[place presentation subject title text here]

[place presentation subject title text here]

Nov. 2013 doc.: IEEE 802.11-13/1392r0 Methodology of Calibrating System Simulation Results Date: 2013-11-11 Authors: Name Yan Zhang Affiliations Marvell Semiconductor Address 5488 Marvell Lane, Santa Clara, CA, USA Yakun Sun Phone Email 408-2220975 [email protected] om

408-2223748 [email protected] l.com Hongyuan Zhang Jinjing Jiang Liwen Chu Hui-Ling Lou Mingguang Xu David Yang Jiaying Zhang Submission Huawei Technologies Slide 1 Yan Zhang, et. Al. Nov. 2013 doc.: IEEE 802.11-13/1392r0 Introduction Highly desirable to have aligned system simulation results [1] from all participating companies when evaluating the candidate

HEW technologies. System simulation results alignment is very challenging. Both PHY/MAC and upper layers are involved many abstractions A large number of APs and STAs simulated simultaneously make it less tractable. Calibration at the early stage is the only approach to this target. Eliminate the deviation from the beginning before no one can explain. Calibration needs to be well defined. Calibration goes way beyond using the same simulation scenario. What can/should be calibrated? How to calibrate? Submission Slide 2 Yan Zhang, et. Al. Nov. 2013 doc.: IEEE 802.11-13/1392r0 Why is Calibration Challenging? A simple example of system simulator Simulation scenario [2] needs to thoroughly cover all setup for the network and each entity within. A large number of AP and STA are randomly dropped in the plane of simulation and then

associated by certain methods Are they dropped and associated the same way? How to define the same way? Dynamic network behavior are abstracted in many aspects. Where to abstract and how to abstract vary dramatically across companies. How to evaluate the impact of abstractions? How to align the impact brought by abstractions? After all, this is only calibration. How to balance the efforts and the accuracy? Submission Slide 3 Yan Zhang, et. Al. Nov. 2013 doc.: IEEE 802.11-13/1392r0 What to Calibrate? Simulation scenarios [2] Harder to calibrate Topology, PHY and MAC parameters,

Traffic models Calibration metrics Simulation scenario compliance STA/AP drop/association static radio characteristics related to PHY layer (large scaling fading, received time domain signal power, long-term S/N or S/I, etc.) Channel modeling dynamic statistics of PHY layer (SNR/SIR/SINR per tone, etc.) PHY abstraction (receiver performance) alignment MAC behavior and system performance (total/average/5%-tile PHY throughput and MAC goodput, average user delay/jitter, etc) Submission Slide 4 Yan Zhang, et. Al. Nov. 2013 doc.: IEEE 802.11-13/1392r0 How to Calibrate? All-in-one vs. step-by-step? Note that the difficulty of calibration for different metrics vary. Calibration should be done step by step.

A guideline is needed to describe the test environments, selected deployment scenarios and evaluation configurations. In order to generate meaningful performance results for each scenario, it is important to calibrate the basic (both static and dynamic) PHY characteristics such as path loss, at the first step. PHY abstraction of receiver performance should be done in parallel via link level simulations. Further calibrations can be done only after majority of the companies results are aligned. Submission Slide 5 Yan Zhang, et. Al. Nov. 2013 doc.: IEEE 802.11-13/1392r0 Step by Step Calibration Currently a HEW simulation scenarios template is served as a guideline [2]. It describes the test environments, selected deployment scenarios and some evaluation configurations. However there are still a lot missing parts to fill in before we can start simulating the system for calibration. Filling in the missing parts of evaluation configurations is the first step to be accomplished Calibration should be started from static radio characteristics alignment For example, the distributions of path loss from STAs to associated APs can be plotted for

each simulation scenario, to make sure participating companies are simulating the same channel, i.e., using the same large scale parameters, and using the same methods of choosing associated APs. Further calibration should include fast fading channel characteristics (small scale channel parameters) alignment PHY layer performance depends on knowledge of fast fading channel. It will be very hard to directly compare the values of sampled (random) fading channel taps. The distribution of per STA average SINR can be plotted assuming most basic setup, e.g., SISO system, round robin scheduling and full buffer traffic. Submission Slide 6 Yan Zhang, et. Al. Nov. 2013 doc.: IEEE 802.11-13/1392r0 Step by Step Calibration (2) PHY abstraction should be done in parallel. It is preferable for companies to agree on a common PHY abstraction technique. Effective SNR to PER lookup table/curves also need to be unified among companies. This can be done via link-level simulations, and done for only once.

If each company wants to apply its own PHY abstraction technique, at least the companies of common preference should align. Throughputs of most basic system should be calibrated using aligned PHY abstraction. PHY abstraction replaces actual decoding in system level simulation. Specify the PHY abstraction technique in the simulation assumptions at the calibration stage, if more than one PHY abstraction methods can be used. After PHY abstraction is aligned, the PHY throughput results of some most basic system should be aligned fairly easily (without or with minimal MAC). MAC layer calibration should be done after PHY layer calibration. MAC layer calibration will be done per simulation scenario with a default/minimal set of existing (e.g. 11ac) features. MAC abstraction should be calibrated now, and advanced simulations can start from this point to evaluate HEW technologies. Submission Slide 7 Yan Zhang, et. Al. Nov. 2013 doc.: IEEE 802.11-13/1392r0 Work Load of Calibration Calibration per scenario Full-blow (at least for a basic scenario) A subset of calibrations

Basic scenario for full-blown calibration A stand-alone scenario? A basic sub-scenario per simulation scenario? Directly use the default params in each simulation scenario? PHY layer calibration is actually minimal amount of work PHY abstraction calibration needs to be done only once. Submission Slide 8 Yan Zhang, et. Al. Nov. 2013 doc.: IEEE 802.11-13/1392r0 Summary It is essential to calibrate the simulation results before evaluating the candidate HEW technologies. Step by step calibration methodology should be adopted in HEW evaluations to accelerate the process. PHY Abstraction Method Evaluation/Downselection PHY Abstraction Method Calibration

PHY statistics (Freq-domain SINR distribution) Simulation Scenario Radio statistics (S/I distribution) Submission MAC calibration PHY Tput calibration Slide 9 Yan Zhang, et. Al. Nov. 2013 doc.: IEEE 802.11-13/1392r0 References [1] 11-13-1051-01-0hew-evaluation-methodology [2] IEEE 802.11-13/0722r13, HEW SG Simulation Scenarios Submission

Slide 10 Yan Zhang, et. Al. Nov. 2013 doc.: IEEE 802.11-13/1392r0 Appendix PHY Abstraction Calibration Link-level simulation for each MCS Simulate over multiple channel types (TGn/ITU/etc) For each channel type: Simulation over a range of SNR (with fine resolution) for each MCS For each SNR, simulate over multiple (for example, 1000) channel realizations For each realization, collect the effective SNR (1 scalar ) and 1 bit flag of decoding outcome (success or failure) for the specific coding scheme used (BCC or LDPC) 1x2 vector as output Combine the large collections of [SNR_eff, flag] over all realizations and SNRs, and quantize to effective SNR vs. PER table PER = (# of successful decodings in a SNR_eff bin) / (# of packets in this SNR_eff bin) This table is coding scheme specific. Submission Slide 11 Yan Zhang, et. Al.

Nov. 2013 doc.: IEEE 802.11-13/1392r0 Appendix PHY Abstraction Calibration (2) Effective SNR vs. PER curves for different channel types should be close (and also close to AWGN results). PHY abstraction method is channel independent. A small spread is acceptable, and an average PER performance over channels (or directly AWGN results) will be used in the end. Companies should show that the PHY abstraction method they prefer is channel independent. Companies should align the effective SNR vs. PER curves for each MCS, if the same PHY abstraction method is used. The agreed PHY abstraction method and the effective SNR vs. PER curves will be used in system simulations. Submission Slide 12 Yan Zhang, et. Al.

Recently Viewed Presentations

  • University of Northern British Columbia DC Pension Plan

    University of Northern British Columbia DC Pension Plan

    fund name number of members number of members b.g. balanced fund (default fund) blk bond index fund slf money market cc&l group cdn q growth cc&l group income&growth ci sig income & growth blk us equity index sla 1yr guaranteed...
  • Sequence 3 : Poles Et Aires De Puissance

    Sequence 3 : Poles Et Aires De Puissance

    - à une façade maritime majeure : New York, Tokyo, Singapour. Correction Document 3 « La définition actuelle des villes mondiales est (...) Aujourd'hui, ce sont les centres de commandement de l'économie-monde. S'y empilent les sièges sociaux, les bourses, les...
  • Lesson 8 Electromagnet An Electromagnet We know that

    Lesson 8 Electromagnet An Electromagnet We know that

    We know that all magnets have two ends, usually marked "north" and"south," And that magnets attract things made of steel or iron We also know the fundamental law of all magnets: Opposites attract and likes repel. An electromagnet works in...
  • Our Journey Together: A Historical and Relational Perspective

    Our Journey Together: A Historical and Relational Perspective

    keep blacks in their place--separate but equal or so we thought (wink, wink), THROUGH wrangling over ordination of women and now how we will fully or . not so fully include LGBTQ sisters and brothers, THROUGH the new, ever- ......
  • NCA: Thematic working sessions Introduction Mark Gough, Executive

    NCA: Thematic working sessions Introduction Mark Gough, Executive

    for working on biodiversity? And on the materiality of it? Is it possible to attribute criteria for each risk type, in other words which businesses are more sensitive to certain risks? Is it only sector related or also location related?...
  • Money - Wattsburg Area School District

    Money - Wattsburg Area School District

    Role of Money. Complete the worksheet provided. Use your textbook pages 210 - 215 as a resource. ... What are the three uses of Money? What are the 6 characteristics of Money? "Ours is the worst possible system, except for...
  • Data generation - WUR

    Data generation - WUR

    Negative correlation with gene length and degree of codon bias. Codon bias is more extreme in highly expressed genes. Genes with longer introns show higher bias in codon usage. The overall codon usage matches the known bias. The expectations. Return...
  • Cert Exam Prep: Exam 70-346: Managing Office 365 Identities ...

    Cert Exam Prep: Exam 70-346: Managing Office 365 Identities ...

    Because Microsoft must respond to changing market conditions, it should not be interpreted to be a commitment on the part of Microsoft, and Microsoft cannot guarantee the accuracy of any information provided after the date of this presentation. MICROSOFT MAKES...