IEEE 802'15 TG3a Philips CFP Presentation

1
Project IEEE P802.15 Working Group for Wireless
Personal Area Networks (WPANs) Submission Title
Philips TG3a CFP Presentation Date Submitted
02 March, 2003 Source Charles Razzell,
Dagnachew Birru, Bill Redman-White, Stuart Kerry
Company Philips Address 1109 McKay Drive, San
Jose, CA 95131, California, USA Voice(408)
474-7243, FAX [phone],
E-Mailcharles.razzell_at_philips.com Re IEEE
P802.15-02/372r8 IEEE P802.15 Alternate PHY Call
For Proposals dated 17 January,
2003 Abstract This presentation gives an
overview of the Philips proposal for an
alternative physical layer for IEEE P802.15.3a
based on a multi-band UWB approach. Purpose Phi
lips requests that the task group considers the
merits of the following physical layer proposal
and evaluates the content in conjunction with
other responses to the Call for
Proposals. 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 contributor
acknowledges and accepts that this contribution
becomes the property of IEEE and may be made
publicly available by P802.15.
2
Philips TG3a CFP Presentation

• An alternate high-rate PHY for Wireless Personal
  Area Networks

3
Outline of Presentation

• Summary of Proposal
• Review of Spectral Keying Modulation
• Example choice of parameters to obtain required
  data rates
• Advantages of low pulse repetition rate
• FEC approach
• Pico-net isolation techniques
• Transmitter power control
• Receiver Implementation Issues
• Parallel/Serial dimensioning of receiver
  structure
• Scalability for low cost
• Conclusions

4
Summary of Proposal

• Scalable data rates to 480 Mbps and beyond
• Spectral KeyingTM modulation
• Compliant with FCC 02-48, UWB Report Order
• Multiband system, scalable from 4-10 bands,
  occupying 2 - 6 GHz bandwidth
• Supports 4 co-located piconets

Modulation scheme originally proposed by General
Atomics who own this trademark
5
Multi-Band Modulation(1) pulsed OFDM
f6
f5
f4
Information capacity proportional to number of
sub-carriers.
f3
f2
f1
Peak/mean ratio too high
6
Multi-band Modulation (2) staggered pulses with
fixed order
f6
f5
Fixed order information capacity proportional to
number of sub-carriers
f4
f3
f2
f1
Peak/mean ratio is much reduced!
7
Spectral Keying(1) Introduce Sequence Keying
Use sequence as information bearing
parameter. Information capacity is proportional
to sub-carriers log2(n!)
8
Spectral Keying(2) Sequence Keying in addition
to PSK
45
Sequence
QPSK
Use of sequence bits approx. doubles number of
bits per pulse of QPSK system when 8 sub-carriers
are used.
Total
40
35
30
25
of bits
20
15
10
5
0
1
2
3
4
5
6
7
8
9
10
of frequencies
9
Coding gain of Spectral Keying (n5)
Spectral Keying vs. BPSK
0
10
SK(n5)
BPSK
-2
10
-4
10
BER
Eb/No dB
10
Summary of Spectral Keying

• Use of sequence to carry information
  significantly increases information per pulse
• This allows the interval between successive
  pulses to be increased.
• This extra guard time can be used for ISI
  reduction and/or possible insertion of
  time-interleaved piconets
• The modulation remains robust, even with 4 bits
  per symbol on each frequency band.

11
Example Parameters for Required PHY Data Rates
12
Example Parameters Continued
13
Advantages of low pulse repetition rate
14
Advantages of Low Pulse Repetition Rate
High ISI zone
CM2 represents NLOS 0-4m CM3 represents NLOS 4-10m
Next impulse response starts here (11MHz)
Next impulse response starts here (22MHz)
How much spreading gain would be required to
achieve similar ISI reduction?
15
Forward Error Correction Approach(based on
results from IMECs T_at_MPO core)

• Parallel Concatenated Convolutional Turbo Codes
• 8-state Recursive Systematic Convolutional Codes
• Collision-free interleaving patterns1
• Parallel SISO units process sub-frames
• Early stop criterion minimizes energy usage
• Decoding latency of 5ms per block

1 A. Gulietti et al. Parallel Turbo Code
Interleavers Avoiding collisions in access to
storage elements. Electronics Letters vol. 38 No
5.
16
Channelization for Pico-nets

• Propose to use a combination of frequency
  interleaving and time interleaving

Time division into even and odd time slots
?2
17
Concept for Time Division of Pulse Interval
Power decay profile from 2 interleaved pico-nets
(CM2)
Net 1
0
Net 2
-5
Two 110Mbps piconets are time interleaved with
11MHz pulse rate each.
-10
average power (dB)
-15
-20
-25
0
10
20
30
40
50
60
70
80
90
delay (ns)
Passive scanning (probe)
18
Evaluation of Candidate Sub-pulse Slots by
Measuring Preamble Quality
Phase bits1
Phase bits0
Preamble may be designed to allow MUI to be
estimated in all four quadrants of a pulse
interval. The best slot is chosen for
transmission of the immediately following frame.
19
Closed loop power control

• An additional use of the pre-amble quality
  feedback mechanism is to allow for closed power
  control
• Preamble acknowledgement word could include 2
  power control and 2 sub-slot selection bits.
• Closed loop power control is considered essential
  for dense deployment scenarios (e.g. multiple
  laptops in a conference room)
• Additional quality information can be gathered
  during the data-bearing part of the packet to
  assist with power control and sub-slot selection.

20
Rx - High Performance Implementation
Quadrature LOs may be low-spec on-chip
oscillators with low area and power. LOs Shared
with Tx Path
Common front-end can reduce power with SK duty
cycle
3-4 bits ADC helped by AGC per band
21
Rx - Low Power/Cost Implementation
ADC Rake Rx replaced by analog detector with AGC
Common front-end can reduce power with SK duty
cycle
Saving of 60mW power from ADCs Rake Rx
22
Tx Implementation
Low spec LO shared with RX channels
23
Consideration of Serial vs. Parallel Receiver
Structures

• Since the sequence of frequencies is unknown and
  cant be anticipated, parallel frequency-selective
  branches are needed
• However, most of the duplicated circuits are very
  small on chip (oscillators, mixers).
• Receiver branches may need to sample a
  significant time window
• to allow for different propagation delays in
  different sub-bands under non-LOS condititions
• to allow for collection of multipath echoes for
  RAKE combining

24
Consideration of Serial vs. Parallel Receiver
Structures (cont.)

• Entirely serial reception of the sequence of
  tones is likely to cause performance loss in
  multipath conditions due to sparse sampling of
  the energy in each sub-band.
• Entirely parallel energy collection in the
  different sub-bands need not be the only
  alternative
• Partial serialization of the Spectral Keying
  modulation can be used to realize an excellent
  compromise

25
Partial Serialization of Spectral Keying
f6
f5
f4
18ns
18ns
26
Partial Serialization of Spectral Keying

• No information loss w.r.t. fully parallel
  Spectral Keying
• Same ML decoding algorithm can be applied
• The number of receiver branches can be halved
• Further degrees of serialization can be
  considered

27
Partial Serialization is Compatible with
Time-division of Pico-nets
f6
f3
Net-2
f5
f2
f4
f1
f3
f6
Net-1
f2
f5
f1
f4
18ns
18ns
28
Comments on Scalability

• RAKE combining need not be used.
• Low cost receiver implementation possible using
  analog correlator with single-bit sampling.
• Full serialization may be used when the number of
  sub-bands is low and the pulse repetition
  interval is sufficiently long (for low cost and
  low data rate applications).

29
Conclusions

• Spectral Keying provides a high order modulation
  scheme with inherent robustness (coding gain
  w.r.t. BPSK).
• The increased number of bits per pulse allows us
  to use a low pulse repetition rate which reduces
  ISI.
• In addition to lower ISI, we obtain energy
  conservation related to Spectral Keyinglow duty
  cycle.
• Pico-net isolation can be achieved in the
  frequency and time dimensions (2 channels in each
  dimension).
• Transmitter power control is strongly recommended
  to enable dense deployment scenarios
• Different degrees of parallel/serial tradeoff can
  be implemented to divide the number of receiver
  branches by 2 or more.
• Maintaining some degree of parallelism leads to
  more optimum reception in multipath conditions.
• Can use high performance digital receiver, or low
  cost/power version.