GI Overhead/Performance Impact on Open-Loop SU-MIMO

Slides:



Advertisements
Similar presentations
Doc.: IEEE /0630r0 Submission May 2015 Intel CorporationSlide 1 Verification of IEEE ad Channel Model for Enterprise Cubical Environment.
Advertisements

Doc.: IEEE /0489r1 Submission May 2010 Alexander Maltsev, IntelSlide 1 PHY Performance Evaluation with 60 GHz WLAN Channel Models Date:
Interdigital Communications Submission doc.: IEEE /1333r1 November 2015 Feasibility of SU-MIMO under Array Alignment Method Date: Slide.
InterDigital, Inc. Submission doc.: IEEE /0911r1 July 2016 Link Level Performance Comparisons of Open Loop, Closed Loop and Antenna Selection.
Doc.: IEEE /1209r0 Submission Hotel lobby SU-MIMO channel modeling: 2x2 golden set generation Date: September 2016 Alexander Maltsev,
Doc.: IEEE /0632r1 Submission May 2016 Intel CorporationSlide 1 Performance Analysis of Robust Transmission Modes for MIMO in 11ay Date:
11ac 80MHz Transmission Flow
EDMG Header Encoding and Modulation
Beam Tracking for ay Date: Authors: January 2017
WUR Link Budget Analysis Follow-up: Data Rates and SIG Bits Protection
PHY Design Considerations for af
Closed Loop SU-MIMO Performance with Quantized Feedback
Open Loop vs Closed Loop SU-MIMO for 11ay
Month Year doc.: IEEE yy/xxxxr0 November 2016
On the Channel Model for Short Range Communications
EDMG Header-B Encoding and Modulation for SC PHY in 11ay
Comparisons of Simultaneous Downlink Transmissions
Channel Estimation Field for EDMG OFDM PHY in 11ay
L-Header spoofing and bit reuse
Rate 7/8 LDPC Code for 11ay Date: Authors:
Symbol Blocking and Guard Interval Definition for SC MIMO in 11ay
DCM SQPSK for Channel Aggregation in 11ay
Maximum Tone Grouping Size for ax Feedback
Further Discussion on Beam Tracking for ay
Ack and Block Ack in bonded channels
PHY abstraction and performance for outdoor channel models
Hybrid Beamforming Protocol Design Details
Protocols for Hybrid Beamforming in ay
Analog and Baseband Beam Tracking in ay
Analog and Baseband Beam Tracking in ay
Maximum Tone Grouping Size for ax Feedback
Preamble design aspects for MU-MIMO support
OFDMA Performance Analysis
MU-MIMO channel access flow for 11ay
Maximum Tone Grouping Size for ax Feedback
Ack and Block Ack in bonded channels
On the Single Carrier Waveforms for 11ay
OFDMA Performance Analysis
DCM SQPSK for Channel Aggregation in 11ay
Further Discussion on Beam Tracking for ay
WUR Dual SYNC Design Follow-up: SYNC bit Duration
Hybrid Beamforming Protocol Design Details
Update on β€œChannel Models for 60 GHz WLAN Systems” Document
Protocols for Hybrid Beamforming in ay
Discussion on WUR Multi-Antenna Transmission
160 MHz Transmission Flow Date: Authors: September 2010
May 2016 doc.: IEEE /XXXXr0 May 2016
11ac Explicit Sounding and Feedback
Single User MCS Proposal
PHY designs for NGV Date: Authors:
DCM SQPSK for Channel Aggregation in 11ay
Preliminary design of EDMG PHY headers
19, Yangjae-daero 11gil, Seocho-gu, Seoul , Korea
CSI Feedback Scheme using DCT for Explicit Beamforming
OFDMA Performance Analysis
Strawmodel ac Specification Framework
Ack and Block Ack in bonded channels
19, Yangjae-daero 11gil, Seocho-gu, Seoul , Korea
19, Yangjae-daero 11gil, Seocho-gu, Seoul , Korea
Discussion on Rank Adaptation
PHY Performance Evaluation with 60 GHz WLAN Channel Models
19, Yangjae-daero 11gil, Seocho-gu, Seoul , Korea
Fixed Inter Frame Spacing for BRP in ay
Channel Modeling with PAA Orientations
Control Trailer Clarifications
19, Yangjae-daero 11gil, Seocho-gu, Seoul , Korea
EDMG STF and CEF Design for SC PHY in 11ay
20 MHz transmission in NGV
Additional SC MCSs in clause 20 (DMG PHY)
PHY Signaling for Adaptive Repetition of 11p PPDU
Presentation transcript:

GI Overhead/Performance Impact on Open-Loop SU-MIMO Month Year doc.: IEEE 802.11 GI Overhead/Performance Impact on Open-Loop SU-MIMO Date: 2016-09-10 Authors:

Month Year doc.: IEEE 802.11 Introduction 802.11ad uses 64 chip guard interval (GI) for single carrier (SC) PHY. Should 11ay use the same GI length? TGay has agreed on a EDMG PPDU format which includes several non-legacy fields (EDMG-header-A, STF, CEF, Header-B) [1]. These additional fields increase the overhead of the data transmission. LoS is the dominant path in several uses cases. Narrow beams resulting from PAA pairs with a large number of elements reduce the delay spread of a point-to-point channel. This contribution investigates the use of shorter GI in specific scenarios. Performance/overhead results show that in certain scenarios the use of a shorter GI is justified.

PPDU Format in 802.11ay Current EDMG PPDU format for SC PHY[1]: EDMG preamble part introduces extra overhead Even though multi data stream transmissions can be applied to data part, EDMG transmission may not be always be as efficient as legacy DMG transmission. Overhead reduction is desirable for EDMG PPDU

Guard Interval In 802.11ad, SC data blocks (448 symbols ) are separated by guard intervals (64 symbols). The 64 GI symbols are modulated symbols from a Golay sequence. The usage of GI: GI is a time period to mitigate inter-block interference GI functions as a cyclic prefix which allows the use of frequency domain equalizer (FDE) at the receiver GI is a periodic known sequence to assist with AGC and phase tracking However, GI is extra overhead for data transmission. Is 64 GI always necessary? In this contribution, a different GI size is evaluated using link level simulation. Overhead comparison is also provided. We focus on the impact from the inter-block interference assuming an FDE 64 448 symbols

GI Evaluation Methodology Link level simulation For GI=32, extra 32 symbols are used for data (480 data symbols). The block length remains 512 symbols. Config #4, Nss=2 Overhead analysis For a fixed packet size, we determine the PPDU duration by taking into account the MCS, number of data streams, preamble format as well as GI size. 𝐸𝑓𝑓𝑒𝑐𝑑𝑖𝑣𝑒 π‘‘π‘Žπ‘‘π‘Ž π‘Ÿπ‘Žπ‘‘π‘’= π‘π‘Žπ‘π‘˜π‘’π‘‘ 𝑠𝑖𝑧𝑒 π‘ƒπ‘ƒπ·π‘ˆ π‘‘π‘’π‘Ÿπ‘Žπ‘‘π‘–π‘œπ‘›

Link level simulation

Simulation Assumptions Based on 11ad SC PHY Spatial stream parser: MCS index is the same for all streams per PPDU, and a single CRC is used per PPDU MMSE receiver with FDE Ideal channel estimation at receiver Enterprise cubicle scenario in 11ay/ad channel model [2] STAs are randomly placed in the cubicle 1 in the center of the CR, 0.9m above the floor AP is positioned at x=2.8, y=6, z=2.9m on the ceiling Detailed assumptions can be found in the appendix PSDU size is 8192 bytes Stream 1 Encoder Output Bits b1 b3 b5 b1 b2 b3 b4 b5 b6 Stream 2 b2 b4 b6

PER performance (MCS5/8) MCS5 (BPSK) and MCS8 (QPSK), there are little or no differences in PER performance NLOS channel improves PER at high SNR. This gain is from frequency diversity such that it is less likely that all frequency tones are stuck in similarly ill-conditioned channels

PER performance (MCS12) For LOS scenario, short and long GI have similar performances For NLOS scenario, with short GI at high SNR, ISI becomes dominant, but the SNR difference is less than 2 dB for PER = 1% NLOS multipath degrades performance at low SNR but improves performance at high SNR

Overhead analysis

Overhead Analysis Parameters GI/data block size: GI=64: 448 data symbols with 64 GI symbols GI=32: 480 data symbols with 32 GI symbols Packet size: Small packet: 1200 Bytes Large packet: 8192 Bytes Channel bandwidth: 2.16Ghz Number of data streams (Nss) 2 data streams for EDMG PPDU Single data stream for DMG PPDU PPDU format: EDMG PPDU and DMG PPDU EDMG STF duration: 512 * Tc EDMG CEF duration: 1152 * Tc EDMG Header-B is not considered Tc is SC chip time, 0.57 ns MCS: 1-12 (including SC BPSK, QPSK and 16QAM)

Small Packet Overhead Analysis MCS Gain (%) (GI32-GI64)/GI64 1 5.7 2 6.1 3 6.9 4 3.8 5 6 4.5 7 4.9 8 5.1 9 0.3 10 5.6 11 12 0.4 GI32 vs GI64 GI 32 shows up to 6.9 percent gain over GI 64 in effective data rate. In general, the higher the MCS, the lower the gain due to GI. This is because with higher MCS, fewer number of SC blocks are required to carry the information bits, resulting in less gain from GI reduction. EDMG vs DMG DMG single data transmission outperforms EDMG two stream transmission at higher MCSs. This is because with higher MCSs, the ratio of data part over the entire PPDU becomes smaller. Thus the savings from the data part cannot compensate the loss from the preamble part.

Large Packet Overhead Analysis MCS Gain (%) (GI32-GI64)/GI64 1 6.6 2 6.7 3 6.5 4 5.7 5 6.1 6 7 5.9 8 5.1 9 5.4 10 4.2 11 4.9 12 5.6 GI32 vs GI64 GI 32 shows up to 6.7 percent gain over GI 64 in effective data rate. In general, the higher the MCS, the lower the gain due to GI. This is because with higher MCS, fewer number of SC blocks are required to carry the information bits, resulting in less gain from GI reduction. EDMG vs DMG With large packet sizes, EDMG two data stream transmission always outperforms DMG single stream transmission.

Conclusions EDMG preamble adds additional overheads in a PPDU Month Year doc.: IEEE 802.11 Conclusions EDMG preamble adds additional overheads in a PPDU short EDMG frame with high MCS is not efficient Using GI length of 32 symbols is sufficient for some of the indoor scenarios. (32 GI, 480 data) block for 2.16GHz channel should be considered as an option.

Straw Poll Should TGay study the option of shorter GI for SC PHY?

Month Year doc.: IEEE 802.11 References Carlos Cordeiro, β€œSpecification Framework for TGay”, IEEE 802.11-15/01358r5 A. Maltsev, et al, β€œChannel models for ieee 802 11ay”, IEEE doc. 11-15/1150r6 R. Maslennikov, et al, β€œImplementation of 60 GHz WLAN Channel Model,” IEEE doc. 11-10/0854r3.

Appendix

Channel parameters For channel with LOS components [3], TX/RX analog beamforming for both polarizations of PAA#i are based on the LOS direction between TX PAA#i ↔ RX PAA#i For channel without LOS components Beam forming based on the AoD/AoA of strongest signal path between TX PAA#i ↔ RX PAA#i Channel bandwidth 1.76 GHz, center frequency 60GHz Each PAA has 2 elements Distance between antenna elements 0.0025m Distance between center of PAAs 10cm For AP-STA scenario, STA is placed at a plane 2m below AP in the cubicle 1. Random rotation around z-axis between STA/AP.