IEEE /121r2 Submission May 2003 Shaomin Mo, Panasonic -- PINTLSlide 1 Project: IEEE P Working Group for Wireless Personal Area Networks (WPANs) Submission Title: [Data Whitening in Base-band to Reduce PSD of UWB Signals] Date Submitted: [5 May, 2003] Source: [Shaomin Mo] Company [Panasonic -- PINTL] Address [Two Research Way, Princeton, New Jersey, USA] Voice:[ ], FAX: [ ], Re: [IEEE P Alternative PHY Call For Proposals, IEEE P /372r8] Abstract:[Base-band processing of whitening data to reduce power spectral density of UWB signals in IEEE systems ] Purpose:[Proposal of base-band processing of whitening data to reduce power spectral density of UWB signals in IEEE systems.] Notice:This document has been prepared to assist the IEEE P It is offered as a basis for discussion and is not binding on the contributing individual(s) or organization(s). The material in this document is subject to change in form and content after further study. The contributor(s) reserve(s) the right to add, amend or withdraw material contained herein. Release:The contributor acknowledges and accepts that this contribution becomes the property of IEEE and may be made publicly available by P

IEEE /121r2 Submission May 2003 Shaomin Mo, Panasonic -- PINTLSlide 2 Data Whitening in Base-band to Reduce PSD of UWB Signals Shaomin Mo Panasonic Information and Networking Technologies Laboratories

IEEE /121r2 Submission May 2003 Shaomin Mo, Panasonic -- PINTLSlide 3 Overview Power Spectra Density (PSD) issue in UWB Analysis of PSD of UWB signals Mechanisms to reduce PSD –Phase reversion to reduce PSD –Architecture of using Linear Feedback Shift Register –Phase reversion for SYNC Conclusion

IEEE /121r2 Submission May 2003 Shaomin Mo, Panasonic -- PINTLSlide 4 PSD is an Important Issue in UWB Communication Systems FCC limited authorization of UWB technology, Feb 14, 2002 Use in restrict spectrum at restrict power Do not interfere with other wireless systems Other agencies still have some reservations about whether UWB will interfere with other wireless systems such as cellular, air navigation and landing systems

IEEE /121r2 Submission May 2003 Shaomin Mo, Panasonic -- PINTLSlide 5 Emission Levels for GSM & TDMA in the Cellular Receiver Bands TechnologyFrequency Range (mobile RX) (MHz) Emission Level (dBm), Bandwidth (kHz) Average Level (dBm/MHz) Part 15 Limit (dBm/MHz) TDMA869 – – dBm, 30kHz indoor hand-held GSM869 – – dBm, 100kHz -71 dBm, 100kHz indoor hand-held Source: “Ultra-Wideband Radio – The New Part 15”, Microwave Journal, February 2003

IEEE /121r2 Submission May 2003 Shaomin Mo, Panasonic -- PINTLSlide 6 Containing PSD is an Important Part in UWB System Design Repeat pulse trains may generate strong line spectra and high PSD Traditional scramblers are not sufficient to contain PSD PSD suppression leads to –Prevention of interference to existing systems –Potential increase in rate, Tx power (distance)

IEEE /121r2 Submission May 2003 Shaomin Mo, Panasonic -- PINTLSlide 7 Signal model Probability function of a n Model of Repeat Pulse Train

IEEE /121r2 Submission May 2003 Shaomin Mo, Panasonic -- PINTLSlide 8 Ps is determined by w(t) and Tc Ps is not affected by Pr{a n } Total PSD is determined by w(t) and Tc Total PSD is not affected by Pr{a n } PSD of Repeat Pulse Train

IEEE /121r2 Submission May 2003 Shaomin Mo, Panasonic -- PINTLSlide 9 PSD of repeat pulse trains consists of S c (f) – continuous component S d (f) – discrete component PSD of Repeat Pulse Train W(f) Tc p

IEEE /121r2 Submission May 2003 Shaomin Mo, Panasonic -- PINTLSlide 10 W(f) – pulse shape & Tx power Tc – clock period or pulse rate p – probability in distribution function –Does not affect total PSD –Changes distribution of PSD between continuous and discrete components Parameters that Determine PSD

IEEE /121r2 Submission May 2003 Shaomin Mo, Panasonic -- PINTLSlide 11 Simplified Form of PSD

IEEE /121r2 Submission May 2003 Shaomin Mo, Panasonic -- PINTLSlide 12 Relationship between Continuous and Discrete Components

IEEE /121r2 Submission May 2003 Shaomin Mo, Panasonic -- PINTLSlide 13 Because total PSD is constant A(f) = B(f) Max(S c (f)) = Max(S d (f)) Relationship between Continuous and Discrete Components

IEEE /121r2 Submission May 2003 Shaomin Mo, Panasonic -- PINTLSlide 14 Total continuous PSD equals total discrete PSD The continuous distributes on all frequencies The discrete distributes on those discrete frequencies separated by 1/Tc. Continuous PSD is lower than that of discrete PSD on the same frequency components Relationship between Continuous and Discrete Components

IEEE /121r2 Submission May 2003 Shaomin Mo, Panasonic -- PINTLSlide 15 PSD with Different p Has Same Envelop but Different Level PSD of single pulseP = 0.25 P = 0.5P = 1.0 Line spectra peak = 15 peak = 9 peak = 3

IEEE /121r2 Submission May 2003 Shaomin Mo, Panasonic -- PINTLSlide 16 Contain PSD Reduce or eliminate discrete component of PSD  reduce PSD across whole spectrum Make Objective of Design

IEEE /121r2 Submission May 2003 Shaomin Mo, Panasonic -- PINTLSlide 17 TDMA Systems Traditional communication systems require randomness inside a frame for timing recovery, equalization, etc.

IEEE /121r2 Submission May 2003 Shaomin Mo, Panasonic -- PINTLSlide 18 New Requirements to UWB Communication Systems Traditional: randomness in X direction UWB: randomness in both X & Y directions

IEEE /121r2 Submission May 2003 Shaomin Mo, Panasonic -- PINTLSlide 19 PSD Analysis: if data is not evenly distributed in Y direction, line spectra appear Phase Waveform of single pulseWaveform of data PS of single pulsePSD of data Original stream: line spectra & peak = 17

IEEE /121r2 Submission May 2003 Shaomin Mo, Panasonic -- PINTLSlide 20 Propose 1: Phase Reversion to Reduce PSD A random sequence {b n } is generated with c n = a n ^ b n. It can be proved that {c n } is used as the new data for transmission.

IEEE /121r2 Submission May 2003 Shaomin Mo, Panasonic -- PINTLSlide 21 Using proposed scheme, line spectra is eliminated and PSD is reduced Waveform of single pulseWaveform of data PS of single pulsePSD of data Proposed 1: PSD of c n, Line spectra gone peak reduced to 8

IEEE /121r2 Submission May 2003 Shaomin Mo, Panasonic -- PINTLSlide 22 Signal model Model of Repeat Pulse Train of Multi-band

IEEE /121r2 Submission May 2003 Shaomin Mo, Panasonic -- PINTLSlide 23 {w m } is a set of waveforms on sub-bands Probability function of a n, same on all sub- bands Model of Repeat Pulse Train (cont.)

IEEE /121r2 Submission May 2003 Shaomin Mo, Panasonic -- PINTLSlide 24 PSD of repeat pulse trains consists of S c (f) – continuous component S d (f) – discrete component PSD of Repeat Pulse Train W(f) Tc p pnpn

IEEE /121r2 Submission May 2003 Shaomin Mo, Panasonic -- PINTLSlide 25 W(f) – pulse shape & Tx power Tc – clock period or pulse rate p m – probability in distribution function of sub- bands p – probability in distribution function of waveforms –Does not affect total PSD –Changes distribution of PSD between continuous and discrete components Parameters that Determine PSD

IEEE /121r2 Submission May 2003 Shaomin Mo, Panasonic -- PINTLSlide 26 Contain PSD Reduce or eliminate discrete component of PSD Make Objective of Design Contain PSD Reduce or eliminate discrete component of PSD Make

IEEE /121r2 Submission May 2003 Shaomin Mo, Panasonic -- PINTLSlide 27 To Make {w m } Evenly Distributed – Rotationally w1w1 w4w4 w3w3 w2w2 w8w8 w7w7 w6w6 w5w5 W m is waveform on sub-band m, 1  m  M

IEEE /121r2 Submission May 2003 Shaomin Mo, Panasonic -- PINTLSlide 28 To Make {w m } Evenly Distributed – Randomly Another way to make {w m } evenly distributed is to randomly and evenly choose w m so that

IEEE /121r2 Submission May 2003 Shaomin Mo, Panasonic -- PINTLSlide 29 To Make {a n } Evenly Distributed A random sequence {b n } is generated with c n = a n ^ b n. It can be proved that {c n } is used as the new data for transmission.

IEEE /121r2 Submission May 2003 Shaomin Mo, Panasonic -- PINTLSlide 30 Waveforms in multi-bandOriginal data, p = 1 PS of waveformsResult data peak = 4 Line spectra peak = 21 PSD of BPSK Data with p=1 & rotationally

IEEE /121r2 Submission May 2003 Shaomin Mo, Panasonic -- PINTLSlide 31 PSD of BPSK Data with p=0.25 & rotationally Waveforms in multi-bandOriginal data, p = 0.25 PS of waveformsResult data peak = 4 Line spectra peak = 17

IEEE /121r2 Submission May 2003 Shaomin Mo, Panasonic -- PINTLSlide 32 PSD of BPSK Data with p=0.4 & rotationally Waveforms in multi-bandOriginal data, p = 0.4 PS of waveformsResult data peak = 4 Line spectra peak = 9

IEEE /121r2 Submission May 2003 Shaomin Mo, Panasonic -- PINTLSlide 33 Waveforms in multi-bandOriginal data, p = 1 PS of waveformsResult data peak = 5 Line spectra peak = 22 PSD of BPSK Data with p=1 & randomly

IEEE /121r2 Submission May 2003 Shaomin Mo, Panasonic -- PINTLSlide 34 PSD of BPSK Data with p=0.25 & randomly Waveforms in multi-bandOriginal data, p = 0.25 PS of waveformsResult data peak = 5 Line spectra peak = 15

IEEE /121r2 Submission May 2003 Shaomin Mo, Panasonic -- PINTLSlide 35 PSD of BPSK Data with p=0.4 & randomly Waveforms in multi-bandOriginal data, p = 0.4 PS of waveformsResult data peak = 5 Line spectra peak = 10

IEEE /121r2 Submission May 2003 Shaomin Mo, Panasonic -- PINTLSlide 36 PSD of QPSK Data with p=1 & rotationally Waveforms in multi-bandOriginal data, p = 1 PS of waveformsResult data peak = 4 Line spectra peak = 19

IEEE /121r2 Submission May 2003 Shaomin Mo, Panasonic -- PINTLSlide 37 PSD of QPSK Data with p=0.25 & rotationally Waveforms in multi-bandOriginal data, p = 0.25 PS of waveformsResult data peak = 4 Line spectra peak = 12

IEEE /121r2 Submission May 2003 Shaomin Mo, Panasonic -- PINTLSlide 38 PSD of QPSK Data with p=0.4 & rotationally Waveforms in multi-bandOriginal data, p = 0.4 PS of waveformsResult data Line spectra peak = 7 peak = 4

IEEE /121r2 Submission May 2003 Shaomin Mo, Panasonic -- PINTLSlide 39 PSD of QPSK Data with p=1 & randomly Waveforms in multi-bandOriginal data, p = 1 PS of waveformsResult data peak = 5 Line spectra peak = 18

IEEE /121r2 Submission May 2003 Shaomin Mo, Panasonic -- PINTLSlide 40 PSD of QPSK Data with p=0.25 & randomly Waveforms in multi-bandOriginal data, p = 0.25 PS of waveformsResult data peak = 4 Line spectra peak = 13

IEEE /121r2 Submission May 2003 Shaomin Mo, Panasonic -- PINTLSlide 41 PSD of QPSK Data with p=0.4 & randomly Waveforms in multi-bandOriginal data, p = 0.4 PS of waveformsResult data Line spectra peak = 7 peak = 5

IEEE /121r2 Submission May 2003 Shaomin Mo, Panasonic -- PINTLSlide 42 Major Challenge in Implementing Phase Reversion Simple way to generate random sequence Easy way to synchronize random number generators in both transmitters and receivers

IEEE /121r2 Submission May 2003 Shaomin Mo, Panasonic -- PINTLSlide 43 Propose 2: Architecture of LFSR LFSR stands for Linear Feedback Shift Registers Easy implementation Very suitable for semiconductor implementation

IEEE /121r2 Submission May 2003 Shaomin Mo, Panasonic -- PINTLSlide 44 LFSR is loaded with a RN per frame & updated per pulse

IEEE /121r2 Submission May 2003 Shaomin Mo, Panasonic -- PINTLSlide 45 Synchronization of LFSR Initial system channel access –Random vectors are generated in advance & stored in an array –Transmitters & receivers keep same array –Index to a vector in the array is put in data to transmit Initial traffic channel access –Sequence number can be used

IEEE /121r2 Submission May 2003 Shaomin Mo, Panasonic -- PINTLSlide bit LFSR vs. Idea Low Bound LFSR is too short Strong line spectra exist Phase controlled by RNs as reference of low bound Proposed LFSR implementation

IEEE /121r2 Submission May 2003 Shaomin Mo, Panasonic -- PINTLSlide bit LFSR vs. Idea Low Bound LFSR is long enough Line spectra is suppressed Very close to reference Phase controlled by RNs as reference of low bound Proposed LFSR implementation

IEEE /121r2 Submission May 2003 Shaomin Mo, Panasonic -- PINTLSlide 48 Propose 3: Phase Reversion on SYNC Three mechanisms can be used: Phase reversion on the whole SYNC SYNC is divided into symbols & phase reversion on symbols Phase reversion & scrambling on symbols

IEEE /121r2 Submission May 2003 Shaomin Mo, Panasonic -- PINTLSlide 49 Phase Reversion on SYNC/symbols can eliminate line spectra but not ripples in PSD One cycle of symbols PSD with phase reversion PSD without phase reversion Waveform of symbols Propose 3: line spectra gone Original: strong line spectra

IEEE /121r2 Submission May 2003 Shaomin Mo, Panasonic -- PINTLSlide 50 Scramble Symbols

IEEE /121r2 Submission May 2003 Shaomin Mo, Panasonic -- PINTLSlide 51 Phase Reversion & Scrambling on SYNC/symbols can smooth ripples & eliminate line: snap shot at 10, runs Proposed 3: PSD of symbol-based phase reversion & scrambling Very close to reference Phase controlled by RNs as reference of low bound

IEEE /121r2 Submission May 2003 Shaomin Mo, Panasonic -- PINTLSlide 52 Conclusion Phase reversion can effectively reduce PSD Phase reversion can be applied to PAM, PPM, Time-Hopping to reduce PSD LFSR is an easy way to generate RNs with good performance Scrambling can enhance performance by smoothing ripples in PSD with extra processing & can be extended beyond SYNC

IEEE /121r2 Submission May 2003 Shaomin Mo, Panasonic -- PINTLSlide 53 Thank you