Nov 2012 doc IEEE 802 11 121376 r

Nov. 2012 doc. : IEEE 802. 11 -12/1376 r 0 TSF Timer Freq. Management and Measurement Procedure (TFM 2 P) Date: 2012 -11 -13 Authors: Submission Slide 1 Shusaku Shimada Yokogawa Co.

Nov. 2012 doc. : IEEE 802. 11 -12/1376 r 0 Abstract The detailed three procedures of enhanced power saving function which employs the proposed TFM 2 P (TSF timer Frequency Management & Measurement Procedure) is presented. TFM 2 P can be used with existing Power Saving mechanisms to allow STA waking up precisely and sleeping longer, and some sort of access control mechanisms for following operational conditions; (1) numerous numbers of sensors or meters, with lower traffic at each STA, requiring battery conservation. (use case 1 a/c/d/e/f) (2) access control numerous numbers of sensors or meters using wake-up timing control schemes by TSF timer synchronization, rather than simple ALOHA. (RAW, TWT, PS-mode, etc. ) Submission Slide 2 Shusaku Shimada Yokogawa Co.

Nov. 2012 doc. : IEEE 802. 11 -12/1376 r 0 Principle of PS feature Submission Slide 3 Shusaku Shimada Yokogawa Co.

Nov. 2012 doc. : IEEE 802. 11 -12/1376 r 0 Wake-up synchronization Simple AP announcement of TSF accuracy (1) Wake-up Timing margin depends on TSF timer freq. accuracy △; AP (e. g. TSF master) TW notified scheduled wake-up time (ideal case) TW (IEEE 802. 11 -2012) Tolerance ± △· (TW –TS) ± 100 ppm Wake-up margin -△· (TW – TS) ≈ STA (e. g. TSF slave) TS actual sleep duration sleep again STA awake < 11 -12/130 r 0 “Beacon Reception of Long Sleeper” > (1) AP is supposed to announce TSF accuracy △, (△<100 ppm) △ includes accuracy of both AP & STA (2) STA is able to wake up at (TW –TS)(1 - △) +TS TS : TSF timer value just after last time it was synchronized Submission Slide 4 Shusaku Shimada Yokogawa Co.

Nov. 2012 doc. : IEEE 802. 11 -12/1376 r 0 Wake-up synchronization Simple AP announcement of TSF accuracy (2) Awake period of STA may become much longer than actual Communication. scheduled wake-up time (ideal case) AP (e. g. TSF master) TW ± △· (TW –TS) actual TW notified ≈ STA (e. g. TSF slave) Wake-up margin - △· (TW –TS) communication actual sleep duration sleep again STA awake TW-actual wake-up point of time Communication may happen within green window. ± △·( TW – TS) STA have to be awake during entire blue STA awake period while actual communication duration Submission may be a part of awake period. Slide 5 Shusaku Shimada Yokogawa Co.

Nov. 2012 doc. : IEEE 802. 11 -12/1376 r 0 Wake-up sync. using TFM 2 P AP announcement of TSF timer stability (1) Wake-up Timer Stability information (±ε) as well as △; announced AP ( TSF master) scheduled wake-up time (ideal case) ±ε △measured wake-up margin AP advertise △worst and ε ≈ compensated by measured TSF frequency Tw-compen < TFM 2 P involves two parameters, i. e. △ and ε > (1) (2) measured AP side point of time (by STA) ± △ · TW Tw notified after TSF frequency measurement Receiver side measured STA (e. g. TSF master) TW -ε STA awake sleep again STA to wake up at, (TW-compen –TS)(1 - ε)+TS ≃ (TW –TS)(1 + △measured - ε) +TS Submission Slide 6 Shusaku Shimada Yokogawa Co.

Nov. 2012 doc. : IEEE 802. 11 -12/1376 r 0 Wake-up sync. using TFM 2 P AP announcement of TSF timer stability (1) Wake-up Timer Stability information (±ε) as well as △ ; scheduled wake-up time (ideal case) announced AP ( TSF master) measured point of time TW ± △ · TW -ε △measured ≈ Receiver side measured STA (e. g. TSF master) actual Tw notified after TSF frequency measurement communication compensated by measured TSF frequency Tw-compen STA awake STA to wake up at sleep again TW-actual point of time (TW-compen –TS)(1 - ε)+TS ≃ (TW –TS)(1 + △measured - ε)+TS after once TFM 2 P has carried out. Submission Slide 7 Shusaku Shimada Yokogawa Co.

Nov. 2012 doc. : IEEE 802. 11 -12/1376 r 0 Comparison of Wake-up synchronization (1) Simple Accuracy Announcement and TFM 2 P (frequency measurement) scheduled wake-up time Tw AP (e. g. TSF master) actual communication ± △advertised · ( TW –TS) Informed Tw is used with △advertised STA w/o TFM 2 P (e. g. TSF slave) Informed Tw and εadvertised is used with measured frequecy STA w/t TFM 2 P (e. g. TSF slave) Submission wake-up margin using accuracy information (△advertised ) sleep awake up sleep again (Tw - TS) (1 - △advertised ) + TS Less wake-up margin by TSF freq. offset compensation and freq. stability information sleep Slide 8 awake sleep again (TW – TS)(1 + △measured - εadvertised) + TS Shusaku Shimada Yokogawa Co.

Nov. 2012 doc. : IEEE 802. 11 -12/1376 r 0 Comparison of Wake-up synchronization (2) Proposed three procedures of TFM 2 P for Power Saving Simple accuracy announcement (broadcast) Time Stamp announcement for TFM 2 P (broadcast) Time Stamp handshake for TFM 2 P (node by node) Broadcast (uni-directional) Unicast handshake (node by node) AP accuracy △AP accuracy STA Stability ε accuracy STA Receiving broadcasted accuracy information, then calculate wake-up margin, △AP+STA Submission AP B 1+B 1 timestamp B 2+B 2 timestamp STA B 2+B 2 timestamp B 1+B 1 timestamp STA B 1+B 1 timestamp B 2+B 2 timestamp STA Receiving four broadcasted time stamp for measuring TSF freq. , then calculate wake-up margin, △measured , ε Slide 9 AP Stability ε M 1+Ack M 6+Ack M 2+Ack STA M 3+Ack M 7+Ack M 8+Ack M 5+Ack STA M 4+Ack STA Handshaking two time measurement to determine each precise offset and freq. , then calculate wake-up margin, △measured , ε Shusaku Shimada Yokogawa Co.

Nov. 2012 doc. : IEEE 802. 11 -12/1376 r 0 Comparison of Wake-up synchronization (3) Scheme IEEE 802. 11 2012 (Conventional) Timer accuracy notification [11 -12/130 r 0] Broadcast or Handshake Inaccuracy information None Pre-defined by Std. Battery life improvement Update Required mechanism PHY/MAC Support None TSF synch only Not required Not Required TSF timer freq. accuracy advertisement MAC: required (ex. estimated) ± 100 ppm Reference (1. 0) +offset Broadcast w/o handshake Broadcast (Time-Stamp Announcement) w/o handshake TFM 2 P Node by node w/t bi-directional handshake Submission Resulting Wake-up Accuracy by ± 20~50 ppm AP announcement +offset by ± 2~10 ppm direct TSF frequency measurement + AP stability advertisement 1. 6 times (1. 2~2. 0) Conditiona lly preferred +offset 2. 5 times (1. 5~4. 0) ± 1~5 ppm null offset Slide 10 Conditiona lly Required TSF timer freq. accuracy advertisement + Two time measurements + Calculation & compensation MAC: required PHY: optional MAC: required PHY : preferable Shusaku Shimada Yokogawa Co.

Nov. 2012 doc. : IEEE 802. 11 -12/1376 r 0 Typical mechanism of TFM 2 P using Broadcast (1) AP as Clock master broadcasts Time Stamp Announcement with no handshake. • Full beacons with DTIM always carry To. D time stamp for TFM 2 P. • All To. D time stamps correspond to its N-times previous DTIM beacon. • Each pair of successive To. D time stamps may be used for TSF frequency estimation with corresponding previous pair of To. A time stamps. N-times DTIM Interval ( N ≥ 1 ) ≈ ≈ Beacon Interval DTIM TIM ≈ Beacon Transmissions ( can be short beacon ) Full Beacon DTIM N-times previous To. D time stamp Submission Busy medium other transmissions TFM 2 P frequency measurement pair Slide 11 Full Beacon DTIM N-times previous To. D time stamp Shusaku Shimada Yokogawa Co.

Nov. 2012 doc. : IEEE 802. 11 -12/1376 r 0 Typical mechanism of TFM 2 P using Broadcast (2) AP as Clock master broadcasts Time Stamp Announcement with no handshake. Sending STA(f 1) Receiving STA(f 2) t 1=To. D(B 1) B 1 B 0 -timestamp t 5=To. D(B 2) B 2 B 1 -timestamp t 6=To. A(B 2) t 1 are known t 9=To. D(B 3) B 3 B 2 -timestamp t 2=To. A(B 1) t 10=To. A(B 3) t 5 are known B 4 B 3 -timestamp Estimation in this figure, t 9 and t 10 is not used. Submission Slide 12 Shusaku Shimada Yokogawa Co.

Nov. 2012 doc. : IEEE 802. 11 -12/1376 r 0 Typical mechanism of TFM 2 P using Broadcast (3) Sending STA(f 1) Receiving STA(f 2) t 1=To. D(B 1) B 1 B 0 -timestamp t 5=To. D(B 2) B 2 B 1 -timestamp t 6=To. A(B 2) t 1 are known t 9=To. D(B 3) B 3 B 2 -timestamp t 2=To. A(B 1) t 10=To. A(B 3) t 5 are known B 4 B 3 -timestamp Estimation in this figure, t 9 and t 10 is not used. Submission δ 2 (e. g. ppm) should be the calibration factor of f 2 to schedule Tw , wake-up time. Slide 13 Shusaku Shimada Yokogawa Co.

Nov. 2012 doc. : IEEE 802. 11 -12/1376 r 0 TFM 2 P mechanism by node-by-node handshake (1) Handshake can be between AP/MP & STA, STA & STA or MP & MP. Sending STA(f 1) t 1=To. D(M 1) t 4=To. A(Ack) t 5=To. D(M 2) t 8=To. A(Ack) Receiving STA(f 2) M 1 t 2=To. A(M 1) Ack t 3=To. D(M 1) M 1 timestamp t 1 and t 4 are known Ack M 2 timestamp Ack t 6=To. A(M 2) t 7=To. D(M 2) t 5 and t 8 are known offset 1 ⧋ [(t 2 -t 1)-(t 4 -t 3)]/2 offset 2 ⧋ [(t 6 -t 5)-(t 8 -t 7)]/2 Submission Slide 14 Shusaku Shimada Yokogawa Co.

Nov. 2012 doc. : IEEE 802. 11 -12/1376 r 0 TFM 2 P mechanism by node-by-node handshake (2) How all STAs synchronizes each other is out of scope of this standard. Sending STA(f 1) Receiving STA(f 2) t 1=To. D(M 1) t 4=To. A(Ack) M 1 t 2=To. A(M 1) Ack t 3=To. D(M 1) M 1 timestamp Ack offset 1=[(t 2 -t 1)-(t 4 -t 3)]/2 t 5=To. D(M 2) t 8=To. A(Ack) M 2 t 6=To. A(M 2) Ack t 7=To. D(M 2) M 2 timestamp Ack offset 2=[(t 6 -t 5)-(t 8 -t 7)]/2 Submission Slide 15 Shusaku Shimada Yokogawa Co.

Nov. 2012 doc. : IEEE 802. 11 -12/1376 r 0 TFM 2 P mechanism by node-by-node handshake (3) How all STAs should synchronize each other after node-by-node calibration can be achieved, is out of scope of this standard. However, node-by-node TFM 2 P is expected to be instrumental because of following reasons, (1) By utilizing existing 11 v timing measurement scheme identical to PTP/IEEE 1588, the best time and frequency accuracy of TSF for wake up can be used with the precise timing offset nulling. (2) This also means that the quick frequency estimation can be possible using shorter time interval of two time measurements. (3) IBSS, MBSS without AP can still utilize TFM 2 P for wake up. (4) To perform such sort of applications, for example, timing sensitive control using DLS, TFM 2 P works. (5) Forward looking applications may be facilitated by precise synch. . Submission Slide 16 Shusaku Shimada Yokogawa Co.

Nov. 2012 doc. : IEEE 802. 11 -12/1376 r 0 Procedure (1) : General ( IE in full beacon body ) • STA can examine TFM 2 P availability in Extended Capabilities Element by acquiring full beacon [ bit xx-xx+1 : TBD ]. If AP provides TFM 2 P, STA are able to select and perform any of TFM 2 P service available, i. e. simple accuracy announcement, TFM 2 P time stamp announcement (APbroadcast), or TFM 2 P node-by-node handshake. • Even if all STAs can use simple TSF accuracy information only without frequency measurement, AP should still provide accuracy announcement and stability information of corresponding services in TFM 2 P IE carried by frame body of full beacon including the detailed parameters. It is up to STA’s decision if any frequency measurement is performed or not. • As like existing TSF timer advertisement, STA shall correct its TSF timer offset with AP timer, and this corrected timer value has to be stored as TS , which is the origin of wake up timing calculation. Submission Slide 17 Shusaku Shimada Yokogawa Co.

Nov. 2012 doc. : IEEE 802. 11 -12/1376 r 0 Procedure (2) : Time Stamp Announcement ( AP-broadcast ) • If AP provides TFM 2 P and a STA selects the timestamp announcement (AP-broadcast) service to calculate its wake up margin, the STA has to obtain AP timer stability information (±εadvertised ) and number of times ( N ) of full beacon carrying DTIM to measure AP TSF timer frequency. Usually, N should be more than 1 sec = 1 million times of 1 us TSF. • Then the STA acquires three consecutive full beacons N-times apart each other and takes To. A information of first two reception by STA PHY itself. Furthermore, STA collects the To. D information corresponding to first two full beacon carried by last two beacons. • Now the STA has two set of To. D-To. A pairs from three full beacon and can estimate the frequency correction coefficient (δ 2 ; ppm). The timing resolution of stamps may be always 1 us or defined by higher layer [TBD]. • Eventually STA determines the wake up margin from, △measured , εadvertised and STA specific stability ( εSTA ; see “implementation practice”) if required. Submission Slide 18 Shusaku Shimada Yokogawa Co.

Nov. 2012 doc. : IEEE 802. 11 -12/1376 r 0 Procedure (3) : Node-by-node handshake ( STA-unicast ) Submission Slide 19 Shusaku Shimada Yokogawa Co.

Nov. 2012 doc. : IEEE 802. 11 -12/1376 r 0 Procedure (3) : Node-by-node handshake ( STA-unicast ) …. Continued from previous slide. • The data format of timestamp should be same as 802. 1 AS structure with 1 ns resolution. struct Timestamp { UInteger 48 seconds; UInteger 32 nanoseconds; }; • Now the STA has two set of To. D-To. A pairs from two times repetition of timing measurement handshake and can estimate the frequency correction coefficient (δ 2 ; ppm), using peer correction coefficient (δ 1 ; ppm). • Eventually STA determines the wake up margin from, △measured , εadvertised , δ 1 and STA specific stability factors ( εSTA ) if required. ( refer to “implementation practice” ) Submission Slide 20 Shusaku Shimada Yokogawa Co.

Nov. 2012 doc. : IEEE 802. 11 -12/1376 r 0 Addition to Extended Capabilities IE [TBD] • Extended capability Element ID Length Capabilities Octets : 1 n • Element ID = 127 • Capability bit = xx - xx+1 [ TBD, e. g. 49 -50 ] bit xx-xx+1 Submission Information Notes TFM 2 P Service availability Set to 0, when unavailable. TSF accuracy = ± 100 ppm Set to 1, when simple accuracy announcement of TSF and TFM 2 P timestamp announcement (AP-broadcast) are available. (for AP only) Set to 2, when simple accuracy announcement of TSF and/or TFM 2 P node by node handshake are available. (for AP (and), for STA(and/or)) Set to 3, when all TFM 2 P timestamp announcement (AP-broadcast), node by node handshake and simple accuracy announcement are available. (for AP only) Slide 21 Shusaku Shimada Yokogawa Co.

Nov. 2012 doc. : IEEE 802. 11 -12/1376 r 0 New IE for TFM 2 P (1) [TBD] • Information Element ID Length TFM 2 P capability Accuracy Stability N To. D Time stamp AP-broadcast reserved Octet : 1 1 10 • Element ID = xxx [ TBD, e. g. 175 ] • Capability Note, N : number of times of DTIM beacon for the interval between two time measurements • 0 for TFM 2 P unavailable. Accuracy has to be ± 100 ppm and stability has to be ± 0. • 1 for TFM 2 P simple accuracy announcement of TSF and TFM 2 P timestamp announcement (APbroadcast) are available as well. (usually for AP only) • 2 for TFM 2 P simple accuracy announcement of TSF and TFM 2 P node by node handshake are available as well. (for both AP and STA) • 3 for all TFM 2 P simple accuracy announcement of TSF, timestamp announcement (AP-broadcast) and node-by-node handshake are available. (usually for AP only) • 4 for TFM 2 P simple accuracy announcement of TSF is only available. (usually for STA only) • 5 for TFM 2 P node-by-node handshake is only available. (usually for STA only) • 6 -255: reserved. Submission Slide 22 Shusaku Shimada Yokogawa Co.

Nov. 2012 doc. : IEEE 802. 11 -12/1376 r 0 New IE for TFM 2 P (2) [TBD] • Information Element ID Length TFM 2 P capability Accuracy Stability To. D Time stamp AP-broadcast reserved Octet : 1 1 • Accuracy : 2 times integer in ppm (means ±value = 0 to absolute max. ) i. e. resolution of half ppm • Stability : 4 times integer in ppm (means ±value = 0 to absolute max. ) i. e. resolution of quarter ppm • To. D Time stamp announcement in case of AP-broadcast struct Timestamp { UInteger 64 microseconds; UInteger 16 nanoseconds; }; Submission This is different from 802. 1 AS structure, because TSF timer resolution of 1 us has to be maintained. Slide 23 Shusaku Shimada Yokogawa Co.

Nov. 2012 doc. : IEEE 802. 11 -12/1376 r 0 Implementation practice for TFM 2 P mechanism How much wake up margin should set at each STA is out of the scope of Standard. However following implementation practice should work for typical implementation of usual sensor nodes including all of use case 1 applications in general. Wake up margin at the frequency estimating STA may be the sum of peer stability information and actual latest fluctuation of measurement by itself. (1) The difference ( i. e. fluctuation ) between latest measured frequency correction coefficient and previous coefficient ( △ δ 2 ) can be summed up with the advertised stability information (±εadvertised ), in addition to the STA’s stability coefficient value. (2) This estimating STA side stability coefficient value with the fluctuation ( △ δ 2 ) can be always updated and maintained for next δ 2 estimation using TFM 2 P. ( εSTA ) (3) If the fluctuation (△ δ 2 ) is small, εSTA will be minimum. Submission Slide 24 Shusaku Shimada Yokogawa Co.

Nov. 2012 doc. : IEEE 802. 11 -12/1376 r 0 Straw poll (1) • Do you support to include the TSF timer frequency measurement function into SFD. – Yes – No – Abstain Submission Slide 25 Shusaku Shimada Yokogawa Co.

Nov. 2012 doc. : IEEE 802. 11 -12/1376 r 0 Straw poll (2) • Do you support to include proposed TFM 2 P procedure as the TSF timer frequency measurement function, into SFD. – Yes – No – Abstain Submission Slide 26 Shusaku Shimada Yokogawa Co.

Nov. 2012 doc. : IEEE 802. 11 -12/1376 r 0 References [1] 11 -12/130 r 0 “Beacon Reception of Long Sleeper” [2] IEEE 802. 11 -2012 [3] IEEE 1588/PTP [4] 11 -11/0905 r 5” TGah Functional Requirements and Evaluation Methodology Rev. 5” [5] PAR and 5 C Submission Slide 27 Shusaku Shimada Yokogawa Co.

Nov. 2012 doc. : IEEE 802. 11 -12/1376 r 0 Appendix : PHY-assist rules for time stamp • Timing Measurement Procedure: IEEE 802. 11 -2012 • Standardized mechanism of To. D/To. A time stamp • Proposed Measurement Point for both To. D/To. A • Either end of STF or start of LTF : t. LTF • Proposed To. A validation by Sig with no CRC error • Every detection of t. LTF is stored (over written) if CRC passed. • By TFM 2 P Procedure • To. A time stamp of frame destined to the STA itself only be used. Submission Slide 28 Shusaku Shimada Yokogawa Co.

Nov. 2012 doc. : IEEE 802. 11 -12/1376 r 0 Examples Submission Slide 29 Shusaku Shimada Yokogawa Co.

Nov. 2012 doc. : IEEE 802. 11 -12/1376 r 0 Frequency Measurement (example 1) No frequency error Propagation Delay=0 Sending STA(f 1) t 1 t 4 Receiving STA(f 2) M 1 Ack offset 1= [(t 2 -t 1)-(t 4 -t 3)]/2 =[(1234578901 -1234567890)-(1234667890 -1234678901)]/2 = 11011 t 2 t 3 offset 2=[(t 6 -t 5)-(t 8 -t 7)]/2 =[(1235627477 -1235616466)-(1235716466 -1235727477)]/2 = 11011 t 5 t 8 M 2 Ack Submission t 6 t 7 Slide 30 Shusaku Shimada Yokogawa Co.

Nov. 2012 doc. : IEEE 802. 11 -12/1376 r 0 Frequency Measurement (example 2) No frequency error Propagation Delay=1 u. Sec Sending STA(f 1) t 1 t 4 Receiving STA(f 2) M 1 Ack t 2 t 3 offset 1= [(t 2 -t 1)-(t 4 -t 3)]/2 =[(1234578902 -1234567890)-(1234667892 -1234678902)]/2 = (11012+11010)/2=11011 offset 2=[(t 6 -t 5)-(t 8 -t 7)]/2 =[(1235627478 -1235616466)-(1235716468 -1235727478)]/2 = (11012+11010)/2=11011 t 5 t 8 M 2 Ack Submission t 6 t 7 Slide 31 Shusaku Shimada Yokogawa Co.

Nov. 2012 doc. : IEEE 802. 11 -12/1376 r 0 Frequency Measurement (example 3) f 2 frequency offset ≈ 4 ppm Propagation Delay=1 u. Sec Sending STA(f 1) t 1 t 4 Receiving STA(f 2) M 1 Ack t 2 t 3 offset 1= [(t 2 -t 1)-(t 4 -t 3)]/2 =[(1234578902 -1234567890)-(1234667892 -1234678902)]/2 = (11012+11010)/2=11011 offset 2=[(t 6 -t 5)-(t 8 -t 7)]/2 =[(1235627482 -1235616466)-(1235716468 -1235727482)]/2 = (11016+11014)/2=11015 t 5 t 8 M 2 Ack Submission t 6 t 7 Slide 32 Shusaku Shimada Yokogawa Co.

Nov. 2012 doc. : IEEE 802. 11 -12/1376 r 0 Frequency Measurement (example 4) f 1, f 2 frequency offset ≈ 4 ppm Propagation Delay=1 u. Sec Sending STA(f 1) t 1 t 4 Receiving STA(f 2) M 1 Ack t 2 t 3 offset 1= [(t 2 -t 1)-(t 4 -t 3)]/2 =[(1234578902 -1234567890)-(1234667892 -1234678902)]/2 = (11012+11010)/2=11011 offset 2=[(t 6 -t 5)-(t 8 -t 7)]/2 =[(1235627482 -1235616466)-(1235716468 -1235727482)]/2 = (11016+11014)/2=11015 t 5 t 8 M 2 Ack Submission t 6 t 7 Slide 33 Shusaku Shimada Yokogawa Co.

Nov. 2012 doc. : IEEE 802. 11 -12/1376 r 0 End Submission Slide 34 Shusaku Shimada Yokogawa Co.