Further Investigation on WUR Performance Month Year doc.: IEEE 802.11-yy/xxxxr0 September 2016 Further Investigation on WUR Performance Date: 2016-09-12 Authors: Name Affiliation Address Phone Email Eunsung Park LG Electronics 19, Yangjae-daero 11gil, Seocho-gu, Seoul 137-130, Korea   esung.park@lge.com Jinsoo Choi js.choi@lge.com Dongguk Lim dongguk.lim@lge.com HanGyu Cho hg.cho@lge.com Eunsung Park, LG Electronics John Doe, Some Company

Recap on [1] Uncoded PER performance for the wake-up receiver is shown September 2016 Recap on [1] Uncoded PER performance for the wake-up receiver is shown Wake-up packet is computed by the OOK modulation OOK symbol is generated using 802.11 OFDM transmitter K subcarriers are used (K = 1, 13, 64) Coefficients of all available subcarriers are set to 0 (off) and 1 (on) for information “0” and “1”, respectively Two decoding methods are applied to the receiver Coherent decoding Non-coherent decoding Eunsung Park, LG Electronics

September 2016 Overview (1/2) In this contribution, WUR performance is further investigated by applying some schemes such as coding and symbol repetition Coherent decoding As shown in [1], the coherent decoding method offers good performance even in the uncoded case, but the performance can be degraded due to several impairments such as CFO and TO Thus, we may need to further enhance the performance, and in this contribution, we additionally check on the performance for the coded case as a reference BCC with ½ code rate using random interleaver is applied Eunsung Park, LG Electronics

Overview (2/2) Non-coherent decoding September 2016 Overview (2/2) Non-coherent decoding WUR performance can be worse than the L-SIG performance as shown in [1], and impairment factors may lead to a further performance degradation Thus, the performance enhancement is imperative in order to achieve a similar performance with the L-SIG (i.e., align the transmission range between the wake-up packet and conventional 802.11 packet) Furthermore, we should consider the coexistence issue with other 802.11 devices (or potentially with non-802.11 devices) Consecutive OFF symbols in a wake-up packet may have a coexistence problem To this end, we apply several schemes as follows Manchester code Symbol repetition We will demonstrate PER performance for both decoding methods The number of information bits is set to 48 (e.g. MAC address) Eunsung Park, LG Electronics

OOK Modulation with Manchester Code (1/3) September 2016 OOK Modulation with Manchester Code (1/3) If Manchester code is applied to the OOK modulation, envelope transition occurs from on/off to off/on in the middle of the symbol time This process can prevent consecutive OFF symbols Then, we can consider that each OOK symbol (information “0” or “1”) is composed of two 1.6us sub-symbols and 0.8us guard interval 4us symbol time is the same as the uncoded case in [1] and thus the data rate is also the same Each sub-symbol for each information is as follows We denote 1.6us sub-symbols “0” and “1” as OFF and ON sub-symbols, respectively, and we will give an example how to generate them using 802.11 OFDM transmitter in slide 6 Also, two options for a symbol structure according to the guard interval will be introduced in slide 7 Information First sub-symbol (1.6us) Second sub-symbol (1.6us) “0” 1 (ON) 0 (OFF) “1” Eunsung Park, LG Electronics

OOK Modulation with Manchester Code (2/3) September 2016 OOK Modulation with Manchester Code (2/3) OFF sub-symbol The signal for the 4us OFF symbol is described in [1] and either the first or second half of this signal except the guard interval can be used (i.e. OFF during 1.6us) ON sub-symbol Set the coefficients as follows Every other available subcarrier : 1 Others : 0 E.g.) indices of available subcarriers are -6 to 6 By doing this, we can generate a 3.2us signal that has 1.6us periodicity Choose either the first or second 1.6us part -32 -31 -30 … -8 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8 … 30 31 Subcarrier index Eunsung Park, LG Electronics

OOK Modulation with Manchester Code (3/3) September 2016 OOK Modulation with Manchester Code (3/3) Symbol structure Option 1 : 0.8us guard interval is used at the front of each symbol Option 1 may lead to a severe inter sub-symbol interference especially in an outdoor environment Option 2 : 0.4us guard interval is used at the front of each sub-symbol Option 2 can reduce the inter sub-symbol interference compared to option 1 CP Information “0” 0us 0.8us 2.4us 4us Information “1” 0us 0.8us 2.4us 4us CP GI OFF sub-symbol GI ON sub-symbol OFF sub-symbol ON sub-symbol CP Information “0” 0us 0.4us 2.0us 2.4us 4us CP Information “1” 0us 0.4us 2.0us 2.4us 4us GI OFF sub-symbol GI OFF sub-symbol GI ON sub-symbol GI ON sub-symbol Eunsung Park, LG Electronics

September 2016 Symbol Repetition For non-coherent decoding, we also consider a symbol repetition in the time domain for a better performance Two symbols are used to indicate information “0” and “1” The more symbols for each information, the better performance However, given a latency issue, we should minimize the overhead the most, and thus we only use two symbols for each information In order to avoid consecutive OFF symbols, we use different symbols between two symbols as follows ON/OFF symbols are depicted in [1] Information First symbol (4us) Second symbol (4us) “0” 1 (ON) 0 (OFF) “1” Eunsung Park, LG Electronics

Simulation Environment September 2016 Simulation Environment Preamble : 11 symbols as in [1] Information bits : 48 bits (e.g. MAC address) Number of available subcarriers : 13 Channel : TGn D, UMi NLOS No CFO/TO Decoding methods Coherent Uncoded as in [1] Coded : BCC w/ ½ code rate and random interleaver We do not consider Manchester code Non-coherent Two options for Manchester code Symbol repetition We do not apply two options for Manchester code to each symbol for the symbol repetition method in the simulation Eunsung Park, LG Electronics

Coherent Decoding TGn D UMi NLOS September 2016 Eunsung Park, LG Electronics

Non-coherent Decoding September 2016 Non-coherent Decoding TGn D UMi NLOS Eunsung Park, LG Electronics

September 2016 Conclusion We investigated the PER performance for the WUR by using several schemes In coherent decoding, we verified that the uncoded case already has a better performance than the L-SIG case and the coded case can further enhance the performance It seems that we don’t have to apply the channel code which causes large overhead and high power consumption However, we need to check on the performance by taking into account several factors which yield performance degradation In non-coherent decoding, we verified that the WUR with several schemes can obtain a better performance than the L-SIG case In both channels, the symbol repetition method offers the best performance at the cost of the overhead Option 2 for Manchester code can provide a comparable performance comparing with the L-SIG performance in the TGn D channel However, both options for Manchester code have poor performance in the UMi channel, and thus we can confirm that they are vulnerable to the inter symbol/sub-symbol interference Therefore, it seems that we need to apply the symbol repetition method at least even considering the impairment factors Or, we can consider other options for Manchester code which have a better performance with a slight overhead increase (i.e. longer guard interval for ISI reduction) Note that those options for Manchester code and symbol repetition can avoid the coexistence problem with other 802.11 and non-802.11 devices Eunsung Park, LG Electronics

September 2016 References [1] IEEE 802.11-16-0865-01-0wur-performance-investigation-on-wake-up-receiver Eunsung Park, LG Electronics