Project: IEEE P802.15 Working Group for Wireless Personal Area Networks (WPANs)Project: IEEE P802.15 Working Group for Wireless Personal Area Networks (WPANs)

Submission Title: [TG4a-SandLinks-CFP-Presentation]Date Submitted: [4 Jan, 2005]Source: [Dani Raphaeli, Gidi Kaplan] Company: [SandLinks]Address: [Hanehoshet 6, Tel Aviv, Israel]E-Mail: [danr@eng.tau.ac.il]

Re: [802.15.4a Call for proposal]

Abstract: [A proposal for the P802.15.4a alt-PHY standard]

Purpose: [Response to WPAN-802.15.4a 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.

Low-Rate UWB Alternate Physical Layer

Proposal Submissionfor TG 802.15.4a

Jan ‘05 Meeting

Dr. Dani Raphaeli & Dr. Gideon KaplanSandLinks

Outline• General Overview• Signal and Packet design• Communication Performance

– Sensitivity, Acquisition– Interference & Coexistence– Aggregate Rate

• Ranging• MAC Protocol Considerations• Block Diagrams and Technical Feasibility• Cost/Complexity• Scalability• Power Consumption• Summary

Technical Requirements of TG-4a

• Low complexity and cost• Low power consumption • Precision location (highly desired – relative rangingrelative ranging)• Extended range• Robustness (against MP, against interference)• Mobility • Low bit rate for each individual link• High Aggregated rate at a collector node• Random, ad-hoc, topology• Work under current 15.4 MAC

General Overview of Proposal

• Symbol Interleaved Impulse Radio• 500Mhz bandwidth in UWB band

– Optional: 80Mhz in 2.4GHz, 200Mhz in 5.2 Ghz• May choose (program) one of several Center

Frequencies • Use of Round Trip Delay for ranging• Low data rate per device allows to obtain PER and

Ranging within substantial distances, for various channel models

• High total (aggregate) rate • Suitable for very low-cost (small die size)

implementation in a standard process• Robust, Flexible and Scalable solution.

Symbol Interleaved Impulse Radio Basic principle: Use pulse trains with constant large separation

between them. Each pulse train represents one symbol. Pulse train (or sequence) is used instead of single pulse to

decrease peak to average, which serves to:• Simplify implementation

• Meet FCC peak power constraint in the UWB band

Pulse sequence polarity corresponding to the 11 bit barker sequence 10110111000

~100ns

~20s

Symbol Interleaved Impulse Radio (cont.)

Many users can transmit concurrently without interference:(each color represents a different packet from a different user).

~20s

Substantial aggregate rate can be achieved (see in the sequel); the transmission management mechanism of 15.4 is appropriate.

Benefits

There is no need for a difficult and slow synchronization process (incurred if several / long sequences are used)

Easy implementation

Passes FCC rules

Reduced sensitivity to Multipath (see figure below)

Near-Far Problem is minimized.

Signal (Pulse) Design• A look on an actual pulse train symbol (fc=4GHz)

• Zoom on a single pulse

• For average and peak powers- see Appendix A

8.5 9 9.5 10 10.5 11 11.5 12 12.5 13 13.5 14 14.5 150.2

0

ns

0 10 20 30 40 50 60 70 80 90 100 1100.2

0.1

0

0.1

0.2

ns

Signal (Pulse) Design• A look on an actual pulse train symbol (fc=4GHz) in

the frequency domain, Pt=-15dbm

3 3.1 3.2 3.3 3.4 3.5 3.6 3.7 3.8 3.9 4 4.1 4.2 4.3 4.4 4.5 4.6 4.7 4.8 4.9 570

60

50

40

GHz

dbm

/MH

z

51

41fc 0.25 fc 0.25

Packet Structure Design

• Preamble (un-modulated) part enables to synchronize on received signal and for receiver acquisition and training.

• Data part uses PPM (binary, possibly M-ary) to convey message [SPDU]. Message lengths – between 7 to 128 Octets (MAC limit). Nominal symbol rate is 50Ksym/sec.

• Response (un-modulated) part allows for synchronous Ack (see in the following) plus data response.

• Total packet length – typically 10 to 20 msec.

Packet Structure

PPM

Preamble DATA (MAC fields)Response Period (optional)

Unmodulated Unmodulated

The Response Period

ACK Preamble

Response Period

ACK DATA

The ACK is transmitted during the response period of the original Packet.

DATA

The Synchronous ACK

• The ACK is transmitted during the response period of the original

packet thereby allowing synchronization of the response to measure the channel round trip delay.

• The Response Period duration is minimally equal to the ACK preamble duration, and at maximum lasts for the entire ACK

• The response (the ACK) is transmitted at a fixed (known) delay relative to the RP pulses. The Node receiving the ACK can measure the RTD and calculate the distance accordingly.

• The symbols of the RP are used for synchronizing the response – This allows the use of low accuracy clocks, which serves to:

REDUCE THE COST MINIMIZE SYSTEM COMPLEXITY (MAC/higher layer not

involved in generating accurate time base)– Since the ACKs are transmitted at a fixed delay, ACK collisions

are avoided as long as the original packets were not colliding

Topology & Types of Devices

The 802.15.4 defines two types of devices:• The low complexity RFD (Reduced Function Device) which

can be only a leaf on the network. • The full complexity FFD (Full Function Device).

• A typical topology composed of manymany RFDs as the sensors or tags and fewfew FFDs as coordinators and data concentrators.

• The topology may change in the network.

PANcoordinator

FFD

RFD

Types of Devices (cont)

• We propose asymmetric PHY: FFD with higher functionality and higher

cost and RFD with lower functionality and cost.

• The ultra low cost RFD (Reduced Function Device) is notnot required to be able to receive multiple packets. It will be capable of:

– Responding to FFD requests.– Sending packets to a FFD– Requesting for a pending packet

• The FFD (Full Function Device) is expected to be able to receive simultaneous multiple packets concurrently. It will be capable of:

– Receiving many packets at the same time and responding each of them with ACK.

– Calculating the distance to each node it received ACK from– Responding to RFD data requests.

Communication Performance – PER vs. Eb/No

• The chosen modulation is PPM• Coding scheme is still TBD. We use simple (63,57) Hamming code

(and hard decision decoding) for the current presentation; however obviously other codes, still simple to implement, exist with a higher coding gain.

• For 32 octets, to get PER of 1%, the BER should be

BER <= 0.01/(32*8)=4e-5• In the next slide, the theoretical results show that Es/N0=11.5dB is

required

• In AWGN channel, for 50Ksym/sec, d=100m is achieved with ~6dB of margin.

Communication Performance – Theoretical BER vs. Eb/No

4 5 6 7 8 9 10 11 12 13 141 10

10

1 109

1 108

1 107

1 106

1 105

1 104

1 103

0.01

0.1

1

PPM (uncoded)PPM (coded)

Theoretical BER performance

Es/N0 (dB)

BE

R

Link Budget (AWGN channel)Nominal Proposed parameters

Peak payload bit rate 50.00 Ksym/secAvg Tx power -14.30 dBmTx Ant Gain 0.00 dBiFc 4.00 GhzPath loss at 1meter 44.48 dBd 100.00 meterPath loss at d meter 84.48 dBRx Ant Gain 0.00 dBiTx Backoff 3.00 dBRx loss 0.00 dBRx rcv power -101.78 dBm

Avg noise power per bit -127.01 dBm (per Rb)Rx Noise figure 7.00Avg thermal noise power -120.01 dBm

Min Es/No 11.50 dBImpl loss 1.00 dB

Link Margin 5.73 dBProposed Sensitivity -107.51 dBm

Pathloss+Margin 90.21 dB

Performance under Multipath

• From the link budget: Receiver Sensitivity is -107.5 dBm; or, total path loss <=90dB.

• Achievable distances for the 9 channel models defined by the TG4a channel modeling subgroup, are shown in the next slide.

• PER performance on these channels was checked by system simulation. The simulation includes: Acquisition Tracking Adaptation Demodulation Decoding Packet processing

• The PER results for several channel models (presented next) show good match with the theoretical predictions.

Distance vs. Channel Models[50Ksym/sec]

CMType of ChannelDistance [m]

1Resident. LOS394

2Resid. hard NLOS (concrete walls*)

8.2

3Office LOS1610

4Office NLOS20.6

5Outdoor LOS421

6Outdoor NLOS75

7Industrial LOS421

8Industrial NLOS27

9Farm393.5

*The high atten. In 15-04-0290-02-004a taken from 802.15-02/444

PER curves

• Simulation on Channels Packet length=512

bits, Hamming code

0

0.01

0.02

0.03

0.04

0.05

11 12 13

Es/N0

PE

R

CM1

CM2

CM3

CM4

CM5

Acquisition

• We assume the super-frame structure includes a Beacon transmission

• In a steady-state, all devices synchronize to the Beacon transmissions of the PAN coordinator

• A quick re-acquisition (in a short length window), to re-align the timing, is performed per each received Beacon.

• The device then listens in the address message space to check if data is waiting; otherwise (if the device does not need to transmit) – it goes back to sleep

• The quick acquisition is performed over the standard 4 octets preamble of the Beacon packet

• All normal transmission packets will also include a 4 octets preamble, used for fine timing acquisition + channel model learning.

Acquisition (cont.)

• In case a new RFD/FFD device joins an existing network, it has first to synchronize to the super-frame structure (namely to the Beacons transmissions)

• One possible mechanism is passive association• Assuming that the power consumption dictates no more than about

1% duty cycle over long periods, this passive process will be relatively slow in time.

• If active association is used, faster synchronization can be achieved.

Interference & Coexistence

• Protection against WLAN and other out of band signals (in 2.4Ghz, 5.3Ghz) aided by a 3rd order Band-Pass filter in the receiver (or an equivalent LPF after down conversion)

• For narrow-band interference (in-band), – High processing gain inherent in the technique (500MHz/50KHz=40dB)– Adaptive or programmable interference rejection mechanism (with mild

requirements) may be employed

• A real life effect which should be considered, is the transmission of “wide band noise” (OOB) by other devices, which covers the same freq band as the UWB device

• • The result show that at most 1m separation insures meting the

criteria of PER<=1%, for UWB signal level 6dB above sensitivity level

• For detailed analysis see spreadsheets in submitted material.

Interference & Coexistence (cont.)

• Under extreme interference cases, a change of the active band may be undertaken (under higher layer command).

• Coexistence with other devices (802 type, Vsats,..) is achieved with a small distance separation, due to the low average power density level of UWB transmission (detailed analysis in submitted material)

• Co-existence with other Piconets – possibly co-located – may be simply achieved by selection of different active frequency bands for the Piconets (up to 3).

• The band select filter provides more than 20dB attenuation, even for the adjacent bands of 4Ghz (centered at 3.5Ghz, 4.5Ghz).

• Further simulation results will be provided later on.

Band Plan

• The analysis (e.g. Link Budget) was made with a Fc=4Ghz (Fl=3.6Gh, Fh=4.4Ghz for -10dB points)

• The UWB freq range can be divided to multi-bands, coordinated with other uses defined by the ITU and IEEE bodies

• Typically a device may be programmable to one of 3 bands in the range 3-5GHz (and additional bands in 6-10GHz when higher speed processes will be cost effective)

• This enhances the robustness of the design and may serve to improve acceptance by regulation bodies worldwide

• Outside the USA, device will operate in 2.4GHz or 5.2GHz until UWB will be approved worldwide.

• Nevertheless, since the high aggregate rate (~10Mbps) enables virtually all multiple uses in the same area, the standard should allow for lower cost devices to be fabricated for one fixed band.

Aggregate Rate Considerations

• Recall the Interleaved pulsed transmission proposed • There are N=200 virtual time slots (of Ts= 100nsec), totaling

20usec, between each transmitted symbols of a single packet

• The transmitting / answering devices can chose one of the N virtual time slots, to transmit their packet

• This choice is kept throughout the packet

• Due to the possible spatial layout of the answering devices, round trip delay differences can be larger than Ts.

• Thus the basic model is multi-channel (N) un-slotted Aloha

• The throughput vs. offered load of such a channel is known, and its peak is 1/2e (per slot).

Aggregate Rate (cont.)

• The ALOHA model assumes that if more than one transmission uses the same slot, than there is a collision and none gets through

• Recall the Barker sequence (of length 11) Processing Gain, allowing for more than one reception in a time slot, if their sequences are in shift

• However some issues like Near-Far (power ratio) and also channel multipath come into play

• First analysis estimates that the effective PG is about 3; further simulations are needed to justify this estimate.

• Thus the scheme has 3N effective slots, so the maximum aggregate rate is

3*200*(1/2e)*1/50usec = 5.5Msym/sec.

Aloha Curve(s)

0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0 0.5 1 1.5 2

Unslotted Aloha

CSMA-CA, a=0.3

Throughput

Load

Aggregate Rate (cont.)

• For a ALOHA channel, insuring stability is of importance, by employing simple anti-congestion (“back-off”) mechanisms

• Usage of Guaranteed Time Slots (GTS) can further improve the capacity, as these will operate at close to 100% efficiency;However this mode is applicable especially to relatively long transmissions.

• Employing a collision avoidance (or CCA) mechanism, performance is improved in the (contention-based) Aloha slots as well as the stability

• With CCA employed, for a propagation delay of ~30nsec, and transmission of 100nsec, theoretical capacity grows up to to

Capacity = 9.6Mbs

• The transmitting / answering devices hear only a partial population of all devices, thus the actual performance improvement of CCA will be assessed via a simulation (per specific channel and node locations).

Ranging

• Basic method proposed is Round Trip Delay measurement (by a FFD).

• Why should we choose RTD for 15.4a?– No need for fixed expensive infrastructure.– No need to generate a very accurate time base.– The only one that can be used in RelativeRelative systems. – Each node makes its own measurement autonomously.– Easy to handle Multipath (take the earliest component).– Straightforward to implement.– Can handle distance measurement with a single node in case

x,y,z coordinate is not necessary.

Ranging (cont.)• Ranging is performed at same distance coverage as is for

communications• The ranging algorithm uses between 30 to 50 symbols for averaging

of the signal

• Simulation results: for LOS channel models (residential, office, outdoor), the ranging accuracy is on the order of 0.3 to 0.5 meter. [Assuming uncorrelated errors at both measurements of the round trip delay, 1.4nsec is equivalent to (1-way) distance error of 30cm]

• For NLOS channel models that were presented, the first path delay varies randomly in a certain range, in the model realizations; thus, ranging has a large error in some of the models.For CM=4 (office NLOS – probably a “soft” NLOS model), the std deviation is about 3nsec (0.66m).

• The random arrival of first cluster in the model needs further discussion.

Max Ranging Error Results LOS channel models, N=50 symbols

Max magnitude of Ranging Error in 90% of cases [m]

SNR 10 11.5 13 15 [dB]

CM 1 1.15 0.32 0.32 0.36

CM 3 1.08 1.52 0.69 0.59

CM5 1.15 0.78 0.52 0.34

Ranging (cont.)

• Considerations for mobile nodes:

• Time for ranging is between 600usec to 1msec.• For mobility values on the order of 1meter/sec (on a mobile luggage

conveyer, for example), the displacement affected while location is measured is negligible – on the order of 0.1 cm. This is also negligible compared to the wave length (~8cm).

• Assuming coherence time requirement of 5ms the maximum doppler rate is ~200Hz, which translates to about 15m/s max speed.

MAC considerations

• Network includes FFD and RFD devices

• Packet structure adheres to 15.4

• Supports the full set of 15.4 MAC functions

• Ranging result – just another parameter transferred from Phy to Mac layer after a single transaction

• Supporting anti-congestion mechanisms at both type of devices.

Receiver Block Diagram

FromAntenna

BPFDown

Conversion +Baseband

Correlator

I

QLNA

LO

Dem odulation+

Acquisition

D ecod ingP acket

hand ling

Hardware

M ay be im plem ented inSoftware

Ranging

Transmitter Block Diagram

B P F P ulseG enera tor

To A ntennaB arker

S equencegenera tor

T im ing andcontro l

Tx dataC hanne l S e lect

P owerA m plifie r

Technical Feasibility

• The analog (RF) part can be implemented by either SiGe or 0.13u CMOS processes.– The former has a higher bandwidth / more accurate

models for high frequencies– The latter is about 30% lower in cost per mm2.– Both technologies are in use today for similar

frequencies (e.g. 802.11a)– The other high speed elements are also based on

existing technology and modules

• All in all, the die size estimation is 6.3 mm2 (see next slide).

Estimated Size and Power (RFD)

Estimated

Die Size [mm2 ]

Estimated Power (mW)

Analog Blocks2.02.5

Analog To Digital 0.53

Digital Blocks, uP, RAM, ROM 3.37.5

Pads 0.5

Total6.313.0

Power Consumption

• The low power consumption is due to – Activating the components only when a transmission is expected (note

the advantage of a short pulse sequence!)

– Low power consumption design methodologies of all the parts

• Each device typically listens only to the Beacons and rest of time is in sleep mode, thus the effective average power consumption will be reduced by a large factor (e.g. 1%), enabling long battery life

• When in acquisition, a search for a symbol over few hypothesis is made.

Scalability

• Higher (peer to peer) data rates can be achieved by 1. interleaving few packets from same source, which essentially

mean lower separation between consecutive symbols. 2. Using higher order PPM

• For example: Interleaving 10 packets and using 16-ary PPM results in 50Kbps*10*4=2Mbps

• ALL RATES ARE COMPATIBLE AND COEXISTENT!

• Lower (peer to peer) data rates can also be achieved (by using lower coding rates, and increasing preamble length accordingly to accommodate lower SNR), but not recommended

• ‘Hooks’ for a cognitive radio can be added in the future, for example to add programmable notch filters in the transmitter.

Summary

• The Symbol Interleaved Impulse Radio system is a sound, complete system proposal that simultaneously answers all the technical requirements of TG-4a of 802.15 and all minimum SCD criteria

• Offers large advantages Offers large advantages (vs. conventional DS solutions)(vs. conventional DS solutions)

in terms of in terms of RangeRange, , PowerPower, , Aggregate rateAggregate rate and and CostCost

• It enables both a robust design in various channels and scenarios, flexibility to a multitude of applications, and a very low-cost solution

• Good distance performance on most channel models

• We will be happy to cooperate with every one that is interested in this direction, in order to further improve its parameters.

Appendix A: Average and Peak Powers

• Regulation:– Average transmission power is limited to -41.3 dBm/Mhz, or

-14.3dBm for a 500Mhz bandwidth

– The peak power per 50Mhz is limited to 0dBm.

• Recall the 11-sequence Barker pulsed transmission (eleven ~2nsec pulses, with 10nsec intervals)

• To achieve the max. Average power, the peak power of each 2nsec pulse will be

-14.3+10*log (20usec/22nsec) = 15dBm• Now check the peak power measured through a 50Mhz wide filter; it

has a time constant of about 20 -30nsec, thus the resultant power is

15 + 10*log (2nsec/10nsec) + 10*log(50/500)= 15-7-10= -2dBm

so that the FCC peak power limit is met.

Appendix B: Interference Spreadsheet (1)Interferer 802.11.15.3 RemarksCenter freq 2.40 GhzTx power (at its center freq) 20.00 dBmTx ant gain 0.00BW 11.00 Mhz

Tx PSD (out of band)) -41.00 dBm/Mhz 1Path loss at 1meter 44.50 dBEst added loss at 1m distance 0.00 dBRx power -85.50 dBm/Mhz

PSD red in freq 25.00 dB 2Rcvd interf power density -110.50 dBm/Mhz

Proposer Fc 4.00 GhzFreq seperation 1.60 GhzPath loss at 1meter 40.06 dBEst added loss at 1m distance 0.00 dBFilter attenuation of Interf signal 58.00 dB 3Narrow band interf rejection 10.00 dBRx power (from filtered interf sig) -92.50 dBmAvg Rx power (from interf sig, over 500Mhz)) -119.49 dBm/Mhz

Total effective Interf noise density -109.98 dBm/MhzProposer thermal noise level (incl NF) -107 dBm/Mhz

Difference of noise levels 2.98 dBNoise level effective increase 1.77 dBMargin (vs 6dB allowed) 4.23 dB

Remarks1) Probably the actual PSD is lower than the FCC allowed levels2) For a large freq separation, due to finite BW of the Power Amp, power is reduced3) A 3rd order Butterworth Band Pass filter attenuation at 2.4Ghz

Appendix B: Interference Spreadsheet (2)Interferer 802.11a RemarksCenter freq 5.30 GhzTx power (at its center freq) 15.00 dBmTx ant gain 0.00 dBiBW 16.60 Mhz

Tx PSD (out of band)) -41.00 dBm/Mhz 1Path loss at 1meter 44.50 dBEst added loss at 1m 0.00 dBPSD red in frequency 20.00 dB 2Rx power (from OOB noise) -105.50 dBm/Mhz

Proposer Fc 4.00 GhzFreq seperation -1.30 GhzPath loss at 1 meter 46.94 dBEst added loss at 1m 0.00 dBFilter attenuation of Interf signal 43.00 dB 3Narrow band interf rejection 10.00 dBRx power (from filtered interf sig) -84.94 dBmAvg Rx power (from interf sig, over 500Mhz)) -111.93 dBm/Mhz

Total effective Interf noise density -104.61 dBm/MhzProposer thermal noise level (incl NF) -107 dBm/Mhz

Effective Noise level Increase 4.37 dBMargin (vs. 6dB allowed) 1.63 dB

Remarks1) The actual PSD is typically substantially lower than the FCC allowed levels2) For a large freq separation, due to finite BW of the Power Amp, power is reduced3) For a 3rd order Butterworth band pass filter

App. B: Co-Existence Example

Co-Existance Analysis with C-band VsatRemarks

Victim receiver C-band VsatInterferer 802.15.4aCenter freq 4.00 GhzTx power (at its center freq) -14.30 dBmTx ant gain 0.00 dBiBW 500.00 Mhz

Path loss at 1 meter 44.48 dBd 20.00 mAtten of building 10.00 dBTx power at distance d -94.80 dBm

Rcvr center freq 4.00 GhzC-band antenna sidelobe gain -0.53 1Rx power per 36Mhz Transponder -105.70 dBm

DVB-S Receiver sensitivity -91.00 dBm 2

1) Assuming the Vsat Antenna has a 20 deg angle2) See document [15-04-609]