1. Radio Propagation

2. References Wireless Communications: Principles and Practice, Chapters 3 and 4, T. Rappaport, Prentice Hall, 1996. Principles of Mobile Communication, Chapter 2, G. Stüber, Kluwer Academic Publishers, 1996. Slides for EE535, K. Chugg, 1999. Spread Spectrum Systems, Chapter 7, R. Dixon, Wiley, 1985 (there is a newer edition). Wideband CDMA for Third Generation Mobile Communications, Chapter 4, T. Ojanpera, R. Prasad, Artech, House 1998. Propagation Measurements and Models for Wireless Communications Channels, Andersen, Rappaport, Yoshida, IEEE Communications, January 1995.

3. Why Look at Degradation In Cellular Networks? —Signal degradation affects system performance and capacity —During design and planning of a network we must provide for effects of degradation —Need to understand how to model them to develop software to handle system design and planning —Needed for tuning and optimizing networks —For providing professional consultancy to the telcommunication industry —To provide the basis for understanding cellular communication standards on noise performance

4. Types of Degradation In Cellular Networks Noise Multiple Access Interference (MAI) Fading

5. 5 Frequencies for mobile communication VHF-/UHF-ranges for mobile radio —simple, small antenna for cars —deterministic propagation characteristics, reliable connections SHF and higher for directed radio links, satellite communication —small antenna, beam forming —large bandwidth available Wireless LANs use frequencies in UHF to SHF range —some systems planned up to EHF —limitations due to absorption by water and oxygen molecules (resonance frequencies) weather dependent fading, signal loss caused by heavy rainfall etc.

6. Frequencies and regulations ITU-R holds auctions for new frequencies, manages frequency bands worldwide (WRC, World Radio Conferences)

7. General Intuition Two main factors affecting signal at receiver —Distance (or delay)  Path attenuation —Multipath  Phase differences Green signal travels 1/2 farther than Yellow to reach receiver, who sees Red. For 2.4 GHz, (wavelength) =12.5cm.

8. Objective Invent models to predict what the field looks like at the receiver. —Attenuation, absorption, reflection, diffraction... —Motion of receiver and environment… —Natural and man-made radio interference... —What does the field look like at the receiver?

9. Models are Specialized Different scales —Large scale (averaged over meters) —Small scale (order of wavelength) Different environmental characteristics —Outdoor, indoor, land, sea, space, etc. Different application areas —macrocell (2km), microcell(500m), picocell

10. Outline Introduction and some terminology Propagation Mechanisms Propagation models

11. Radio Propagation Mechanisms Free Space propagation Refraction —Conductors & Dielectric materials (refraction) Diffraction —Fresnel zones Scattering —“Clutter” is small relative to wavelength reflectionscatteringdiffraction shadowing refraction

12. Free Space Propagation Model Received power at distance d is —where P t is the transmitter power in Watts —a constant factor K depends on antenna gain, a system loss factor, and the carrier wavelengthconstant factor

13. Large Scale Models Path loss models Outdoor models Indoor models

14. 14 What does “dB” mean? dB stands for deciBel or 1/10 of a Bel The Bel is a dimensionless unit for expressing ratios and gains on a log scale Gains add rather than multiply Easier to handle large dynamic ranges

15. 15 dB 2 Example : Attenuation from transmitter to receiver. —P T =100, P R =10 —attenuation is ratio of P T to P R —[P T /P R ] dB = 10 log(P T /P R ) = 10 log(10) = 10 dB Useful numbers: —[1/2] dB  -3 dB —[1/1000] dB = -30 dB

16. 16 dB 3 dB can express ratios, but what about absolute quantities? Similar units reference an absolute quantity against a defined reference. —[n mW] dBm = [n/mW] dB —[n W] dBW = [n/W] dB Ex: [1 mW] dBW = -30 dBW

17. 17 Free Space Free space power flux density (W/m 2 ) —power radiated over surface area of sphere —where G t is transmitter antenna gain By covering some of this area, receiver’s antenna “catches” some of this flux

18. Free Space Path Loss Path Loss is a measure of attenuation based only on the distance to the transmitter Free space model only valid in far-field; —Path loss models typically define a “close-in” point d 0 and reference other points from there:

19. 19 2-Ray Ground Reflection For d >> h r h t,For d >> h r h t —low angle of incidence allows the earth to act as a reflector —the reflected signal is 180  out of phase —P r  1/d 4 (  =4) R T htht hrhr Phase shift!

20. Log-Distance Path Loss Model Log-distance generalizes path loss to account for other environmental factors Choose a d 0 in the far field. Measure PL(d 0 ) or calculate Free Space Path Loss. Take measurements and derive ß empirically.

21. Log-Distance 2 Value of  characterizes different environments Rappaport, Table 3.2, pp. 104

22. Log-Normal Shadowing Model Shadowing occurs when objects block LOS between transmitter and receiver A simple statistical model can account for unpredictable “shadowing” —Add a 0-mean Gaussian RV to Log-Distance PLAdd —Markov model can be used for spatial correlation

23. 23 IN real world : where X  is a zero-mean Gaussian RV (dB) with standard deviation   and computed from measured data

24. Indoor Path Loss Models Indoor models are less generalized —Environment comparatively more dynamic Significant features are physically smaller —Shorter distances are closer to near-field —More clutter, scattering, less LOS

25. Indoor Modeling Techniques Modeling techniques and approaches: —Log-Normal,  <2 for LOS down corridor —Log-Normal shadowing model if no LOS —Partition and floor attenuation factors —Computationally intensive “ray-tracing” based on 3-D model of building and attenuation factors for materials

26. 26 Partition Losses Partition losses: same floor —Walls, furniture, equipment —Highly dependent on type of material, frequency Hard partitions vs soft partitions —hard partitions are structural —soft partitions do not reach ceiling “open plan” buildings

27. 27 Partition Losses 2 Partition losses: between floors —Depends on building construction, frequency —“Floor attenuation factor” diminishes with successive floors —typical values: 15 dB for 1st floor 6-10 dB per floor for floors 2-5 1-2 dB per floor beyond 5 floors

28. 28 Materials Attenuation values for different materials

29. Outline Introduction and some terminology Propagation Mechanisms Propagation models —Large scale propagation models —Small scale propagation (fading) models

30. 30 Propagation Models: Large Large scale models predict behavior averaged over distances >> —Function of distance & significant environmental features, roughly frequency independent —Breaks down as distance decreases —Useful for modeling the range of a radio system and rough capacity planning

31. Approaches to Modelling —Models valid for far-field, apply to a range of distances —large scale models: concerned with gross behavior as a function of distance —small scale (fading) models: concerned with behavior during perturbations around a particular distancefading Micro-cells may require different models than most of the work featured here —Smaller transmit range —Likely to be near reflectors: on desk or floor. On the other hand, at smaller scales things are less smooth: “ground reflection” may turn into scattering

32. 32 Propagation Models: Small Small scale (fading) models describe signal variability on a scale of —Multipath effects (phase cancellation) dominate, path attenuation considered constant —Frequency and bandwidth dependent —Focus is on modeling “Fading”: rapid change in signal over a short distance or length of time.

33. Factors Influencing Fading Motion of the receiver: Doppler shift Transmission bandwidth of signal —Compare to BW of channel Multipath propagation —Receiver sees multiple instances of signal when waves follow different paths —Very sensitive to configuration of environment

34. Effects of Multipath Signals Rapid change in signal strength due to phase cancellation Frequency modulation due to Doppler shifts from movement of receiver/environment Echoes caused by multipath propagation delay

35. 35 Types of Fading Fast fading —Rapid changes in strength over distances about half wavelength 900MHz wavelength is 0.33m 20-30dB Slow fading —Slower changes due to user passing different height buildings, gaps in buildings etc. —Over longer distances than fast fading Flat fading —Non-selective —Affects all frequencies in same proportion Selective fading —Different frequency components affected differently

36. There are two major categories of fading —(1) small-scale fading - caused by superposition of multi-path signals speed or RX or TX bandwidth of transmitted signal —(2) large-scale fading -called path loss and depends on the distance between TX and RX also known as ‘log-normal’ fading or shadowing

37. 37 Small-Scale Fading Also known by other names such as ‘Fading’; multipath and Rayleigh fading Rayleigh fading is a result of constructive and destructive interference between several versions of the same signal at the receiver, leading to attenuation of signal power or amplitude —Usually over a fraction of the signal wavelength —Attenuation between 20 to 30 dB —Multi-path fading manifests as time spreading or time variation of the signal (due to motion, foliage, reflections and scattering)

38. Characteristics of Small-Scale Fading Small-scale fading occurs as either of 4 types: frequency selective fading in which the bandwidth of the signal is greater than the coherence bandwidth and the delay spread is greater than the symbol rate; Signals at some frequency components experience more fading than others - (caused by multi-path delay spread) flat fading when the bandwidth of the signal is less than the coherence bandwidth and the delay spread is less than the symbol rate - (caused by multi-path delay spread) fast fading when the Doppler spread is high and the coherence time is less than the symbol period and slow fading with a low Doppler spread and coherence time is greater than the symbol period - (caused as well by Doppler spread)

39. The Multipath Channel One approach to small-scale models is to model the “Multipath Channel” —Linear time-varying function h(t,  ) Basic idea: define a filter that encapsulates the effects of multipath interference —Measure or calculate the channel impulse response (response to a short pulse at f c ): h(t,  )   t

40. 40 signal at sender signal at receiver LOS pulses multipath pulses

41. Channel Sounding “Channel sounding” is a way to measure the channel responseChannel sounding —transmit impulse, and measure the response to find h(  ). —h(  ) can then be used to model the channel response to an arbitrary signal: y(t) = x(t)  h(  ). —Problem: models the channel at single point in time; can’t account for mobility or environmental changes h(t,  )   SKIP

42. Characterizing Fading * From the impulse response we can characterize the channel: Characterizing distortion —Delay spread (  d ): How long does the channel ring from an impulse? —Coherence bandwidth (B c ): Over what frequency range is the channel gain flat? —  d  1/B c In time domain, roughly corresponds to the “fidelity” of the response; sharper pulse requires wider bandtime domain

43. Delay Spread Definition The standard deviation of the distribution of multi-path signal amplitudes is called delay spread. —Delay spread varies with the terrain with typical values for rural, urban and suburban areas:

44. Coherence Bandwidth Doppler frequency shift causes the signal bandwidth to change. What then is the real bandwidth of the signal? The concept of coherence bandwidth is used to address this question Definition: Coherence bandwidth is defined to be the statistical measure of the range of frequencies over which the channel is considered constant or flat. It is the bandwidth over which two frequencies have a strong potential for amplitude correlation.

45. Estimation of Coherence Bandwidth Typical values of delay spreads for various types of terrain:

46. Effect of Delay Spread * Does the channel distort the signal? —if W << B c : “Flat Fading”Flat Fading Amplitude and phase distortion only —if W > B c : “Frequency Selective Fading”Frequency Selective Fading If T <  d, inter-symbol interference (ISI) occurs For narrowband systems (W  1/T), FSF  ISI. Not so for wideband systems (W >> 1/T) For a system with bandwidth W and symbol time T...

47. Two Pulses in Time-Variant Multipath

48. Detector Output for Pulsed Source Received Pulse V e ( t ) e j  t t t Ve(t)Ve(t) Envelope of received voltage Q(t)Q(t) t Power delay profile Transmitted Pulse p(t) e j  t t

49. Qualitative Delay Spread RMS Delay spread (  ) Mean excess delay Noise threshold Delay  Power(dB)  Typical values for  : Indoor: 10-100 ns Outdoor: 0.1-10  s

50. 50 Flat Fading T >>  d and W << B C  minimal ISI 0 TsTs 0  0 Ts+Ts+ fcfc fcfc fcfc t t t f f f s(t) r(t) h(t,  ) Time domain (convolve) Freq domain (filter) = = Delay spread Coherence BW

51. 51 Frequency Selective Fading T > B C  ISI 0 TsTs 0  0 Ts+Ts+ fcfc fcfc fcfc t t f f f s(t) r(t) h(t,  ) Time domain (convolve) Freq domain (filter) = = Delay spread Coherence BW TsTs

52. Characterizing Fading 2 * Characterizing Time-variation: How does the impulse response change with time? —Coherence time (t c ): For what value of  are responses at t and t+  uncorrelated? (How quickly is the channel changing) —Doppler Spread (f d ): How much will the spectrum of the input be spread in frequency? —f d  1/t c

53. Effect of Coherence Time * Is the channel constant over many uses? —if T << t c : “Slow fading” Slow adaptation required —if T > t c : “Fast fading” Frequent adaptation required For typical systems, symbol rate is high compared to channel evolution For a system with bandwidth W and symbol time T...

54. 54 Fast Fade As a mobile user travels they are moving in and out of constructive and deconstructive areas of interference. If they are at a point where signals are in phase they will add. When moving a distance of ¼ wavelength that increases the direct path there will be an equivalent reduction of the reflected path resulting in a 180 degree phase shift and cancellation.

55. Doppler Overview Apparent shifts in frequency of transmitted signal due to motion of transmitter/receiver or both. Shift depend on the relative velocity of the transmitter and receiver. Non-relativistic motionRelativistic motion : Cellular communication hampered by multipath fading effects and receiver movement (non- relativistic Doppler).

57. Small Scale Fading Rapid fluctuations in receiving conditions due to small movement of the receiver. Some causes: —Multipath Fading (Rayleigh and Rician) —Frequency shift due to movement – Doppler

58. Doppler Fading (1/3) For a vehicle moving in a straight line at constant velocity v, the Doppler frequency shift, fd is given by : Typical frequency range : —Most Cellular - 800 to 1500MHz —UHF – 300 to 3000MHz (used by TV, PCS etc.) Typical Doppler shifts : —5Hz to 300 Hz For example, at for a carrier frequency of 2GHz and a mobile speed of 68 mph, max fd = 200Hz

59. Doppler Fading (2/3) Doppler Spread (B D ) – The difference between the maximum and minimum values of fd. Coherence Time (T C ) — Statistical measure of the time duration over which the channel is invariant. —Defined as 1/ B D. Doppler spread and Coherence time characterizes fading speed and its frequency selectiveness.

60. Received Power Spectrum with Doppler (1/3) Assumptions : —Isotropic antenna with unity gain and receiving average power p (without Doppler). —PDF of the direction of waves reaching the receiver is uniformly distributed between 0 and 2 . —Waves coming in from different directions add up to give a PSD S(f). Received signal frequency, f = f 0 + f d The PSD for signals in the range f to f+df corresponds to the waves coming in the direction given by +/- (  +d  ). =>S(f)df = 2* d  *(p/2  ) = d  *p/ 

61. 61 Received Power Spectrum with Doppler (2/3) Also, df = -fm*sin  where fd = fm*cos  But sin  =sqrt(1 – cos^2(  )) =sqrt(fm^2 – (f- f 0 )^2)/ fm So, df = - sqrt(fm^2 – (f- f 0 )^2) Substituting back, we get |S(f)| = p/(  *sqrt(fm^2 – (f- f0)^2))

62. 62 Received Power Spectrum with Doppler (3/3) Doppler Power Spectrum :

63. Statistical Fading Models Fading models model the probability of a fade occurring at a particular location —Used to generate an impulse response —In fixed receivers, channel is slowly time-varying; the fading model is reevaluated at a rate related to motion Simplest models are based on the WSSUS principle

64. WSSUS * Wide Sense Stationary (WSS) —Statistics are independent of small perturbations in time and position —I.e. fixed statistical parameters for stationary nodes Uncorrelated Scatter (US) —Separate paths are not correlated in phase or attenuation —I.e. multi-path components can be independent RVs Statistics modeled as Gaussian RVs Central limit theorem

65. Common Distributions Rayleigh fading distribution —Models a flat fading signal —Used for individual multipath components Ricean fading distribution —Used when there is a dominant signal component, e.g. LOS + weaker multipaths —parameter K (dB) defines strength of dominant component; for K=- , equivalent to Rayleigh

66. Rayleigh Distribution If the impulse response h( , t) of the mobile radio station is time invariant and has zero mean, then the envelope of the impulse response has a Rayleigh distribution given as: where   is the total power in the multipath signal

67. Rice Fading If however the impulse response has a non zero mean then there is a significant component of the direct path (line of sight, specular component) signal and the magnitude of the impulse response has a Ricean distribution Ricean distribution is the combination of Rayleigh signal with the direct line of sight signal. The distribution is: s 2 is the power of the line of sight signal and I 0 is a Bessel function of the first kind

68. Application of WSSUS Multi-ray Rayleigh fading: —The Rayleigh distribution does not model multipath time delay (frequency selective) —Multi-ray model is the sum of two or more independent time-delayed Rayleigh variables s(t) R1R1 R2R2  r(t)  Rappaport, Fig. 4.24, pp. 185.

69. 69 Convolution Integral Convolution is defined by this integral: Indexes relevant portion of impulse response Scales past input signal

70. 70 The impulse response “box” notation implies the convolution operator,  —Convolution operates on a signal and an impulse response to produce a new signal. —The new signal is the superposition of the response to past values of the signal. —Commutative, associative

71. 71 y(t) Convolutions 2 y(t) is the sum of scaled, time-delayed responses x(t)  h(t) = + Each component of the sum is scaled by the x(t)dt at that point; in this example, the response is scaled to 0 where x(t) = 0.

72. t y(vt,t) =  x(  ) h 1 (vt,t -  ) d  -  This is a time varying system with impulse response of h(t,  )

73. پاسخ کانال با يک فرايند گوسی مختلط بيان می گردد توان دريافتی روی تاخيرهای متفاوت که روی زمان متوسط گيری شده باشد تعريف: زمان منتجه که اصولا مقادير غیر صفر دارد را گستردگی تاخير اضافی می ناميم Exess Delay Spread  Q(  ) TMTM

74. ممان مرتبه دوم اين تابع را مقدار موثر گستردگی تاخير کانال می نامند  Different Intensity-Delay Profile have different effect on quality of system

75. Saleh & Valenzuela (1987) Measured same-floor indoor characteristics —Found that, with a fixed receiver, indoor channel is very slowly time-varying —RMS delay spread: mean 25ns, max 50ns —With no LOS, path loss varied over 60dB range and obeyed log distance power law, 3 > ß > 4 Model assumes a structure and models correlated multipath components. Rappaport, pp. 188

76. Saleh & Valenzuela 2 Multipath model —Multipath components arrive in clusters, follow Poisson distribution. Clusters relate to building structures. —Within cluster, individual components also follow Poisson distribution. Cluster components relate to reflecting objects near the TX or RX. —Amplitudes of components are independent Rayleigh variables, decay exponentially with cluster delay and with intra-cluster delay

77. Channel Sounding 2 Several “Channel Sounding” techniques can measure the channel response directly: —Direct RF pulse (we hinted at this approach) —Sliding correlator —Frequency domain sounding

78. Channel Sounding 3 Direct RF Pulse —Xmit pulse, scope displays response at receiver —Can be done with off-the-shelf hardware —Problems: hard to reject noise in the channel —If no LOS must trigger scope on weaker multipath component may fail to trigger lose delay and phase information

79. Channel Sounding 4 Sliding correlator —Xmit PseudoNoise sequence —Rcvr correlates signal with its PN generator —Rcvr clock slightly slower; PN sequences slide —Delayed components cause delayed correlations —Good resolution, good noise rejection

80. Channel Sounding 5 Frequency domain sounding —Sweep frequency range —Compute inverse Fourier transform of response —Problems not instantaneous measurement Tradeoff between resolution (number of frequency steps) and real-time measurement (i.e. duration as short as possible)

81. Note: fades occur at distances of ½ wavelength.

82. Fading Remedies => Increase TX power. Ex. 20 dB is multiplying TX pout by 100. Good for base but handheld at 700 mw is impractical. =>Frequency Diversity. 2 channels in place of 1 in each direction. Impractical due to bandwidth requirement. =>Spread Spectrum. Distribution of signal information over a range of frequencies. Fading of a NARROW channel causes a small loss of data that can be corrected by error correction. =>CDMA. Performs well in presence of multi-path. Several reflected data streams can be received at different times. Rake receiver combines power from various streams. =>Space Diversity.Appropriate location of antennas to reduce multi- path.

83. Multipath Model —Multi-path is modeled as a linear time varying filter with impulse response h(t,  ) where  is the multi-path delay in the channel for a fixed time t a low pass filter approximation is used in practice signal components are modeled relative to the component that arrived first with delay   = 0 components arriving latter are separated at discrete times with delays in N equally spaced time intervals of width  components in bin with delay   = i  are thus lumped together as one

84. Doppler Shifts Doppler Effect —A moving object causes the frequency of a received wave to change —In a cellular communication environment the measured frequency increases as the mobile moves towards a base station —As it moves away from the base station, the frequency decreases Effects of Doppler shifts —bandwidth of the signal could increase or decrease leading to poor and/or missed reception —For a mobile phone in an object (eg. car) moving at a speed of v m/s, the Doppler shift is where  is the angle made by the signal path to the base station and the ground plane

85. Effects of Doppler Frequency Shift The effect in time is coherence time variation and signal distortion —Coherence time is the time duration over which two signals have strong potential for amplitude correlation —Coherence time expressions —where f m is the maximum Doppler shift, which occurs when  = 0 degrees —To avoid distortion due to motion in the channel, the symbol rate must be greater than the inverse of coherence t

86. Propagation - Practical Models —Propagation out door is difficult to predict and as such, empirical models, without real analytical basis are applied —Most of the models used are accurate to within 10 to 14 decibels in urban and suburban areas —They tend to be less accurate in rural areas because most of the data used may have been collected in the urban and suburban areas —In practice there are huge variations in the types of terrain and environment to cover. —Heights of antenna, clutter, tree density, beamwidth, wind speed, season and multi-path, vary widely and affect mobile phone waves. —Hence complex models are required for such situations. They are used to predict propagation loss.

87. 87

88. 88