1. IEEE 802.15.13 Introduction into Pulsed Modulation PHY
May 2015
doc.: IEEE /0496r1
May 2018
IEEE Introduction into Pulsed Modulation PHY
Date: Place: Warsaw, Poland
Authors:
Volker Jungnickel (Fraunhofer HHI)
Edward Au (Marvell Semiconductor)

2. May 2015
doc.: IEEE /0496r1
May 2018
Abstract
This presentation provides an overview of the pulsed modulation PHY in doc /r2 discussed in TG13.
Volker Jungnickel (Fraunhofer HHI)
Edward Au (Marvell Semiconductor)

3. Table of Content Physical layers in TG13 Pulsed modulation PHY
May 2015
doc.: IEEE /0496r1
May 2018
Table of Content
Physical layers in TG13
Pulsed modulation PHY
Numerology
PPDU format
Fields
Synchronization
Channel estimation
Header
MIMO reference signals
Payload
Volker Jungnickel (Fraunhofer HHI)
Edward Au (Marvell Semiconductor)

4. Shannon’s theorem: C=B*log2(1+SNR)
May 2015
doc.: IEEE /0496r1
May 2018
Physical Layers in TG13
High-bandwidth OFDM PHY
LB OFDM PHY
Pulsed Modulation PHY
Spectral Efficiency
Bandwidth
Shannon’s theorem: C=B*log2(1+SNR)
C: Channel capacity, B: Bandwidth, SNR: electrical Signal-to-noise ratio
Slide 4
Volker Jungnickel (Fraunhofer HHI)
Edward Au (Marvell Semiconductor)

5. Pulsed modulation PHY Why? Pulsed Modulation PHY (PM PHY)
May 2015
doc.: IEEE /0496r1
May 2018
Pulsed modulation PHY
Why?
100 Mbit/s = 10 MHz * 10 bps/Hz needs a very high SNR ~ dB
100 Mbit/s = 100 MHz * 1 bps/Hz needs 3dB x 9 =27 dB less electrical SNR
Higher bandwidth  20x less optical power (Uplink, Internet of Things)
Pulsed Modulation PHY (PM PHY)
Target is 10…100 Mbit/s with low power for enhanced coverage
High bandwidth useful with special LED drivers
Lower SNR  lower spectral efficiency
Serial modulation with Reed-Solomon Coding, 2-PAM and 8B10B line coding or M-PAM with Hadamard-Coded Modulation (HCM)
Rate adaptation through variable M and variable number of HCM codes
Support for MIMO  spatial diversity, reliability
Slide 5
Volker Jungnickel (Fraunhofer HHI)
Edward Au (Marvell Semiconductor)

6. Descriptive Text Pulsed Modulation PHY May 2018
doc.: IEEE /0496r1
May 2018
Descriptive Text
Pulsed Modulation PHY
Pulsed Modulation PHY enables moderate data rates from 1 Mbit/s to some 100 Mbit/s. The main approach is to achieve high data rates by using high bandwidth while keeping spectral efficiency low. This approach offers higher reach where power efficiency is an issue, e.g. in uplink scenarios and the Internet of Things (IoT). 2-PAM with 8B10B line coding and variable optical clock rate or M-ary PAM with Hadamard-Coded Modulation (HCM) are used together with RS FEC. Controlled by higher layers, the PM PHY includes means to adapt the data rate to varying channel conditions i) by varying the optical clock rates, ii) by varying the modulation alphabet size M for PAM and the number of codes used in HCM and iii) by selecting the most appropriate set of transmitters.
Volker Jungnickel (Fraunhofer HHI)
Edward Au (Marvell Semiconductor)

7. May 2015
doc.: IEEE /0496r1
May 2018
Numerology
Opt. clock rate /MHz
Opt. clockc ycle/ns
Tseq/ns
TCP/ns
Nseq/optical clock cycles
NCP/optical clock cycles
MCS
Data rate/
Mbit/s
Channel estimation sequence (Appendix
6.25
160
5120 32 1
2-PAM
8B10B
RS(256,248)
4.7
A32
12.5 80 64 2
9.4
A64 25 40
128 4 19
A128 50 20
256 8 38
A256
100 10
512 16 75
A512
200 5
1024
150
A1024
Optical clock rates (OCR) in Table 1 are obtained from a common reference clock of 100 MHz available from low-cost off-the-shelf crystal oscillators as 100 MHz/2n where n=-1…4. The reference clock can also be obtained via Ethernet using the precision time protocol (PTP) defined in IEEE std. 1588v2. Jitter can be improved by adding synchronous Ethernet (synchE) defined in ITU-T recommendation G.8262.
Volker Jungnickel (Fraunhofer HHI)
Edward Au (Marvell Semiconductor)

8. PPDU format Synchronization header Channel estimation
May 2015
doc.: IEEE /0496r1
May 2018
PPDU format
Preamble
Channel estimation
SHR
PHY header
HCS
Optional Fields
PHR
PSDU
PHY payload
Synchronization header
Channel estimation
Physical layer header with header check sequence
Optionial fields for MIMO channel estimation
PSDU for payload
Volker Jungnickel (Fraunhofer HHI)
Edward Au (Marvell Semiconductor)

9. Synchronization preamble
May 2015
doc.: IEEE /0496r1
May 2018
Synchronization preamble
AN is pseudo-noise : Gold sequence of length N
a) Preamble length L = 2xN (N=31, proposed by ETRI)
double-word and autocorrelation is standard approach
b) Preamble length L = 6xN (N=16, 32, 64, proposed by HHI)
6-fold repetition  shorter peak  improved detection
Pulsed modulation requires zero timing error
P2xN = [AN AN] 
P6xN = [AN AN -AN AN -AN -AN] 
Volker Jungnickel (Fraunhofer HHI)
Edward Au (Marvell Semiconductor)

10. Channel estimation May 2018
doc.: IEEE /0496r1
May 2018
Channel estimation
Channel estimation (CE) is used for equalization and subsequent detection of header information and data.
The CE sequence allows frequency-domain equalization by adding a cyclic prefix (CP). There is a base sequence AN with N=Nseq in Table 1and a cyclic prefix CP.
By increasing the OCR, the number of optical clock cycles for the sequence and CP, i.e. Nseq and NCP, increase proportionally, see Table 1.
The CE sequence is finally passed through a 2-PAM modulator.
Volker Jungnickel (Fraunhofer HHI)
Edward Au (Marvell Semiconductor)

11. May 2015
doc.: IEEE /0496r1
May 2018
PHY header
The PHY header has a fixed length and defines the following fields.
Field
Octet
Bits
Description FT
[7:0]
Frame type
PSDU_length
1-2
[15:0]
Length of PSDU in optical clock cycles
RS_type 3
Type of RS
NRS 4
Number of RS
Time stamp
5-8
[31:0]
Depends on the Frame type (FT)
MCS
9-12
Modulation and Coding Vector for PSDU
Volker Jungnickel (Fraunhofer HHI)
Edward Au (Marvell Semiconductor)

12. PHY header fields May 2018 FT defines the frame type
doc.: IEEE /0496r1
May 2018
PHY header fields
FT defines the frame type
FT=0: Transport frame (used in MAC e.g. for Data, RTS, CTS, ACK, Feedback, Control)
FT=1: Probe frame (used for MIMO channel estimation)
FT>1 Reserved
The PSDU length scales from 0 up to aMaxPHYFrameSize.
RS_type defines the use of time- or frequency-domain RS.
NRS is the number of RS.
Time stamp is a number counting time (e.g. in 10 ns units per second). Time per second is obtained from the one-pulse-per-second (1 PPS) signal from GPS or PTP grandmaster. Additional time information can be obtained via higher layers.
MCS defines the modulation and coding scheme (see next slide).
Volker Jungnickel (Fraunhofer HHI)
Edward Au (Marvell Semiconductor)

13. May 2015
doc.: IEEE /0496r1
May 2018
MCS field
Field
Octet
Bits
Values
Stream 1 9
[0]
Line coding
0:8B10B, 1:HCM
[3:1]
Modulation
0:2-PAM …
3:16-PAM
>3: reserved
[7:4]
NHCM
0: NHCM=0
15: NHCM=15
Stream 2-4
10-12
[31:8]
MCS defines the used modulation and coding schemes defined by the MAC layer.
For single-stream transmission, MCS is a number.
For spatial multiplexing transmission, MCS is a vector where each element contains the MCS per stream.
Volker Jungnickel (Fraunhofer HHI)
Edward Au (Marvell Semiconductor)

14. Header check sequence May 2018
doc.: IEEE /0496r1
May 2018
Header check sequence
The header check sequence (HCS) uses CRC-16 as defined in Annex C. The HCS bits shall be processed in the transmitted order. The registers shall be initialized to all ones.
Note that „Annex C“ refers to the Annex in TG13 Draft D2.
Volker Jungnickel (Fraunhofer HHI)
Edward Au (Marvell Semiconductor)

15. May 2015
doc.: IEEE /0496r1
May 2018
Optional fields
Optional fields contain reference symbols (RS) for multiple-input multiple-output (MIMO) channel estimation. Repetitions, FEC, line coding and HCS do not apply for MIMO RS. MIMO RS can be defined in time- and frequency domain. The use of time- or frequency-domain RS is configured by the MAC layer.
At lower OCR, time-domain MIMO RS are appropriate.
At higher OCR, frequency-domain MIMO RS apply.
Volker Jungnickel (Fraunhofer HHI)
Edward Au (Marvell Semiconductor)

16. May 2015
doc.: IEEE /0496r1
May 2018
Time-domain MIMO RS
Time-domain (TD) MIMO RSs are constructed as follows:
for the ith data stream/transmitter use the ith row of the NxN Hadamard matrix HK where N=Nseq. HK is given by
increment k from k=1…K with N=2K. The resulting sequence is scrambled by logical XOR operation with the base sequence AN. A cyclic prefix is finally inserted.
All pairs of sequences in HK are mutually orthogonal. The XOR operation with AN does not change the orthogonality of sequences but improves cross-correlation properties which is beneficial in case of multi-path [5, 6].
Note that the sequence for the first stream/transmitter always contains AN.
Volker Jungnickel (Fraunhofer HHI)
Edward Au (Marvell Semiconductor)

17. Frequency-domain MIMO RS
May 2015
doc.: IEEE /0496r1
May 2018
Frequency-domain MIMO RS
Frequency-domain (FD) MIMO RSs apply for transmissions at higher OCR using FDE.
FD MIMO RS are orthogonal in the frequency domain.
FD MIMO RSs are a set of NRS OFDM symbols constructed by using the base sequence AN where N=Nseq according to Table 1 as follows.
A specific comb of subcarriers identifies a particular stream or transmitter.
Comb spacing Δ is defined by higher layers taking Δ≤Nseq/NCP into account.
The definition of Δ is contained in ERS_type and IRS_type.
There are Ncomb=Nseq / Δ non-zero signals (tines) in the comb. The base sequence AN where N=Ncomb yield an appropriate definition of the signals on these tines.
For the first stream/transmitter, the comb starts at the first subcarrier following the DC subcarrier onto which the first element of AN is mapped, while other subcarrier carry the other elements of AN consecutively.
By using a single FD MIMO RS, up to Δ-1 streams/transmitters can be identified by a cyclic shift of the comb by Nshift=0…Δ-2 of subcarriers.
The MAC layer shall reserve the shift Nshift = Δ-1 for noise estimation at the receiver.
Volker Jungnickel (Fraunhofer HHI)
Edward Au (Marvell Semiconductor)

18. Frequency-domain MIMO RS (2)
May 2015
doc.: IEEE /0496r1
May 2018
Frequency-domain MIMO RS (2)
Any subset of streams/transmitters smaller than Δ-1 can be identified by a single FD MIMO RS. When using more streams or transmitters, add more RSs.
Higher layers shall indicate the use of MIMO RS by variables Δ and NRS, where index RS means ERS and IRS, accordingly.
In order to keep RSs for multiple subsets of streams/transmitters mutual orthogonal, the mth RS is obtained by multiplication of the appropriate RS with the respective elements from the mth row of the MxM Hadamard matrix HK identifying the mth subset of RSs using the same cyclic comb shift.
HK is obtained as follows
where k=1…K and M=2K=NRS is defined by higher layers.
Volker Jungnickel (Fraunhofer HHI)
Edward Au (Marvell Semiconductor)

19. Header encoding and modulation
May 2015
doc.: IEEE /0496r1
May 2018
Header encoding and modulation
Scrambling is optional to randomize uncoordinated interference. For error protection, the header can be repeated. Next, 8B10B line encoding applies to the header.
Header encoding uses RS(36,24) code.
In this way, only the systematic part of the binary output code word (24 bits) is well balanced. For maintaining a constant average light output for the entire sequence, also the redundant part of the binary code word (36-24=12 bits) has to be passed through the 8B10B line encoder. Both parts are concatenated in a multiplexer.
Next the signal is passed through the bit-to-symbol mapper for 2-PAM modulation.
The bit-to-symbol mapper is using 2-PAM. Each input bit is mapped in one symbol. Symbols are mapped to levels as {0, 1} to {0, 1}, respectively. A constant value of 0.5 is always subtracted to make the mapper output DC free. Setting the modulation amplitude and the bias signal of the LED is due to the analogue optical frontend.
Finally, a spatial pre-coder selects what transmitters will sent out the packet and how.
Volker Jungnickel (Fraunhofer HHI)
Edward Au (Marvell Semiconductor)

20. Payload encoding and modulation
May 2015
doc.: IEEE /0496r1
May 2018
Payload encoding and modulation
Scrambling is optional to randomize uncoordinated interference.
Next, 8B10B line encoding applies to the payload.
Payload encoding uses RS(256,248) code. In this way, only the systematic part of the binary output code word (248 bits) is well balanced. For maintaining a constant average light output for the entire sequence, also the redundant part of the binary code word ( =8 bits) has to be passed through the 8B10B line encoder. Both parts are concatenated in a multiplexer and passed through the bit-to-symbol mapper for 2-PAM modulation.
In combination with Hadamard Coded Modulation (HCM) other than the trivial mode HCM(1,1), 8B10B line coding is not used while M-PAM with M≥2 can be used.
Finally, a spatial pre-coder selects what transmitters will sent out the packet and how.
Volker Jungnickel (Fraunhofer HHI)
Edward Au (Marvell Semiconductor)

21. Bit-to-symbol mapper May 2018 May 2015 doc.: IEEE 802.11-15/0496r1
Volker Jungnickel (Fraunhofer HHI)
Edward Au (Marvell Semiconductor)

22. Hadamard-coded modulation (HCM)
May 2015
doc.: IEEE /0496r1
May 2018
Hadamard-coded modulation (HCM)
Hadamard Coded Modulation (HCM) is an extension of the bit-to-symbol mapper. Besides removing the need for line coding, HCM allows the use of M-PAM with variable M, despite the high-pass characteristics of the channel, together with a variable number of codes.
Volker Jungnickel (Fraunhofer HHI)
Edward Au (Marvell Semiconductor)

23. Transmission modes for PM-PHY
May 2015
doc.: IEEE /0496r1
May 2018
Transmission modes for PM-PHY
PAM level/ spectral efficiency [bit/s/Hz]
FEC RS(n,k)
Line code
HCM
Optical Clock Rates/MHz
Data Rate/Mbps
2 / 1
(256, 248) for payload
(36,24) for header
8B10B
(1,1)
100/2n with n=-1…4
use Table 1 for HCM(1,1)
and take into account
spectral efficiency for M-PAM
spectral efficiency of line coding
the appropriate code rate of the FEC
factor m/16 by using m codes for HCM
1B1B
(1-15,16)
4 / 2
8 / 3
16 / 4
The table lists possible transmission modes by combining line coding, FEC, HCM and OCR. In combination with Table 1, it is possible to obtain the data rate for each transmission mode.
Using RS(256,248) with 2-PAM, 8B10B and n=4 (6.25 MHz) yields 4.8 Mbit/s.
Using RS(256,248) with 16-PAM, m=15 for HCM and n=0 (100 MHz) yields 363 Mbit/s.
Volker Jungnickel (Fraunhofer HHI)
Edward Au (Marvell Semiconductor)

24. May 2015
doc.: IEEE /0496r1
May 2018
Spatial precoder
In general, the spatial precoder is a matrix-vector operation P·x operating symbol-wise when using time-domain RS and subcarrier-wise when using frequency-domain RS.
If FT=0 (probe frame), the transmitter multiplies the 1x1 scalar stream of header symbols x with the NERSx1 vector P which contains all ones.
All transmitters broadcast the same header information (global transmission). The master coordinator in the infrastructure network sends the header information to all transmitters. All transmitters send in a synchronous manner. How to realize synchronization of multiple distributed OWC transmitters is out of scope for this standard.
If FT=1 (transport frame), the transmitter multiplies the 1x1 stream of header information symbols x with the NERSx1 precoding vector P which contains ones for all active transmitters in a coordinated transmission cluster and zeros elsewhere.
All transmitters in the cluster broadcast the same header information (regional transmission). The master coordinator in the infrastructure network sends header information to all active transmitters in a coordinated transmission cluster. All transmitters send in a synchronous manner. How to realize synchronization of multiple distributed OWC transmitters is out of scope for this standard.
Volker Jungnickel (Fraunhofer HHI)
Edward Au (Marvell Semiconductor)