Review of UltraWideband Radio Channel Modeling

1
Review of Ultra-Wideband Radio Channel Modeling

• Martin Weisenhorn
• Zurich Research Laboratory
• E-mail mwe_at_zurich.ibm.com

2
Outline

• Measurement of UWB Channel Impulse Response
• Modeling of Observed Effects
• A Method for Modeling MIMO Channels
• Channel Properties Affecting the Performance of
  Communication Systems
• Time Variance of UWB Channels
• Bandpass or Passband Representation

3
UWB Channel Impulse Response

• Large signal bandwidth 500 MHz 7.5 GHz in the
  band 3.1-10.6 GHz (according to FCC rules)

Possible scheme to acquire the impulse response
of a channel
500 MHz
f , t
Source Multispectral Solutions
Source Multispectral Solutions
4
UWB Channel Impulse Response
Pulse duration 2 ns - 0.13 ns
Pulse spread 60 cm - 4 cm
Propagation scenario
LOS and NLOS impulse responses
LOS channel
B7.5 GHz
10 20
30
NLOS channel
Receiver
Transmitter
B7.5 GHz
t ns
5
UWB Channel Impulse Response
Impact of antenna characteristics on impulse
response
Channel impulse response h(t) from feed point to
feed point depends on antenna impulse responses

• Antenna impulse response
• Frequency
• Elevation ? and azimuth ?

In most measurement campaigns antennas with quite
uniform characteristics, as an approximation to
the isotropic radiator were employed
?
As a consequence, the resulting channel models
are in principle not valid for antennas with
other characteristics
6
UWB Channel Impulse Response
Example of a power delay profile
Observed effects
frequency band 1 - 11 GHz
dB
Source ACORDE Univ. of Cantabaria
7
Statistical Modeling of the UWB Channel
Grouping of effects

• Small sale effects
• Shape of path impulse response
• Arrival time of path impulse responses
• Small changes of arrival times of path impulse
  responses have large impact on the amplitude of
  superimposed rays
• Large scale effects
• Time decay constant of power delay profile
• Path loss
• Frequency selective path loss

8
Path Loss Characterization
4
LOS
NLOS
SNR values achieved in practice
Example
1
1
1
LOS
NLOS
cdf
d20 m
d 4 m
0
SNR dB
-20 0 20 40
9
Frequency Selective Path Loss
A frequency dependent path loss is observed from
the power spectral density of channel impulse
responses
A theoretical explanation includes the smaller
effective aperture and higher material dependent
attenuation for higher signal frequencies
The effect can be modeled in the frequency domain
with a filter characteristic
where the m - paramter is in the range (0.8,
1.2), see 1
Source IMST
10
Frequency Selective Path Loss
B7.5 GHz
baseband representtion
The impulse response of this filter shows that
only weak correlation is introduced between rays
it should therefore be sufficient to consider the
frequency selective attenuation effect as being
included in the path loss description.
11
Statistical Modeling of Impulse Response
Overview on the modeling of different channel
properties
12
Statistical Modeling of Impulse Response
Many channel models were proposed, suggesting
different statistics to describe channel impulse
responses
Primary reasons (depend on extraction procedure
of model parameters)

• Different measures were acquired,?, resulting
  statistical parameters cannot be compared, e.g.
• Antenna characteristics included or excluded
• All rays independent on amplitude considered or
  only rays considered with amplitudes beyond a
  threshold

Secondary reasons (depending on channel
measurement procedure)

• Different measurement setups, building types,
  antenna characteristics were used, different
  measurement principle introducing different
  systematic errors e.g.
• Impulse response samples considered or square
  root of energy captured during time bins
  considered.

13
A Method for Modeling MIMO Channels
Virtual source concept

• Each virtual source causes an individual path
  impulse response

• Delays of path impulse responses depend on
  transmitter and receiver positions

• Channel impulse responses are functions of the
  transmitter and receiver position

• Mixed, deterministic statistical channel model
• Virtual source positions depend deterministically
  on room dimensions
• Path impulses responses are realizations of a
  stochastic process

14
Channel Properties Affecting the Performance of
Communication Systems
Cross correlation function
B7.5 GHz
?d cm

• Diagonal lines stem from identical path impulse
  responses contained in both received impulse
  responses
• ACF is concentrated in time
• CCF is concentrated in space

? ns
15
Channel Properties Affecting the Performance of
Communication Systems
Cross correlation function
B500 MHz

• Diagonal lines stem from identical path impulse
  responses contained in both received impulse
  responses
• ACF is concentrated in time
• CCF is concentrated in space

16
Channel Properties Affecting the Performance of
Communication Systems
Small scale statistics of impulse response energy

• Responsible channel properties
• Fading characteristics of ray amplitudes
• Number of resolvable rays
• UWB channel properties
• Number of resolvable rays is large and increasing
  with the channel bandwidth
• High diversity offering, increasing with channel
  bandwidth

cdf
energy
17
Channel Properties Affecting the Performance of
Communication Systems

• Received energy
• Almost constant under small scale motion
• Unlike for narrowband systems, no additional
  transmitted power required to overcome deep small
  scale fades
• High diversity offering due to large number of
  resolvable paths
• ACF
• Very concentrated in time when compared to delay
  spread
• ISI robust systems feasible, the more rays
  captured the more robust
• CCF
• Concentrated in space
• Potential reduction of crosstalk from undesired
  users
• Small dimensions allowed for antenna arrays (MIMO)

18
Time Variance of UWB Channels

• No measurements of real time varying UWB channels
  are available

• In a first measurement campaign, static
  measurements were made to assess the impact of
  obstruction, caused by a single human when
  crossing the LOS path

19
Time Variance of UWB Channels
Source Steve Schell Bitzmo, Inc 3
20
Time Variance of UWB Channels
Channel similarityfrom position to position

• Results for 2000 MHz bandwidth
• mean,min,max,etc over all f of
  prediction_error(position)

Source Steve Schell Bitzmo, Inc 3
21
Passband or Bandpass Representation

• Narrowband Systems with mixers
• Baseband representation is a natural choice of a
  higher layer description
• The symbol clock lies on a grid of equal carrier
  phases
• Channel description is more compact in baseband
  representation, signals with much lower
  bandwidths involved
• UWB Systems not employing mixers
• In UWB transmitters the carrier frequency is not
  uniquely defined.
• The symbol clock cannot be related to the carrier
  phase
• Baseband representation requires that every
  transmitted symbol has individual complex phase
• Signal bandwidth can be reduced only by a factor
  in the order of 10
• ? The choice of passband or baseband
  representation should possible depend on aspects
  of the system implementation

22
Conclusion

• A principal difficulty of determining the
  generally admitted channel impulse response, is
  caused by the directivity of antennas and the
  channel itself.
• The currently most accepted UWB channel models
  statistically describe effects that can be
  observed in the time domain exponential power
  decay with time, amplitude statistics, arrival
  times
• A mixed deterministic - stochastic approach can
  be used for the modeling of MIMO UWB Channels
• The energy of UWB channel impulse responses shows
  small variations, when compared to narrowband
  channels, correlation functions of channel
  impulse responses are concentrated in time and
  space.
• A first step into the direction of modeling time
  variant UWB channels shows that channel tracking
  will be a requirement for certain types of
  applications
• The choice of channel modeling in either baseband
  or passband should depend on the system
  implementation

23
References
1 J. Kunisch and J. Pamp, Measurement results
and modeling aspects for the UWB radio channel,
in Proc. 2002 IEEE Conf. on Ultra Wideband
Systems and Technologies, Baltimore, May 2002,
pp. 19-23.
2 Jeff Foerster, Channel Modeling
Sub-committee Report Final, doc. IEEE
802.15-02/368r5-SG3a, 18 November, 2002,
online. Available http// grouper.ieee.org/grou
ps/802/15/pub/2002/Nov02/
3 Steve Schell, Analysis of Time Variance of a
UWB Propagation Channel doc. IEEE
802.15-02/452r0-SG3a, 5 November 2002, online.
Available http//grouper.ieee.org/groups/802/15/p
ub/2002/Nov02/
4 S. S. Ghassemzadeh and R. Jana and C. W. Rice
and W. Turin and V. Tarok, A Statistical Path
Loss Model for In-Home UWB Channels, IEEE
UWBST-2002 May, 2002
24
Amplitude Distribution
Comparison of distributions (1-20m)
LOS
NLOS
Source Intel Research and Development
25
Passband or Bandpass Representation
Baseband and passband representation of a channel
model are related by a one to one mapping, having
different statistical descriptions.
Example Baseband Rayleigh fading channel
Uniformal distributed
Gaussian distributed
Rayleigh distributed
26
Angle and Time of Arrival
Source R. Jean-Marc Cramer, Robert A. Scholtz,
Moe Z. Win TRW Space and Electronics
27
Typical Normalized Antenna Azimuth and Elevation
Patterns (omni-directional antennas)
Soyrce Time Domain
Source Time Domain
28
Space Time Picture of Propagating Waves
Source IMST