UWB link design

1
Project IEEE P802.15 Working Group for Wireless
Personal Area Networks (WPANs) Submission
Title Designers Guide to TG3a UWB Link
Margins Date Submitted 13 January 2004 r1 on
18 January 2004 r2 on 21 January 2004 Source
Kazimierz Kai Siwiak Company
TimeDerivative, Inc. Address PO Box
772088, Coral Springs, FL 33071 Voice
1-954-937-3288 E-Mail k.siwiak_at_ieee.org
Re Link Margins for UWB from the system
designers point of view Abstract This
contribution describes UWB system link margins
from a customers point of view, and contrasts
those margins with the SG3a / TG3a selection
criteria. Purpose UWB Link margins in the
selection process were determined for the purpose
of comparing the relative merits of various UWB
approaches. While suitable for that purpose, the
results are optimistic for practical system
designs. This contribution documents areas of
additional practical link losses, and is a first
step in practical link design. The additional
losses are different for different UWB
PHYs. 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
System Designers Guide to UWB Link Margins

• Kai Siwiak
• IEEE Submission
• Vancouver IEEE January 2004

3
Revisions

• Ive gotten many enquiries indicated that more
  information is desired on this topic. These
  revisions add information.
• rev 1
• added more detail on OATS signals
• rev 2
• DS-UWB signal shape on OATS added
• more detail added to the multipath losses issue
• embedded antenna losses estimated
• modulation efficiency added
• statistical design for reliability

4
Introduction

• The data rate and range capabilities of TG3a UWB
  PHYs derived from Selection Criteria 03/031r5 are
  excessively optimistic
• The Selection Criteria are designed for PHY
  selection and are NOT useful for system designs
• Several factors detracting from the link margin
  are presented here need to be considered
• Different results are seen for different UWB PHYs

5
Selection Criteria is Not a Design Tool

• The Selection Criteria Link Calculation is
• Optimistic in free space by 5 to 11 dB
  depending on the variety of UWB used
• Optimistic by 11 to 17 dB in multipath for 50
  reliable link
• Additional factors like realistic antennas (5
  dB?) and statistical link design (5 to 9 dB?) can
  detract further from the link
• The actual link is margin-starved!
• System Designers dilemma How good is the link,
  really?

6
The Selection Criteria

• Selection criterion is a convenience
• Was a suitable basis for 15.3a PAR
• Calculation is almost equal for all, but
  artificial
• Result is contrived, but generally adequate for
  PHY selection
• Link margin for design must be found more
  accurately
• Noise BW error is corrected
• EIRP is corrected based on FCC OATS measurement
  method
• Multipath propagation model included
• Effects of multipath must be included

7
Channel Noise BW Error

• Rb term in 03/031r5 is throughput, not channel BW
• True channel noise BW is Rb/(FEC Rate)
• Effect is link SNR overestimated by the amount
  of the FEC rate

8
Effect of FEC in the Free Space Selection
Criteria Scenario

• In AWGN FEC can be a net loss at low Eb/N0
• In multipath...
• Need Monte Carlo simulations
• FEC drives BER curve to the AWGN value

Eb/N0 pre-FEC operating point
with FEC
BER
BPSK or QPSK
Eb/N0
9
EIRP The FCC Way

• Selection criterion uses -41.3 dBm/MHz
• FCC says
• Derate full anechoic chamber results by 4.7dB
  see FCC RO 02-48
• Or, measure in semi-anechoic chamber or certified
  OATS (Open Air Test Site)
• 4.7dB accounts for a constructive coherently
  adding ground reflection (in FCC semi-anechoic
  chamber)
• The net effect similar for both systems because
  receiver BW is 1 MHz

10
PSD Measurements on FCC OATS
test sense antenna ground to 3 m search for peak
D (h1-h2)2d2
R (h1h2)2d2
DUT
DR-D
ED 1
D
D 3222 - 3 0.61 m
if wavelet is shorter than about 0.61 m than the
two paths add as power, otherwise, add as
voltage IF the test receiver BW is large enough!
It is NOT
ER 0.718
1 m
metal ground plane
R
3 m
11
MBOFDM Signal Spectrum Analyzer Signal on FCC
OATS
12
Moving the Sense Antenna just moves around the
Peaks
... a few cm up
2
amplitude
1.5
need to de-rate EIRP by this amount 4.7 dB!
S
1
i
2
amplitude
0.5
1.5
0
0
20
40
60
80
100
120
140
frequency
S
1
i
i/3
11 dB peak to dip ... this is an indicator of
Rayleigh fading problem
0.5
a few cm down ...
0
0
20
40
60
80
100
120
140
frequency
i/3
13
DS-UWB Signal PSD on FCC OATS
need to de-rate EIRP by this amount 4.7 dB!
DS-UWB signal, free space
DS-UWB signal on FCC OATS
2
2
amplitude
amplitude
1.5
1.5
1
1
0.5
0.5
0
0
3
3.5
4
4.5
5
5.5
3
3.5
4
4.5
5
5.5
frequency, GHz
frequency, GHz
14
DS-UWB Signal on FCC OATS ...

• Coherence length is the chip length, however
• Test receiver BW is 1 MHz, hence coherence
  length is much larger than the chip length
• Net result reflection from OATS ground plane
  adds coherently, even for impulses
• Signal behaves same as OFDM same amplitude
  profile across the band!
• Actual effect on wide-band victim receivers much
  more benign for DS-UWB vs. OFDM, but EIRP is
  affected in the same way

15
Path Loss

• TG3a channel model does not consider propagation
  attenuation properly
• Median loss not taken into account
• it is NOT 1/r2 at 10m
• strongest path breaks to 1/r3 near 3m
• One model (SBY model) of the additional loss is
• L10 log(1-e-dt/d) where dt3 m, d10 m
• L5.9 dB

Reference K. Siwiak, H. Bertoni, and S. Yano,
Relation between multipath and wave propagation
attenuation, Electronic Letters, Vol. 39, No. 1,
Jan. 9, 2003, pp. 142-143.
16
Operation In Fading

• Actual average antenna loss for 0 dBi antenna
  1.8 dB
• Typical embedded antenna losses are in the 5 to 8
  dB range
• Fading Effect on 10 m link
• DS-UWB -1.0 dB CDMA symbol energy is equally
  distributed across the 1.4 GHz signal, hence
  Rayleigh fading is never experienced even though
  selective fading could place a null at some
  frequencies for sinewave or long persisting
  signals
• MBOFDM -6.0 dB Slide 11 herein substantiates
  an 11 dB peak-to-dip value, or 11-4.76.3 dB
  mean-to-dip value for the 2-ray multipath
  scenario on the FCC OATS site
• ref 15-03-0344-03-0003a
• lesser values also claimed, however references
  are not available

17
MBOFDM and DS-UWB Statistics
1
25 of the signal levels are 6 dB below mean
for MBOFDM 1 dB below mean for DS-UWB 5 dB
difference
0.25
5 dB
0.1
8 dB
Probability the signal energy is less than X
10 of the signal levels are 9.5 dB below mean
for MBOFDM 1.5 dB below mean for DS-UWB 8 dB
difference
Rayleigh faded signal
need to operate somewhere below the 25 signal
loss level for FEC to operate properly
1.4 GHz BW
4 MHz BW
0.01
-20
-5
-15
-10
0
5
X, signal energy relative to mean level, dB
18
Operating in Fading

• A sufficient fraction of the signal energy must
  be above a given level so that FEC can work all
  systems
• In OFDM
• the faded components must be sufficiently
  interleaved so that the fading looks random
  rather than burst error
• enough tones must be available for FEC to work
• DS-UWB spreads all bits over all BW

19
Signal Above the Mean Level

• MBOFDM can (but doesnt) use more bits on
  stronger tones (water filling - a very complex
  problem)
• energy is recovered, but not utilized!
• DS-UWB data is spread across entire band
• all energy contains all data

20
Modulation Efficiencies in AWGN
Because the link is starved for margin, use of
modulation with the highest efficiency is
important ... M-BOK modulations tend to better
efficiency as M increases, PAM and QAM
modulations tend to become less efficient as M
increases
Source K. Siwiak, D. McKeown, Ultra-wideband
Radio Technology, Wiley UK, 2004
21
Modulation Efficiencies in AWGN
Better
Worse
Source K. Siwiak, D. McKeown, Ultra-wideband
Radio Technology, Wiley UK, 2004
22
Summary Link Margin Effect
lesser values also claimed various proposal
report between about 5 and about 6 dB of margin
at 10 m and 110 Mb/s
23
More dBs needed Statistical Link Design

• Average stated so far the 50 probability
  that the link will be closed
• If we require more than 50 (90 typically)
  probability that a link be successful in some
  scenario then
• need to add zs dB
• s is the lognormal standard deviation in dB, s4
  dB for shadowing (more for other effects)

Source K. Siwiak, Radiowave Propagation and
Antennas for personal Communications, Second Ed.,
Artech MA, 1998
24
Conclusions

• UWB link is MARGIN-STARVED
• Other issues remain real antennas (5 dB?) and
  statistical link design (5 to 9 dB?)
• NEED to review available improvements
• Modulation efficiency (need better efficiency
  than BPSK/QPSK) M-BOK rather than M-QAM
• FCC emission measurement method
• Cost of diversity improvements need to be
  explored
• NEED to review the application space