mar. 2014 doc.: IEEE 802.11-14/0335r0 Simplification of Interference Modeling Explicitly modeling each interferers channel is costly. Suggest to approximate some interference as Gaussian channel. Skip generating a large amount of the fading channels Without introducing inaccuracy on received SINR and PHY performance.

A common practice for complexity reduction. [6] Question: how to select an interference to be approximated? Long Term SIR thresholding If the long term received power from an interferer relative to that of the desired transmitter is lower than a threshold, approximate its signal to be Gaussian. A static decision for each drop Submission Slide 17 Yakun Sun, et. al. (Marvell) Mar. 2014 doc.: IEEE 802.11-14/0335r0 Simplification of Interference Modeling (2) Specifically, the interference on a particular tone 2 TX n TX n I f PRX hRX f PRXTX m N 0 n0 m1 TX n TX desire 0 n : PRX PRX all interference to the current frame TX m TX desire 1 m : PRX PRX all interference to the current frame Delta = inf : explicitly model the fading channels of all seen interferers for the frame Delta = 10dB: explicitly model the fading channels of all seen interferers whose received power is within 10dB of the desire transmitter, model the rest of seen interferers as AWGN by their received power Step-2 calibration is a perfect stage to study the threshold Choose a threshold that does not impact the SINR distribution. Use scenario3 and 4 as an example. Submission

Slide 18 Yakun Sun, et. al. (Marvell) Mar. 2014 doc.: IEEE 802.11-14/0335r0 Simulation Assumptions (Scenario 3) Parameter Value Environment BSSs in Hexagon (figure 5), simulated BSS in 1 channel (figure 6) BSS radius: R=7m Number of STAs 30 STAs per BSS Channel Model TGn D (AP-AP, AP-STA), TGn B (STA-STA) Penetration Loss None BW 20MHz at 2.4GHz. Each simulated BSS selects the same channel. TX Power AP: 17dBm, STA: 15dBm Association 100% STA associated with the strongest AP Submission Slide 19 Yakun Sun, et. al. (Marvell) Mar. 2014 doc.: IEEE 802.11-14/0335r0

Instantaneous UL SINR Per Tone CDF (%) 100 90 Threshold = inf (complete modeling) 80 Threshold = 10dB 70 Threshold = 30dB 60 50 40 30 20 10 0 -20 -10 0 10 20 30 40 50 60 70 80 SINR per tone for unicast frame (dB) Reasonably small deviation between complete interference modeling and SIR thresholding of 30 and 10dB . Using 10dB threshold put 96% channels into AWGN Using 30dB threshold put 65% channels into AWGN Submission

Slide 20 Yakun Sun, et. al. (Marvell) Mar. 2014 doc.: IEEE 802.11-14/0335r0 Effective SINR Per Frame 100 100 90 Threshold = inf (64QAM) 80 Threshold = 10dB (64QAM) 80 Threshold = 10dB (64QAM) 70 Threshold = 30dB (64QAM) 70 Threshold = 30dB (64QAM) 60 Threshold = inf (256QAM) 60 Threshold = inf (256QAM) 50 Threshold = 10dB (256QAM) 40 Threshold = 30dB (256QAM) CDF (%) Threshold = inf (64QAM) CDF (%)

90 Threshold = 10dB (256QAM) 50 Threshold = 30dB (256QAM) 40 30 30 20 20 10 10 0 -10 -5 0 5 10 15 20 25 30 Effective SINR Per Unicast Frame (dB) Submission 0 0 1 2 3 4

5 6 7 RBIR Per Unicast Frame (bit) Slide 21 Yakun Sun, et. al. (Marvell) 8 Mar. 2014 doc.: IEEE 802.11-14/0335r0 Simulation Assumptions (Scenario 4) Parameter Value Environment BSSs in Hexagon (figure 8), ICD = 130m Number of STAs 30 STAs per BSS (50% outdoor, 50% indoor) Channel Model UMi (AP-AP, AP-STA, STA-STA) Penetration Loss 20dB (outdoor-indoor) BW 20MHz at 2.4GHz. Each simulated BSS selects the same channel. TX Power AP: 30dBm, STA: 15dBm Association 100% STA associated with the strongest AP Submission

Slide 22 Yakun Sun, et. al. (Marvell) Mar. 2014 doc.: IEEE 802.11-14/0335r0 Instantaneous UL SINR Per Tone 100 90 80 CDF (%) 70 60 50 40 Threshold = 10dB 30 Threshold = 30dB 20 10 0 -80 Submission -60 -40 -20 0 20 SINR Per Tone (dB) Slide 23 40 60 80 Yakun Sun, et. al. (Marvell)

Mar. 2014 doc.: IEEE 802.11-14/0335r0 Effective SINR Per Frame 100 100 Threshold = 10dB, 64QAM 90 Threshold = 30dB, 64QAM 80 Threshold = 10dB, 64QAM 90 Threshold = 30dB, 64QAM 80 Threshold = 10dB, 256QAM Threshold = 10dB, 256QAM 70 Threshold = 30dB, 256QAM 60 CDF (%) CDF (%) 70 50 Threshold = 30dB, 256QAM 60 50 40 40 30 30

20 20 10 10 0 0 -5 0 5 10 15 20 Effective SNR Per Unicast Frame (dB) Submission 25 30 Slide 24 0 1 2 3 4 5 RBIR Per Unicast Frame (bit) 6 7 Yakun Sun, et. al. (Marvell) 8

Mar. 2014 doc.: IEEE 802.11-14/0335r0 Summary Two options of instantaneous SINRs calibration are proposed. Suggestion1: Use Option 1 (SINR per tone) given its convenience and readiness. Option 2/2a can be revisited in the latter steps of calibrations. Suggestion2: Using SIR-thresholding to approximate some interference as AWGN Exact threshold can be also chosen through calibration. Submission Slide 25 Yakun Sun, et. al. (Marvell) Mar. 2014 doc.: IEEE 802.11-14/0335r0 References [1] 11-13-1392-00-0hew-methodology-of-calibrating-system-simulationresults [2] 11-14-0053-00-0further-considerations-on-calibration-of-systemlevel-simulation [3] 11-14-0116-01-0Long-Term-SINR-Calibration-for-System-Simulation [4] 11-14-0336-00-0Calibration-of-Long-Term-SINR-for-SystemSimulation [5] 11-13-1001-06-0hew-HEW-evaluation-simulation-scenariosdocument-template [6] 11-13-0043-02-0PHY-abstraction-in-system-level-simulation-forHEW-study Submission Slide 26 Yakun Sun, et. al. (Marvell)

Recently Viewed Presentations