Professional Documents
Culture Documents
Coverage Planning: Contents
Coverage Planning: Contents
Coverage Planning
Contents :
• Introduction & Definition of Terms
• The reference sensitivity performance
• Radio Frequency Propagation and Power
• Diffraction, Reflection, ducting
• System Interference
• Radio-frequency power
• Radio Wave Propagation Models
• Link budget calculation
• Suitable prediction models for Macro, Micro and Pico-cells
• Base Site Separation
• Location Probability
• Criteria for antenna selection parameters
• Fading
• Noise Figure calculations
• Principles of Planning Tools and their usage
• Measurement Tools supporting Cell Planning
• Effect of Sectorisation on the coverage
• Exercises
The minimum receiver sensitivity levels in UL and DL are defined in GSM 05.05:
3 80 – (<160 ) W 5 – ( < 10 ) W
5 20 – (<40 ) W
6 10 – (< 20 ) W
7 5 – (< 10 ) W
8 2.5 – ( < 5 ) W
Table 2 Maximum output power (before combiner input) for normal BTS / TRX of different
power classes.
TRX power class GSM 900 micro BTS GSM 1800 micro BTS
Table 3 Maximum output power (per carrier, at antenna connector, after all stages of
combining) for micro BTS/ TRX of different power classes.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.
The reference sensitivity performance, as defined in GSM 05.05 for the GSM 900, in
terms of frame erasure, bit error, or residual bit error rates (whichever appropriate) is
specified in the table below, according to the type of channel and the propagation
condition.
GSM 900
Propagation conditions
Type of channel
static TU 50 TU 50 RA 250 HT 100
(no FH) (ideal FH) (no FH) (no FH)
FACCH/H (FER) 0.1 % 6.9 % 6.9 % 5.7 % 10.0 %
SCH (FER) 1% 16 % 16 % 15 % 16 %
class II (RBER) 2% 8% 8% 7% 9%
Note1:
The specification for SDCCH applies also for BCCH, AGCH, PCH, SACCH. The actual
Definition of terms:
EVSIDR: Erased Valid SID frame rate (frames marked with (SID=0) or (SID=1) or ((BFI or
UFI)=1) if a valid SID frame was transmitted)
ESIDR: Erased SID frame rate (frames marked with SID=0 if a valid SID frame was
transmitted).
BER: Bit error rate on the Air-interface. Determined by the value of RX_QUAL.
BFI: Bad Frame Indicator. A parameter within the TRAU frame. The value of the BFI
indicates to the voice decoder if a TRAU frame contains valid data (BFI=0) or not (BFI=1).
Depending on that information, the voice decoder uses or discards a TRAU frame. Note: for
FCCH frames, BFI always equals 1, beacuse they contain signalling data.
SID: Silent descriptor. A 2-bit-long parameter that indicates whether a TRAU frame in the
uplink is a SID frame (SID=2) or a regular speech frame (SID=0). A SID frame contains only
data for the purpose of producing comfort noise but no speech information.
RBER, BFI=0: Residual bit error rate (defined as the ratio of the number of errors detected
over the frames defined as "good" to the number of transmitted bits in the "good" frames).
RBER, (BFI or UFI)=0: Residual bit error rate (defined as the ratio of the number of
errors detected over the frames defined as "reliable" to the number of transmitted bits
in the "reliable" frames).
RBER, SID=2 and (BFI or UFI)=0: Residual bit error rate of those bits in class I
which do not belong to the SID codeword (defined as the ratio of the number of errors
detected over the frames that are defined as "valid SID frames" to the number of
transmitted bits in these frames, under the condition that a valid SID frame was sent).
RBER, SID=1 or SID=2: Residual bit error rate of those bits in class I which do not
belong to the SID codeword (defined as the ratio of the number of errors detected
over the frames that are defined as "valid SID frames" or as "invalid SID frames" to
the number of transmitted bits in these frames, under the condition that a valid SID
frame was sent).
Note 2:
1 <=α<= 1.6. The value of α can be different for each channel condition but must
remain the same for FER and class Ib RBER measurements for the same channel
condition.
Note 3:
FER for CCHs takes into account frames which are signalled as being erroneous (by
the FIRE code, parity bits, or other means) or where the stealing flags are wrongly
interpreted.
NOTE 4:
Ideal FH case assumes perfect decorrelation between bursts. This case may only be
tested if such a decorrelation is ensured in the test. For TU50 (ideal FH), sufficient
decorrelation may be achieved with 4 frequencies spaced over 5 MHz.
occurs when the path is blocked by an object with large dimensions relative to and
sharp irregularities (edges).
Secondary “wavelets” propagate into the shadowed region
3.2 Reflection
3.3 Scattering
Occurs when a wave impinges upon an object with dimensions on the order of λ or
less, causing the reflected energy to spread out or “scatter” in many directions.
Small objects such as street lights, signs, & leaves cause scattering.
Figure 3 Scattering
fm =V
λ
where V(m/s), fm(HZ) and λ (m).
Coherence Time is the time duration over which two received signals have a strong
potential for amplitude correlation.
Two signals arriving with a time separation greater than TC are affected
differently by the channel.
If the symbol rate is greater than 1/TC, the channel will not cause distortion due to
motion
Co-channel interference occurs when there are two or more transmitters within a
cellular system, or even a neighbouring cellular system that are transmitting on the
same frequency (channel). This type of interference is usually generated because
channel sets have been assigned to two cells that are not far enough apart; their
signals are strong enough to cause interference to each other.
Hint:
Co-channel interference , when it occurs, is a by-product of the basic tenet of cellular
system design: frequency reuse.
Though the basic principle of cellular system design is to reuse assigned frequencies
over again throughout a system, it is also very important to ensure that the
interference occurs between identical frequencies (i.e, channel set) . therefore , there
with ensuring that the cell sites are placed far enough apart geographically, the
appropriate power levels must be maintained at cell base station throughout a system
Hint:
Cells that are co channels must never be direct neighbours of each other. Other
An “infant” cellular system can start out in its early developmental stages with no cells
that are cochannel. But as the subscriber base grows, additional capacity is needed
in the system. New cells and radio channels are added until a point is reached where
cochannel interference:
The carrier to interference ration (C/I) is a measure of the desired signal the cell or
mobile phone ‘sees” relative to interfering signals.
Hint:
Ideally, the goal in RF design is to have a C/I ration of 18 dB or better throughout a
cellular system to avoid cochannel interference. The 18 dB level was chosen by the
There should be an 18 dB difference between any given cell and all other cells (and
mobile phones) throughout a cellular system. The carrier to interference ratio is also
known as the signal to interference ratio. The frequency-reuse plan is a tool used to
keep the C/I ratio at the ideal level of 18 dB or better.
Hint:
As more cells are added to a cellular system, a migration occurs from a noise limited
system to an interference limited system. The system has more potential to produce
cochannel interference and/or adjacent channel interference (described below).
Adjacent channel interference is caused by the inability of a mobile phone to filter out
the signals (frequencies) of adjacent channels assigned to side by side cell sites (e.g.
Channel 361 in cell A, channel 362 in cell B, where cells A and B are next to each
other). Adjacent channel interference occurs more frequently un small cell clusters
and heavily used cells.
Good system design can minimize adjacent channel interference temporarily by
preventing adjacent channel assignments in cells that are next to each other.
There are differing views as to why specifications for mobile phones are so poor
concerning their inability to filter out adjacent frequencies.
There are others types of interference that occasionally plague cellular systems. The
most common form of interference, other than cochannel and adjacent channel
interference , is Intermodulation interference (IM). If a cell site is collocated with other
radio-based services, Intermodulation interference may result; competent
engineering practices should overcome this interference .
Intermodulation interference describes the effect of several signals mixing together to
produce an unwanted signal, or even no signal at all; another type of IM is created by
mobile phones themselves. If a customer is on a call in close proximity to a cell site
on the opposite band (A band / B band), the power from all of the radio channels in
the cell can cause the receiver in the mobile phone to overload. When this happens
the result will be a dropped call. This problem is a direct result of mobile phone
manufacturers’ reducing the cost of producing the phones. To resolve this situation,
both cellular carriers operating in the market may have to place cells sites near each
other so that a stronger signal is maintained in the mobile phone. Then, the mobile
phone will not overload and drop the call.
3.8 Ducting
4. Radio-frequency power
Effective Radiated Power (ERP) is determined by multiplying the gain of the antenna
(mobile or base station) times the power delivered to the base of the antenna. ERP is
measured in watts.
Example:
10 W of radio energy directed into a 10 dB gain cellular antenna equals 100W of
ERP.
Note:
Formulas used to determine ERP can be very granular, as they represent
logarithmic relationships.
62 dBm 58 dBm/MHz
Table 4 Base station Transmit Power
levels are known as station class marks. mobile telephones (car installations) have higher
power; they emit more power because they run off car batteries.
For time-varying fields, the differential form of these equations in cgs is:
∇ .E = 4 πρ
1 ∂ B
∇ × E = −
c ∂ t
∇ .B = 0
4 π 1 ∂ E
∇ × B = J +
c c ∂ t
where :
ρ
∇ .E =
ε
°
∂B
∇ × E = −
∂t
∇ .B = 0
∂E
∇ × B = µ J + ε µ
° ° ° ∂t
Where ε0 is the permittivity of free space and µ0 is the permeability of free space.
The Deterministic models are based on simplifying assumption for the general
problem. This can be a mathematical approximation of the original problem (like the
finite difference model). Or it can be a simple model for a special situation of the
general problem (like the knife edge model). Deterministic model can reach a very
high precision, but they suffer from a very high complexity.
The Semi empirical models are a combination of empirical models with deterministic
models for special situations (like knife edge models).
Let us call by PT the Transmitted power and PR received power. The received power
is given by:
P =P ∗ C
R T dn
During the propagation, the Radio wave is affected by attenuation: The Path Loss.
Path Loss:
P
− 10.log R = L = −10.log(C) + 10.n.log(d ) = − A − α.log(d)
PT
P
R = C.d − n
P
T
5.4 Examples
P 2
R = L = λ
P 4πd
T
Where:
C
λ= : Wavelength in vacuum.
f
C= 2.9979.108 m/s: speed of light in vacuum.
F: frequency in MHZ.
D: distance in Km.
2
1 C
L = .
4Π d f
2
1 C C
So L(db) = 10.log . = 20.log( ) + 20.log(f) + 20.log(d)
4πd f 4π
Finally:
2-Ray Models and Fresnel Zones are useful for predicting indoor and microcell
coverage :
• The distance between the transmitter and receiver is often less than the
breakpoint.
• Fresnel zones not practical for distances beyond the breakpoint.
E1: the Electromagnetic field received through the direct path d1.
E2: the Electromagnetic field received through the indirect path d2 (the influence of
the surface “reflexion” is completely neglected.
ET: the total Electromagnetic field received through both the direct and indirect path.
Ee = E0.exp[j2 π ft]
E1 = E0.exp[j2 π f(t-d1/c) ]
E2 = E0.exp[j2 π f(t-d2/c) ]
P= |E|2
(hBS − hMS ) 2
d1 = d 2 + (hBS − hMS ) 2 ≈ d +
2d
d1 = d 2a + d 2b
(hBS + hMS ) 2
d 2 = d 2 + (hBS + hMS ) 2 ≈ d +
2d
hMS .hBS
d 2 − d1 = 2
d
2
λ 2 kh h 2πf
PR ≈ .4.sin MS BS ;k =
4πf d c
kh h
L(db) = 32.44 + 20log(f) + 20log(d) - 6.02 - 20log sin( MS BS )
d
For large (the MS is far away from BS), we can write: d>>>khMShBS,
kh h kh h 2.πfh MS hBS
sin MS BS ≈ MS BS =
d d cd
Finally:
where:
f: frequency in MHz.
d: distance in km.
The model is valid for hBS > 50m and d in the range of km or for LOS microcell
channels in urban areas.
-n
d
PR ≈
d0
d
L(db) = L d0 + 10.n.log( )
d0
Where :
d0: reference distance ca. 1km for macro cells or in the range of 1m -100m for micro
cells; should be always in the far field of the antenna
L d0 : reference path loss; to be measured at the reference distance.
Environment Exponent n
Free space 2
Okumura developed a set of famous curves, often used by other model developers:
• Okumura’s model is in graphical form
• Based on extensive measurements in the Tokyo area
• Model is valid:
over the frequency range150-1920 MHZ.
over distances from 1-100 km.
for an urban environment over quasi-smooth terrain.
The Okumura Hata model assumes a quasi flat surface, i.e. obstacles like buildings
are not explicitly taken into account. Thus the Okumura Hata model is isotropic. The
different types of surfaces (big cities, small cities, suburban and rural) are
distinguished by different correction factors in this model.
Distance d= 1… 20km.
L urbain = 69.55 + 26 .16 .log(f) − 13 .82 .log(h BS ) - d(h MS ) - c + [44.9 - 6.55log(h BS ) ]log( d )
Where:
• f: frequency in MHz.
• d: distance in km.
Application
• Masaharu Hata took graphical data from the Okumura Model and created
some equations for the frequency range 150-1500 MHz.
The Hata Model and its COST-231 extension is one of the most widely used in
cellular/PCS due to it’s simplicity and accuracy.
Where
The major difference between the Okumura Hata model is a modified dependence on
-3 city center
C 2.[log(f/28)]2 + 5.4 suburban areas
For f= 1800MHz, hBS= 30m, hMS= 1,5m, the correction term for the dependence on
hMS can again be neglected. For the other terms of COST Hata model the insertion of
the values serves:
c = -3 city center
The ITU (or CCIR) model was originally developed for radio broadcasting. It is based
on measurements in the UHF and VHF range which are summarized in graphs (ITU-
R 370-7, ) for the field strength.
The different topographic situations are described by the parameters hBSeff and ∆h .
The ITU model describes the radio wave propagation for the ranges
Definition:
hBSeff is the antenna height above the mean elevation of the terrain measured in a
range from 3km to 15 km along the propagation path.
∆h is the mean irregularity of the terrain in the range from 10km to 50 km along
the propagation path, i.e. 90% of the terrain exceed the lower limit and 10% of the
terrain exceed the upper limit of the band defined by ∆h .
The curves for the field strength are given for different hBSeff and ∆h = 50m. The
correction for other values of Dh is given in an additional graph.
Since local effects of the terrain are not taken into account the deviation between
predicted and actual median field strength may reach 20dB for rural areas. In urban
areas this value may be well exceeded.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.
• The Walfisch Ikegami model is valid between 800 and 2,000 MHz and over
distances of 20 m to 5 km.
• Useful for dense urban canyon-style environments where antenna height is lower
than the average building height.
• The Walfisch Ikegami Model includes a diffraction constant and the street width.
• Signals are ‘guided’ along the street, like an urban canyon.
Further parameter:
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. .
With LOS between BS and MS (base station antenna below roof top level):
With NON-LOS:
LNLOS=
L0 Lrts + Lmsd ≤0
• Lrts : roof top to street diffraction and scatter loss: -10 + 0.354.φ, 0<φ<35°
0 hBS≤ ∆h
54, hBS> ∆h
18, hBS> ∆h
kd=
Although designed for BS antennas placed below the mean building height the COST
Walfisch Ikegami model show often considerable inaccuracies.
This is especially true in cities with an irregular building pattern like in historical grown
cities. Also the model was designed for cities on a flat ground. Thus for a hilly surface
the model is not applicable.
This model is based on the assumption that the path loss is correlated with the total
depth B of the building blocks along the propagation path. This results in an extra
contribution to the LOS attenuation:
L = L LOS (d) + α(B)
For both L LOS (d) and α(B) can be read off graphs based on extensive measurements.
This model is not very precise and large errors occur in the following situation:
When the prediction point is on the main street but there is no LOS path.
When the prediction point is in a side street on the same side of the main street
as the BS.
Diffraction models apply for configurations were a large obstacle is in the propagation
path and the obstacle is far away from the transmitter and the receiver, i.e: h>>λ and
h<<d1,d2.
Huygens principle: all points of a wave front can be considered as a source for a
secondary wavelet ⇒sum up the contributions of all wavelets starting in the half
plane above the obstacle.
Phase differences have to be taken into account (constructive and destructive
interferences).
h 2 (d1 + d 2 )
The excess path length: ∆ =
2d1d 2
2π∆ π 2 2(d1 + d 2 )
Phase difference: ϕ = = u with u = h. Fresnel Kirchoff diffraction
λ 2 λd1d 2
parameter.
ED 1+ i − iπ u 2
2 ∫
L D (v ) = −20log( ) = −20 .log( exp( ) du )
E0 2
E0 : field strength obtained by free field propagation without diffraction (and ground effects).
ED : diffracted Field strength.
LD(v)=
Fresnel Zone
Theoretically, propagation can be predicted with the help of a 2-ray model and
Fresnel Zones: An electromagnetic wave front can be divided into zones of
concentric circles, separated by λ/2. These zones are called Fresnel Zones.
1st Fresnel Zone defines a propagation breakpoint, do.
Before do, the propagation is approximately free space propagation.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. .
The attenuation of several obstacles is computed obstacle by obstacle with the single
knife edge method, i.e. first diffraction path:l1I2, second diffraction path: I2I3.
2(d1 + d 2 ) 2(d 2 + d3 )
With: v 1 = h1 . and v 2 = h2 .
λd 1 d 2 λd 2 d 3
≈ 0, v<-0.78,
LD(v)=
This model is rather imprecise. The error grows with the number of obstacles.
First the attenuation of the main obstacle is computed (in this example O1 with the
path I1s1). In the second step the possible (main) obstacles along the paths to and
from the main obstacle are computed (here O2 with l2l3).
This procedure is continued until all obstacles are taken into account.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.
The recursion procedure is the same as for the Deygout model. Instead of taking a
correction term in the attenuation the receiver is considered at an effective position at
a height heff.
2(d1 + d 2 + d3 ) 2(d 2 + d 3 )
v 1 = h 1. and v 2 = h 2 .
λd1 (d2 + d3 ) λd2 d 3
d1 d3
h1 = h1 − .h eff and h eff = h 2 − .(H2 − H1 )
d1 + d 2 + d3 d2
The attenuation predicted by this model is between the values obtained from the
Epstein Peterson model and the Deygout model without the correction term.
2(d1 + d 2 + d3 )
v 1 = h1 .
λd1 (d 2 + d 3 )
2(d 2 + d3 )
v 2 = h2 .
λd 2 d3
Correction term: [
C(O1,O2)= 12−20log( 2 ).( p )2p
1−α/π q
]
2(d1 + d 2 + d 3 ) d (d + d 2 + d3 ) 2(d1 + d 2 + d3 )
p = h1 . , α = arctan( 2 1 , q = h2 .
λd1 (d2 + d3 ) d1d3 λd3 (d2 + d1 )
The correction term is chosen such that the result coincides in a good approximation
with an exact solution. After n steps this models may cover up to 2n-1 obstacles.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. .
Semi empirical model combine deterministic models like knife edge models with
empirical models like Okumura Hata or COST Hata.
The mentioned empirical models are only valid for a quasi flat surface. In combination
with knife edge models they can be extended to hilly surface or a mountain area.
The combination of empirical and deterministic models requires usually additional
correction terms. For the specific combination of models and their correction terms
most user develop their own solution which they calibrate with their measurements. .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
However, there are continuous improvements for hardware, software and algorithms.
With this method a very high precision for the prediction of the path loss can be
obtained:
However, as for the ray launching and ray tracing method, there are continuous
improvements for hardware, software and algorithms.
Link budget tells us the maximum allowable path loss on each link, and which link is
the limiting factor.
This maximum allowable path loss will set our maximum cell size.
Instead of solving for propagation loss in the prediction equations, we can take the
maximum allowable loss from the link budget and calculate the cell radius, d, from
the propagation model.
The link budget is simply a balance sheet of all the gains and losses on a
transmission path. The link budget usually includes a number of product gains/losses
and “margins”
With the extension of cellular services to building interiors, signal penetration to and
from these buildings must be considered. Thus, it is possible to determine unwanted
interference levels, or even consider the possibility of extending the services provided
by exterior microcells to the interior of certain buildings.
possible to determine penetration as a function of the striking angle of the wave and
the average window size.
Uplink Equation
Lpu = Pm + Gm + Gb + Gd - Ld - Lj – Sb
Downlink Equation
Lpd = Pb - Ld - Lj - Ltf + Gb + Gm – Sm
Where:
MS Feeder Loss Lm dB 0 0
Area Classification
Symb. Units UL DL UL DL UL DL
BS Antenna Height Hb m 30 30 45 45 60 60
Feeder Loss per m Lf/f dB/m .0646 .0646 .0646 .0646 .0477 .0477
Body Attenuation Ab dB 2 2 2 2 2 2
Vehicle Attenuation Av dB 0 0 0 0 6 6
Total Path Loss Lpt dB 136.5 141.7 138.5 143.7 144.5 149.8
The lower Lpt value plugs into the empirical models to yield the cell radius
Example of an UL link budget (GSM 900 MHz MS power class 4, BS with tower
mounted amplifier, frequency hopping on, receive diversity used):
33 33 39 dBm
Max. Output power
0 0 -2 dB
Feeder Loss
0 0 +2 dBi
Antenna Gain
MS
Path loss dB
Interference Margin -2 -2 -2 dB
7.1 Macrocells
A macrocell is a large cell that provides coverage over a range of several kilometres.
This definition is rather broad, as it covers many different propagation environments:
rural areas, mountain areas, motorways, suburban residential areas and urban
residential areas. The definition of macrocell still applies even in urban areas with
high building densities that the base station is located above the mean height of the
surrounding buildings, normally on rooftop masts.
Clearly, the application of a single model to such widely differing environments is not
the optimum approach, but it is typical in most commercial applications. It is common
to use empirical models which consider statistical variations according to the
morphology of the terrain. The most commonly chosen is the Okumura-Hata model
with the modifications introduced by COST-231. Normally a term is added to this
formulation in order to account for multiple edge diffraction following some of the
classic models -Epstein-Peterson, Deygout, Giovanelly, etc.- These methods adopt
different approaches to the problem of assessing multiple edges loss on the basis of
the classic expression of Fresnel for a single edge.
Given the statistical nature of this type of methods it is obvious that they require
tuning, i.e. they need to be adapted to the particular environment in which they are
going to be applied. Commercial tools differ in the by-default configuration of the
typical models -it is typical that the model has been optimised to work in the typical
propagation conditions encountered in one given country, that are likely to differ
strongly with respect to another one. Also, different tools provide varying levels of
flexibility for the user to adjust the operation of the model, as fine tuning by the
operator -basing on measurement data- is only facilitated in a few cases.
7.2 Microcells
Microcells are by definition urban cells covered by base stations with antennas
located below the surrounding buildings roof tops. Guided propagation of waves
through the streets is achieved by reflection, corner diffraction, and occasionally over
rooftops. The coverage range is normally less than 1 kilometre.
There are two alternative approaches to solving the electromagnetic problem behind
the coverage prediction in this type of environments. The first entails the use of
empirical methods, in the same line as in macrocells. The accuracy of the results
provided is worse here than in macrocells, due to the smaller range of this cells, in
which statistical averaging is not valid any more. The second approach relies on
physical models that provide accurate results not requiring tuning. The disadvantage
in this case, however, is given by the large computation times.
7.3 PICOCELLS
Picocells are generally smaller than microcells and are located inside buildings. It will
therefore become necessary to have means of radio coverage prediction in interior
environments.
There are two ways of approaching this problem, based on statistical or deterministic
models. The former are based on statistical information regarding the mean
characteristics of the dispersors in environments of this kind, while the latter use
realistic information. Obviously the latter cease to be applicable when the information
on characterization of dispersors is insufficient to give an adequate level of detail.
ITU + - -
Okumura Hata + 0 -
COST Hata + 0 -
Epstein Peterson + + -
Deygout + + -
Giovanelli + + -
Finite difference 0 + +
The following tables show the maximum cell to mobile distances and site separations
(to provide continuous coverage) for a number of environments, calculated using a
theoretical GSM model with actual equipment parameters.
Tri-sectored sites were assumed. In each case the results are given for both outdoor
and indoor quality coverage, assuming a building penetration loss of 17 dB.
Maximum outdoor
2.1 km 5.8 km 6.6 km 16.7 km
quality range
Maximum indoor
0.7 km 1.5 km 2.1 km 5.4 km
quality range
9. Location Probability
As will be discussed later, these values correspond to the following cell area
probabilities:
Band : Choose frequency band for your antenna. Selecting a band will filter the
below choices.
Beam width : Choose a nominal horizontal beam width to narrow down the
selection.
Gain : Choose a nominal gain to further narrow down the selection. Gain is
expressed in dBi.
Name: Choose an antenna by name. The names are on the form family-band-
beamwidth-gain-downtilt-antenna connector.
El-tilt : Choose electrical downtilt. Fixed antenna downtilt will only have one option
here which will be chosen by default. For variable electrical tilt antennas (AEDT), the
available antenna patterns for each measured tilt will be listed.
Frequency : Choose what measurement frequency to use for the chosen antenna.
By default the first frequency in the list of available frequencies will be chosen.
Please note that it is important to choose a frequency for multiband antennas in order
for the coverage calculation to calculate for the right frequency band.
BTS Parameters : The following BTS parameters can be changed; Mechanical tilt,
Antenna Orientation, Antenna Input power, Propagation model, Tilt Method.
Mechanical Tilt : Set the mechanical or /and electrical tilt in whole degrees. Uptilt or
downtilt.
Power : The power setting indicates the power measured in dBm at the antenna
input terminal. The power setting should hence correspond to the base station output
power minus any external combiner and feeder losses in the normal BTS case. The
ERP (Effective Radiated Power) will be power setting plus antenna gain.
Propagation Model
Choose propagation model to use for this option when calculating image in plot
window. Choose the same model for all options unless you deliberately want to
compare models.
11. Fading
Fading occurs on different scales due to different causes.
Fading appears statistically but different fading types obey different probability
distributions.
Propagation models predict only the average value of the receive level.
An extra margin has to be added due the fading effect.
The common question for all fading effects is: how big to choose the margin such that
the received level drops not below a given limit with a specified probability?
• Fast Fading:
(High Doppler Spread) and (Coherence time < symbol period)
• Slow Fading
(Low Doppler spread) and (Coherence time > symbol period)
• Flat Fading
(BW of signal < Coherence BW) and (Delay spread < symbol period)
V V V V2 + V2
f(V ) = R .I ( R1 R ).exp( − R1 R)
R P 0 P 2P
N N N
Where
including
other noise sources like man made noise.
2
For VR1 >> 1 : the Rice distribution can be approximated by a Gauß distribution:
PN
1 (V − V )2
f(V ) = .exp( − R R1 )
R 2πP 2P
N N
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Rayleigh fading is the other important special case of the Ricean fading.
Rayleigh fading describes the situation were there is no dominant path, i.e. a non
LOS situation.
All contribution to the received signal are comparable in strength and arrive
statistically distributed.
V V2
f(V ) = 2 R .exp( − R )
R V 2 V2
R R
Where
1 2 1 P
P0 = VR : the averaged received power: f(P0 ) = .exp( − 0 )
2 P0 P0
Integrated probability for the power to be below a fading marging for a Rayleigh
distribution:
All described types of fast fading have as characteristic length scale the wavelength
of the signals.
Slow fading denote the variation of the local mean signal strength on a longer time
scale.
The most important reason for this effect is the shadowing when a mobile moves
around (e.g. in a city).
Measurements have shown that the variation of the mean receive level is a normal
distribution on a log scale log normal fading.
L(d) = L (d) + X σ
1 (P - P ) 2
X σ (P) = .exp[ − ]
2πσ 2σ 2
Let Pm be a minimal receive level, what is the probability Pr that the receive level is
higher than the minimal receive level, i.e. Pr( Pr(d)>Pm)=?
To compute the probability that the receive level exceeds a certain margin the
Gaussian distribution has to be integrated. This leads to the Q function:
∞ 1 x2 1 z
Q(z) = ∫ .exp( − )dx = .(1 − erf( ))
z 2π 2 2 2
Jake’s formula
Jake’s formula gives a relation for the probability that a certain value Pm at the cell
boundary at radius R is exceeded and the corresponding probability for the whole
cell.
d
P (d) = P − L (d ) + 10.n.log( )
R R 0 d
0
1 1 − 2ab 1 − ab
Pr (Pm ) = 1 − erf(a) + exp( ).1 − erf( )
cell 2 b2 b 2
where
(Pm − P (R))
• a= R .
2σ
10.n.log(e )
• b= .
2σ
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . .
Environment σ (0) σ (i )
LNF LNF
Dense urban 10 dB 9 dB
Urban 8 dB 9 dB
Rural 6 dB 8 dB
To achieve a certain cell edge probability s LNF must be multiplied with a factor given
in the following table:
(Cell edge probability means the probability to have coverage at the border of the
cell)
With the given path loss specifications, its very beneficial to compute the percentage
of area with signal level greater than the threshold level, given a known likelihood of
Where:
The following set of curves show the percentage of useful coverage area, U(g) ,
where γ is a threshold level of the received signal plotted against s/n. If s/n = 4, then
with 75% boundary coverage provides 93% area coverage; with s/n = 6, area
coverage is 82% with 65% boundary coverage, etc.
Integrating the Gaussean distribution function over the whole cell area delivers cell
area probabilities. Some example results are given in the following table:
50 77
75 91
90 97
95 99
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.
Thermal Noise:
Every object which is at a temperature T > 0°K emits electromagnetic waves (thermal
noise). Therefore, electromagnetic noise can be related to a temperature.
P = s * e* A * T4
Noise Factor:
The Noise Factor can be calculated from the Noise Temperature as follows:
Noise Factor = Noise Temperature / 290°K + 1
Noise Figure:
The noise figure is the value of the Noise Factor given in dB:
Noise Figure = 10 * log (Noise Factor)
Conversion table
RS = NF + Nt + C/N
Where :
Noise Figure (NF).
Thermal Noise Power (Nt).
Minimum C/N (Carrier to Noise).
Not much can be done to change Nt or C/N, but NF can be reduced to improve the
receiver sensitivity of the BTS, thereby extending the range of the cell.
A Low-Noise Amplifier (LNA) at the top of a BTS tower ‘manipulates’ the characteristics of
LNAs have a very low Noise Figure, allowing us to lower the receive sensitivity on the
uplink (typically the limiting link).
Feeder length of 50 meters of 7/8-in. air dielectric coaxial cable with 6.06 dB/100
meter loss
The gain of the LNA is 12 dB and the NF is 2 dB
NF1 = 2 dB = 1.58
NF2 = 3.03 dB = 2
G1 = 12 dB= 15.85
NFTotal = 10*log {1.58+((2-1)/15.85)} = 2.15 dB
Therefore, the tower mounted LNA offers a better Total NF, and thus a better receive
sensitivity at the BTS.
Exercise:
What is the maximum outdoor, indoor cell radius and in-car cell radius:
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . .
The cell planning tools require as one input digital map data (which are often based
on paper maps, satellite photos,..). These digital map data should contain information
about, the land usage (so called “Clutter” information), about the height of obstacles
and they should also contain so called vector data (like rivers, streets,…).
Before working with these digital data, some pre-processing of the data may be
required. Some ideas are sketched on the following pages.
Projections
Geodetic datum:
simplified mathematical representation of the size and shape of the earth
1. Local geodetic datum — best approximates the size and the shape of the
particular part of the earth.
2. Geocentric datum — best approximates the size and shape of the earth as a whole
The GPS uses a geocentric datum to express its position because of its global extent.
a. Equirectangular projection.
b. Mercator projection.
c. Lambert‘s cylindrical equal area.
d. Gall‘s sterographic cylindrical.
e. Miller cylindrical projection.
a. Cassini projection.
b. Transverse Mercator.
c. Transverse cylindrical equal area projection.
True along some parallel somewhere between the equator and a pole and distortion
increases away from this standard
True only at their centre point, but generally distortion is worst at the edge of the map
1. Gall‘s projection.
2. Miller projection.
3. Robinson projection.
4. Van der Grinten Projection.
• Remember that e.g. different nations may use different geodetic datum.
• For daily work, try to use the same geodetic datum: in your planning tool(s), for
your GPS systems, and for your paper maps.
14.8 Examples
14.8.1 MACROCELLS
• In the case of the terrain height, the most sensitive calculation is diffraction.
Errors can range from 0.3 to 1 dB per meter of error in the height of the
obstruction. In the case of multiple obstructions the error multiplies .
• In the case of morphology the propagation models incorporate correction
factors for land use, ranging from 0 to 30 dB. Any imprecision in the database
can cause incorrect application of these factors and lead to errors in the
coverage prediction .
14.8.2 MICROCELLS
Precise modelling of propagation conditions in microcellular environments requires
terrain databases with resolutions as fine as 1 m. At the same time, not only must the
resolution be higher, but much more detailed information is needed. It is essential to
have three-dimensional data on the environment and characterization of the
component materials of potential dispersors, such as buildings. All this information is
not available commercially and is therefore costly.
• Coverage planning.
• Capacity planning.
• Frequency planning.
• Link Budget calculations.
• Propagation predictions.
• Propagation model fine tuning.
• Co- and adjacent channel interference analysis.
• Macro, micro cell planning.
• Handling of multi-layer structures.
• Repeater system handling.
• Microwave planning.
Tools using empirical propagation models require map data with less resolution
compared to tools working with deterministic propagation models.
• Digital data with high resolution are required (often very expensive).
• Typical pixel size: 2m x 2m to 10m x 10m.
• Mostly used for big cities only.
Depending not only on the hardware used but also on the algorithms behind the
software, the computational time required by different tools varies significantly.
Planning tools do not run fully automatically but always require some input and an
intelligent and creative usage.
Remember:
• Since propagation models do not necessarily describe exactly the real situation, a
fine tuning of the models is necessary (e.g. clutter data may vary from country to
country).
• This tool tuning should be done in the start phase of the network planning (i.e.
before a detailed plan is performed).
• A test transmitter is located at typical site locations; a test receiver measures the
RX_LEV along predefined measurement routes. These measured values are
taken as input for the tool fine tuning.
Figure 23 Celular Network in an urban area, including the predicted path Loss from each
cell.
The technique of sectorisation is well known, and consists in dividing the cell into
several equal angular sectors, usually 3 or 6. The coverage of a sector is ensured by
a directional antenna, which radiates mostly to the area of that sector and not to the
others. By doing this, the number of interfering Base Stations (BSs), is reduced and
the co-channel interference decreases. The most common sectorisation is the
division into three sectors of 120º each. However, the use of real antennas (i.e., of
nonideal radiation patterns) has an impact in the performance of the network, and
needs to be evaluated.
Sectorisation was implemented by dividing into 3 sectors. For antennas with an ideal
radiation pattern, there is a maximum of three interferers (the axes of the sectors
oriented to 60º, 180º and 300º). When using real antennas, the radiation pattern is
not confined to the 120º sector, and an overlap occurs between adjacent sectors;
hence, there will be a maximum of six interferers, although not all of them reaching
the receiver with the same interfering power.
Some simulations were made with this cell configuration. The use of 60º, 90º and
120º antennas was studied, as well as their influence on the calculation of C/I,
being also compared with the ideal case.
Exercise
Compare the coverage of an omni-cell (antenna gain = 10 dBi) and the coverage of a
three sectorcell configuration (antenna gain 18 dBi).
Trunking gain (especially interesting for those networks having only a few
frequencies).
Omni antennas are usually less bulky than sector antennas.
Suitable in those areas, where the surrounding terrain limits the coverage (before
the maximum omni cell radius is reached).
Exercises
Ex 1.
Consider:
• an extended cell with 100 km cell radius covering a sea area (clutter term: 30 dB),
• a 900 MHz mobile station of power class 4,
• a BS with the GSM minimum receiver sensitivity,
• an (BS) antenna gain of 15 dBi.
Ex 2.
Consider:
For a satellite carrying the BS, what would be the maximum radius for the satellite
orbit.
Ex 3.
How many sites can be saved in principle if TMAs with 6 dB gain are used in the
network? Use typical values and Hata’s propagation formula for calculation.