15-04-0011-00-004a

1
Project IEEE P802.15 Working Group for Wireless
Personal Area Networks (WPANs) Submission Title
STM_CEA-LETI_CWC_AETHERWIRE_MITSUBISHI_FTRD
15.4aCFP response Date Submitted January 4th,
2005 Source Ian Oppermann (1), Mark Jamtgaard
(2), Laurent Ouvry (3), Philippe Rouzet (4),
Andreas F. Molisch(5), Philip Orlik(5), Zafer
Sahinoglu (5), Rick Roberts (6), Vern Brethour
(7), Adrian Jennings (7) , Patricia Martigne (8),
Benoit Miscopein (8), Jean Schwoerer
(8) Companies (1) CWC-University of Oulu,
Tutkijantie 2 E, 90570 Oulu, FINLAND (2) Æther
Wire Location, Inc., 520 E. Weddell Drive,
Suite 5, Sunnyvale, CA 94089, USA (3) CEA-LETI,
17 rue des Martyrs 38054, Grenoble Cedex, FRANCE
(4) STMicroelectronics, CH-1228, Geneva,
Plan-les-Ouates, SWITZERLAND (5) MERL, 201
Broadway, Boston, USA (6) Harris, (7)Time
Domain, Hansville, Alabama (8) FT RD, 28 Chemin
des vieux chênes, BP98, 38243 Meylan Cedex Voice
(1) 358 407 076 344, (2) 408 400 0785 (3) 33 4
38 78 93 88, (4) 41 22 929 58 66 (5) 1 617 621
7500 E-Mail (1) ian_at_ee.oulu.fi, (2)
mark_at_aetherwire.com(3) laurent.ouvry_at_cea.fr, (4)
philippe.rouzet_at_st.com, (5) molisch, porlik,
zafer_at_merl.com, rrober14_at_harris.com,
vern.brethour, adrian.jennings_at_timedomain.com,
(8) patricia.martigne, benoit.miscopein,
jean.schwoerer_at_francetelecom.com Abstract
UWB proposal for 802.15.4a alt-PHY Purpose
Proposal based on UWB impulse radio for the IEEE
802.15.4a CFP Notice This document has been
prepared to assist the IEEE P802.15. 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 contributors
acknowledge and accept that this contribution
becomes the property of IEEE and may be made
publicly available by P802.15
2
List of Authors

• CWC Ian Oppermann, Alberto Rabbachin (1)
• AetherWire Mark Jamtgaard, Patrick Houghton (2)
• CEA-LETI Laurent Ouvry, Samuel Dubouloz,
  Sébastien de Rivaz, Benoit Denis, Michael
  Pelissier, Manuel Pezzin et al. (3)
• STMicroelectronics Gian Mario Maggio, Chiara
  Cattaneo, Philippe Rouzet al. (4)
• MERL Andreas F. Molisch, Philip Orlik, Zafer
  Sahinoglu (5)
• Harris Rick Roberts (6)
• Time Domain Vern Brethour, Adrian Jennings (7)
• France Telecom RD Patricia Martigne, Benoit
  Miscopein, Jean Schwoerer (8)

3
Outline

• Introduction
• Background
• Transmitted Signal
• Receiver Architectures
• Bandwidth Usage
• Optional Aspects
• System performances
• Link budget
• Framing, throughput
• Power Saving
• Ranging and Delay Estimation
• Feasibility
• Conclusions

4
Proposal Main Features

1. Impulse-radio based (pulse-shape independent)
2. Support for different receiver architectures
   (coherent/non-coherent)
3. Flexible modulation format
4. Support for multiple rates
5. Enables accurate ranging/positioning
6. Support for multiple SOP

5
Motivation for (2-4)

• Supports homogenous and heterogeneous network
  architectures
• Different classes of nodes, with different
  reliability requirements (and cost) must
  inter-work

6
UWB Technology

• Impulse-Radio (IR) based
• Very short pulses ? Reduced ISI
• Robustness against fading
• Episodic transmission (for LDR) allowing long
  sleep-mode periods and energy saving
• Low-complexity implementation

7
Modulation Features

• Simple, scalable modulation format
• Flexibility for system designer
• Modulation compatible with multiple
  coherent/non-coherent receiver schemes
• Time hopping (TH) to achieve multiple access

8
Commonalities with Other Proposals

• The present document is the result of preliminary
  merging efforts among the new CWC/STm/LETI/AetherW
  ire/MERL/Harris/FTRD/TDC group. Work is still
  ongoing for refining and consolidation of some of
  the parameters described in this proposal.
• Discussions are under way for further
  collaborations and merging in 802.15.4a.

9
Outline

• Introduction
• Background
• Transmitted Signal
• Receiver Architectures
• Bandwidth Usage
• Optional Aspects
• System performances
• Link budget
• Framing, throughput
• Power Saving
• Ranging and Delay Estimation
• Feasibility
• Conclusions

10
Definitions

• Coherent RX The phase of the received carrier
  waveform is known, and utilized for demodulation
• Differentially-coherent RX The carrier phase of
  the previous signaling interval is used as phase
  reference for demodulation
• Non-coherent RX The phase information (e.g.
  pulse polarity) is unknown at the receiver
• -operates as an energy collector
• -or as an amplitude detector

11
Pros () and cons (-) of RX architectures

• Coherent
• Sensitivity
• Use of polarity to carry data
• Optimal processing gain achievable
• - Complexity of channel estimation and RAKE
  receiver
• - Longer acquisition time
• Differential (or using Transmitted Reference)
• Gives a reference for faster channel
  estimation (coherent approach)
• No channel estimation (non-coherent approach)
• - Asymptotic loss of 3dB for transmitted
  reference (not for DPSK)
• Non-coherent
• Low complexity
• Acquisition speed
• - Sensitivity, robustness to SOP and interferers

12
Time Hopping Impulse Radio (TH-IR) - Principle
1
Tc
Tf
Ts
-1

• Each symbol represented by sequence of very
  short pulses (see also Win Scholtz 2000)
• Each user uses different sequence (Multiple
  access capability)
• Bandwidth mostly determined by pulse shape

13
Transmitted Reference (TR)

• TR schemes simplify the channel estimation
  process
• Reference waveform available for synchronisation
• Potentially more robust (than non-coherent) under
  SOP operation
• Supports both coherent/differentially-coherent
  demodulation
• Implementation challenges
• Analogue Implementing delay value,
• delay mismatch, jitter

14
Transmitted Reference
data
Td
1
Tc
Tf
reference
Ts
-1

• First pulse serves as template for estimating
  channel distortions
• Second pulse carries information
• Drawback Waste of 3dB energy on reference pulses

15
Outline

• Introduction
• Background
• Transmitted Signal
• Receiver Architectures
• Bandwidth Usage
• Optional Aspects
• System performances
• Link budget
• Framing, throughput
• Power Saving
• Ranging and Delay Estimation
• Feasibility
• Conclusions

16
Design Parameters (1)

• Motivation
• Flexible waveform
• Simple
• Compatible with multiple coherent/non-coherent
  receiver schemes
• Large Bandwidth
• () Higher transmit power
• () improved time resolution
• (-) Increased design complexity
• (-) Less stringent requirements on out of band
  interference filtering
• Signal BW of 500 MHz - 2 GHz in Upper bands
• Signal BW of 700 MHz in 0 to 960 MHz Lower band
  (low band)
• Long Pulse Repetition Period
• () more energy per pulse (easier to detect
  single pulse)
• () Lower inter-pulse interference due to channel
  delay spread
• (-) Higher peak voltage requirements at
  transmitter
• (-) Longer acquisition time
• ? Frame duration between 40ns (first realization)
  and 125ns (second realizations). Higher values
  for the frame duration have been mentioned.
  Further discussions are required to fix the values

17
Design Parameters (2)

• Simple modulation schemes
• BPPM combined with Transmitted Reference
• Channelization
• Coherent schemes Use of TH codes and polarity
  codes
• Non-coherent schemes Use of TH codes (polarity
  codes for spectrum smoothing only)
• Long TH code length
• () higher processing gain, robustness to SOP
  operation
• (-) Lower bit-rate
• (-) Longer acquisition time, shorter frame size
  (synch. phase)
• TH code length 8 or 16
• TH code binary position (delay of 0 or t? ),
  bi-phase
• For first realization, higher-order TH with
  shorter chip duration (multiples of 2ns) can be
  used. This is under discussion

18
Transmission

• Basic idea use modulation scheme that allows
  coherent, differentially coherent, and incoherent
  reception
• Combine BPPM with more sophisticated TR scheme
• Non-coherent receiver sees BPPM with pulse stream
  per bit
• More sophisticated receiver sees BPPM (1 bit)
  plus bits carried in more sophisticated
  modulation scheme (e.g. extended TR)
• Advantages
• Coherent, differential and non-coherent receiver
  may coexist
• reference can be used for synch and threshold
  estimation
• Concept can be generalized to N-ary TR system

19
Waveform Design

• Coexistence of coherent and non-coherent
  architectures
• Combine BPPM with BPSK
• Divide each symbol into two 125 ns BPPM slots
  (250 ns symbol)
• In either slot transmit a signal that can be
  received with a variety of receivers
  differentially coherent or coherent receivers.
• Non-coherent receivers just look for energy in
  the early or late slots to decode the bit.
• Other receivers understand the fine structure of
  the signal.

20
Waveform Design

• Two possible realizations
• The whole symbol (consisting of N_f frames) is
  BPPM-modulated.
• Have a 2-ary time hopping code, so that each
  frame has BPPM according to TH code

21
First Realization
22
Second Realization
Ts
2-PPM 16 chips 2-ary TH code
(coherent decoding possible)
This is a time-hopping that can be exploited by
non-coherent receiver Time hopping code is (2,2)
code of length 8 or 16 Effectively 28 or 216
codes to select for channelization for
non-coherent scheme
23
Mitigation of peak voltage through multi pulses
TfPPI
ppV peak-to-peak voltage
M 1
IS  EQUIVALENT  TO
TfPPI
M 4
ppV/2
TfPPI
M 2
ppV/sqrt(2)
24
Coexistence of Different Receiver Architectures

• Want waveform that allows TR reception without
  penalizing coherent reception
• That is achieved by special encoding and waveform
  shaping within each frame. Does not affect the
  co-existence of coherent/non-coherent receivers

25
Basic Properties

• Use of Doublets with memory from previous bit.
  (Encoding of reference pulse with previous bit)
• Agreed on 20ns separation between pulses
• Extensible to higher order TR for either reducing
  the penalty in transmitting the reference pulse
  or increasing the bit rate?
• Also allows the use of multi-DOUBLET

26
Differential Encoding of Bits
b0
b4
b2
b3
b1
b5
b-1
Tx Bits
0 0 1 1
0 0
1
Reference Polarity
-1 -1 1
1 -1
-1
1 -1 1
-1 1 -1
Ts
Note This slide is meant to describe the
encoding of data on the reference pulse and data
pulse in the basic modulation format. For
simplicity we have omitted the multipulse/multifra
mes per symbol structure.
27
Total Modulation Scheme (First Realization)
THE KEY SLIDE OF THE PROPOSAL this is the
modulation format that allows Coherent,
differentially coherent, and non-coherent
demodulation at once
28
Higher-order modulation
TH Pattern
TH Code 1,1 1,1
0,1 0,0 1,0
0,1
Data 1,1 1,1
1,1 1,1 1,1
0,0
29
Comments on Transmitted Signal

• Frame period for solution 2 is Tframe (Np D)
  t? tdelay
• ?delay is some allowance for channel delay spread
• Frame period could be dynamic modified dependant
  on
• the estimated channel delay spread or
• ability of receiver to cope with delay spread
• Symbol period is length of the TH code x Tframe
• Upper Band Nominally 250 ns x 16 4 µs
• Lower Band Nominally 500 ns x 8 4 µs
• Realistic Receiver structures exist for
  multi-pulse TR schemes (see back-up slides)

30
Outline

• Introduction
• Background
• Transmitted Signal
• Receiver Architectures
• Bandwidth Usage
• Optional Aspects
• System performances
• Link budget
• Framing, throughput
• Power Saving
• Ranging and Delay Estimation
• Feasibility
• Conclusions

31
Proposed RAKE -- Coherent Receiver
Channel Estimation
Rake Receiver Finger 1
Rake Receiver Finger 2
Sequence Detector
Demultiplexer
Convolutional Decoder
Summer
Data Sink
Rake Receiver Finger Np

• Addition of Sequence Detector Proposed
  modulation may be viewed as having memory of
  length 2
• Main component of Rake finger pulse generator
• A/D converter 3-bit, operating at symbol rate
• No adjustable delay elements required

32
Proposed Transmitted Reference Receiver
Differentially Coherent

• Addition of Matched Filter prior to delay and
  correlate operations improves output signal to
  noise ratio and reduces noise-noise cross terms

Matched Filter
Convolutional Decoder

Td
SNR of decision statistic
33
Differentially-Coherent/Non-Coherent Receiver
Architecture Basic Mode and Enhanced Mode 1
BPPM Demodulation branch
Controlled Integrator
r(t)
LNA
Dump Latch
x2
RAZ
DUMP
RAZ
Tracking Thresholds setting
Ranging branch
Block index for acquisition reference
Energy Analyzer
Leading-edge refinement search
Range info
Recyle this branch for Enhanced Data Rate Modes
34
De-spreading TH Codes
TH Sequence Matched Filter
r(t)
Bit Demodulation
LNA
Case I - Coherent TH despreading
TH Sequence Matched Filter
Bit Demodulation
b(t) soft info
r(t)
LNA
Case II Non-coherent / differential TH
despreading
35
Outline

• Introduction
• Background
• Transmitted Signal
• Receiver Architectures
• Bandwidth Usage
• Optional Aspects
• System performances
• Link budget
• Framing, throughput
• Power Saving
• Ranging and Delay Estimation
• Feasibility
• Conclusions

36
Bandwidth Usage (1/4)

• Flexible use of (multi-)bands
• Signal bandwidth may be 500 MHz to 2 GHz
• Bandwidth may change depending on application and
  regulatory environment
• Optional sub-GHz band 140 MHz to 800 MHz with a
  center frequency of 470 MHz
• Use of TH and/or polarity randomization for
  spectral smoothing
• Different bandwidth use options being considered

37
Bandwidth Usage 2GHz option (2/4)
ISM Band
ISM Band
Upper Band 3
Upper Band 1
Upper Band 2
Lower band
0.96 3.1
5.1 6.0
8.0 8.1 10.1
GHz
38
Bandwidth Usage -500 MHz Option (3/4)
ISM Band
ISM Band
Upper Bands 1 - 4
Upper Bands 5 - 12
Lower band
0.96 3.1
5.1 6.0
8.0 8.1 10.1
GHz
39
Bandwidth Usage Variable Option (4/4)
ISM Band
ISM Band
Upper Bands 1 - 4
Upper Bands 5 - 12
Lower band
0.96 3.1
5.1 6.0
8.0 8.1 10.1
GHz
40
Outline

• Introduction
• Background
• Transmitted Signal
• Receiver Architectures
• Bandwidth Usage
• Optional Aspects
• System performances
• Link budget
• Framing, throughput
• Power Saving
• Ranging and Delay Estimation
• Feasibility
• Conclusions

41
Spectral Shaping Interference Suppression
(Optional)

• Basis pulse use simple pulse shape gaussian,
  raised cosine, chaotic, etc.
• Drawbacks
• Possible loss of power compared to FCC-allowed
  power
• Strong radiation at 2.45 and 5.2 GHz

Monocycle, 5th derivative of gaussian pulse
Power spectral density of the monocycle
10log10P(f)2 dB
frequency (Hz)
42
Linear Pulse Combination

• Solution linear combination of delayed, weighted
  pulses
• Adaptive determination of weight and delay
• Number of pulses and delay range restricted
• Can adjust to interferers at different distances
• (required nulldepth) and frequencies
• Weight/delay adaptation in two-step procedure
• Initialization as solution to quadratic
  optimization problem (closed-form)
• Refinement by back-propagating neural network
• Matched filter at receiver ?good spectrum helps
  coexistence and interference suppression

43
Spectral Shaping Polarity Scrambling
Td 10 ns
Td 20 ns
W/ Polarity Scrambling
W/O Polarity Scrambling
44
Adaptive Frame Duration

• Advantage of large number of pulses per symbol
• Smaller peak-to-average ratio
• Increased possible number of SOPs
• Disadvantage
• Increased inter-frame interference
• In TR also increased interference from reference
  pulse to data pulse
• Solution adaptive frame duration
• Feed back delay spread and interference to
  transmitter
• Depending on those parameters, TX chooses frame
  duration

45
Outline

• Introduction
• Background
• Transmitted Signal
• Receiver Architectures
• Bandwidth Usage
• Optional Aspects
• System performances
• Link budget
• Framing, throughput
• Power Saving
• Ranging and Delay Estimation
• Feasibility
• Conclusions

46
PER in 15.4a Channel Model Non-Coherent (Energy
Collection) BPPM
Framing format

• Simulations over 1000 channel responses
• BW 2GHz Integration Time 80ns
• Implementation loss Noise figure margin 11
  dB
• Max range is determined from
• Required Eb/N0,
• Implementation margin
• Path loss characteristics

X1 (CM8) X2 (CM1) X3 (CM5) X4 (CM9)
Required Eb/N0 19.5 dB 20 dB 21 dB 21.5 dB
Max Range (I) 10.78 m 84.61 m 86.72 m 58.67 m
Max Range (II) 7.33 m 53.25 m 54.15 m 34.72 m
Case I 250 kbps PRP 250 ns with 16
pre-integrations 4 µs
Case II 250 kbps PRP 500 ns with 8
post-integrations
47
PER/BER in 15.4a Channel Model DBPSK (RAKE)
Theoretical BER Curves Integration Time 50 ns
Implementation loss and Noise figure margin 11
dB
X1 (CM8) X2 (CM1) X3 (CM5) X4 (CM9)
Required Eb/N0 18 dB 17.5 dB 18.5 dB 18.5 dB
Max Range (I) 12.66 m 116.70 m 120.27 m 90.84 m
Max Range (II) 9.18 m 79.34 m 81.23 m 58.67 m
Case I 250 kbps PRP 250 ns with 16
pre-integration 4 µs
Case II 250 kbps PRP 500 ns with 8
post-integrations
48
Outline

• Introduction
• Background
• Transmitted Signal
• Receiver Architectures
• Bandwidth Usage
• Optional Aspects
• System performances
• Link budget
• Framing, throughput
• Power Saving
• Ranging and Delay Estimation
• Feasibility
• Conclusions

49
Link Budget Non-Coherent (Energy Collection)
BPPM
Parameter Mandatory Value (PRP 4 µs) Optional Value(PRP 500ns - 8 integrations)
Peak Payload bit rate (Rb) 250 kb/s 250 kb/s
Average Tx Power Gain (PT) -10.64 dBm -10.64 dBm
Tx antenna gain (GT) 0 dBi 0 dBi
fc (geometric frequency) 3.873 GHz 3.873 GHz
Path Loss _at_ 1m L1 20log10(4.?.fc / c) 44.20 dB 44.20 dB
Path Loss _at_ d m L2 20log10(d) 29.54 dB _at_ d 30 m 29.54 dB _at_ d 30 m
Rx Antenna Gain (GR) 0 dBi 0 dBi
Rx Power (PR PT GT GR L1 L2) -84.38 dBm -84.38 dBm
Average noise power per bit N -174 10log10(Rb) -120.02 dBm -120.02 dBm
Rx noise figure (NF) 7 dB 7 dB
Average noise power per bit (PN N NF) -113.02 dBm -113.02 dBm
Minimum Eb/N0 (S) 14 dB 17.6 dB
Implementation Loss (I) 5 dB 5 dB
Link Margin (M PR - PN S I) 9.64 dB 6.04 dB
Proposed Min. Rx Sensitivity Level -94.02 dBm -90.42 dBm
50
Link Budget DBPSK (RAKE)
Parameter Mandatory Value (PRP 4 µs) Optional Value(PRP 500ns - 8 integrations)
Peak Payload bit rate (Rb) 250 kb/s 250 kb/s
Average Tx Power Gain (PT) -10.64 dBm -10.64 dBm
Tx antenna gain (GT) 0 dBi 0 dBi
fc (geometric frequency) 3.873 GHz 3.873 GHz
Path Loss _at_ 1m L1 20log10(4.?.fc / c) 44.20 dB 44.20 dB
Path Loss _at_ d m L2 20log10(d) 29.54 dB _at_ d 30 m 29.54 dB _at_ d 30 m
Rx Antenna Gain (GR) 0 dBi 0 dBi
Rx Power (PR PT GT GR L1 L2) -84.38 dBm -84.38 dBm
Average noise power per bit N -174 10log10(Rb) -120.02 dBm -120.02 dBm
Rx noise figure (NF) 7 dB 7 dB
Average noise power per bit (PN N NF) -113.02 dBm -113.02 dBm
Minimum Eb/N0 (S) 13 dB 16 dB
Implementation Loss (I) 5 dB 5 dB
Link Margin (M PR - PN S I) 10.64 dB 7.64 dB
Proposed Min. Rx Sensitivity Level -95.02 dBm -92.02 dBm
51
Outline

• Introduction
• Background
• Transmitted Signal
• Receiver Architectures
• Bandwidth Usage
• Optional Aspects
• System performances
• Link budget
• Framing, throughput
• Power Saving
• Ranging and Delay Estimation
• Feasibility
• Conclusions

52
Framing 802.15.4 Compatible
53
Throughput
Data Frame (32 octet PSDU)
ACK Frame (5 octet PSDU)
Tdata
Tack
T_ACK
IFS

• Numerical example (high-band)
• Preamble SFD PHR 6 octets
• Tdata 1.216 ms
• T_ACK 50 ms (turn around time requested by IEEE
  802.15.4 is 192ms)
• Tack 0.352 ms
• IFS 100µs
• Throughput 32 octets/1.718 ms 149 kb/s
• Average data-rate at receiver PHY-SAP 250 kb/s
  (Basic Mode)

54
Outline

• Introduction
• Background
• Transmitted Signal
• Receiver Architectures
• Bandwidth Usage
• Optional Aspects
• System performances
• Link budget
• Framing, throughput
• Power Saving
• Ranging and Delay Estimation
• Feasibility
• Conclusions

55
Saving Power

• Power Saving techniques achieved by combining
  advantages offered at 3 levels
• Technology (best if CMOS)
• Architecture (flexible schemes provided by the
  THpulse modulation)
• System level (framing, protocol usage)
• Selected techniques used in one existing
  realization (see proof of concept slides)
• Low-duty cycle Episodic transmission/reception
• Scheduled wake-up
• 80ms RTOS tick
• Ad-hoc networking using multi-hop
• Special rapid acquisition codes / algorithm
• Matchmaking further reduces acquisition time
• Multi-stage time-of-day clock
• Synchronous counter / current mode logic for
  highest speed stages
• Ripple counter / static CMOS for lowest speed
  stages
• Compute-intensive correlation done in hardware

56
Outline

• Introduction
• Background
• Transmitted Signal
• Receiver Architectures
• Bandwidth Usage
• Optional Aspects
• System performances
• Link budget
• Framing, throughput
• Power Saving
• Ranging and Delay Estimation
• Feasibility
• Conclusions

57
Ranging

• Motivation
• Benefit from high time resolution (thanks to
  signal bandwidth)
• Theoretically 2GHz provides less than 20cm
  resolution
• Practically Impairments, low cost/complexity
  devices should support 50cm accuracy with simple
  detection strategies (better with high resolution
  techniques)
• Approach
• Use Two Way Ranging between 2 devices with no
  network constraint (preferred) no need for time
  synchronization among nodes
• Use One Way Ranging and TDOA under some network
  constraints (if supported)

58
TOA Delay Estimation - Non-Coherent

• Use bank of integrators to determine coarse
  synchronisation uncertainty region
• Symbol synchronisation uncertainty region given
  by coarse synchronisation ( e.g., 4ns-20ns)
• A refinement search is applied onto the
  uncertainty region by either
• further dividing it into narrower non overlapping
  regions for non-coherent refinement (e.g., 1ns gt
  4ns) or
• Coherent search with a template correlation

Integrator outputs
Detects the coarse uncertainty region
Energy Analyzer
Performed within the selected uncertainty region
Leading Edge Search Refinement
TRB the length of uncertainty region
Range info
59
TOA Delay Estimation - Non-Coherent (contd)

• The algorithm selects the maximum value
  integration window index and then it searches
  backward to find the first integration value
  which crosses an adaptively set threshold.
• If there are no values crossing the threshold,
  the peak position is used for the TOA estimation.

MES-SB based TOA Estimate
Searchback window
Strongest Path, energy block
Threshold based TOA Estimate
Threshold
N
0
1
2
Actual TOA
Contains leading edge
MES Maximum Energy Search TC Threshold
comparison SB Search Back
MES based TOA Estimate
60
ENERGY Spread in CM1

• PDF of TOA estimation errors are illustrated for
  MES at various EbN0
• CM1, integration interval 4ns, Tf200ns (results
  will be updated for Tf240ns)

61
Ranging Simulation Settings
Notations and Terms Definition Value in Simulations
Tf Pulse repetition interval, frame 200ns
Nb Number of blocks within a Tf 50
Nc Number of refinement intervals within a Tf 400
TH Time hopping sequence in chips h1, ...., h5
POL Polarity codes p1, ..., p5
N1 Number of frames in the 1st-step 50
N2 Number of frames in the refinement 30
BW Bandwidth 2GHz
C Number of correlators (refinement stage) 10
Note Results are to be provided when Tf is set
to 240ns.
62
Ranging Results

• IEEE 802.15.4a CM1-Residential LOS

True Distance (m) One-way ranging error (confidence level)
25m 8cm (97)
30m 8cm (90)

• Round Trip ranging error
• (with no drift compensation)
• 16cm (0.088ms), no clock drift
• 17.1cm (1ppm)
• 20.1cm (4ppm)
• 26cm (10ppm)
• 56cm (40ppm)

63
Two Way Ranging (TWR)
Is the frequency offset relative to the nominal
ideal frequency

• Range estimation is affected by
• Relative clock drift between A and B
• Prescribed response delay
• Clock accuracy in A and B
• Channel response (weak direct path)

250 kbps, 38 bytes PPDU 250 kbps, 38 bytes PPDU 500 kbps, 9 bytes PPDU 500 kbps, 9 bytes PPDU
Df/f \ Treply (max error) 1408 ms 1226 ms 336 ms 154 ms
4 ppm 1.69 m 1.47 m 0.40 m 0.18 m
25 ppm 10.56 m 9.19 m 2.52 m 1.15 m
40 ppm 16.9 m 14.7 m 4.0 m 1.8 m
Example using Imm-ACK SIFS of 15.4 and 15.3 of
respectively 192us and 10 us and PPDU size of
respectively 38 and 9 bytes

• Simple immediate TWR made unusable with
  reasonnable crystal accuracies. Solution is
• Performing fine drift estimation/compensation
• Benefiting from cooperative transactions
  (estimated clock ratios )

64
Outline

• Introduction
• Background
• Transmitted Signal
• Receiver Architectures
• Bandwidth Usage
• Optional Aspects
• System performances
• Link budget
• Framing, throughput
• Power Saving
• Ranging and Delay Estimation
• Feasibility
• Conclusions

65
Antenna Feasibility Capacitive Dipole and
Various Bowtie Antennas
Bowtie antenna
66
Proof-of-Concept (1) Non-coherent Transceiver
Non-coherent, Energy Collection Receiver
5 Mbps BPPM 350 ps pulse train with long
scrambling code
67
Proof-of-Concept (2) Non-coherent Transceiver
UWB-IR BPPM Non-Coherent Transceiver
Implementation
UWB Transmitter 400 µm x 400 µm 0.35 µm CMOS
UWB Transceiver Test architecture lt10 mm2 0.35 µm
SiGe Bi-CMOS
68
Proof-of-Concept (3) Transmitter - Lower Band
UWB Transmitter chip for generating impulse
doublets
69
Proof-of-Concept (4) Receiver - Lower Band
Coherent UWB Receiver with multiple time
integrating correlators
70
Proof-of-Concept (5) High Speed Coherent
Circuit Elements
RF front end chips in CMOS 0.13mm, 1.2V
20 GHz digitizer for UWB
20 GHz DLL for UWB
3-5 GHz LNA Chip and layout
71
Outline

• Introduction
• Background
• Transmitted Signal
• Receiver Architectures
• Bandwidth Usage
• Optional Aspects
• System performances
• Link budget
• Framing, throughput
• Power Saving
• Ranging and Delay Estimation
• Feasibility
• Conclusions

72
Conclusions

• Proposal based upon UWB impulse radio
• High time resolution suitable for precise ranging
  using TOA
• Modulation
• Pulse-shape independent
• Robust under SOP operation
• Facilitates synchronization/tracking
• Supports multiple coherent/non-coherent RX
  architectures
• System tradeoffs
• Modulation optimized for several aspects
  (requirements, performances, flexibility,
  technology)
• Trade-off complexity/performance RX
• Flexible implementation of the receiver
• Coherent, differential, non-coherent (energy
  collection)
• Analogue, digital
• Fits with multiple technologies
• Easy implementation in CMOS
• Very low power solution (technology,
  architecture, system level)

73
Backup Slides
74
BER Performance in AWGN Channel
MRC Solution (coherent)
Differential Solution
Energy Collection solution in OOK
Transmitted Reference (one pulse)
-3 dB the reference is not in the same PRP !
PER 1 with 32 bytes PSDU ? acceptable BER
4x10-5 with no channel coding
75
BER Performance in AWGN Channel
76
Antenna Practicality

• Bandwidth 3 GHz-10 GHz
• Form factor
• Omni-directional

77
Positioning from TDOA
3 anchors with known positions (at least) are
required to find a 2D-position from a couple of
TDOAs
Specific Positioning Algorithms
78
TR BPPM Scheme Comparison
79
Assumptions and Notes

• Results are theoretical calculations
• Assumes ideal impulse UWB pulses in AWGN
  channel
• Different TR-BBP options are considered with
  different number of pulses per pulse train
• Multipath fading simulations can be performed to
  back up theory

80
Pulse repetition structures TR BPPM with
doublets (Scheme 1)
81
Pulse repetition structuresTR BPPM single
reference (Scheme 2)
82
Pulse repetition structuresAuto Correlation
BPPM with doublets (Scheme 3)
83

Pulse repetition structuresAuto Correlation
BPPM single reference (Scheme 4)
84
Pulse repetition structuresAuto Correlation
BPPM alternate (Scheme 5)
85
Parameters

• PPI slot - slot inside each TH chip containing a
  burst of pulses including reference pulses (ref.
  slides from Laurent / CEA)
• Np represents the number of pulses in each PPI
  slot
• The energy E per PPI slot is kept constant
• The pulse energy Ep E/Np
• TW represent the time-bandwidth product

86
Ep/N0 degradation versus number of pulses per
pulse train
87
Ep/N0 degradation versus Time/Bandwidth product
88
Ep/N0 degradation versus number of pulses per
pulse train
89
Conclusions

• Scheme 5 - AC Alternate performs better then
  all the other pulse repetition structures.
• AC generally performs better than TR
• AC alternate and AC with doublets have the
  advantage of requiring only a single delay line.
• Scheme 5 - AC Alternate, was proposed at
  Monterey meeting in January.
• Criticism was given based on accumulated noise
  in noise-cross-noise-cross-noise... Products.
• Seems to outperform other schemes with simple
  analysis
• Also more readily implementable since fixed delay
  line can be used.

90
Channel / Delay Estimation Coherent Approach
91
Channel / Delay Estimation Coherent Approach

• Swept delay correlator
• Principle estimating only one channel sample per
  symbol.
• Similar concept as STDCC channel sounder of Cox
  (1973).
• Sampler, AD converter operating at SYMBOL rate
  (1.2 Msamples/s)
• Requires longer training sequence
• Two-step procedure for estimating coefficients
• With lower accuracy estimate at which taps
  energy is significant
• With higher accuracy determine tap weights
• Silence periods for estimation of interference

92
Optimal Energy-Threshold Analysis (CM-1)

• Optimal normalized threshold (normalized with
  respect to the difference between the maximum and
  minimum energy blocks) changes with Eb/N0 and
  block size.
• Smaller thresholds are required in general at
  high Eb/N0, while larger thresholds at lower SNR
  values

MAE Mean Absolute Error in detecting leading
energy block with simple threshold crossing (1000
channel realizations)
Eb/N0 8 --- 26dB