july, 19984 - 1rf100 (c) 1998 scott baxter physical principles of propagation chapter 4 section a
TRANSCRIPT
July, 1998 4 - 1RF100 (c) 1998 Scott Baxter
Physical Principles of Propagation
Physical Principles of Propagation
Chapter 4 Section A
July, 1998 4 - 2RF100 (c) 1998 Scott Baxter
Introduction to Propagation
Propagation is the heart of every radio link. During propagation, many processes act on the radio signal.
• attenuation– the signal amplitude is reduced by various natural mechanisms. If there
is too much attenuation, the signal will fall below the reliable detection threshold at the receiver. Attenuation is the most important single factor in propagation.
• multipath and group delay distortions– the signal diffracts and reflects off irregularly shaped objects, producing a
host of components which arrive in random timings and random RF phases at the receiver. This blurs pulses and also produces intermittent signal cancellation and reinforcement. These effects are overcome through a variety of special techniques
• time variability - signal strength and quality varies with time, often dramatically• space variability - signal strength and quality varies with location and distance• frequency variability - signal strength and quality differs on different
frequencies To master propagation and effectively design wireless systems, you must know:
• Physics: understand the basic propagation processes • Measurement: obtain data on propagation behavior in area of interest• Statistics: analyze known data, extrapolate to predict the unknown• Modelmaking: formalize all the above into useful models
July, 1998 4 - 3RF100 (c) 1998 Scott Baxter
Frequency and Wavelength: Implications
Radio signals in the atmosphere propagate at almost speed of light
= wavelength
C = distance propagated in 1 second
F = frequency, Hertz The wavelength of a radio signal
determines many of its propagation characteristics
• Antenna elements size are typically in the order of 1/4 to 1/2 wavelength
• Objects bigger than a wavelength can reflect or obstruct RF energy
• RF energy can penetrate into a building or vehicle if they have apertures a wavelength in size, or larger
/2
C / F
for AMPS: F= 870 MHz
0.345 m = 13.6 inches
for PCS-1900: F = 1960 MHz
0.153 m = 6.0 inches
July, 1998 4 - 4RF100 (c) 1998 Scott Baxter
Propagation Effects of Earth’s Atmosphere
Earth’s unique atmosphere supports life (ours included) and also introduces many propagation effects -- some useful, some troublesome
Skywave Propagation: reflection from Ionized Layers
• LF and HF frequencies (below roughly 50 MHz.) are routinely reflected off layers of the upper atmosphere which become ionized by the sun
• this phenomena produces intermittent world-wide propagation and occasional total outages
• this phenomena is strongly correlated with frequency, day/night cycles, variations in earth’s magnetic field, 11-year sunspot cycle
• these effects are negligible for wireless systems at their much-higher frequencies
July, 1998 4 - 5RF100 (c) 1998 Scott Baxter
More Atmospheric Propagation Effects
Attenuation at Microwave Frequencies
• rain droplets can substantially attenuate RF signals whose wavelengths are comparable to, or smaller than, droplet size
• rain attenuations of 20 dB. or more per km. are possible
• troublesome mainly above 10 GHz., and in tropical areas
• must be considered in reliability calculations during path design
• not major factor in wireless systems propagation Diffraction, Wave Bending, Ducting
• signals 50-2000 MHz. can be bent or reflected at boundaries of different air density or humidity
• phenomena: very sporadic unexpected long-distance propagation beyond the horizon. May last minutes or hours
• can occur in wireless systems
Refraction by air layers
Ducting by air layers
>100 mi.
“Rain Fades” onMIcrowave Links
July, 1998 4 - 6RF100 (c) 1998 Scott Baxter
Dominant Mechanisms of Mobile Propagation
Most propagation in the mobile environment is dominated by these three mechanisms:
Free space
• No reflections, no obstructions
– first Fresnel Zone clear
• Signal spreading is only mechanism• Signal decays 20 dB/decade
Reflection
• Reflected wave 180out of phase
• Reflected wave not attenuated much
• Signal decays 30-40 dB/decade Knife-edge diffraction
• Direct path is blocked by obstruction
• Additional loss is introduced
• Formulae available for simple cases We’ll explore each of these further...
Knife-edge Diffraction
Reflection with partial cancellation
B
A
d
D
Free Space
July, 1998 4 - 7RF100 (c) 1998 Scott Baxter
Free-Space Propagation
The simplest propagation mode• Antenna radiates energy which spreads in space• Path Loss, db (between two isotropic antennas)
= 36.58 +20*Log10(FMHZ)+20Log10(DistMILES )• Path Loss, db (between two dipole antennas)
= 32.26 +20*Log10(FMHZ)+20Log10(DistMILES )
• Notice the rate of signal decay:• 6 db per octave of distance change, which is
20 db per decade of distance change Free-Space propagation is applicable if:
• there is only one signal path (no reflections)• the path is unobstructed (i.e., first Fresnel zone is
not penetrated by obstacles)
First Fresnel Zone ={Points P where AP + PB - AB < }Fresnel Zone radius d = 1/2 (D)^(1/2)
1st Fresnel Zone
B
A
d
D
Free Space “Spreading” Loss energy intercepted by receiving antenna is proportional to 1/r2
r
July, 1998 4 - 8RF100 (c) 1998 Scott Baxter
Reflection With Partial Cancellation
Mobile environment characteristics:
• Small angles of incidence and reflection
• Reflection is unattenuated (reflection coefficient =1)
• Reflection causes phase shift of 180 degrees Analysis
• Physics of the reflection cancellation predicts signal decay of 40 dB per decade of distance
Heights Exaggerated for Clarity
HTFT HTFT
DMILES
Comparison of Free-Space and Reflection Propagation ModesAssumptions: Flat earth, TX ERP = 50 dBm, @ 1950 MHz. Base Ht = 200 ft, Mobile Ht = 5 ft.
Received Signal in Free Space, DBM
Received Signal inReflection Mode
DistanceMILES
-52.4
-69.0
1
-58.4
-79.2
2
-64.4
-89.5
4
-67.9
-95.4
6
-70.4
-99.7
8
-72.4
-103.0
10
-75.9
-109.0
15
-78.4
-113.2
20
Path Loss [dB ]= 172 + 34 x Log (DMiles )- 20 x Log (Base Ant. HtFeet)
- 10 x Log (Mobile Ant. HtFeet)
SCALE PERSPECTIVE
July, 1998 4 - 9RF100 (c) 1998 Scott Baxter
Signal Decay Rates in Various Environments
We’ve seen how the signal decays with distance in two basic modes of propagation:
Free-Space
• 20 dB per decade of distance
• 6 db per octave of distance Reflection Cancellation
• 40 dB per decade of distance
• 12 db per octave of distance Real-life wireless propagation
decay rates are typically somewhere between 30 and 40 dB per decade of distance
Signal Level vs. Distance
-40
-30
-20
-10
0
Distance, Miles1 3.16 102 5 7 86
One Octave of distance
(2x)
One Decadeof distance (10x)
July, 1998 4 - 10RF100 (c) 1998 Scott Baxter
Knife-Edge Diffraction Sometimes a single well-defined
obstruction blocks the path, introducing additional loss. This calculation is fairly easy and can be used as a manual tool to estimate the effects of individual obstructions.
First calculate the diffraction parameter from the geometry of the path
Next consult the table to obtain the obstruction loss in db
Add this loss to the otherwise-determined path loss to obtain the total path loss.
Other losses such as free space and reflection cancellation still apply, but computed independently for the path as if the obstruction did not exist
H
R1 R2
attendB
0
-5
-10
-15
-20
-25
-4 -3 -2 -1 0 1 2 3-5
= -H2
1 1
R1 R2
July, 1998 4 - 11RF100 (c) 1998 Scott Baxter
Local Variability: Multipath Effects
The free-space, reflection, and diffraction mechanisms described earlier explain signal level variations on a large scale, but other mechanisms introduce small-scale local fading
Slow Fading occurs as the user moves over hundreds of wavelengths due to shadowing by local obstructions
Rapid Fading occurs as signals received from many paths drift into and out of phase
• the fades are roughly /2 apart in space:
7 inches apart at 800 MHz., 3 inches apart at 1900 MHz
• fades also appear in the frequency domain and time domain
• fades are typically 10-15 db deep, occasionally deeper
• Rayleigh distribution is a good model for these fades
these fades are often called “Rayleigh fades”
Multi-path Propagation
A
d
10-15 dB
Rayleigh Fading
July, 1998 4 - 12RF100 (c) 1998 Scott Baxter
Combating Rayleigh Fading: Space Diversity
Fortunately, Rayleigh fades are very short and last a small percentage of the time
Two antennas separated by several wavelengths will not generally experience fades at the same time
“Space Diversity” can be obtained by using two receiving antennas and switching instant-by-instant to whichever is best
Required separation D for good decorrelation is 10-20
• 12-24 ft. @ 800 MHz.
• 5-10 ft. @ 1900 MHz.
Signal received by Antenna 1
Signal received by Antenna 2
Combined Signal
D
July, 1998 4 - 13RF100 (c) 1998 Scott Baxter
Space Diversity Application Limitations
Space Diversity can be applied only on the receiving end of a link.
Transmitting on two antennas would:• fail to produce diversity, since the
two signals combine to produce only one value of signal level at a given point -- no diversity results.
• produce objectionable nulls in the radiation at some angles
Therefore, space diversity is applied only on the “uplink”, i.e.., reverse path
• there isn’t room for two sufficiently separated antennas on a mobile or handheld
Signal received by Antenna 1
Signal received by Antenna 2
Combined Signal
D
July, 1998 4 - 14RF100 (c) 1998 Scott Baxter
Using Polarization DiversityWhere Space Diversity Isn’t Convenient
Sometimes zoning considerations or aesthetics preclude using separate diversity receive antennas
Dual-polarized antenna pairs within a single radome are becoming popular
• Environmental clutter scatters RF energy into all possible polarizations
• Differently polarized antennas receive signals which fade independently
• In urban environments, this is almost as good as separate space diversity
Antenna pair within one radome can be V-H polarized, or diagonally polarized
• Each individual array has its own independent feedline
• Feedlines connected to BTS diversity inputs in the conventional way; TX duplexing OK
Antenna A
Antenna B
Combined
A B A B
V+Hor\+/
July, 1998 4 - 15RF100 (c) 1998 Scott Baxter
The Reciprocity PrincipleDoes it apply to Wireless?
Between two antennas, on the same exact frequency, path loss is the same in both directions
But things aren’t exactly the same in cellular --
• transmit and receive 45 MHz. apart
• antenna: gain/frequency slope?
• different Rayleigh fades up/downlink
• often, different TX & RX antennas
• RX diversity Notice also the noise/interference
environment may be substantially different at the two ends
So, reciprocity holds only in a general sense for cellular
-148.21 db@ 835.03 MHz
-151.86 db@ 870.03 MHz
-148.21 db@ 870.03 MHz
July, 1998 4 - 16RF100 (c) 1998 Scott Baxter
Propagation ModelsPropagation Models
Chapter 4 Section B
July, 1998 4 - 17RF100 (c) 1998 Scott Baxter
Types Of Propagation Models And Their Uses
Simple Analytical models • Used for understanding and
predicting individual paths and specific obstruction cases
General Area models• Primary drivers: statistical• Used for early system
dimensioning (cell counts, etc.) Point-to-Point models
• Primary drivers: analytical• Used for detailed coverage
analysis and cell planning Local Variability models
• Primary drivers: statistical• Characterizes microscopic level
fluctuations in a given locale, confidence-of-service probability
Simple Analytical• Free space (Friis formula)• Reflection cancellation• Knife-edge diffraction
Area• Okumura-Hata• Euro/Cost-231• Walfisch-Betroni/Ikegami
Point-to-Point• Ray Tracing
- Lee’s Method, others• Tech-Note 101• Longley-Rice, Biby-C
Local Variability• Rayleigh Distribution• Normal Distribution• Joint Probability Techniques
Examples of various model types
July, 1998 4 - 18RF100 (c) 1998 Scott Baxter
General Principles Of Area Models
Area models mimic an average path in a defined area
They’re based on measured data alone, with no consideration of individual path features or physical mechanisms
Typical inputs used by model:• Frequency• Distance from transmitter to
receiver• Actual or effective base
station & mobile heights• Average terrain elevation • Morphology correction loss
(Urban, Suburban, Rural, etc.) Results may be quite different
than observed on individual paths in the area
RSSI, dBm
-120
-110
-100
-90
-80
-70
-60
-50
0 3 6 9 12 15 18 21 24 27 30 33
Distance from Cell Site, km
FieldStrength,dBµV/m
+90
+80
+70
+60
+50
+40
+30
+20
Green Trace shows actual measured signal strengths on a drive test radial, as determined by real-world physics.
Red Trace shows the Okumura-Hata prediction for the same radial. The smooth curve is a good “fit” for real data. However, the signal strength at a specific location on the radial may be much higher or much lower than the simple prediction.
July, 1998 4 - 19RF100 (c) 1998 Scott Baxter
The Okumura Model: General Concept
The Okumura model is based on detailed analysis of exhaustive drive-test measurements made in Tokyo and its suburbs during the late 1960’s and early 1970’s. The collected date included measurements on numerous VHF, UHF, and microwave signal sources, both horizontally and vertically polarized, at a wide range of heights.
The measurements were statistically processed and analyzed with respect to almost every imaginable variable. This analysis was distilled into the curves above, showing a median attenuation relative to free space loss Amu (f,d) and correlation factor Garea (f,area), for BS antenna height ht = 200 m and MS antenna height hr = 3 m.
Okumura has served as the basis for high-level design of many existing wireless systems, and has spawned a number of newer models adapted from its basic concepts and numerical parameters.
Med
ian
Att
enu
atio
n A
(f,d
),
dB
1
2
5
40
70
80
100
100 3000500Frequency f, MHz
10
50
70Urban Area
d,
km
30
850
26
35
100 200 300 500 700 1000 2000 3000Frequency f, (MHz)
5
10
15
20
25
30
Co
rrec
tio
n f
acto
r, G
area
(d
B)
9 dB
850 MHz
Open area
Quasi open area
Suburban area
July, 1998 4 - 20RF100 (c) 1998 Scott Baxter
Structure of the Okumura Model
The Okumura Model uses a combination of terms from basic physical mechanisms and arbitrary factors to fit 1960-1970 Tokyo drive test data
Later researchers (HATA, COST231, others) have expressed Okumura’s curves as formulas and automated the computation
Path Loss [dB] = LFS + Amu(f,d) - G(Hb) - G(Hm) - Garea
Free-Space Path Loss Base Station
Height Gain= 20 x Log (Hb/200)
Mobile StationHeight Gain
= 10 x Log (Hm/3)
Amu(f,d) Additional Median Lossfrom Okumura’s Curves
Med
ian
Att
enu
atio
n A
(f,d
), d
B
1
2
5
40
70
80
100
100 3000500
Frequency f, MHz10
50
70Urban Area
d, k
m
30
850
26
Morphology Gain0 dense urban5 urban10 suburban17 rural
35
100 200 300 500 700 1000 2000 3000Frequency f, (MHz)
5
10
15
20
25
30
Co
rrec
tio
n f
acto
r, G
area
(d
B)
850 MHz
Open area
Quasi open area
Suburban area
July, 1998 4 - 21RF100 (c) 1998 Scott Baxter
The Hata Model: General Concept
The Hata model is an empirical formula for propagation loss derived from Okumura’s model, to facilitate automatic calculation.
The propagation loss in an urban area is presented in a simple general format A + B x log R, where A and B are functions of frequency and antenna height, R is distance between BS and MS antennas
The model is applicable to frequencies 100 MHz-1500 MHz, distances 1-20 km, BS antenna heights 30-200 m, MS antenna heights 1-10 m
The model is simplified due to following limitations:
• Isotropic antennas
• Quasi-smooth (not irregular) terrain
• Urban area propagation loss is presented as the standard formula
• Correction equations are used for other areas Although Hata model does not imply path-specific corrections, it has significant
practical value and provide predictions which are very closely comparable with Okumura’s model
July, 1998 4 - 22RF100 (c) 1998 Scott Baxter
Hata Model General Concept and Formulas
Formulas for median path loss are:(1) - Standard formula for urban areas(2) - For suburban areas(3) - For rural areas
Formulas for MS antenna ht. gain correction factor A(hm)(4) - For a small to medium sizes cities(5) and (6) - For large cities
f - carrier frequency, MHz
hb and hm - BS and MS antenna heights, m
d - distance between BS and MS antennas, km
(1) LHATA (urban) [dB] =69.55 + 26.16 x log ( f ) + [ 44.9 - 6.55 x log ( hb ) ] x log ( d ) -13.82 x log ( hb ) - A ( hm )
(2) LHATA (suburban) [dB] = LHATA (urban) - 2 x [ log ( f/28 ) ]2 - 5.4
(3) LHATA (rural) [dB] =LHATA (urban) - 4.78 x [ log ( f ) ]2 - 18.33 x log ( f ) -40.98
(4) A ( hm ) [dB] = [ 11 x log ( f ) - 0.7 ] x hm - [ 1.56 x log ( f ) - 0.8 ]
(5) A ( hm ) [dB] = 8.29 x [ log ( 1.54 x hm ) ]2 - 1.1 (for f<= 300 MHz.)
(6) A ( hm ) [dB] = 3.2 x [ log ( 1175 x hm ) ]2 - 4.97 (for f > 300 MHz.)
Environmental Factor C0 dense urban-5 urban-10 suburban-17 rural
July, 1998 4 - 23RF100 (c) 1998 Scott Baxter
The COST-231 model was developed by European COoperative for Scientific and Technical Research committee. It extends the HATA model to the 1.8-2 GHz. band in anticipation of PCS use.
COST-231 is applicable for frequencies 1500-2000 MHz, distances 1-20 km, BS antenna heights 30-200 m, MS antenna heights 1-10 m
Parameters and variables:• f is carrier frequency , in MHz• hb and hm are BS and MS antenna heights (m)• d is BS and MS separation, in km• A(hm) is MS antenna height correction factor
(same as in Hata model)• Cm is city size correction factor: Cm=0 dB for
suburbs and Cm=3 dB for metropolitan centers
The EURO COST-231 Model
LCOST (urban) [dB] = 46.3 + 33.9 x log ( f ) + [ 44.9 - 6.55 x log ( hb ) ] x log ( d ) + Cm -13.82 x log ( hb ) - A ( hm )
EnvironmentalFactor C1900 -2 dense urban -5 urban-10 suburban-26 rural
July, 1998 4 - 24RF100 (c) 1998 Scott Baxter
Examples of Morphological Zones Suburban: Mix of
residential and business communities. Structures include 1-2 story houses 50 feet apart and 2-5 story shops and offices.
Urban: Urban residential and office areas (Typical structures are 5-10 story buildings, hotels, hospitals, etc.)
Dense Urban: Dense business districts with skyscrapers (10-20 stories and above) and high-rise apartments
Suburban SuburbanSuburban
UrbanUrbanUrban
Dense Urban Dense UrbanDense Urban
Although zone definitions are arbitrary, the examples and definitions illustrated above are typical of practice in North American PCS designs.
July, 1998 4 - 25RF100 (c) 1998 Scott Baxter
Example Morphological Zones
Rural - Highway: Highways near open farm land, large open spaces, and sparsely populated residential areas. Typical structures are 1-2 story houses, barns, etc.
Rural - In-town: Open farm land, large open spaces, and sparsely populated residential areas. Typical structures are 1-2 story houses, barns, etc.SuburbanSuburban
RuralRural
Suburban
Rural
Rural - HighwayRural - HighwayRural - Highway
Notice how different zones may abruptly adjoin one another. In the case immediately above, farm land (rural) adjoins built-up subdivisions (suburban) -- same terrain, but different land use, penetration requirements, and anticipated traffic densities.
July, 1998 4 - 26RF100 (c) 1998 Scott Baxter
The MSI Planet General Model
Pr - received power (dBm)Pt - transmit ERP (dBm)Hb - base station effective antenna heightHm - mobile station effective antenna heightDL - diffraction loss (dB) K1 - intercept K2 - slopeK3 - correction factor for base station antenna height gainK4 - correction factor for diffraction loss (accounts for clutter heights)K5 - Okumura-Hata correction factor for antenna height and distanceK6 - correction factor for mobile station antenna height gainKc - correction factor due to clutter at mobile station locationKo - correction factor for street orientation
Pr = Pt + K1 + k2 log(d) + k3 log(Hb) + K4 DL + K5 log(Hb) log(d)
+ K6 log (Hm) + Kc + Ko
This is the general model format used in MSI’s popular PlaNET propagation prediction software for wireless systems. It includes terms similar to Okumura-Hata and COST-231 models, along with additional terms to include effects of specific obstructions and clutter on specific paths in the mobile environment.
July, 1998 4 - 27RF100 (c) 1998 Scott Baxter
Typical Model Results Including Environmental Correction
-2-5-10-26
TowerHeight,
m
EIRP(watts)
C,dB
Range,kmf =1900 MHz.
Dense Urban Urban
Suburban Rural
30303050
200200200200
2.523.504.8
10.3
COST-231/Hata
f = 870 MHz.
Okumura/Hata
0-5-10-17
TowerHeight,
m
EIRP(watts)
C,dB
Range,km
Dense Urban Urban
Suburban Rural
30303050
200200200200
4.04.96.7
26.8
July, 1998 4 - 28RF100 (c) 1998 Scott Baxter
Propagation at 1900 MHz. vs. 800 MHz.
Propagation at 1900 MHz. is similar to 800 MHz., but all effects are more pronounced.
• Reflections are more effective
• Shadows from obstructions are deeper
• Foliage absorption is more attenuative
• Penetration into buildings through openings is more effective, but absorbing materials within buildings and their walls attenuate the signal more severely than at 800 MHz.
The net result of all these effects is to increase the “contrast” of hot and cold signal areas throughout a 1900 MHz. system, compared to what would have been obtained at 800 MHz.
Overall, coverage radius of a 1900 MHz. BTS is approximately two-thirds the distance which would be obtained with the same ERP, same antenna height, at 800 MHz.
July, 1998 4 - 29RF100 (c) 1998 Scott Baxter
Walfisch-Betroni/Walfisch-Ikegami Models
Ordinary Okumura-type models do work in this environment, but the Walfisch models attempt to improve accuracy by exploiting the actual propagation mechanisms involved
Path Loss = LFS + LRT + LMS
LFS = free space path loss (Friis formula)
LRT = rooftop diffraction loss
LMS = multiscreen reflection loss Propagation in built-up portions of cities is
dominated by ray diffraction over the tops of buildings and by ray “channeling” through multiple reflections down the street canyons
-20 dBm-30 dBm-40 dBm-50 dBm-60 dBm-70 dBm-80 dBm-90 dBm-100 dBm-110 dBm-120 dBm
Signal Level
Legend
Area View
July, 1998 4 - 30RF100 (c) 1998 Scott Baxter
Statistical TechniquesDistribution Statistics Concept
An area model predicts signal strength Vs. distance over an area
• This is the “median” or most probable signal strength at every distance from the cell
• The actual signal strength at any real location is determined by local physical effects, and will be higher or lower
• It is feasible to measure the observed median signal strength M and standard deviation
• M and can be applied to find probability of receiving an arbitrary signal level at a given distance
Median Signal Strength
,dB
Occurrences
RSSI
Normal Distribution
RSSI, dBm
Distance
Model is tweaked to produce “Best-Fit” curve
Signal Strength Predicted Vs. Observed
Observed Signal Strength
50% of observeddata is above curve
50% of observeddata is below curve
July, 1998 4 - 31RF100 (c) 1998 Scott Baxter
Statistical TechniquesPractical Application Of Distribution Statistics
General Approach:• Use favorite model to predict Signal
Strength• Analyze measured data, obtain:
– median signal strength M(build histogram
of observed vs. measured data)
– standard deviation of error, (determine from
histogram) • add an extra allowance into model
– drop curve so a desired % of observations are above model predictions Median
Signal Strength
,dB
Occurrences
RSSI
Normal Distributio
n
RSSI, dBm
Distance
25% of locations exceed blue curve
50% exceed red
75% exceed black
SIGNAL STRENGTH vs DISTANCE
Min signal req’d for operation
Cell radius for 75% reliability
at edge
Cell radius for 75% reliability
at edgeCell radius for 90% reliability
at edge
July, 1998 4 - 32RF100 (c) 1998 Scott Baxter
Cell Edge Area Availability And Probability Of Service
Overall probability of service is best close to the BTS, and decreases with increasing distance away from BTS
For overall 90% location probability within cell coverage area, probability will be 75% at cell edge
• Result derived theoretically, confirmed in modeling with propagation tools, and observed from measurements
• True if path loss variations are log-normally distributed around predicted median values, as in mobile environment
• 90%/75% is a commonly-used wireless numerical coverage objective
• Recent publications by Nortel’s Dr. Pete Bernardin describe the relationship between area and edge reliability, and the field measurement techniques necessary to demonstrate an arbitrary degree of coverage reliability
Statistical View ofCell Coverage
Area Availability:90% overall within area
75%at edge of area
90%
75%
July, 1998 4 - 33RF100 (c) 1998 Scott Baxter
Application Of Distribution Statistics: Example Let’s design a cell to deliver at least -95
dBm to at least 75% of the locations at the cell edge (This will provide coverage to 90% of total locations within the cell)
Assume that measurements you have made show a 10 dB standard deviation
On the chart:• To serve 75% of locations at the cell
edge , we must deliver a median signal strength which is .675 times stronger than -95 dBm
• Calculate:- 95 dBm + ( .675 x 10 dB ) = - 88 dBm
• So, design for a median signal strength of -88 dBm!Standard Deviations from
Median (Average) Signal Strength
Cumulative Normal Distribution
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
-3 -2.5 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 2.5 3
75%
0.675
July, 1998 4 - 34RF100 (c) 1998 Scott Baxter
Statistical Techniques:Normal Distribution Graph & Table For Convenient Reference
Cumulative Normal Distribution
Standard Deviation from Mean Signal Strength
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
-3 -2.5 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 2.5 3
CumulativeProbability
0.1%1%5%
10%
StandardDeviation
-3.09-2.32-1.65-1.28-0.84 20%-0.52 30%
0.675 75%
0 50%0.52 70%
0.84 80%1.28 90%1.65 95%2.35 99%3.09 99.9%3.72 99.99%4.27 99.999%
July, 1998 4 - 35RF100 (c) 1998 Scott Baxter
Building PenetrationStatistical Characterization
Statistical techniques are effective against situations that are difficult to characterize analytically
• Many analytical parameters, all highly variable and complex
Building coverage is modeled using existing outdoor path loss plus an additional “building penetration loss”
• Median value estimated/sampled
• Statistical distribution determined
• Standard deviation estimated or measured
• Additional margin allowed in link budget to offset assumed loss
Typical values are shown at left
Building penetration
Typical Penetration Losses, dBcompared to outdoor street level
EnvironmentType
(“morphology”)
MedianLoss,
dB
Std.Dev., dB
Urban Bldg. 15 8
Suburban Bldg. 10 8
Rural Bldg. 10 8
8 4Typical Vehicle
Dense Urban Bldg. 20 8
Vehicle penetration
July, 1998 4 - 36RF100 (c) 1998 Scott Baxter
Composite Probability Of ServiceAdding Multiple Attenuating Mechanisms
For an in-building user, the actual signal level includes regular outdoor path attenuation plus building penetration loss
Both outdoor and penetration losses have their own variabilities with their own standard deviations
The user’s overall composite probability of service must include composite median and standard deviation factors
COMPOSITE = ((OUTDOOR)2+(PENETRATION)2)1/2
LOSSCOMPOSITE = LOSSOUTDOOR+LOSSPENETRATION
Building
Outdoor Loss + Penetration Loss
July, 1998 4 - 37RF100 (c) 1998 Scott Baxter
Composite Probability of ServiceCalculating Fade Margin For Link Budget
Example Case: Outdoor attenuation is 8 dB., and penetration loss is 8 dB. Desired probability of service is 75% at the cell edge
What is the composite ? How much fade margin is required?
Composite Probability of ServiceCalculating Required Fade Margin
EnvironmentType
(“morphology”)MedianLoss,
dB
Std.Dev., dB
Urban Bldg. 15 8
Suburban Bldg. 10 8
Rural Bldg. 10 8
8 4Typical Vehicle
Dense Urban Bldg. 20 8
BuildingPenetration
Out-DoorStd.Dev., dB
8
8
8
8
8
CompositeTotal
AreaAvailabilityTarget, %
90%/75% @edge
90%/75% @edge
90%/75% @edge
90%/75% @edge
90%/75% @edge
FadeMargin
dB
7.6
7.6
7.6
6.0
7.6
COMPOSITE = ((OUTDOOR)2+(PENETRATION)2)1/2
= ((8)2+(8)2)1/2 =(64+64)1/2 =(128)1/2 = 11.31 dB
On cumulative normal distribution curve, 75%
probability is 0.675 above median. Fade Margin required =
(11.31) (0.675) = 7.63 dB. Cumulative Normal Distribution
Standard Deviations from Median (Average) Signal Strength
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
-3 -2.5 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 2.5 3
75%
.675
July, 1998 4 - 38RF100 (c) 1998 Scott Baxter
CommercialPropagation Prediction
Software
CommercialPropagation Prediction
Software
Chapter 4 Section C
July, 1998 4 - 39RF100 (c) 1998 Scott Baxter
Point-To-Point Path-Driven Prediction Models
Use of models based on deterministic methods
• Use of terrain data for construction of path profile• Path analysis (ray tracing) for obstruction, reflection analysis• Appropriate algorithms applied for best emulation of underlying
physics• May include some statistical techniques• Automated point-to-point analysis for enough points to appear to
provide large “area” coverage on raster or radial grid Commonly-used Resources
• Terrain databases
• Morphological/Clutter Databases
• Databases of existing and proposed sites
• Antenna characteristics databases
• Unique user-defined propagation models
July, 1998 4 - 40RF100 (c) 1998 Scott Baxter
Path-Driven Propagation Prediction Tools Data Structure
Geographic “Overlay” Format: Output Map(s) on screen or plotter
• Coverage– field strengths @ probability– probabilities @ field strength
• Best-Server• C/I (Adjacent Channel & Co-
Channel) Cell locations, cell grid Terrain elevation data
• USGS & Commercial databases• Satellite or aerial photography
Clutter data• Roads, rivers, railroads, etc.• State, county, MTA, BTA boundaries
Traffic density overlay Land use overlay
July, 1998 4 - 41RF100 (c) 1998 Scott Baxter
The World as “seen” by a Propagation Prediction Tool
Propagation tools use a terrain database, clutter data for land use, and vectors to represent features and traffic levels. The figure at right is a 3-D view of such databases in the area of this demonstration. Notice the granularity of the data and the very mild terrain undulations in the area, exaggerated 8 times in this view.
July, 1998 4 - 42RF100 (c) 1998 Scott Baxter
Survey Of Commercially Available Tools
A wide variety of software tools are available for propagation prediction and system design
Some tools are implemented on PC/DOS/Windows platforms, others on more powerful UNIX platform
Capabilities and user interfaces vary greatly
Several of the better-known tools for cellular RF engineering are shown in the table at right
RF Prediction Software Tools•Qualcomm
•QEDesign CDMA Tool(Unix)
•MSI•PlaNet (Unix)
•LCC•CellCad (Unix)•ANet (DOS PC)
•CNET•Wings (Unix)•Solutions (mainframe)
•ComSearch•IQSignum(Unix)
•AT&T•PACE (DOS PC)
•Motorola•proprietary (Unix)
•TEC Cellular: Wizard (DOS)•Elebra: CONDOR, CELTEC•Virginia Tech MPRG
•SMT-Plus Indoor Site Planning Tool
July, 1998 4 - 43RF100 (c) 1998 Scott Baxter
Composite Coverage Plot A composite coverage plot shows
the overall coverage produced by each sector in the field of view
The color of each pixel corresponds to the signal level of the strongest server at that point
Such plots are useful for identifying coverage holes and overall coverage extent
July, 1998 4 - 44RF100 (c) 1998 Scott Baxter
Equal Power Handoff Boundaries Plot
A Best Server Plot or in CDMA terms, an Equal Power Handoff Boundaries plot paints each pixel with a unique color to identify the best-serving sector at that point
• the boundaries shown are the equal-power points between cells
This type of plot is extremely useful in creating initial neighbor lists and identifying areas of no dominant server
Some tools (MSI Planet) can generate automatic neighbor lists from such a plot
July, 1998 4 - 45RF100 (c) 1998 Scott Baxter
Qualcomm’s QEDesign
Qualcomm’s commercial tool QEDesign offers a number of features targeted at CDMA system design and analysis. The figures above show the output of its microcell propagation analysis tool in the Washington, DC area, and a three-dimensional view of an antenna pattern. Other features of this package include live cursor mode in which the user can drag the cursor about and see in near-real-time the line-of-sight area visible from the selected location, or a coverage footprint calculated from that location.
July, 1998 4 - 46RF100 (c) 1998 Scott Baxter
General Survey Of Tool Features
Universal Basic Features of Most Tools
Automatically calculates signal strength at many points over a geographic area
• Use databases of terrain data, environmental conditions, land use, building “clutter”, estimated geographic traffic distribution, etc.
• User-definable 3-dimensional antenna patterns
• Automatically analyzes paths, selects appropriate algorithms based on path geometry
• Produces plots of coverage, C/I, etc. Used for analysis of sites, interference,
frequency planning, C/I evaluation, etc. Drawback: requires significant computation
power, time and RF staff special training
-20 dBm-30 dBm-40 dBm-50 dBm-60 dBm-70 dBm-80 dBm-90 dBm
-100 dBm-110 dBm-120 dBm
Signal Level
Legend
C/ILegend
>20 dB<20 dB<17 dB<14 dB
July, 1998 4 - 47RF100 (c) 1998 Scott Baxter
General Survey Of Tool Features, Continued
Popular Features of Advanced Tools
Accepts measurement input, can automatically generate predicted-vs-measured statistics and map displays
Automatic hexagon-manipulation grid utility
Maintains cell sites in relational database
• Easy manipulation, import, export Flexible user interface allows
multitasking Allows multiple user-defined
propagation models Three dimensional terrain view Roads, boundaries, coastline easily
overlaid onto any display
A
A
AA
A A
AA
A AA
A A
A
A
Pred. MeasMean -76 -72Std. Dv 9 12Samples 545 545
Area Name: DALLAS
Site Name
Subs: 100,000
Site # LatitudeLongitudeType Capacity
Number of Sites5 Total Capacity (Erlangs)221
SITE - 1SITE - 2SITE - 3SITE - 4SITE - 5
A1A2A3A4A5
33/17/4633/20/0833/16/5033/10/2833/25/21
96/08/3396/11/4996/12/1496/11/5196/03/53
S322S211S332S1101
77379188
Date: Initial Service
7
8
9
1
3
2
1
3
24
5
6
7
8
9
7
8
9
1
3
2
6
4
6
10
11
July, 1998 4 - 48RF100 (c) 1998 Scott Baxter
General Survey Of Tool Features, Continued
More Popular Advanced Features
Produces plots of server boundaries, C/I plots, handoff boundaries, etc.
Allows interactive change of antenna number, type, orientation, power and tilt
Using growth-scaleable traffic input mask, can predict traffic carried by each site, # channels required
• Can automatically highlight cells not meeting specified grade of service
Algorithms for automatic frequency planning and optimization
User can define or “mask” cells to be changed or unchanged during automatic optimization
43
2
56
17
43
2
56
17
CELL ERL Channels14 8.3 1722 2.1 526X 1.7 426Y 23 3126Z 14 20
July, 1998 4 - 49RF100 (c) 1998 Scott Baxter
General Survey Of Tool Features, Continued
More Popular Advanced Features
Identification of server and interferer signal levels in live cursor mode upon graphical coverage display
Generates bin C/I & coverage statistics for system evaluation
Predicted handoff analysis
• Statistical analysis of most likely handoff candidates
• Automatic generation of neighbor cell lists
• Percentage probability of handover
Runs on powerful workstations to minimize computation time
Cell 51 -82 dBmCell 76 -97 dBmC/I +15 dB
Cell 18Cell 24 48%Cell 16 22%Cell 17 18%Cell 05 8%Cell 22 4%
C/I Pct. of Area>20 dB 93.0%<20 dB 7.0%<17 dB 2.2%
July, 1998 4 - 50RF100 (c) 1998 Scott Baxter
Resolution Of Terrain Databases
Elevation data in terrain databases can be stored in any of several formats:
• Contour vectors: lines of constant elevation in vector segment form, digitized from topographic maps
• Elevation sample points on rectangular grids with fixed spacing
• Elevation sample points on latitude-longitude grids with spacing of a fixed number of arc-seconds
• Data can be converted from one format to another
10m
10m
3 arc-seconds
3 arc-seconds
July, 1998 4 - 51RF100 (c) 1998 Scott Baxter
Resolution Of Terrain Databases, Continued
It is useful to know the horizontal spacing in feet between sample points in a terrain database using arc-seconds, i.e., latitude-longitude spacing
North-South spacing is constant, everywhere on the planet
• 1 arc-second = 101.34 feet
• 1 degree = 69.096 miles East-West sample spacing varies with
the cosine of the North Latitude
• = 101.34 feet/arcsecond at the Equator
• = 0 feet/arcsecond at Poles
• = 101.34 ft. * Cos (N Lat) per arcsecond,
everywhere
N30º
N60º
(North Pole) N90º
(Equator) 0º
S30º
S60º
(South Pole) S90º
Latitude
0º Greenwich, UK
W 30º
W 60º
W 90º
W 120º
Longitude
1sec.
101.34 ft
101.34 ft * Cos (N Latº )
July, 1998 4 - 52RF100 (c) 1998 Scott Baxter
CommercialMeasurement Tools
CommercialMeasurement Tools
Chapter 4 Section D
July, 1998 4 - 53RF100 (c) 1998 Scott Baxter
Propagation Data Collection Philosophy
RF testing of sites is usually performed for one of two reasons: Drive Testing for model calibration
• Prior to cell design of a wireless system, accurate models of propagation in the area must be developed for use by the prediction software. A significant number of typical sites are evaluated using the test transmitter and receiver to determine signal decay rates and to get a fairly accurate understanding of the effects of typical clutter in the area.
• Tests are also conducted to evaluate the additional attenuation which the signal suffers during penetration of typical buildings and vehicles.
• The focus is on developing models generally applicable to the area, not on the performance of specific individual sites.
Drive Testing for site evaluation
• Although propagation models for an area already have been refined, coverage of a particular site is so critical, or its environment so variable due to urban clutter, that it is essential to actually measure the coverage and interference it will produce. The focus is on this specific site.
July, 1998 4 - 54RF100 (c) 1998 Scott Baxter
CW or Modulated Test Signals? Can measurements of unmodulated RF carriers provide adequate
propagation data for system design, or is it advisable to use a modulated RF signal similar to the type which will be radiated by actual BTS in the contemplated system?
• CW (continuous wave, i.e., unmodulated carriers) transmitters are moderately priced ($10K-$25K). CW-only receivers are priced from $5K to over $20K.
• Technology-specific GSM or CDMA modulated test transmitter-receiver systems are available, at costs in the $100,000-$275,000 range per TX-RX system.
Multiple Sites Simultaneously
Multipath Characteristics
Modulated Systems CW Systems
FER, BER statistics
Too expensive!
Delay Spread
Yes
Yes
Usually Not. However, DSP post-processing can yield some multipath data using various transforms. (Not
commercially available yet.)
No
Propagation Loss Mapping Yes Yes
July, 1998 4 - 55RF100 (c) 1998 Scott Baxter
Summary of Available Commercial Tools
Measurement data can be collected manually, but it is simply too tedious to obtain statistically useful quantities by hand
There are many commercial data collection systems available to automate the collection process
Many modern propagation prediction software packages have the capability to import measurement data, compare it with predicted values, and generate statistical outputs (mean error, standard deviation, etc.).
Commercial Measurement Systems•Grayson Electronics:
•Inspector32, Spectrum Tracker•Wireless Measurement Instrument•Handheld Logger
•MLJ, Berkeley Varitronics•CW test transmitters, receivers
•Qualcomm•Mobile Diagnostic Monitor•CDMA test TX-RX & analyzer
•SAFCO•SmartSAM , SmartSAM Plus*, PROMAS*, CDMA OPAS32
•COMARCO•NES-150, NES-250, NES-350
•LCC•RSAT; “Walkabout”, RSAT 2000 w/expansion chassis* TDMA/AMPS, GPS
•ZKSAM - AMPS tools•Rohde & Schwarz: GSM Tools
July, 1998 4 - 56RF100 (c) 1998 Scott Baxter
Elements of Typical Measurement Systems
WirelessReceiver
PC or Collector
GPSReceiver
DeadReckoning
Main Features Field strength measurement
• Accurate collection in real-time• Multi-channel, averaging
capability Location Data Collection Methods:
• Global Positioning System (GPS)
• Dead reckoning on digitized map database using on-board compass and wheel revolutions sensor
• A combination of both methods is recommended for the best results
Ideally, a system should be calibrated in absolute units, not just raw received power level indications
• Record normalized antenna gain, measured line loss
July, 1998 4 - 57RF100 (c) 1998 Scott Baxter
Typical Test Transmitter Operations
Typical Characteristics
• portable, low power needs
• weatherproof or weather resistant
• regulated power output
• frequency-agile: synthesized Operational Concerns
• spectrum coordination and proper authorization to radiate test signal
• antenna unobstructed
• stable AC power
• SAFETY:
– people/equipment falling due to wind, or tripping on obstacles
– electric shock
– damage to rooftop
July, 1998 4 - 58RF100 (c) 1998 Scott Baxter
A Typical Mobile Test Receiver
Receivers and decoders are installed only for the appropriate technologies and frequency bands
Internal GPS or external GPS may be used, with or without dead-reckoning capabilities
Internal GPSReceiver,
if used
Up to 4 technology-specific
decoder boards:AMPS, TDMAGSM, CDMA
Paging
Up to 4 technology andband-specific receivers:800 MHz. cellular150, 450, 800 Paging1900 PCS
Up to 2 handsetsmay be connectedfor GSM or CDMAat 800 or 1900 MHz.
inputs to internal RXs
MainOn/Off
RF toInt. GPS
July, 1998 4 - 59RF100 (c) 1998 Scott Baxter
Selecting and Tuning Propagation Models
Parameters of propagation models must be adjusted for best fit to actual drive-test measured data in the area where the model is applied
The figure at right shows drive-test signal strengths obtained using a test transmitter at an actual test site
Tools automate the process of comparing the measured data with its own predictions, and deriving error statistics
Prediction model parameters then can be “tuned” to minimize observed error
July, 1998 4 - 60RF100 (c) 1998 Scott Baxter
Measured Data vs. Model Predictions
Is the propagation model approximately correct?• Is the data scatter small enough to justify use of a model?• correct slope to match data• correct position up/down on Y-axis?
July, 1998 4 - 61RF100 (c) 1998 Scott Baxter
Analysis of Measured vs. Predicted
Several tools produce histograms showing the distribution of the differences between measured and predicted values
The mean of the difference between predicted and measured is a very important quantity. It should be small (on order of a few dB).
The standard deviation of the difference also should be small. If it is substantially larger than 8 dB., then either:
• the environment is very diverse (perhaps it should be broken into pieces with separate models for better fit) or
• the slope of the model is significantly different than the observed slope of the measurements (review the Sig. vs. Dist. graph)
July, 1998 4 - 62RF100 (c) 1998 Scott Baxter
Displaying Error Distribution by Location
Suppose a major hill blocked the signal in one direction, or the antenna pattern had an unexpected minimum in that direction
This would cause the data in the shadowed region to differ substantially from data in all remaining directions
Some tools can display the error values on a map like the one at right, to provide quick visual evidence for recognizing this type of problem