doc.: IEEE 802.11-11/1482r4 Submission January 2012 Sameer Vermani, Qualcomm Preamble Format for 1 MHz Date: 2012-01-16 Authors: N am e A ffiliations A ddress Phone em ail Sam eerV erm ani Qualcom m 858-845-3115 svverman@ qualcom m.com Eugene Baik Qualcom m Lin Y ang Qualcom m H em anth Sam path Qualcom m V K Jones Qualcom m Richard V an N ee Qualcom m R on Porat Broadcom N iharJindal Broadcom V inko Erceg Broadcom H ongyuan Zhang M arvell Y ong Liu M arvell Raja Banerjea M arvell Slide 1
Transcript
doc.: IEEE 802.11-11/1482r4
Submission
January 2012
Sameer Vermani, Qualcomm
Preamble Format for 1 MHz
Date: 2012-01-16Authors:
Name Affiliations Address Phone email Sameer Vermani
Name Affiliations Address Phone email Eldad Perahia Intel
Tom Tetzlaff Intel
Tom Kenney Intel
Jianhan Liu MediaTek
Vish Ponnampalam MediaTek
ChaoChun Wang MediaTek
James Wang MediaTek
James Yee MediaTek
Yongho Seok
LG Electronics
Seunghee Han
LG Electronics
Jinsoo Choi
LG Electronics
Jinsam Kwak LG Electronics
George Calcev Huawei
Bin Chen Huawei
Young Hoon Kwon Huawei
Junghoon Suh Huawei
Tianyu Wu Huawei
Slide 2
doc.: IEEE 802.11-11/1482r4
Submission
Abstract• Based on motions in document 1294r1, the group adopted the following:
– An 802.11ah STA shall support reception of 1 MHz and 2 MHz PHY transmissions.
• Subsequently, in document 1311r0, the group motioned in:– The 2 MHz PHY transmission shall be an OFDM based waveform consisting of a total of 64
tones (including tones allocated as pilot, guard and DC). – The tone spacing for all other bandwidth modes shall be same as the tone spacing in the 2
MHz mode
• Above implies that the 1MHz mode is based on 32 sub-carriers• This presentation is a follow-up to document 1482r2 where we first
introduced a preamble format for 1MHz– Did not run strawpolls/motions in order to give more time for discussions
• Outline of this presentation (Italics is essentially a review of 1482r2)– Lowest rate for 1MHz
• 2x repetition is needed for the 1MHz mode to provide good in-home coverage
– Propose a preamble format for the 1MHz mode which can support 2x repetition• Robust packet detection• LTF section to provide accurate channel estimation• Simple mode detection between 1MHz and 2MHz modes• SIG Field Design
Sameer Vermani, Qualcomm
January 2012
Slide 3
doc.: IEEE 802.11-11/1482r4
Submission
Lowest rate for 1MHz: Need for Repetition• Based on 900 MHz in-home measurement results shown in appendix, need 105
dB of pathloss for good in-home coverage– 2 out of 3 homes we did measurements in showed 105 dB as the worst case path-loss
• Making the following assumptions for link budget,– 0 dBm output power (without PA)– -4dBi TX antenna gain– 105dB pathloss
• Worst case pathloss seen in 2 out of 3 homes during measurements– -3dBi RX antenna gain– kTBW = -115dBm– NF = 4dB
• To close the link, the lowest MCS should be able to operate at -1 dB SNR or better in AWGN– SNR = 0 – 4 – 105– 3 - (-115 + 4) = -1dB
• Means we need 2x repetition for robust coverage– 2x repetition of 11ac MCS0 has 10^(-1) PER around -2 dB SNR in AWGN
Sameer Vermani, Qualcomm
January 2012
Slide 4
doc.: IEEE 802.11-11/1482r4
Submission
Impact of not having repetition
• 11ah will be subjected to similar complaints about indoor coverage that exist in current WiFi standards– In the SEP2.0 Market Requirement document, created with heavy input from utility
companies, range of Wi-Fi networks was deemed insufficient in the case of some larger home configurations and in Multi Dwelling Units
• Without repetition, the following options may be needed for improving range:– Require customers to install higher-power sensors for “hard to reach places”
• Not practical to sell two types of sensors, those for the ‘easy to reach places’ and those for the ‘hard to reach places’
• “High-power sensors” cannot work with coin-cell batteries due to peak current drain– Force customers to deploy multiple 11ah APs
• Choosing location & configuration of the APs non-trivial for average person
Sameer Vermani, Qualcomm
January 2012
Slide 5
doc.: IEEE 802.11-11/1482r4
Submission
Observations
• We believe a 2x repetition mode should be included in the 11ah standard– Additional 1MHz preamble overhead required is manageable in our view, given a well-
designed preamble.
• Harder to justify 4x repetition due to burden placed on 1MHz normal rate high-MCS transmissions, caused by the shared preamble length overhead– If 4x repetition is desired, best approach would be to separate 1MHz normal rate and low-
rate repetition mode preambles which leads to proliferation of modes
• Rest of the presentation proposes a 1MHz preamble format assuming a 2x repetition rate is present– Common preamble for repetition and non-repetition MCSs
Sameer Vermani, Qualcomm
January 2012
Slide 6
doc.: IEEE 802.11-11/1482r4
Submission
Packet Detection• Detection needs to be sensitive enough to support 2x repetition
• There is possibility of power boost on STF– PAPR of STF is much lower than data
• STF Proposal– Power boosted 4 symbol STF for 2x repetition MCS only [See appendix for details]
• Quantization error not an issue at rep-2 MCS, so receiver can be oblivious of the power boost– Tone allocation to ensure same periodicity in 1MHz and 2MHz STFs
• For 2MHz: {±4 ±8 ±12 ±16 ±20 ±24}• For 1MHz: {±4 ±8 ±12}
Sameer Vermani, Qualcomm
January 2012
Slide 7
doc.: IEEE 802.11-11/1482r4
Submission
LTF section• Use 4 LTF symbols
– In order to extract ~ 3.0 dB gain out of 2x repetition, we need 4 LTF symbols [See Appendix]
– Add separate GI for 3rd and 4th LTF symbols to align LTF symbols with the SIG of 2MHz
Sameer Vermani, Qualcomm
LTF section(4 repetitions of LTF1)
DGI LTF symbol LTF symbol GI LTF symbol GI LTF symbol
January 2012
Slide 8
doc.: IEEE 802.11-11/1482r4
Submission
Mode Detection: 1 vs 2 MHz• Propose to have 2 methods of mode detection• Method 1: Orthogonal LTF sequences for 1MHz and 2MHz [2]
– Can keep the 2MHz LTF the same– The only “requirement” on the 1 MHz LTF is that it should be nearly orthogonal with the
2MHz LTF within the overlapping band– Confirmed through simulations that freq domain cross-correlation approach is robust [See
Appendix for results]• Method 2: QBPSK rotation on the 1st symbol of 2 MHz SIG
– Same as 11n GF HTSIG field modulation– Differentiate from the 3rd LTF1 symbol (composed by ±1) of 1MHz
Sameer Vermani, Qualcomm
January 2012
Slide 9
STF1(Same periodicity
as STF2, 3 dB boosted for rep 2)
Repetition coded SIG Field
4 LTFs (Orthogonal to 2MHz LTF in
freq)Data
SIG Field(2 symbols
QBPSK rotated
2 LTFs DataSTF2Additional
LTFs if MIMO 2 MHz, 64FFT, SU open loop
1 MHz, 32FFT, SU open loop
SIG-A Field1st symbol
QBPSK rotated
2 LTFs DataSTF2 VHT-STF VHT-LTFs VHT-SIG-B2 MHz, 64FFT, MU
doc.: IEEE 802.11-11/1482r4
Submission
SIG Field
• SIG goes at BPSK-rate ½ -rep 2• No MU transmissions for the 1MHz mode• No AID supported
– Details on next slide
• Leaves 6 reserved bits assuming tail is present
Sameer Vermani, Qualcomm
SIG Field Bits Comments
STBC 1 Same as in 11ac
Num SS 2 Number of spatial streams for SU
SGI 1 Short Guard Interval
Coding 2 1st bit is coding type (LDPC/BCC), 2nd bit is for LDPC Nsym ambiguity
MCS 4 MCS
Aggregation bit 1 Signals use of AMPDU
Length 9 Length field (in symbols when aggregation is ON, is in bytes when aggregation is OFF, Mandate AMPDU for packet sizes > 511 bytes
Reserved 6 (TBD) Some possible uses are MAC bits or any other new features etc. Details TBD
CRC 4 4 bits of CRC should be enough
Tail 6 (TBD) Tail-biting can be explored
Total 36
SIG Field Contents
January 2012
Slide 10
doc.: IEEE 802.11-11/1482r4
Submission Sameer Vermani, Qualcomm
Why no AID in 1 MHz ?
• For transmissions without aggregation, receivers can terminate at the MAC header– Terminating few symbols after SIG-field leads to only a small hit in the power save
compared to terminating on the SIG based on PAID
• For the AMPDU case, the receiver can always chose to save power by doing the following:– Terminate reception if the first MAC address does not match: similar power-save hit
(compared to PAID) as in the non-aggregated case• Chances of a later MPDU being correct once the first one fails are very low
• AMPDU transmissions are likely from extended-range WiFi use-case – These are likely to use 2 MHz transmissions that can have a PAID [1]
January 2012
Slide 11
doc.: IEEE 802.11-11/1482r4
Submission Sameer Vermani, Qualcomm
Length field
• Save bits in the SIG by having dual-interpretation length field
• Length is in number of symbols when AMPDU is used as delimiters carry the exact byte length
• Force AMPDU to be used whenever number of bytes in the packet is greater than 511– For high MCSs, the overhead of delimiters is not a problem– For low MCSs, the limit allows for long enough PPDUs without aggregation– For packets larger than 511 bytes, delimiter overhead is low
January 2012
Slide 12
doc.: IEEE 802.11-11/1482r4
Submission
Conclusions
• We proposed 2x repetition as the lowest rate for 1MHz• We also proposed a preamble format for SU open loop packets as shown above
– Robust and simple packet detection for up-to 2x repetition– Accurate channel estimation which enables repetition mode to provide ~3 dB gain over MCS0– Simple differentiation from 2MHz mode during LTFs and through QBPSK rotation of 2MHz SIG– SIG sent using 2x repetition and spanning 5-6 symbols
Sameer Vermani, Qualcomm
STF1(Same periodicity as 2 MHz,3 dB power boost for rep-2)
Repetition coded SIG Field
LTF1 (Orthogonal in freq domain to
2MHz LTF)
Repetition or non-repetition encoded Data
4 symbols 5-6 symbols4 symbols=160 ms
LTF2 to LTFN(for MIMO)
Preamble Format for 1 MHz
DGI LTS LTS GI LTS GI LTS
January 2012
Slide 13
doc.: IEEE 802.11-11/1482r4
Submission
Straw-poll 1
• Do you agree to adopt MCS0 rep 2 as the lowest rate for 1 MHz ?
Sameer Vermani, Qualcomm
January 2012
Slide 14
doc.: IEEE 802.11-11/1482r4
Submission
Straw-poll 2
• Do you agree with having a 4 symbol packet detection section for the 1 MHz mode ?– A 3 dB power boost is only applied for 2x repetition MCS– Have same periodicity as 2 MHz STF with following tone
allocations:• For 2MHz {±4 ±8 ±12 ±16 ±20 ±24} • For 1MHz {±4 ±8 ±12}.
Sameer Vermani, Qualcomm
January 2012
Slide 15
doc.: IEEE 802.11-11/1482r4
Submission
Straw-poll 3
• Do you agree with the general preamble structure for 1MHz SU open loop packet as in the figure below?– The relationship between NSTS and NLTF is the same as 11n/11ac (for 2 through 4
streams), using the same P matrix (for 1 through 4 streams)
Sameer Vermani, Qualcomm
STF1(Same periodicity as 2 MHz,3 dB power boost for rep-2)
Repetition coded SIG Field
LTF1 (Orthogonal in freq domain to
2MHz LTF)
Repetition or non-repetition encoded Data
4 symbols 5-6 symbols4 symbols
LTF2 to LTFN(for MIMO)
Preamble Format for 1 MHz
DGI LTS LTS GI LTS GI LTS
January 2012
Slide 16
doc.: IEEE 802.11-11/1482r4
Submission Sameer Vermani, Qualcomm
Straw-poll 4
• Do you accept the 1MHz SIG field contents as shown on slide 10 ?
January 2012
Slide 17
doc.: IEEE 802.11-11/1482r4
Submission Sameer Vermani, Qualcomm
APPENDIXMeasurement, Simulation Results
January 2012
Slide 18
doc.: IEEE 802.11-11/1482r4
Submission
In-home Measurement Results• Conducted measurements in 3 homes• Home 1
– Two floor home, 5000 sq. ft– House already equipped with ZigBee devices– Kept the transmitter source where the ZigBee controller was located
• Same location as the primary WLAN AP in that home– Measured pathloss to every location in the house that had a ZigBee sensor
• 23 measurement points in total– Worst case pathloss of 105 dB and 2 other locations with 95 dB pathloss
• Home 2– Three floor home, 2500 sq. ft.– Transmitter on 1st floor and tested pathloss to 7 locations– Measured Pathloss in every room on all 3 floors
• Typically 15 separate measurements were made in each room, moving the antenna ~1ft between each in order to spatially sample the fading environment.
– Worst case pathloss observed was 95 dB
• Home 3– Two floor home, 1900 sq ft– Transmitter on 1st floor and tested pathloss to various rooms on both floors, and outdoor locations– Worst case pathloss of 105 dB seen in backyard (water sprinkler location)
• The propagation path had 3 walls (including a garage wall)
• General observation: ~20-25dB path loss variations seen within a single room in a house
Sameer Vermani, Qualcomm
January 2012
Slide 19
doc.: IEEE 802.11-11/1482r4
Submission Sameer Vermani, Qualcomm
Packet detection Details
• Explanation of 3 dB power boost on 4 symbols for 2x rep– Looking at 2 MHz preamble as basis (since it is direct down-clock)
• 2 OFDM symbols of packet detection section needed to get sufficient detection sensitivity for supporting MCS0 – Same number of symbols as 11n 20 MHz
– 1 MHz preamble needs 4 symbols of packet detection to support MCS0• We need double the number of symbols as the transmission is over 32 tones (less samples per symbol)
– To support 2x repetition at 1 MHz, need 3 dB power-boost on a 4 symbol packet detection section
• Results on right confirm intuition above– Auto-correlation based detector– M is the number of STF periods used for auto-correlation
• 1 OFDM symbol contains 5 STF periods (same as 11n)
– 2x repetition has 10^(-1) PER at -2 dB in AWGN
January 2012
Slide 20
doc.: IEEE 802.11-11/1482r4
Submission
LTF section
• In order to extract ~ 3.0 dB gain out of 2x repetition, we need 4 LTF symbols
Sameer Vermani, Qualcomm
• Simulation parameters– 1 MHz, FFT size of 32, with 24 data tones– AWGN Channel– 250-byte packets– No RF impairments– Include channel estimation using 2, 4, and 8 channel
estimation (LTF) symbols– Results with smoothing over channel estimates
• 2x repetition (ideal CE)gives 3 dB gain at 10% PER
• With 2x rep at 10% PER:– CE based on 2 LTFs is 1.6 dB worse than ideal
CE– With 4 LTFs, this gap reduces to 0.8 dB– Going from 2 to 4 LTFs provides 0.8 dB more
gain
-5 -4 -3 -2 -1 0 1 2 3 4 510
-2
10-1
100
32-pt FFT, AWGN Channel, 1x1 BPSK BCC-1/2 + 2x RepetitionSmoothing is applied to all channel estimations
SNR, dB
Pac
ket
Err
or R
ate
(PE
R)
No Repetition, Ideal CE
No Rep, CE w/ 2 LTF Symbols
2x Rep, Ideal CE2x Rep, CE w/ 2 LTFs
2x Rep, CE w/ 4 LTFs
2x Rep, CE w/ 8 LTFs
January 2012
Slide 21
doc.: IEEE 802.11-11/1482r4
Submission
Mode Detection Simulation Assumptions
• The simulation uses two LTF symbols with ideal combining (3dB gain) • Transmit power for 64FFT (56 tones) assumed to be the same as 32FFT packet
(26 tones).• Plots show the probability of misclassification (not misdetection) 64FFT with
32FFT (Up or Down halves) and vice versa – note there is no need to detect packet presence – that was done during the STF
1 -1 1 1 1 -1 1 -1 -1 -1 1 0 0];– We assumed 26 tones are used – Sequence used has 2.4dB PAPR and good cross correlation properties with the 64FFT LTF sequence
Sameer Vermani, Qualcomm
32p
January 2012
Slide 22
doc.: IEEE 802.11-11/1482r4
Submission
Mode detection Simulation Results
Sameer Vermani, Qualcomm
-8 -7.5 -7 -6.5 -6 -5.5 -510
-3
10-2
10-1
SNR [dB]
Pro
b. M
is-C
lass
ifica
tion
AWGN
64FFT
32FFT U32FFT D
-6 -5 -4 -3 -2 -1 0 110
-3
10-2
10-1
100
SNR [dB]
Pro
b. M
is-C
lass
ifica
tion
SCM Urban Macro
64FFT
32FFT U32FFT D
January 2012
Slide 23
doc.: IEEE 802.11-11/1482r4
Submission
Motion 1
• Move to adopt MCS0 rep 2 as the lowest rate for 1 MHz ?
Sameer Vermani, Qualcomm
January 2012
Slide 24
doc.: IEEE 802.11-11/1482r4
Submission
Motion 2
• Move to have a 4 symbol packet detection section for the 1 MHz mode ?– A 3 dB power boost is only applied for 2x repetition MCS– Have same periodicity as 2 MHz STF with following tone
allocations:• For 2MHz {±4 ±8 ±12 ±16 ±20 ±24} • For 1MHz {±4 ±8 ±12}.
Sameer Vermani, Qualcomm
January 2012
Slide 25
doc.: IEEE 802.11-11/1482r4
Submission
Motion 3 • Move to have the general preamble structure for 1MHz SU open loop
packet as in the figure below?– The relationship between NSTS and NLTF is the same as 11n/11ac (for 2 through 4
streams), using the same P matrix (for 1 through 4 streams)
Sameer Vermani, Qualcomm
STF1(Same periodicity as 2 MHz,3 dB power boost for rep-2)
Repetition coded SIG Field
LTF1 (Orthogonal in freq domain to
2MHz LTF)
Repetition or non-repetition encoded Data
4 symbols 5-6 symbols4 symbols
LTF2 to LTFN(for MIMO)
Preamble Format for 1 MHz
DGI LTS LTS GI LTS GI LTS
January 2012
Slide 26
doc.: IEEE 802.11-11/1482r4
Submission Sameer Vermani, Qualcomm
Motion 4
• Move to adopt the 1MHz SIG field contents as shown on slide 10 ?
