France Telecom – Jean Schwoerer, Benoit Miscopein, Stephane Mebaley-Ekome (1)
CEA-LETI – Laurent Ouvry, Raffaele D’Errico, François Dehmas, Mickael Maman, Benoit Denis, Manuel Pezzin (2)
THALES – Arnaud Tonnerre (3)
University of Surrey, UK – Ehsan Z. Hamadani
INSA-Lyon, FR – Jean-Marie Gorce
Outline
Introduction
PHY proposal
Advantages of UWB
Band Plan and PLL reference diagram
Pulse Repetition Frequency, Preamble, Modulation
Variable bit rate & throughput
Link Budget, Performances
Feasibility examples
MAC elements proposal
Beacon-based mode: Evolution of IEEE 802.15.4 MAC
Beacon-disabled mode: Preamble sampling approach
Upper layers responsibility
Conclusions & References
PHY Proposal
Introduction
Proposal main features:
Based on IEEE802.15.4a-2007 where a very reduced set of mode is selected for the BAN context
UWB Impulse-radio based
Support for different receiver architectures (coherent/non-coherent)
Flexible modulation format
Support for multiple rates
Support for multiple SOP
Low radiated power
Low radiated power
Low PSD, low interference, low SAR
High co-existence with existing 802.x standards
Real potential for low power consumption
Large bandwidth worldwide
Spectrum is worldwide available
Robust to multipath and fast varying channels
Flexible, scalable (e.g. data rates, users)
Low complexity HW/SW solutions in advanced development (eg 802.15.4a)
Band plan (selection from 15.4a)
Band plan versus regulation for UWB
PLL Reference Diagram
Pulse Repetition Frequency
Selected value
15.6 MHz PRF (64.10 ns of pulse repetition period PRP)
Use of binary codes, No bursts => mean PRF = peak PRF
Rationale
Typical channel delay spread for indoor applications below 25 ns in 90% of channel realizations => very limited ISI (or IPI)
Integer relationship with base frequency of the bandplan (499.2 MHz / 32) => no fractional PLL
Maximized Pulse amplitude
=> better for transmitter power consumption (between-pulses duty cycling)
=> better for pulse detectability in the receiver
=> better for threshold crossing receivers
=> compatible with low voltage CMOS technologies without I/O « tricks » (around 780 mVp-p)
A single value => simplicity
Compatible with bit rate scalability with short spreading factors
No high speed clock / long FIRs filters to generate or correlate with bursts of pulses
Preamble
PRF is the same as one of the mean PRF in 15.4a
Sp code duration:
Example with N=31 : Tpsym = 31 * PRP = 1.9871 us
To be checked if enough for a correct (Pd,Pfa) versus complexity
Number of Sp codes in the preamble
Example with Nsp = 16 : Tsync = 16 * Tpsym = 31.7936 us
This is probably a maximum value
To be checked if enough for a correct (Pd,Pfa) versus overhead
Modulation
PRF is ~ the same as one of the mean PRF in 15.4a
Spreading code length Sd
Example with N=7 (BARKER CODE) : Symbol duration Ts = 7 * PRP = 0.4488 us
Modulation : 1 bit per symbol + second bit used for redundancy OR non coherent demod
DBPSK : BPSK with differential encoding (at symbol level)
Sub-optimal but easier to implement and less sensitive to clock drift
DBPSK + PPM : the whole S code can be shifted within the PRP with ½ the PRP value
Does not affect the mean PRF value and the spectrum shape, Is simple to implement (though a little more complex than pure DPBSK), Is compatible with non coherent / threshold crossing detectors, Is ISI compatible in the BAN context
DBPSK + chip-PPM : each chip of the code is shifted according to the chip value => selected option
Bit rate is adjusted with the number of pulses per symbol keeping a constant mean PRF
2.22 Mbit/s is the default uncoded data rate
5.2 and 15.6 Mbit/s are mandatory => No additional complexity & allows to reduce channel use
31.2 Mbit/s is proposed optionally for coherent receiver (uses both PPM & DPBSK)
Proposed FEC : systematic RS (63,55) as in 15.4a (maximum efficiency ~0.87)
The lowest rate is a compromise between “Tx_on time” and range (+clock drift compensation requirements).
Link budget
Performances analysis methodology and channel models
Goal:
Get/discuss performances at a link budget / outage probability level with the different channel models at the default 2.2 Mbps rate before digging into the design level performances
Methodology:
Perform extra measurements at CEA-Leti (for UWB 3-5GHz but also for 2.4 GHz)
Complement the IEEE802.15.6 CM3 UWB channel model with extra measurements and models and compare channel models with each other
Move towards a scenario based approach
Scenario = [at least] given (Tx,Rx) couple + given generic environment + given generic movement
Justified by the huge dispersion of the BAN radio channel
Calculate outage probabilities given the path loss and shadowing statistics
Performances: channel path loss models
Available CM3 UWB channel models as in TG6 document :
A (source NICT)
Indoor and anechoic
B (source IMEC)
Anechoic
C (source Samsung)
Indoor and anechoic
Conclusion
Huge dispersion (tens of dBs) between models
Distance is not a relevant parameter to get a path loss model
Coming back to the scenario based channel characterization is proposed
2-5 GHz, indoor and anechoic, 7 subjects, standing/walking/running, scenarios as depicted below
Log normal path loss model. Shadowing and small scale fading modeled separately.
Reference : D'Errico R.; Ouvry L.,“Time-variant BAN channel characterization” TD(09)879, COST2100, 18-19/05/2009, Valencia, Spain (Measurement set up details available on request)
Performances: extra channel measurements
On previous page:
Error bar @ 1 std of mean channel gain over subjects
Without the slow shadowing std
2.4 GHz and 3-5 GHz for comparison
On this page
UWB 3-5 GHz only
Error bar @ 2 stds of mean channel gain and slow shadowing (95% confidence interval)
Conclusions
2.5% outage probability @ ~-70dB channel loss for the 8 scenarios in indoor conditions
Higher outage in anechoic chamber depending on the scenario (not shown here)
Performances : outage probability
Starting from:
The link budget
PTx + antenna = -20.5 dBm
Sensitivity = -90.5 dBm
Includes 6dB NF, 5dB I.L. and 9dB min required EbN0
70dB total link margin
The different scenario based path loss models
CM3 UWB A back to the scenarios
CM3 UWB C back to the scenarios
CEA-Leti’s measurements
Get the outage probability performance for an EbN0
Probability that the received power is higher than the receiver sensitivity
from the proposed -90.5dBm sensibility (2.2 Mbps)
through to -87.8dBm (5.2 Mbps)
and to -82dBm (15.6 Mbps)
Performances : outage probability
Performances : outage probability
Performances : scenario based outage probability
Tentative consolidation of
CM3 A (NICT)
CM3 C (Samsung)
CEA-Leti’s measurements
for four indoor scenarios :
Hip-left ear
Hip-right wrist
Hip-thigh
Hip-chest
(others available as well, including anechoic chamber cases)
Needs further update to refine comparisons, but aims at opening discussions
Performance : EbN0 requirements (DBPSK)
Conditions: DBPSK, 20 bytes PSDU, CM3 A channel
Same results (with better PER floor in highest rate) for 256 bytes
Conclusions on link performances on the different channel
The proposed link budget and system specification makes the UWB proposal feasible for most of the scenarios
outage of 5% as in TRD
1e-2 PER for 20 bytes PSDU, or
1e-1 PER for 256 bytes PSDU
Actual EbN0 requirement still to refine (current simulation with realistic receiver on CM3 gives 13dB after RS decoding, within the proposed IL values, but the CM3 multipath model is questionable)
However, large variations between the different models (CM3 A is optimistic, CM3 C is pessimistic, CM3 B and CEA-Leti’s measurement are median)
Further analysis in the 7.25-8.5 GHz band is necessary
Background design know-how (see references)
Conclusions
Proposal based upon UWB impulse radio alt-PHY of 15.4a
Advantage
Early implementations exist: experienced proposal
Selection of the most relevant modes and their adaptation to the BAN context (note: 15.4a mandatory mode is NOT the selected option for 15.6)
A standard exists which will speed up the 15.6 standard drafting steps
Modulation:
DBPSK provide robustness for a limited complexity & 2PPM allow several receiver implementations
RS FEC help to improve link budgets and parity bit will improve robustness of the DBPSK receiver
System tradeoffs
Variable bit rates allow to accommodate all applications envisaged in TG6
Minimizing talk time improve energy consumption, SOP performances, and regulatory compliance
Flexible implementation of the receiver
Compatible with a lot of UWB detectors (coherent, differential, energy, threshold crossing)
FEC decoder is optional
Fits with multiple technologies
Compatible with implementation in low voltage CMOS
Very low power integrated solutions already proven (thus to be adapted)
The very low transmit power is a very attractive feature for the UWB PHY adoption
MAC elements proposal
MAC layer
BANs may be coordinated most of the time
The coordinator can allocate a negotiated bandwidth to QoS demanding nodes (data rate, BER, latency)
A beacon based approach is adapted
Some applications imply a lower channel load
A beacon-free mode is more efficient
An IEEE 802.15.4-like MAC layer is a good base for BANs
With a beacon-enabled mode including TDMA and Slotted-ALOHA (for UWB) with relaying
With an enhanced beacon-free mode by using Preamble Sampling
Beacon-based MAC mode
BAN requires above all:
Relaying capability: to cope with low-power emission and severe NLOS conditions
Would be limited to 2-hops in the BAN context
Reliability: high level of QoS for critical / vital traffic flows
Proposed architecture
Mesh network centralized on the gateway
Full mesh topology based on a scheduling tree
Guaranteed access for management and data messages (real TDMA)
Beacon-based MAC mode
Superframe structure
Based on 802.15.4
Inclusion of a control portion for management messages
The control portion shall be large enough to allow dynamic changes of topology
The CAP is minimized, mostly used for association using slotted ALOHA
Beacon-based MAC mode
Fine structure
Superframe is divided equally into slots
Use of Minislots in the Control portion
Provides flexibility: adaptation to different frame durations
Guaranteed Time MiniSlot (GTMS) shall be introduced in CFP
Beacon-based MAC mode
Control portion structure
Beacon period
The beacons are relayed along the Scheduling Tree
The beacon-frame length shall be minimized
Beacon alignment procedure shall be used
Request period
Request period is a set of GTS dedicated to allocation demands
Transmission from the leaves to the coordinator
Topology management period
Hello frames, for advanced link state procedure
Scheduling tree based update
Beacon-free MAC mode
However, some situations might not require a beacon-enabled MAC protocol (see references.)
Symmetric or asymmetric network, low communication rate and small packets
Network set-up, coordinator disappearance, etc
In such conditions, the downlink is an issue : nodes must be powered-up to receive data from the coordinator (e.g. polling/by-invitation MAC scheme)
We propose a Preamble Sampling MAC protocol for UWB
Beacon-free MAC mode
Preamble Sampling principles:
Nodes periodically listen to the channel. If it is clear, they go back to sleep; conversly, they keep on listening until data
Nodes are not synchronized but share the wake up period (TCI)
A packet is transmitted with a preamble as long as TCI
Beacon-free mode
Preamble sampling protocols are known to be the most energy efficient uncoordinated MAC protocols
TCI depends on the traffic load and the TRX consumption (TX/RX/listen modes)
Can be of the order of 100 ms
The preamble is a network specific sequence
Either a typical wake-up signal
Or a preamble with modulated fields (e.g. @, time left to data)
Beacon-free MAC mode
To comply with LDC regulation in the lower band (e.g. Europe), the preamble can be relayed by the BAN nodes (up to 5ms long bursts)
Beacon-free MAC mode
Mode 0: only one burst per device is emitted
Mode 1: burst is re-emitted by a device, under the LDC limit, until a neighbour relays
Mode 2: same as mode 1 + former relays, with emission credits left, can relay again
Upper Layers responsibility
Enabling/disabling beacons switching procedure
From beacon-free to beacon-based mode
BAN formation (coordinator election)
New coordinator election if former coordinator leaves
Can be triggered by a user requesting a link with high QoS
From beacon-based to beacon-free mode
Fallback mode if the coordinator leaves
If the required BW (rate of GTS requests) is below a given threshold and requested QoS is adequate
Conclusions on MAC
Propose a combination of
A Beacon-based true TDMA mode including fast relaying and mesh support
A Beacon-free mode using preamble sampling and extra features adapted to UWB
M. Pezzin, D. Lachartre, « A Fully Integrated LDR IR-UWB CMOS Transceiver Based on "1.5-bit" Direct Sampling », ICUWB 2007, Singapore, September 2007
D.Lachartre, B. Denis, D. Morche, L. Ouvry, M. Pezzin, B.Piaget, J. Prouvée, P. Vincent, « A 1.1nJ/b 802.15.4a-Compliant Fully Integrated UWB Transceiver in 0.13μm CMOS», ISSCC 2009, San Francisco, February 2009
European ICT PULSERS II and ICT EUWB projects deliverables
Timmons, N.F. Scanlon, W.G. "Analysis of the performance of IEEE 802.15.4 for medical sensor body area networking", IEEE SECON 2004, Santa Clara, Ca, October 2004
Questions ?
Back-up slides
Beacon-free MAC mode
In this collaborative scheme, the preamble burst can composed of
Destination adress (compuls.)
Time left to data (compuls.)
Packet length
Source adress
Need for a relaying collision resolution policy
Based on the wake-up instant during the burst (priority is given to the node which has detected the burst first)
Beacon-free MAC mode
Relationship between wake-up instant in the burst and back-off length can be of any kind (linear, logarithmic, exponential…)
Beacon-free MAC mode
Collaborative scheme is very flexible
Nodes can relay once or up to the emission maximum duration (50 ms in Europe)
If the source listens to the relays, it can get the wake-up schedules of its neighboors
The destinator can emit a specific burst to notify the source
Possibility to stop the relay process
Reduction of the latency because the source can anticipate the payload emission
Beacon-free MAC mode
Exemple of latency reduction
Upper layers responsibility
Management of cooperative transmissions
One way relay channel (OWRC)
A single source node transmits data to a single destination node with the help of some relay nodes
Two way relay channel (TWRC)
Two nodes like to communicate to each other through the help of a relay
Upper layers responsibility
Cooperation possibilities
Low data rate transmission
Complexity of involved nodes and power consumption are the main issues
Low complexity cooperative techniques may be used
High data rate transmission
Increasing data rate and channel availability are more important
Decode and forward schemes in OWRC or even network coding with TWRC are more suitable