You are on page 1of 68

CROWN COPYRIGHT 2002

Applications for reproduction should be made to HMSO

ISSUE 1

Removal of discontinuities in the clear-air


model of Recommendation ITU-R P.452

Final Report

This document is the final report of a study conducted into the removal of
discontinuities in the clear-air radio propagation model of ITU-R P.452. The nature of
the principal clear-air models responsible for the discontinuities is examined. Three
methodologies for removal are considered before the adoption of a mathematical
blending technique. Two input papers were submitted into ITU Working Party 3M
following approval by UK Study Group 3. These papers described the background to
the problem and a proposed solution using a four-stage blend of the existing
individual P.452 propagation models. The two input papers are included as annexes
to this report. Simulation runs of the proposed new model changes were compared to
the current model and the results are presented in this report. Attendance of the
WP 3M Meetings and Study Group 3 Meeting in Geneva in May 2002 to support the
proposed revisions to the recommendation were completed successfully. A new
version of P.452 including all of the proposals presented here was agreed by ITU-R
Study Group 3 and is currently undergoing approval by correspondance.
Removal of discontinuities from ITU-R P.452

DOCUMENT CONTROL

Status Issue

Date 13 Aug 2002

Version 1

Client Reference AY 4208 ITT:0492

History First Draft QA review C Cheeseman


Second Draft Review Dave Eden
First Issue

Contributors M F Hamer
RSBG10/pp7
BTexacT Technologies
Adastral Park
Martlesham Heath
IPSWICH
SUFFOLK
IP5 3RE

Tel: +44 1473 642848


Fax: +44 1473 647390
EMail: Malcolm.hamer@bt.com

File reference Final Report Issue 1.doc


Removal of discontinuities from ITU-R P.452

TABLE OF CONTENTS

1 INTRODUCTION 4

2 BACKGROUND 5

3 HISTORICAL PERSPECTIVE 6

4 METHODOLOGY 8
4.1 Single unified model 8
4.2 Bridging model 9
4.3 Blending functions 10

5 SOFTWARE TOOLS USED 17

6 SUBMISSIONS TO WP3M 18
6.1 Removal of Discontinuities, 3M/58-E 18
6.2 Removal of Discontinuities- further proposed changes, 3M/69-E 18

7 SUBSEQUENT RESULTS 20

8 CONCLUSIONS 24

9 RECOMMENDATIONS 25

ANNEX 1 27

ANNEX 2 46
Removal of discontinuities from ITU-R P.452

1 INTRODUCTION

This document is the Final Report of work undertaken by BT for the UK


Radiocommunications Agency into the “Removal of discontinuities in the clear-air
model of Recommendation ITU-R P.452”.

The report examines the nature of the principal clear-air models responsible for
causing the discontinuities when combined into the overall model prediction. The
most significant discontinuities are shown to exist between the line-of-sight and
ducting/layer reflection models. Initially, individual model elements were compared
across these two models in an attempt to find an analytical mixed model solution to
the problems. It was quickly apparent that this approach was not a feasible option in
the allotted time given the innate complexity of the ducting/layer reflection model itself.
Therefore the approach opted for was to blend together the existing model results
from the recommendation at the transitions causing the discontinuities. Several
different marginal changes of input parameters are shown to result in overall model
discontinuities. A series of blends applied in a specific order are shown to remove the
worst-case discontinuities. However, during the testing phase of the initial proposals,
a further unwanted model result was seen. Another blend was proposed to remove
this problem.

The timetable for the work was organised to achieve UK Study Group 3 Meeting
approval of the solution proposed before a submission into international Working
Party 3M proposing a change to the Recommendation. Attendance at the May 2002
Working Party 3M Meeting in Geneva to support a proposed revision of ITU-R P.452
was an integral part of the contracted study. These constraints meant that no more
than about 3 man-months of actual technical work towards a solution could be
undertaken. The remaining time was spent in testing and paper preparation for the
required input meetings.

Two papers describing the proposed solution were prepared for UK Study Group 3.
Each paper provides a detailed description of part of the blending solution required
and has an appendix making specific proposals to alter recommendation ITU-R
P.452. The second of these appendices contains not only the additional changes but
also all of the earlier modifications from the first and should be viewed as the definitive
statement of required changes. Both papers were endorsed, with some editorial
changes, by the UK Study Group and were submitted to International Working Party
3M as input papers, 3M/58 and 3M/69. Annexes 1 and 2, which form the bulk of this
report, contain these two submissions.

At the Working Party Meeting both of the input papers were assigned to Sub Group
3M3, chaired by Dr David Bacon. Drafting Group 3M3B, chaired by the author of this
report, was formed to consider all inputs relating to recommendation ITU-R P.452.
Working Party 3M approved all the proposed changes contained within this report and
produced document 3M/TEMP/69.

The Study Group was held on the 3rd/4th June 2002 and the document containing the
proposed removal of the discontinuities amongst other changes was input as 3/71.
The Study Group agreed the changes to the recommendation and sought approval by
correspondence. Once this process is complete, the major milestones set out within
the initial contract proposal will have been met successfully.

Page 4 of 68
Removal of discontinuities from ITU-R P.452

This final report will not repeat in the main sections the description of the blends and
their implementation contained within the two Annexes. Rather it will concentrate on
the background and preparatory work done to justify the approach adopted. Where
appropriate the two annexes will be introduced and should be read as if they are a
continuation of the main text.

2 BACKGROUND

The clear-air propagation in Recommendation ITU-R P.452 is currently modelled by


the power addition of a number of separate models, each of whose contribution varies
as a function of the characteristics of the propagation path. The propagation path
itself is analysed to belong to one of these three fundamental types: -

• Line-of-sight Path
• Diffraction (including sub-path diffraction) Path
• Transhorizon Path.
Different combinations of four basic propagation mechanisms, which are modelled in
isolation of one another, are assumed to apply to each of these path types. These
mechanisms and their respective models can be characterised as follows: -

• Line-of-sight
• Diffraction
• Ducting/layer reflection
• Tropospheric
Unfortunately these propagation mechanisms do not switch on and off as precisely as
assumed at the boundaries of the path classifications. The true picture is a continuum
of propagation mechanisms that mix in a continuously variable fashion across the
range of path types. To compound the problem still further a number of arbitrary
distance switches have been introduced, particularly into the onset and cessation of
the ducting/layer reflection and diffraction models. Similar, but unrelated parameter
switches, have been introduced within the individual models themselves as they have
matured. For instance, both the over-sea surface duct coupling and terrain roughness
corrections in the ducting/layer reflection models have empirically determined switch
points. However, in nearly all cases, these internal model switches have been
modelled by mathematical functions that, although not continuous in first differential,
are continuous in value.

The challenge is to ensure that both within and across model boundaries the value
being returned is a continuous function of not only continuous parameters such as
frequency, but also discrete parameters such as the path classification. Whilst all of
the path classifications have the potential to exhibit large discontinuities as small input
changes force different classifications, one change has caused more problems than
any other. This is the change from a line-of-sight path to a transhorizon path.

It is this transition, primarily, that has been investigated in this Research Project.

Page 5 of 68
Removal of discontinuities from ITU-R P.452

3 HISTORICAL PERSPECTIVE

One of the main contributory factors to the possibly large discontinuities at the
transition from a line-of-sight path to a transhorizon path is the completely dissimilar
propagation models invoked either side of the transition. Figure 1 shows a schematic
flow chart of the major model elements of both the line-of-sight and ducting/layer
reflection models. This diagram illustrates a number of different features and
concepts that are introduced in subsequent paragraphs.

Both models recognize the need for a small time percentage, p, signal enhancement
but attempt to achieve this in entirely different ways. For its part, the model invoked
for a line-of-sight path is relatively unsophisticated and uses only the one term, Es(p),
to correct for multi-path and focussing effects (layer refraction in the other model).
The expression used has a relatively insensitive distance dependent term only. In
contrast, the ducting/layer reflection model uses a two part term, A(p), for short term
effects. An important point to note, is that one model references the time percentage,
p, to 50.0% time and the other to β % time. This latter reference point is dependent
upon the path location in the world, its roughness and its geometry amongst other
parameters. The time percentage terms in each model only reduce to zero at these
reference time percentages.

At high time percentages, between the variable β % and 50% time, the sub-path
element of the diffraction model also becomes important. With the currently used
Deygout diffraction model, a maximum correction of the order of 12dB is added to the
overall path loss as the path classification moves from marginal line-of-sight (sub-path
diffracted in P.452 terminology) to fully transhorizon. This correction is introduced
gradually as the terrain incursion into the first Fresnel zone increases. At the point
that the terrain obscures the optical line-of-sight path a further transition to the
ducting/layer reflection model ensues.

In contrast to both line-of-sight and diffraction models, the ducting/layer reflection


model is complex by comparison. Firstly three different atmospheric propagation
mechanisms are being treated within the same model. Early experimental work has
shown that surface ducting and both elevated layer reflection and refraction can be
modelled with the one set of equations over the frequency range of the model. Below
the lower model frequency limit of 700MHz, typical duct dimensions are not sufficient
to support the increasing wavelength of the signal and this assumption increasingly
breaks down. If it were not for this fact, the lower frequency limit of the
recommendation could be set at a lower limit below 100MHz, which would go a long
way towards achieving the desired frequency compatibility with ITU-R P.620.

Since the overall approach of both line-of-sight and ducting/layer reflection models is
to calculate a loss at the reference time percentage and then vary this using the actual
time percentage, there is the potential to attempt to find an analytical match. Figure 2
shows the natural form of the two models and the respective reference points. The
ducting/layer reflection model is the solid red curve and its reference point is at the
intersection of the curve and β% time. This intersection point can be moved in the
direction of the arrows by the other term shown in the model in Figure 1. The time
dependent term in the line-of-sight model, Es(p), can only rotate the reference line-of-
sight model value about its intersection with 50% time (one such line, dotted, is shown
for an arbitrary time percentage).

Page 6 of 68
Removal of discontinuities from ITU-R P.452

Line-of-sight KEY:
Lb0(p)
Model Time percentage
reference functions
92.5 20⋅ log( f ) 20⋅ log(d ) Es Ag
Equivalent model
  − d   p  terms
2.61 − exp   ⋅ log  Lba(p)
  10    50.0 

Af Ad Ag Ducting/layer
reflection Model
20⋅ log( f ) 20 ⋅ log(d lt + d lr )
Transfer
102.45 Ast Asr Act Acr γ d ⋅θ ’
function
A(p)
x=r
x=t
Γ
Y ? N Y ? N 5.0 E − 5 ⋅ ae 3 f  p  p
− 12 + (1.2 + 3.7 E − 3d ) ⋅ log  12 ⋅  
θ’’x > 0 β β 

(
20 log 1 + 0.361⋅θ ’’x ⋅ )
f ⋅ d lx + 0.264θ ’’x ⋅ 3 f 0.0
1 .0 E − 3
+ θt +θ r
ae
β 0 ⋅ µ 2 ⋅ µ3
ω >=0.75; dcx<dlx; dcx<5.0 x=t x=r
θx <= 0.1dlx
(
− 3 exp − 0.25d cx ⋅ [1 + tanh(0.07(50.0 − hxs ))]
2
) 0.0 θx Y ? N 0 . 1 ⋅ d lx  500 d2 
α

 ⋅ 
( )
Y ? N
 ae hte + hre 
2
hm > 10

exp[− 4.6 E − 5 ⋅ (hm − 10)(43 + 6d i )] 1.0


1.076
(2.0058− log(β ))1.012
(
⋅ exp 9.51 − 4.8 log(β ) + 0.198(log(β )) ⋅1E − 6 ⋅ d 1.13
2
) − 0.6 − ε ⋅1 E − 9 ⋅ d 3.1τ
min(d − d lt − d lr ,40)

Figure 1. Schematic comparison of the ITU-R P.452 line-of-sight and ducting/layer reflection models

Page 7 of 68
Removal of discontinuities from ITU-R P.452

90
β LOS50

100

110 Es terms

120
Loss (dB)

130
A(p) terms

140

150 Reference
Transmission
Loss
160

170
0.1 1 10 100
Time %

Figure 2. Schematic showing the time percentage reference points in the line-of-
sight and ducting/layer reflection models.

4 METHODOLOGY

Several possible approaches to reducing the discontinuities at the line-of-sight to


transhorizon transition were considered. Figure 1 helps to shape this decision by
making transparent the extent of the work required for each of the choices. However,
each choice must also be assessed for side effects with the other empirical switches
contained within the model and interactions with the other models that have not been
considered at this stage:

4.1 Single unified model

Perhaps the most elegant solution would be a single unified model to describe all
propagation mechanisms over all parameter space. Given the disparity between the
existing models and exhaustive testing requirements, this appeared to be totally
unattainable within the time constraints of the project. For this reason, it was given no
further consideration.

Page 8 of 68
Removal of discontinuities from ITU-R P.452

4.2 Bridging model

Another solution is to introduce a new model to use in the transition region between
the line-of-sight and ducting/layer reflection models, with perhaps, a new path
classification to define the extent of this region. The new model would need to take
the most applicable elements from each of the models on either side of the transition
and be formulated in such a way that it gives a result that exactly matches the other
two at the two new transition points. In order to understand the viability of this
approach, the line-of-sight and ducting/layer reflection models were examined for
commonality.

Figure 1 illustrates those terms that were identified as having some degree of direct
equivalence in the two models. If, by considering each model at its respective
reference time percentage to ignore the time percentage terms for the moment, the
two models reduce to:

Lb 0 (50.0) = 92.5 + 20 log( f ) + 20 log(d ) + Ag (1)

Lba (β ) = 102.45 + 20 log( f ) + 20 log(d lt + d lr ) + Ast + Asr + Act + Acr + Ag + γ d ⋅θ ’ (2)

Transfer
region
coupling
energy

dlt dlr

6dB 6dB
coupling coupling
loss loss

Ducting region

Figure 3a & b. Schematic diagram of the ducting/layer reflection models.

Page 9 of 68
Removal of discontinuities from ITU-R P.452

The log(f) and gaseous absorption terms, Ag, will immediately cancel out to leave a
simplified expression for both equation (1) and (2). Putting aside the site shielding
terms, Ast and Asr and the surface duct coupling terms, Act and Acr for the moment, the
angular correction term, γdθ’, must equate to the changed distance term plus the
changed constant. Figure 3 is a pictorial view of this and indicates why the angular
correction term can be interpreted as a transfer function bridging the portion of the
transhorizon path between the respective terminal horizon distances, dlt and dlr. In
addition the difference between the constants is approximately 12dB, which can be
thought of as the sum of the coupling loss into and out of the ducting transfer region.
It is not coincidental that the surface duct coupling for terminals placed at the coast
peaks at approximately -6dB. Thus in this special case where the coupling into the
duct is maximised, the coupling loss into the duct transfer region is cancelled out and
is considered to be close to zero.

Thus, at the reference points of the two models, the primary terms can be matched
fairly well. The difficulty with this approach arises with the reintroduction of the time
percentage or secondary terms into the simplified schematic. There is little or no
commonality between the Es(p) and A(p) terms in the two models. The reference
point in the line-of-sight model is independent of the global position and type of path
whereas that in the ducting/layer reflection model has a dependency upon virtual
every parameter that is used to describe the path. Figure 2 illustrates the difficulty of
finding an analytical match to bridge these two reference points when one is moving in
both the x and y directions in response to every change in the path.

Although not completely ruling out the possibility of finding a bridging function, the
timescales for this project were insufficient to countenance it. In addition, should such
an equation set be found, it would need further revision if either of the two models
were to change. This is tantamount to locking the recommendation into its current
models, which is undesirable. The third approach is much more pragmatic. A simple
acceptance of the existing model results, well within the boundary conditions under
which they were conceived, is coupled with a decision as to which is still correct in the
transition region. A mathematical blend is then performed to merge one model into
the other.

4.3 Blending functions

Although not as self-contained as other approaches, modification to either or both of


the models to blend the two together as the transition approaches was considered the
most likely to succeed in the time allowed. The resulting model was expected to
exhibit the least changes from the status quo and therefore require less testing and be
more likely to gain acceptance of the international community. Since the line-of-sight
model is so much simpler than the ducting/layer reflection model, it is proposed to
incorporate some of the elements of the line-of-sight equations into the ducting/layer
reflection model with the aim of blending one set into the other. The preceding
section and private conversations with M T Hewitt, one of the main contributors to the
model, indicated that an embedded free space component exists within the
ducting/layer reflection model. This was considered to significantly enhance the
chance of a successful blend.

As a preliminary step to finding a blend, the range of values of small time percentage
enhancements found within the ducting/layer reflection model needs to be explored. It
was also necessary to quantify a marginal line-of-sight path in terms of a readily found

Page 10 of 68
Removal of discontinuities from ITU-R P.452

parameter that exhibits consistent behaviour on either side of the transition. Ideally
this parameter will already be in use elsewhere within the model.

4.3.1 Matching the models

Figure 4 illustrates the line-of-sight model enhancements as a function of time


percentage with distance as a parameter.

LOS50

5
- Enhancement (dB)

Path distance:
Reaches a limit
at 20 to 30 km
10

15

20. 3
1 10 0.01 0.1 1 10 100
Tim e Per centage

Figure 4 Characteristic behaviour of the line-of-sight model enhancement. The


parameter is path length.

A minimum distance is required is required before signal phasing can cause multi-
path. In practice, measurements confirmed that it was fully developed by 8-10km.
Beyond 20 to 30km the lines stagnate and their behaviour becomes less certain. As
the path length increases, however, there is an increasing probability that it will
become transhorizon at some point, at which time the ducting/layer reflection model
enhancements will replace those of Figure 4. Figure 5 is a schematic view of the
enhancements from the two models assuming that the path goes from line-of-sight to
transhorizon between 10 and 30 km. Below 10km, the lines are a re-arrangement of
those of Figure 4. Above 30km, the enhancements are calculated from the
ducting/layer reflection model and can have a wide spread of values dependent upon
all of the path parameters. Between these distances the lines must join through the
crossover region where neither model is adequately describing the physics of the
situation. It is this undefined region that any blend must bridge. In mathematical
terms of the input parameters on either side of the divide it becomes a one to many

Page 11 of 68
Removal of discontinuities from ITU-R P.452

mapping, since path length is the only input parameter on the left and there are at
least ten parameters needed on the right side.

Line-of-sight Transhorizon

10

10 30

8
Enhancement (dB)

0
1 10 100
Distance (km)

Figure 5 Schematic of the enhancements of the line-of-sight and ducting/layer


reflection model enhancements with distance. The parameter is time
percentage

Because of this huge variety of scenarios that can be conceived for the transition
region, the bridging blend needed to be highly flexible. Figure 6 shows the relative
magnitude of the enhancement terms of the two models either side of the point the
path becomes marginal line-of-sight for some very simple path profiles where only the
zone type is changed. These notional values, which should be the same if no
discontinuities are to occur, are plotted as a function of increasing path distance with
time percentage as a parameter. These calculations are made using a flat earth path
profile with both terminals of equal height and adjusted at each point to give the
marginal line-of-sight condition. There is only one family curves for the line-of-sight
model since the enhancement equation is not a function of zone type. The marked
step seen at 60km in the zone 1 ducting/layer reflection model results is caused by
the tanh function in the expression for Act and Acr.

Page 12 of 68
Removal of discontinuities from ITU-R P.452

20

10

0
Enhancement (dB)

10

20

30

40
1 .10
3
1 10 100
Distance (km)
Line-of-sight model
Ducting model (zone 1)
Ducting Model (zone 4)

Figure 6 Actual calculations of the enhancements of the line-of-sight and


ducting/layer reflection model enhancements with distance. The
parameter is time percentage. The path profiles are all flat earth and the
terminal heights are adjusted at each point for the marginal line-of-sight
condition

One difficulty with all of this approach is the variety of scenarios that can be conceived
for the transition. Transhorizon paths either have separate or a single common radio
horizon from each terminal. The path can be over land, sea or a mixture of the two
and thus all zone classifications can be present. In the first instance, it is reasonable
to assume that as the obstacle heights are reduced in a transhorizon path, that the
double horizon will go first to a common horizon before becoming sub-path diffraction
and finally line-of-sight. It was therefore decided to concentrate upon the transition
from line-of-sight to a common horizon transhorizon path in the first instance. Initially
this was studied for the flat earth single zone situation before introduction of the
terrain and zone perturbations.

4.3.2 The choice of transition parameter

In order for any blend function to work at the transition from transhorizon to line-of-
sight path, a parameter that is proportional to the proximity of the marginal line-of-
sight condition is required. This parameter must be readily accessible for any path
type. The path profile is already analysed for a variety of parameters within the
current recommendation. The search for a useful parameter should start with one of
these. Fortunately, there is a parameter amongst this set that varies in a monotonic

Page 13 of 68
Removal of discontinuities from ITU-R P.452

way with the proximity of the path to the marginal line-of-sight condition. The path
angular distance, θ, has the necessary qualities (Figure 7).
FIGURE 7
Example of a (trans-horizon) path profile

ith terrain point

Interfering station (T) Interfered-with


Mean sea level hl station (R)
θt
d
θr
d lt
htg di d lr
hgt h rg
hts hgr
hrs
·a
ae = k 50

Note 1 – The value of θt as drawn will be negative.


0452-07

Figure 7 A diagram taken from P.452 showing the Path Angular Distance
parameter, θ, in relationship to particularly the remaining angle
parameters.

Figure 8 shows how the path angular distance changes as a path changes from
transhorizon, through marginal line-of-sight to sub-path diffracted. The path angular
distance alters from a positive value, through zero to a negative value as the path type
alters. In mathematical terms at the grazing condition, the sum of the two horizon
angles is negative and exactly opposite the angle, ψ, subtended by the path at the
earth’s centre. One end of the blend is thus clearly defined; when the path angular
distance goes to zero, the path becomes marginally line-of-sight and the line-of-sight
model should apply. The other end of the blend is more subjective. The value at the
end point is still defined as the ducting/layer reflection model result but at what angle
should this be reached. An investigation over a range of path distances centred on
20km indicated an angle of 0.6mrad gave sensible transitions.

Page 14 of 68
Removal of discontinuities from ITU-R P.452

Local
θ
horizontal
θt θr

Transhorizon
path

Local
horizontal
θt θr

θ=0

Marginal line-of-sight
path

Local
horizontal
θr
θt

θ
First Fresnel
Zone

Sub-path
Diffracted

Figure 8 Relationship of the path angular distance, θ, to the other angles.

Page 15 of 68
Removal of discontinuities from ITU-R P.452

4.3.3 Maximum signal enhancements

Although a blend based upon the path angular distance will, in principal, eliminate the
discontinuity at the line-of-sight to ducting/layer reflection model boundary, some
additional calculations need to be performed first for two reasons:

• The ducting/layer reflection model result can exceed the line-of-sight model
results by up to 10dB in some circumstances. In considering this effect, it
follows that the signal enhancement cannot have different maximum
magnitudes in the two models otherwise this will be a contributory factor to the
discontinuity at the marginal line-of-sight condition. However, it is very difficult
to envisage a physical scenario, particularly at time percentages below β %,
where an obstructed path is producing a higher signal enhancement than its
notional line-of-sight equivalent. Thus an important assumption is introduced
into the blending process.

The notional equivalent line-of-sight path will have a lower signal


enhancement loss than a similarly partially obstructed one even at small
time percentages.

The simplest way to achieve this is to compare the ducting/layer reflection


model result to the notional line-of-sight model result for a similar length path
and, if it is higher, clamp it to the line-of-sight model result. However, at time
percentages exceeding β% another correction must also be made.

• Once the time percentage exceeds the reference point β %, the sub-path
diffraction model cannot be ignored. A mixture of both anomalous short term
and bulk atmospheric refraction effects occur and any blend has to allow for
this effect.

There are several radio meteorological parameters that have now been
mapped globally that could be considered for use in assessing anomalous
propagation. Not all of these have been available, however, since the
inception of the ducting/layer reflection model. Initially three atmospheric
parameters were used in the model but one, N0, the surface atmospheric
refractivity, has been phased out. The two remaining parameters, ∆N and β 0
address rather different aspects. ∆N, the net refractivity lapse rate through the
lowest 1km of atmosphere applies to a uniformly mixed atmosphere and is
used to derive the standard k factor. The other parameter, β 0, is the
percentage of time that the lapse rate through the first 150m of atmosphere
exceeds 100 N-units/km. This is used as a relative indicator. The derivation of
this parameter has been complicated by some significant advances in
meteorological equipment over a number of years. Original values of β 0 derive
from old Bean maps, which were constructed from slow acting radiosondes.
Studies comparing old and new radiosondes indicate that a calibration factor of
approximately 1.57 is required between the two sets.

It is important to understand the significance of β 0. It only indicates the


percentage of time during which stratification is likely to be present. No
knowledge of how many layers are present or whether they are surface or
elevated can be obtained. The ducting/layer reflection model uses the
standard k factor, derived from the ∆N value, for the centre location of the path

Page 16 of 68
Removal of discontinuities from ITU-R P.452

and the anomalous propagation constant β 0 to derive a reference time


percentage, β. Since a mechanism for blending the bulk atmospheric effects
with the anomalous effects already exists within the present diffraction model,
it would appear equally reasonable to apply the same procedure to set the
maximum enhancement allowed. Thus below β % time, the full line-of-sight
model enhancement can be assumed. In the region between β % time and
50% time, the same inverse cumulative normal curve that is used in the
diffraction model can be applied to smoothly blend out the sub path loss that
otherwise would be present for a marginal line-of-sight path.

5 SOFTWARE TOOLS USED

Most of the preliminary investigation into the discontinuities was conducted using
MathSoft MathCAD v2001 program. Where appropriate, the program was
supported by user written dynamic link libraries, DLL’s, coded using Borland C++
Builder v5.0. The derivation of a path profile from the Globe v1 worldwide DEM
database and the ITU-R P.452 model calls themselves were two such places. Thus
any length path profile can be imported into a MathCad program by making the
appropriate function call with the latitude/longitude of the terminals as parameters.

The implementation of version 10 of the P.452 model was made using a C++ class.
The MathCad DLL provided the wrapper functionality to translate the MathCad data
structures into standard C++ variable types. In order to return a model loss back to
MathCad, a two-stage process was adopted. The first function call from MathCad
consisted of a function that passed in most of the input parameters excepting time
percentage. An instance of the model was then created within the DLL by a call to the
appropriate C constructor. All the necessary supporting path and other analyses were
then made and stored within appropriate class members. The second stage was then
a call to either the full model loss or one of the individual constituent model losses
passing in the time percentage parameter.

This approach gave several advantages. Firstly the MathCad documents examining
particular scenarios were uncluttered by the complex implementation details of the
path analyses and model equations themselves giving compact documents. Secondly
one instance call of the model could be re-used with different P.452 model calls giving
great flexibility. A third benefit was the ability to focus directly on the problem and not
unwanted implementation details. Finally, the C++ class approach allowed the
proposed model changes to be implemented in a derived class of the basic model.
Those functions where the changes occurred were coded as overridden base class
functions. Thus by defining a base class model pointer, the original or proposed
P.452 model could easily be called by merely invoking either the base or derived class
constructor. All of this detail is hidden away in the DLL from the MathCad documents.

The final testing of the new model involved running many thousands of P.452 paths
over a range of different terrains. Again a hybrid approach was adopted which
exploited the individual strengths of the tools used. A C++ simulator was
implemented within the same DLL as the P.452 models. This could make as many
calls as desired (typically 25000) to the P.452 model with random latitude/longitude
inputs. The size of the terrain area, degree of randomness and many other controlling
parameters were passed from MathCad through the interface DLL. Simulation run

Page 17 of 68
Removal of discontinuities from ITU-R P.452

results were written out to a series of files for later analysis by other MathCad
programs.

6 SUBMISSIONS TO WP3M
It is pertinent to examine the the two submissions made to WP3M at this stage.
These can be found in Annexes 1 and 2. Sections 6.1 and 6.2 provide relevant
background material for each annex respectively and should be read prior to each.

6.1 Removal of Discontinuities, 3M/58-E

A limited and substantially summarised amount of the background material covered to


date is included as an introduction to the submission. However, the blending
technique was chosen without the extensive discussion presented in section 4 above.
The specific model switching causing the problems are detailed in sections 2 through
to 5 of the annex for a simple synthetic path profile provided in Figure 1 of the annex.
Whilst this is a stylised profile, it is highly plausible in terms of the major obstacles
employed. Example path losses are provided.

Section 6 of the annex gives a revised procedure for combining the models together
with the necessary equations. Although the angular blend is the fundamental
procedure employed, the additional precursor blending described in section 4.3.3
above is introduced first in order to facilitate the introduction of the necessary
equations in the correct implementation order. Again examples of the precursor
blending are given in figures 4 & 5 of the annex. Finally within the section, an
example of the angular blend is given for a very simple synthetic path profile. This
was used to show clearly the relationship between the path angular distance, θ, and
the other geometric angles involved. Whilst the terminal heights are easily calculated
by analytic means for such simple profiles, this is not the case for other real profiles.
A series of solve blocks were coded within MathCad to iterate to the terminal heights
required to give a specific value of path angular distance, θ. These functions could be
made to reach the constraint by adjusting the terminal heights either symmetrically,
i.e. both together, or asymmetrically, i.e. fixing the receiver height to some value and
adjusting only the transmitter.

The solve blocks were used in section 7 of the annex to show the initial results from
applying the blends described. The initial results showed the effect of the blends
upon the discontinuities given in section 2 of the paper.

In the conclusion of the submission it stated that, although specific examples were
encouraging, a much wider statistical variety of path profiles needed to be tested

6.2 Removal of Discontinuities- further proposed changes, 3M/69-E

It was intended that the second submission to ITU WP3M would be a series of
simulation runs showing the effects of the blends upon some widely varying paths.
This would become the hard evidence that the proposed changes were not
detrimental to the model overall. In the event, initial testing highlighted that in any

Page 18 of 68
Removal of discontinuities from ITU-R P.452

random selection of paths, the line-of-sight to ducting/layer reflection model switching


does not occur that frequently in comparison to one of the other switches that had to a
large extent been overlooked in the first paper. In concentrating upon marginal line of
sight paths it had been shown that the arbitrary removal of the 20km lower limit for the
inclusion of the ducting/layer reflection model calculation was not a problem since the
new blends effectively ignored the ducting/layer reflection model result. The first full-
scale testing showed that this assumption was not valid for transhorizon path profiles
with path angular distances in excess of 2mrad. Thus transhorizon paths that were
far from being close to the marginal line-of-sight boundary showed very large model
loss discontinuities at the 20 km distance boundary. Thus the focus of the second
paper changed from one of pure test results to a further proposed change plus test
results.

Figure 1 of annex 2 shows one of many typical simulation runs made in a terrain block
which covered most of mainland Europe. This area was selected to encompass as
many types of terrain as possible so as not to bias the results. The path lengths in
any one run were fixed and initially concentrated around the value of 20km where the
ducting/layer reflection model was switched on for transhorizon paths. Time permitted
a succession of different terminal heights to be examined and Table 1 in annex 2
analyses the results into numbers of paths in certain categories. These categories
were chosen to distinguish the paths which were predominately affected by the model
changes proposed in appendix1 to annex 1 and those now showing large
ducting/layer reflection model discontinuities at 20km. In all cases, many more paths
full into the second rather than first category. Figures 3 and 5 illustrate these
categories as histograms.

The fundamental cause of the discontinuity at 20km is related to which models are
combined within the overall model result. Near to the marginal line-of-sight condition,
although still present, the effects of removing the 20km switch are masked by the
changes already proposed in annex 1. For a transhorizon path, the overall P.452
model result for paths exceeding 20km long is currently obtained by blending the
result of the diffraction and ducting/layer reflection models. Below 20km the
ducting/layer reflection model is deemed to be inaccurate leaving only the result of the
diffraction model. If the switch at 20km were to be removed, then the overall model
result below 20km will also be the combination the two models.

It is important to understand the comparisons made in the first half of annex 2. The
comparison in each case is between the result of the proposed change to that of the
current P.452 model. This will always show some degree of discontinuity below
20km. Potentially there is no discontinuity in results between paths either side of
20km if the ducting/layer reflection model switch is simply removed but this
presupposes that the ducting/layer reflection model is both accurate and stable below
20km. The ducting/layer reflection model accuracy is certainly not guaranteed below
20km but this will not matter if it is always higher in value than the diffraction model
result and stable since the combined value will be the lower of the two. One of the
reasons for examining in detail some of the paths that show large discontinuities is to
assess these very points.

It is obvious from the limited number of individual cases studied that the ducting/layer
reflection model loss is not always higher than the diffraction result. Indeed the
opposite conclusion is just as likely. This implies that the combined result will be
dominated by the ducting/layer reflection model result below 20km if the switch is
removed. With regard to stability, a clearer picture results. For both very uniform

Page 19 of 68
Removal of discontinuities from ITU-R P.452

synthetic profiles and real path profiles the variation of the ducting/layer reflection
model result from 20 down to 2 km is not very marked once terrain profile point ‘ noise‘
is taken out. The exact details of how this calculation is made are in the annex. This
has important implications in that, although the result is not usable itself, it can be
safely used in any algorithm without the fear that the overall result will be unstable.

Having judged the ducting/layer reflection model result to be unreliable for distances
below 20km, can the same be said of the diffraction model result? Below β 0% time, it
is assumed that the atmospheric stratification resulting in anomalous propagation will
effectively bypass the bulk atmospheric effects that are addressed in the diffraction
model. Hence the diffraction model result calculated at β 0% time is left unchanged for
all lower time percentages. However, there is no validated anomalous short-range
model available for transhorizon paths. In addition, the latest version of P.452 uses a
diffraction model that is based upon selection of the three most significant profile
points of the terrain. At longer distances, with large numbers of profile samples, this
approach has been shown to give low overall differences between model and data.
However, as the number of profile samples decreases on shorter paths, the technique
becomes less effective and wide variations in the result can be expected as each
profile point is removed. In the absence of any other models, at very short distances,
the only model result that can be used is the diffraction one and well above 20km the
ducting/layer reflection model is the more credible.

Given these two extremes, the proposal is to blend these two values over a distance
range from 5 to 40 km with equal proportions occurring at 20km. A modification of the
same hyperbolic tangent function used in annex 1 for the angle blend can be used for
this distance blend. Figure 12 of annex 2 illustrates the histograms for two simulation
runs for paths of 19.99 and 20.01km. Again a comparison with the current P.452
model is made in each case. Whilst the histograms of the model losses of the
transhorizon paths including the two new distance blends at each of these distances
is as expected virtually identical, the histograms of the model losses of the current
P.452 transhorizon paths lie symmetrically on either side.

Since there are now two seemingly fundamental blends, for angle and distance, is the
order in which they are implemented important? Experiment has shown that it is
important that the angle blend is the last one made. If the order is reversed, there is
no longer a continuum of values between the line-of-sight model result and the onset
of the transhorizon results.

Finally, it is very important to re-iterate a part of the conclusion of the annex. The
approach taken throughout these two annexes has been a pragmatic solution to a
problem of discontinuities within the existing overall P.452 model. There is no
intention to endorse any individual model for use outside its accepted input parameter
range. The methods should be seen clearly in the light of mathematical manipulations
to smoothly blend believable individual model results into one another and no more.

7 SUBSEQUENT RESULTS

A number of other simulation runs have been done since the submission of the two
papers into WP3M. The three following figures confirm the stability of the proposed
changes.

Page 20 of 68
Removal of discontinuities from ITU-R P.452

Figure 9 25000 random paths between 3 and 1000km long containing at least 75% water. The probability density function was
selected to give a uniform point density when plotted on a logarithmic axis.

Page 21 of 68
Removal of discontinuities from ITU-R P.452

Figure 10 The current P.452-10 model losses of the 25000 random paths shown in figure 9 analysed into line-of-sight, sub-path
diffracted and transhorizon. The transhorizon paths are further subdivided into the dominating model mechanism.

Page 22 of 68
Removal of discontinuities from ITU-R P.452

90

140
Loss (dB)

190

240

290

1 .10
3
1 10 100
Distance (km)
LOS
Subpath Diffraction
Transhorizon Diffraction
Transhorizon Ducting
Transhorizon Tropospheric

Figure 11 The proposed P.452 model losses of the 25000 random paths shown in figure 9 analysed into line-of-sight, sub-path diffracted
and transhorizon. The transhorizon paths are further subdivided into the dominating model mechanism.
Page 23 of 68
Removal of discontinuities from ITU-R P.452

Figure 9 shows 25000 randomly generated paths over most of mainland Europe. The
paths varied from 3 to 1000km in length and contained a minimum of 75% water. By
increasing the amount of water in the paths, the number of paths being dominated by
the ducting/layer reflection model was increased. Terminal heights were 150 and 10m
respectively and the frequency was 1.0GHz.

Figures 10 and 11 illustrate the P.452 overall model losses before and after the
changes proposed within this report. These losses have been colour coded in the first
instance into the three path classifications used within P.452. These are line-of-sight,
sub-path diffracted and transhorizon. The transhorizon paths have been further
subdivided into which of the three constituent model mechanisms is dominant. It is
clear that the clamping and angle blending being applied to the ducting/layer reflection
model is achieving the desired effects without noticeable side effects.

8 CONCLUSIONS
This report has examined in detail the reasons that underlie some of the largest
discontinuities found in the overall prediction result from ITU-R P.452. The largest
discontinuities have been shown to occur in marginal line-of-sight paths at the
transition between the line-of-sight model and ducting/layer reflection model
boundary. The need to correct discontinuities between the constituent model parts is
something of a misnomer since, if the models are accurately predicting the loss, they
should give very similar answers at their boundary conditions. One conclusion from
this report is that the individual P.452 models are being exercised outside of the input
parameter ranges considered during their development. Some of these boundary
conditions are not well understood, particularly for the anomalous propagation
mechanisms modelled by the ducting/layer reflection model. The original model was
developed to best-fit measurement data from the COST210 data set. All of the paths
in this data set were in excess of 50km in length from NW Europe with uncluttered
terminals on relatively tall towers. The choice of 20km as the lower limit for invoking
the ducting/layer reflection model is arbitrary and, at best, a figure chosen to be well
removed from the shortest paths in the data set.

By exercising the P.452 model well away from the conditions prevailing in the original
data set, the user community is taking a risk that may be unjustified. However, given
the ever-increasing requirement for interference predictions on shorter-range paths
with cluttered terminals, immense pressure has been exerted upon the propagation
experts to find solutions to any problems that manifest themselves. This report does
not foresee the development of new analytical models that will address these new
scenarios in the short term since it would entail a very costly measurement campaign
over several years. The current economic environment within the telecommunications
industry is also a stumbling block to blue sky research.

The report concludes that a pragmatic solution of either a semi-analytic or fully


empirical nature is required for a short-term solution. Both these approaches are
examined in the report. Although there is large degree on commonality between the
line-of-sight model and ducting/layer reflection model at the time percentage reference
points, further progress with a new semi-analytic model is hindered by the completely
dissimilar methods of introducing time percentage. Although not totally
insurmountable given enough time, the only viable solution is a fully empirical
approach.

Page 24 of 68
Removal of discontinuities from ITU-R P.452

The empirical approach in the first instance would seem to offer a worst-case result
but it does have a number of advantages that can be exploited. No new propagation
models need to be developed and more importantly tested against data. By defining
the blends in terms of existing input or derived path profile parameters, no more
parameter extraction is required. The original assumptions that relate to the validity or
not of the original models remain unchanged. As long as the individual P.452 models
are evaluated at a number of key reference points and a few intuitively simple
assumptions are made, an overall model result that varies smoothly through the
marginal line-of-sight condition is easily derived. Since the P.452 models are only
evaluated at particular reference points, the option to alter or replace these models
and re-evaluate the points is left open.

The report has largely ignored the tropospheric model which is also a constituent part
of the transhorizon path calculation. This model is, however, the dominating
mechanism only at long distance and higher time percentages. Even in the flat earth
case it is unlikely that a transition from a transhorizon to a line-of-sight path will occur
in the region where the tropospheric model is dominating unless the terminals are
very high. In any case, this type of transition has not been seen to generate large
discontinuities in practice and it was felt justified to ignore tropospheric discontinuities.

A solution to the discontinuities was only part of the work of the contract. It was of
equal importance to have the proposals adopted within the next version of the
recommendation. Submission of input papers to UK Study Group 3 and subsequently
to international ITU Working Party 3M is the recognised method to achieve such
changes. Two substantive papers were submitted via this route to the Working Party.
Both WP3M and SG3 meetings were attended so that full support could be given to
the proposals. The outcome was completed successfully since the entire proposal
features in the next revision of the recommendation, which is currently undergoing
approval by correspondence.

This report has highlighted the need for additional propagation models. As stated
earlier, the existing models were derived from fits to measured data from paths in
excess of 50km. As the range of radio services has increased, the user community is
attempting to use the recommendation for very much shorter paths, 5km and below.
It is quite understandable that the component models do not give similar results when
operated this far from their intended operational parameters. Undoubtedly, some will
work better than others on these short paths but deciding which answer is correct,
particularly for a transhorizon path, is not going to be possible without a measurement
campaign to provide the data. Since the development of P.452 a number of
worldwide databases of radio meteorological parameters have been published. In
theory, a new anomalous propagation model could be developed based upon the new
maps. However, if it is to be useful, it needs to cover a wider input parameter range
than the existing ducting/layer reflection model, which is a proven entity. Without the
data to validate the new model outside of this range, credibility is always going to be
an issue.

9 RECOMMENDATIONS

The following specific areas of propagation research have been identified within this
document:

Page 25 of 68
Removal of discontinuities from ITU-R P.452

• A new measurement campaign for short-range paths, 2 to 100km, over a


range of frequencies and time percentages.
• A short-range anomalous propagation model to fit this measured data.
• A revision to the line-of-sight model with closer integration to the ducting/layer
reflection model with, possibly a common time percentage reference point.

Page 26 of 68
Removal of discontinuities from ITU-R P.452

10 ANNEX 1

This annex contains the first of two input papers submitted to Working Party 3M following approval
by UK Study Group 3. This paper culminates in an Appendix that gives the specific proposals for
changing ITU-R P.452. This appendix should be disregarded as it is superseded by the appendix
attached to the second of the ITU submissions given in Annex 2.

Annex 1 Page 27 of 68
Removal of discontinuities from ITU-R P.452
INTERNATIONAL TELECOMMUNICATION UNION

RADIOCOMMUNICATION Delayed Contribution


STUDY GROUPS Document 3M/58-E
22 April 2002
English only

Received: 18 April 2002


Subject: Recommendation ITU-R P.452-10

United Kingdom

PROPOSED REVISION OF RECOMMENDATION ITU-R P.452

Removal of discontinuities
1 Introduction

The present clear-air model within ITU-R P.452 is currently modelled by the power addition of a
number of entirely separate models, each of whose contribution varies as a function of the perceived
characteristics of the propagation path. The propagation path itself is classified into one of these
three fundamental types:-

Line of Sight Path:- Characterised by full terrain clearance of the first


Fresnel Zone surrounding the straight line joining the
transmitting and receiving terminals
Line of Sight Path with sub- Characterised by a terrain incursion into the first
path diffraction:- Fresnel Zone surrounding the straight line joining the
transmitting and receiving terminals but remaining
below the line
Trans-horizon:- Characterised by terrain obstruction of the straight
line joining the transmitting and receiving terminals

Different combinations of four basic propagation mechanisms, which are modelled in isolation of
one another, are assumed to apply to each of these path types. These mechanisms and their
respective models are classified within the Recommendation as follows:-
• Line of Sight
• Diffraction
• Ducting / Layer Reflection
• Tropospheric
Unfortunately in the real world these propagation mechanisms do not switch on and off as precisely
as assumed at the boundaries of the path classifications. The true picture is a continuum of
propagation effects that mix in a continuously variable fashion across the range of path types. To
curb the worst of the unwanted side effects of combining the models, a number of arbitrary distance
switches have been introduced. These relate in particular to the onset and cessation of the
ducting/layer reflection and diffraction models. To compound the problem further, similar, but
unrelated, switches have been introduced into the models themselves as they have matured. Many,
but not all of these switches, have been incorporated into the respective models though the use of

Annex 1 Page 28 of 68
Removal of discontinuities from ITU-R P.452

transitional mathematical functions that, although not continuous in the first differential are
continuous in value.

The result of these stepwise improvements to the Recommendation since its inception has been to
create a few situations where an abrupt change in output can be produced by an arbitrarily small
change to an input to the model. Furthermore, in the worst cases, the output step in model loss can
be in the opposite direction to that expected by the changing inputs. This is a particular problem at
the transition from a Transhorizon to a grazing Line of Sight path, however this is brought about.
Several proponents of the Recommendation, as long as six years ago (UKSG3 CP110, 1995), have
correctly identified the specific models and model behaviour that induces the problem. However,
such was the internal complexity of the ducting/layer reflection model, one of the models most
closely identified with the transition, that no simple solution was forthcoming. Since those days,
the Recommendation is being used for an ever-increasing range of situations that are outside the
boundaries of the original test paths conceived in the European COST210 programme (relatively
tall, unobscured terminals on 50 km and upward paths). The move to analyse shorter paths in
particular, has only accentuated the problems. Such problems have serious operational
disadvantages and can severely hamper simulators and other software expecting monotonic model
behaviour.

It is proposed that a new method of blending the outputs of the four models be adopted that makes
use of some of the geometric parameters of the path already calculated. In addition, it is proposed
to abolish the lower distance switch at 20 km for the ducting/layer reflection Model. It has to be
understood that this in no way implies that the currently constituted model is suitable for use in
isolation down to very short distances. Indeed, it is doubtful whether, for small distances, the
atmospheric conditions can be described in the terms used within the model itself. Specific
assumptions relating to the maximum signal level that is to be expected at the Trans-horizon to
grazing Line of Sight transition are made and the resultant of combining the individual model
answers is mathematically constrained to this boundary condition. Two specific path profiles, both
of a synthetic nature are presented in evidence, but it is hoped that the detailed tests on real path
profiles will show no adverse effects.

Appendix 1 contains a specific proposal to change the manner in which the three transhorizon path
models are combined in ITU-R P.452. It is intended that by making such a proposal, the necessary
widespread testing by others can commence in time for other submissions on the topic to the
international WP 3M meeting in Geneva in May.

Annex 1 Page 29 of 68
Removal of discontinuities from ITU-R P.452

2 Current Model Behaviour at the Trans-horizon / Line-of-sight boundary

Synthetic Path Profile


16

14
Three
Txh -> <- Rxh
12

10
Height (m)

0
0 2 4 6 8 10 12 14 16 18 20
Dist (km)
Path profile
Horizon A
Horizon B
Local Horizontal at A
Local Horizontal at B
Single Horizon path
LOS path
Double Horizon path
Smooth earth

FIGURE 1
A synthetic path profile containing a single fixed receiver and a transmitter with three
variable heights

Figure 1 shows a synthetic path profile of length 20 km, which contains two obstructions. The
obstructions are each 10m high and are situated 3 and 17 km along the path respectively. The
smooth earth surface surrounding the obstacles is of zone type SEA. The receiver terminal, Rxh, is
12.7 m high whilst the transmitting terminal is shown in one of three positions. The first of these
positions, a type X path, is 13.23 m high (red solid line) which gives a line of sight path to the
receiver. The second position, a type Y path, is 12.80m high (blue dotted line) and results in a
trans-horizon path with a single common horizon, B, for both terminals at 17 km along the path.
The last position, a type Z path, is 12.38 m high (purple dash line) and again results in a trans-

Annex 1 Page 30 of 68
Removal of discontinuities from ITU-R P.452

horizon path. However, this last path now has two distinct horizons, one at A for the transmitting
terminal and the other at B for the receiving terminal. Whilst this profile is highly stylised and is
shown with a greatly exaggerated y scale, it is not unrealistic in the sense that the major features
determined above are highly plausible. The exact terminal heights chosen above have been selected
to allow the three differing transmitting terminal heights to be distinguished on the diagram. If the
receiving terminal were to be raised by a small amount, the range of transmitting terminal heights
that result in path types X, Y and Z would decrease. In the limit, there exists some receiving
terminal height where an infinitely small step change, δ, in transmitting terminal height would yield
these three path types.
It is interesting to compare the propagation loss resulting from the application of the current P.452
model to the three situations illustrated above. Figure 2 shows the overall propagation loss as a
function of time percentage for the three path types, X, Y and Z. In addition, a fourth case is also
shown. This path is the result of applying the model to one of the trans-horizon paths, (Y), but with
a vanishingly small reduction in the step size of the profile, which results in an overall path length
of 19.999 km.
The remaining test conditions used in the analysis are listed in Table 1. These conditions are typical
for a path from the southern half of the United Kingdom

TABLE 1
The other test conditions used in the P.452 analysis
Frequency 2.1 GHz
Path centre latitude 51.745407 deg
Path centre longitude 0.875429 deg
Annual average lapse rate 48
Beta 0 7.803 %
Zone type (excluding obstructions) Sea
Combined antenna gain 80 dB

The curves in Figure 2 indicate that a large change in output value results from an incrementally
small change in either the terminal heights or path distance. Three separate effects have been
identified, which, although related, will be described separately:-

• Ducting/layer reflection model – line-of-sight model switching


• Horizon distance dependent switching
• Ducting/layer reflection model path length switching

It is reasonable to suppose that as a path type moves from the trans-horizon case with separate
terminal horizon points to a line-of-sight case that it will pass though the intermediate case of trans-
horizon with a single common horizon unless both the path and terminal heights have a high degree
of symmetry associated with them. In this special case, the switch is straight to the line-of-sight
condition.

Annex 1 Page 31 of 68
Removal of discontinuities from ITU-R P.452

3 Ducting/layer reflection model – Line-of-sight model switching


At the transition between a grazing line-of-sight path and a trans-horizon path with either a single or
double horizon point(s), Rec. 452 uses either the clear-air line-of-sight model or the power addition
of the ducting/layer reflection, diffraction and troposcatter models.
Rec452 Model Loss vs Time%
95

β0

105

115
Loss (dB)

125

135

145

155

1 .10
3
0.01 0.1 1 10 100
Time %
Transhorizon Path type Y
LOS path type X
Transhorizon Path type Z
Free Space Loss
Transhorizon Path 19.999km

FIGURE 2
Path losses calculated using the existing P.452 models and the various trans-horizon path
types described in the text

In general one or other of these individual results, depending upon the exceedance time percentage
required, dominates this power addition. At small time percentages below 1 %, the ducting/layer
reflection model will invariably dominate and produce the smallest loss. Whilst both the line-of-
sight and ducting/layer reflection models recognize the need for small time percentage signal
enhancements, they attempt to achieve this in entirely different ways. This results in the large
discrepancy seen between the line-of-sight and trans-horizon curves in Figure 2.

Annex 1 Page 32 of 68
Removal of discontinuities from ITU-R P.452

4 Horizon Distance dependent switching


Path types Y and Z, although both trans-horizon and hence invoking the same P.452 models, still
show more than a 10 dB difference in predicted loss. Most of this difference can be attributed to the
change in combined horizon distance, dlt + dlr. This switches from 20 km to 6 km as the single
horizon point separates into two. Within the ducting/layer reflection model there is a term
20*log(dlt + dlr) in the expression for Af which accounts for most of this change.
5 Ducting / Layer Reflection Model path length switching
Below 20 km for a trans-horizon path, the power addition excludes the ducting/layer reflection
model. Therefore, since troposcatter is low at these distances, the power addition reduces to the
result of the diffraction model. This gives the highest loss of those considered and hence the largest
potential step change of output.
One other point that needs to be made is that the transition, when it occurs, is often the wrong way
round; the obstructed loss is often lower than the line-of-sight loss, which is intuitively incorrect.
6 A revised procedure for combining the P.452 models
In addressing the problems highlighted in the previous section, a number of issues need to be
resolved. In this paper each of these issues is addressed in turn by successively blending the result
of the previous one with the next one. It was found that three separate blending stages were
required to smoothly merge the line-of-sight model to the power addition of the other three models
as the path undergoes the transition from trans-horizon to grazing line-of-sight.

Perhaps first and foremost issue is that of accuracy; when the line-of-sight and ducting/layer
reflection models differ at the grazing angle, which is the most accurate? Although the line-of-sight
model within P.452 is far less complex than the equivalent line-of-sight model in ITU-R P.530,
which caters for fades as well as enhancements to the signal, it is difficult to envisage how the
signal loss for an obstructed trans-horizon path can ever be lower than that of the equivalent line-of-
sight path. We therefore have to assume that the enhancement afforded by the line-of-sight model,
whichever one this might be, represents the lowest loss that a path can experience.

If this criterion is used as the basis for the minimum loss for the transition from line-of-sight to
trans-horizon paths, then it follows that either the internal workings of the ducting/layer reflection
model must be altered or the result from this model must be blended into the line-of-sight result.
Figure 3 illustrates the result of running the P.452 on the path profile scenario described in the
previous section with the exception that the zone type has been changed to INLAND. This shows
that one of the two trans-horizon path types, Y, now results in a loss that is higher than the
enhanced line-of-sight loss. This illustrates the extreme difficulty in making changes to the
ducting/layer reflection model, since virtually every aspect of the geographic and radio-
meteorologic nature of the path is an input. It will be far easier to ensure that in those situations
where the ducting/layer reflection model loss is lower than the enhanced line-of-sight result that the
resultant blend is always higher than or equal to the line-of-sight result at the same percentage time.
Equation 1 will smoothly clamp the ducting/layer reflection model result from dropping below the
line-of-sight model result at some time percentage p:

Annex 1 Page 33 of 68
Removal of discontinuities from ITU-R P.452

Rec452 Model Loss vs Time%


105

β0

115

125

135
Loss (dB)

145

155

165

175

1 .10
3
0.01 0.1 1 10 100
Time %
Transhorizon Path type Y
LOS path type X
Transhorizon Path type Z
Free Space Loss
Transhorizon Path 19.999km

FIGURE 3
The analogous path losses to those given in Figure 2 but with the zone type changed to type
INLAND

  L ( p)   L ( p)
Lminba ( p ) = η * ln exp ba  + exp b 0   (1)
  η   η 
where:
Lminba(p): the revised ducting/layer reflection model loss.
Lba(p): the ducting/layer reflection model loss.
Lb0(p): the line-of-sight model loss.
η: a parameter that controls the sharpness of the blend; set to [2.5] here.

Figure 4 illustrates the effect of the line-of-sight to ducting/layer reflection model blend for two
different curves. These were taken from Figure 3 and show the effect of the blend for ducting

Annex 1 Page 34 of 68
Removal of discontinuities from ITU-R P.452

model losses that both cross and remain below the enhanced line-of-sight line. The corner infill and
proximity of the lines for the former and latter cases respectively are controlled by the parameter η.

LOS to Ducting Model Blend


105

β0 50.0
115

125
Loss (dB)

135

145

155

165

1 .10
3
0.01 0.1 1 10 100
Time %
FSL
Enhanced LOS Model
Ducting Model (2 horizons)
New Blend
Ducting Model (1 horizon)
New Blend

FIGURE 4
The results of applying the first of the blend techniques to two of the ducting/layer reflection
model results on the INLAND zone type path

Whilst this will prevent the ducting/layer reflection model loss from ever becoming lower than the
line-of-sight loss at those smaller time percentages where bulk diffraction effects are being
effectively bypassed by the anomalous signal enhancements from either the line-of-sight or
ducting/layer reflection model, another issue has to be considered at higher time percentages. As
bulk diffraction effects start to become significant, the enhanced line-of-sight line will no longer
represent the lowest signal level that the grazing line-of-sight path can reach.
At 50% time it would seem realistic that at the grazing line-of-sight condition, there would be no
signal enhancement and the path would show a sub-path diffraction loss. This loss would be 6 dB
for a knife-edge model, but is closer to 12 dB for the present Deygout diffraction model. Having
defined the behaviour at the two extremes of time percentage, the question of how to join these two
points together arises. One obvious parameter that can be used is that of β0, which defines the point
incidence of anomalous propagation for the particular path. The time variability of the excess loss

Annex 1 Page 35 of 68
Removal of discontinuities from ITU-R P.452

due to the diffraction mechanism is assumed to be the result of changes in bulk atmospheric radio
refractivity lapse rate. This process is only considered valid down to β0 time. Below this, signal
levels are dominated by the anomalous propagation mechanisms rather than that of the bulk
refractivity characteristics of the atmosphere. We can therefore assume that any sub-path
diffraction losses will be bypassed below β0 time. All that then remains is to define the function
that will blend the 50% of time point into the β0 time point.
Two alternatives were tried at this point. The hyperbolic tangent and inverse cumulative normal
functions were both evaluated as potential candidates. The inverse cumulative normal function was
selected for two reasons. Firstly it gave a better shape to the final curve and secondly it was already
being used in the Recommendation within the Diffraction model itself and therefore had more
credibility. Figure 5 illustrates the modification that this term has upon the enhanced line-of-sight
values above β0 % time.

Sub-path Diffraction Blend at High Time%


105

β0 50.0
115

125
Loss (dB)

135

145

155

165

1 .10
3
0.01 0.1 1 10 100
Time %
FSL
Enhanced LOS Model
Diffraction Model
New Blend

FIGURE 5
The result from applying the second of the blending techniques
The blue thick solid line sets the minimum loss that the ducting model can reach

The solid blue line in Figure 5 now represents the minimum loss that can be obtained at the
transition point from a trans-horizon to grazing line-of-sight path. In calculating a point on this line,
Lminb0(p), use is made of the interpolation factor, Fi, that has already been calculated in equation

Annex 1 Page 36 of 68
Removal of discontinuities from ITU-R P.452

13a) of Section 4.3 on Diffraction. Calculate L Eβ from evaluating the line-of-sight model at β0%
time:
Lb 0 β = Lb 0 ( β 0%) (2)
Calculate Lbd50 from evaluating the diffraction model at 50% time:
Lbd 50 = Lbd (50%) (3)
Finally calculate Lminb0(p) at any time percentage from:
 Lb 0 ( p ) for p < β 0
Lminb 0 ( p ) = 
 Lbd 50 − (Lbd 50 − Lb 0 β )⋅ Fi
(4)
for p ≥ β 0

FIGURE 7
Example of a (trans-horizon) path profile

ith terrain point

Interfering station (T) Interfered-with


Mean sea level hl station (R)
θt
d
θr
d lt
htg di d lr
hgt h rg
hts hgr
hrs
·a
ae = k 50

Note 1 – The value of θt as drawn will be negative.


0452-07

FIGURE 6
A diagram taken from P.452 showing the Path Angular Distance parameter θ in relationship
to the remaining parameters.

Annex 1 Page 37 of 68
Removal of discontinuities from ITU-R P.452

Path profile

Term clr

Height (m) 15

10

0
0 5 10 15 20
Dist (km)

FIGURE 7
A simple synthetic path profile showing a family of trans-horizon paths as the terminal
heights are raised in unison. The last line is also grazing line-of-sight.

An extensive study of the loss values at the ducting/layer reflection model to grazing line-of-sight
transition has shown a wide difference variation in the two values that have to be bridged. A
smoothing function that takes some geographic parameter relating to the path is needed to finally
blend the ducting/layer reflection model result to the line-of-sight model result over a defined range
of that parameter. The parameter selected is the Path Angular Distance, θ, from Table 7 of the
Recommendation. Figure 6, reproduced from P.452 itself, illustrates the parameter in terms of the
path profile.
Figure 7 shows a set of trans-horizon paths as each member moves closer to the grazing line-of-
sight condition. The angle θ changes from positive values for the trans-horizon paths though zero at
grazing line-of-sight to negative values as the sub-path obstruction moves out of the Fresnel zone.
In mathematical terms at the grazing condition, the sum of the two horizon angles is negative and
exactly opposite the angle subtended by the path at the earth’s centre.
This is best illustrated in Figure 8 which shows the major path related angles that derive from the
Path Analysis given in Section 5 of the Recommendation as applied to the simple synthetic profile
shown in Figure 7. The angles are shown as a function of the terminal heights, which are moved
upwards in unison. It is seen that the Path Angular Distance, θ, falls to zero at the grazing height
whereas the terminal horizon angles, which are equal and superimposed on the plot, reach their
most negative values. This figure also provides a clear indication of the final required blend shown
as the S curve used on the legend. The ducting/layer reflection model value is required to migrate
smoothly into the line-of-sight model value over some range of the angle θ. The curve, shown in
the figure as a red dashed line, makes this transition occur over the angular range of θ. from 0.5 to 0

Annex 1 Page 38 of 68
Removal of discontinuities from ITU-R P.452

mrad. This is an arbitrary choice and is open to discussion. More recent studies have shown that a
range of approximately 0.6 mrad gives suitable results on a variety of path lengths from 1 to 100km.

Ducting path angles vs Txh


1

0.5

0.0

0
Angles (mrad)

0.5

1.5

2
10 11 12 13 14 15 16
Txh (m)
thetaT & R
theta T & R Mod
theta Mod
theta
Site Shielding T
Site Shielding R
Local T Horizontal
Local R Horizontal
S curve used

FIGURE 8
A graph of the major angles resulting from the P.452 path analysis procedure as a function of
transmitting terminal height for the scenario in Figure 7.

Again the choice of function to use for the S shaped curve is open to debate. The function chosen
was the hyperbolic tangent, which has already been used at several places within the
Recommendation and is easily adjusted to change the angular range. The function has thus been

Annex 1 Page 39 of 68
Removal of discontinuities from ITU-R P.452

formulated in terms of two additional parameters, Θ & Ξ, which control the angular onset of the
blend and the certainty with which it reaches the final values. Calculate the interpolation factor, Fj:
  Ξ⋅ 
F j = 1.0 − 0.5 ⋅ 1.0 + tanh 3.0 * (θ − Θ )  (5)
  Θ 
where
Θ: a fixed parameter determining the angular range of the transition; set to [0.3]
Ξ: a fixed parameter determining the approach slope at the ends of the range; set to
[0.8]

Finally the overall modified ducting/layer reflection model loss, L‘ba(p), is calculated from:
L‘ba ( p ) = Lminba ( p ) + (Lminb 0 ( p ) − Lminba ( p )) ⋅ F j 6)

Rec452 Model Losses vs Time%


105

β0

115
Loss (dB)

125

135

145

1 .10
3
0.01 0.1 1 10 100
Time %
Transhorizon paths from P.452
LOS path type X
Transhorizon paths from new blend
Free Space Loss

FIGURE 9
Path losses calculated using P.452 before and after the application of the new blending
functions to the ducting/layer reflection model. The zone type is INLAND.

Annex 1 Page 40 of 68
Removal of discontinuities from ITU-R P.452

In the case of a trans-horizon path, the overall model prediction is then calculated from the power
addition of the modified ducting/layer reflection, troposcatter and diffraction models respectively.
In the case of the grazing line-of-sight path, the standard line-of-sight model is used. Significantly,
the switch that turns off the ducting/layer reflection model below 20 km for a trans-horizon path is
removed. This means a value is calculated from this model even for paths as short as 1 km.
7 Results
Figures 9 and 10 show comparisons between a set of trans-horizon paths that culminate in a grazing
line-of-sight path calculated using the existing method in P.452 before and after applying the
blending modifications to the ducting/layer reflection model. The path scenario is the one given in
Figure 1 with the addition of six further equally spaced extra transmitting terminal heights below
the three that are illustrated. Figure 9 is for a zone type of INLAND whereas Figure 10 is for zone
type SEA. This accounts the change in β0 value.

Rec452 Model Losses vs Time%


95

β0

105

115
Loss (dB)

125

135

145

1 .10
3
0.01 0.1 1 10 100
Time %
Transhorizon paths from P.452
LOS path type X
Transhorizon paths from new blend
Free Space Loss

FIGURE 10
Path losses calculated using P.452 before and after the application of the new blending
functions to the ducting/layer reflection model. The zone type is SEA.

Annex 1 Page 41 of 68
Removal of discontinuities from ITU-R P.452

It is apparent that none of the ducting/layer reflection model losses are now lower than the enhanced
line-of-sight model loss. Since the isolated trans-horizon line of path type Y (single common
horizon) has a terminal height nearest to the grazing line-of-sight case - i.e. has the smallest angle θ,
it has undergone the largest blend toward the line-of-sight model result. This results in this
particular blended line being very close to all the other blended curves that arise from the double
horizon cases. This even holds true for Figure 9 where the unblended P.452 result for this particular
terminal height is some 4 dB below the enhanced line-of-sight line at 0.001%.
8 Conclusions
The disparity between the ducting/layer reflection model and the line-of-sight model under specific
path profile conditions has long been known to cause a problem in ITU-R P.452. The two models,
although both attempting provide signal enhancements at small time percentages, have such a
dissimilar internal construction and input parameters that it has proved impossible to get widespread
agreement at the point where one model replaces the other. This paper shows the preliminary
results from applying a new blending technique to the overall prediction in ITU-R P.452.
The new technique does not attempt to alter the specific formulation of each model. Rather it starts
from an assumption that the notional equivalent line-of-sight path will have a lower enhancement
loss than a similarly partially obstructed one even at small time percentages. The ducting/layer
reflection model result is blended to the line-of-sight model loss to a degree, which varies according
to how close the path is to being grazing line-of-sight. The blend is achieved in three stages. First
the ducting/layer reflection model result is prevented from dropping below the enhanced line-of-
sight result. Secondly the minimum value of loss at any particular time percentage including the
sub-path diffraction effect is calculated. Finally the ducting/layer reflection model loss is blended
to this modified enhanced line-of-sight result over a small range of the path angular distance angle,
θ. Finally the lower limit of 20km for performing the ducting/layer reflection model calculation is
abolished. These changes in no way imply that in isolation the ducting/layer reflection model is
suitable for this increased distance range. The post processing of the three model results is merely
an inventive mechanism to combine them into a meaningful result that avoids the worst of the
discontinuities currently experienced.
Results thus far are encouraging but further testing over a wide range of path distances and types is
now needed. The technique has, where possible, made use of existing parameters and functions
contained within the P.452 model overall. Since none of the individual models themselves are
modified, it should be possible to change one or more of them and re-apply the technique to the new
models using the same assumption that the lowest loss will arise from the enhanced line-of-sight
model.
Appendix 1 gives the specific changes to the Recommendation. These changes concentrate upon
the model changes required. A change, such as that proposed, could also impinge upon the
descriptive text in several parts of the document. Whilst some effort has been put into identifying
these, there may be instances where some slight change of words is still required.

Annex 1 Page 42 of 68
Removal of discontinuities from ITU-R P.452

APPENDIX 1 TO ANNEX 1

Proposed revision of Recommendation ITU-R P.452


The following text describes the changes required to the Recommendation to implement the new
method. Where only small changes have been made to larger existing sections of text, for the
benefit of the reader, this text has been underlined.

1 In section 2, at the end of the last paragraph, add:


“This overall prediction is made using a blending technique that ensures for any given path distance
and time percentage that the signal enhancement in the equivalent notional Line-of-sight model is
the highest attainable.”

2 Replace Step 5 in section 3.2.1 as follows:



Step 5: Calculation of propagation predictions

Table 4 indicates, for each type of path, the propagation models that are appropriate. The necessary
equations for these individual propagation mechanism predictions are to be found in the text
sections indicated in the table. In order to build an overall prediction, the individual propagation
mechanism predictions must be calculated and combined in the manner shown in section 4.7. For
trans-horizon paths, elements from both the line-of-sight and diffraction models are re-used within
the combining process. Once this has been achieved for each of the required time percentages, the
prediction is complete.

3 In Table 4, under Trans-horizon Models required, remove the bracketed expression
following the words “Ducting/layer reflection”.
4 In Table 5, under Trans-horizon Actions required, change the text to read:


The overall prediction is obtained in two stages:
The modified ducting/layer reflection model loss, L‘ba ( p), is found by
application of the algorithm in section 4.7.1. The overall prediction can
then be obtained by applying the following ancillary algorithm:
Lb( p) = −5 log (10– 0.2Lbs + 10– 0.2Lbd + 10– 0.2L‘ba) + Aht + Ahr dB (8c)
where Lbs ( p) and Lbd ( p): individual predicted basic transmission loss for
p% of time given by the troposcatter and diffraction propagation models
respectively.
NOTE 1 – Where a model has not been proposed for a path (because the
conditions given in Table 4 were not met), the appropriate term should be
omitted from equation (8c).

Annex 1 Page 43 of 68
Removal of discontinuities from ITU-R P.452

5 Add a new section 4.7 as follows:


4.7 The Overall Prediction


Table 5 gives the actions required to build the overall prediction for each classification of path type.
For paths classified as line-of-sight or line-of-sight with sub-path diffraction, no further pre-
processing of the individual model results are required before applying the required action from the
Table.

4.7.1 Trans-horizon paths


In the case of trans-horizon paths, although the line-of-sight model is not a required model, use is
made of the equivalent notional line-of-sight model loss in the combining process. The overall
prediction is based upon calculation of a modified ducting/layer reflection loss, L‘ba(p), from the
following function prior to the application of equation (8c) in Table 5:
L’ba ( p ) = Lminba ( p ) + (Lminb 0 ( p ) − Lminba ( p )) ⋅ F j (28)
where:
Lminba(p): modified ducting/layer reflection loss:
  L ( p)   L ( p) 
Lminba ( p ) = η ⋅ ln exp ba  + exp b 0   (29)
  η   η 
where:
Lba(p): the ducting/layer reflection loss from equation (16)
Lb0(p): a notional line-of-sight loss for the path evaluated from equation (16).

η = [2.5]
Lminb0(p): the notional minimum propagation loss that the modified
ducting/layer reflection loss can attain:
 Lb 0 ( p ) for p < β 0
Lminb 0 ( p ) = 
 L bd 50 − (Lbd 50 − Lb 0 β ) ⋅ Fi for p ≥ β 0
(30)

where:
Lb0β: the notional line-of-sight loss evaluated at β0% time from equation
(9):
Lb 0 β = Lb 0 ( β 0 %) (31)
Lbd50: the diffraction loss evaluated at 50% time from equation (14):
Lbd 50 = Lbd (50%) (32)
Fi: interpolation factor based on a log-normal distribution of diffraction
loss defined in equation (13c)
Fj: interpolation factor which blends the modified ducting/layer reflection into the
notional line-of-sight loss:
 
F j = 1.0 − 0.5 ⋅ 1.0 + tanh 3.0 *
[0.8]⋅ (θ − Θ) 
 (33)
  Θ 

Annex 1 Page 44 of 68
Removal of discontinuities from ITU-R P.452

where:

Θ = 0.3
θ: path angular distance defined in Table 7.

5 Renumber the remaining equations accordingly.

_____________________

Annex 1 Page 45 of 68
Removal of discontinuities from ITU-R P.452

11 ANNEX 2

This annex contains the second of two input papers submitted to Working Party 3M
following approval by UK Study Group 3. This paper culminates in an Appendix that
gives the specific proposals for changing ITU-R P.452. This appendix should be
regarded as the full set of changes to P.452 and as such supersedes the proposals
given in appendix 1 to annex1

Annex 2 Page 46 of 68
Removal of discontinuities from ITU-R P.452
INTERNATIONAL TELECOMMUNICATION UNION

RADIOCOMMUNICATION Delayed Contribution


STUDY GROUPS Document 3M/69-E
15 May 2002
English only

Received: 14 May 2002


Subject: Recommendation ITU-R P.452-10

United Kingdom

PROPOSED REVISION OF RECOMMENDATION ITU-R P.452

Removal of discontinuities – further proposed changes


1 Introduction

In Document 3M/58, the major discontinuities associated with ITU-R P.452 were described and a
proposal made to alleviate the worst of these effects by a process of model blending prior to the
final power summation to obtain the overall loss. The worst of the discontinuities were all
associated with the ducting/layer reflection model and, in particular, with its behaviour at the
boundaries of its applicability. Three separate effects were identified which were described in
isolation of one another:
• Ducting/layer reflection model – line-of-sight model switching
• Horizon distance dependent switching
• Ducting/layer reflection model path length switching
In Document 3M/58, emphasis was given to obtaining a solution to the first of these discontinuities.
This solution required a three-stage blend to be performed. First the ducting/layer reflection model
result is prevented from falling below the path-equivalent enhanced line-of-sight result. Secondly
the minimum value of loss at any particular time percentage including sub-path diffraction effects is
calculated and called the modified enhanced line-of-sight value. Finally the ducting/layer reflection
model loss is blended to this modified enhanced line-of-sight result over a small range of path
angular distance, θ, which is one of the geometric parameters that result from the initial path
analysis. Significantly, the switch that turns off the ducting/layer reflection model below 20 km for
a transhorizon path was removed as part of this process to allow these blends to be made for paths
as short as 1km. This was done in the knowledge that the blending process would not allow any
erroneous answer that might result from exercising the ducting/layer reflection outside of its
prescribed boundaries to be dominant. In addition, further testing work on the consequences of
these changes was encouraged.
Some further testing has now been undertaken and a further undesirable feature has emerged as a
consequence of the removal of the 20 km switch on transhorizon paths that, in particular, are far
removed from the parameter values where the blending techniques described above are needed.
This document describes this testing and identifies this further problem, which relates to the
removal of the 20km distance switch. A simple solution is proposed as an addition to the blends
already described. Finally a modified version of Appendix 1 of 3M/58 is appended to this paper.
This contains all of the specific proposals from this and the earlier paper to change the manner in
which the three transhorizon model results are combined in ITU-R P.452.

Annex 2 Page 47 of 68
Removal of discontinuities from ITU-R P.452

2 Testing

In 3M/58, highly stylised, but not unrealistic path profiles were used examine the discontinuities
and the proposed model amendments. In devising these path profiles, it is never possible to
conceive of every individual problem path type to exercise the model fully. Generating large
numbers of real path profiles from a terrain database and analysing the results of applying the P.452
model to them before and after the proposed changes is one way to assess the effectiveness of these
changes. Any paths that yield anomalous results can then be analysed in detail. This approach was
used to assess the proposed model changes documented in 3M/58.
In selecting the region from which to draw the path profiles, consideration must be given to
choosing a large enough representative area to include all possible types of land and sea features in
a statistically balanced way. Figure 1 shows the area chosen. The path profiles were taken from the
30 arcsec GLOBE database within a box bounded by –10 to 40 deg longitude and 35 to 60 deg of
latitude. This covers most of mainland Europe and European Russia. Moreover, the terrain
includes both flat coastal land such as found in Holland and high Alpine pastures present in
Switzerland. There is also a sufficiently long coastline to exercise paths of all of the main zone
types: inland, coastal and sea.
The P.452 model changes were implemented as a derived C++ class using Borland C++ Builder and
written to a Dynamic Link Library (DLL) that could be called from MathSoft MathCAD
program. The advantage of this approach is to combine the speed of the native C code with the
flexibility and analytical power of MathCAD. All data generated was also written to ASCII files
for further post processing if required. Since the aim of this paper was to examine primarily the
ducting model, the parameters most closely associated with this model were those stored. All of the
basic and advanced path profile analysis parameters, together with model losses at a number of
different time percentages were recorded. Although capable of generating a complete distribution
of path lengths between a lower and upper limit, the program was set to generate random
orientation paths of one fixed length at a time. Provision was made to force paths to start and/or
stop either on land only or anywhere within the area. The analysis that follows was set up in such a
way that all transmitting terminals started on land but receiving terminals could be anywhere. This
allows for near 100% sea paths to be included where transmitters are close to a coastline.
If the range of terminal heights, frequencies and distances possible are considered, the data
collection quickly becomes an intractable multi-dimensional problem. To simplify this only subsets
of this matrix were taken with emphasis being placed upon those most likely to illustrate the model
changes. Since much of the work involves the analysis of the path profiles, which is largely
independent of frequency (sub-path diffraction excepted), only 1GHz has been examined. Other
frequencies are expected to show the same general trends with changes in the scaling of the ‘y’ loss
axis. The distances chosen were in the main close to and either side of the 20km duct switch, but a
number of distances both lower and higher were also used. There was little point in going beyond
200km for this ducting/layer reflection study since the dominant model mechanism becomes
tropospheric. Results were calculated for a matrix of transmitting and receiving terminal heights
ranging from 100m down to 1m. It was found that terminal height was not a very significant factor
in the conclusions made and only a small number of representative values are shown here.
Each run created 25000 random paths for the analysis.

Annex 2 Page 48 of 68
Removal of discontinuities from ITU-R P.452

FIGURE 1

Region selected for the analysis containing 25000 random paths each 50km in length, starting on land and terminating anywhere

Annex 2 Page 49 of 68
Removal of discontinuities from ITU-R P.452

2.1 Analysis methods


To simplify the analysis, the random seed was reset at the start of each run. This has the
effect of choosing the same ‘random paths’ for any particular set of inputs allowing for an
easier before and after model change analysis. The path classification analysis is the same for
both the before and after model change runs and only need be done once. Scatter plots of
path loss before and after applying the model changes were generated. The points on these
plots were colour coded according to a number of different criteria. The first classification
was to identify paths that were line-of-sight, sub-path diffracted or transhorizon. The
transhorizon paths were further sub-classified into those with a path angular distance, θ, of
less than 1 mrad and those above this limit. In addition histograms and cumulative
distribution plots of the losses before and after model changes were made.
Particular points from certain runs were identified for individual analysis where the scatter
plots indicated abnormal or marginal behaviour. Some attempt was made to identify the
individual characteristics of these paths that resulted in such behaviour. Where any distance
trends were being obscured by small random variations in terrain heights from point to point,
the salient features of the paths were transferred to similar noise-free synthetic profiles. These
were then analysed for distance trends.
3 Results

3.1 General
Figure 2 illustrates a scatter diagram showing the P.452 losses pre- and post- the model
changes being proposed within 3M/58. The paths are all 19.99 km in length and randomly
generated over the area designated within Figure 1. The individual points have been colour
coded into a number of categories to aid understanding. The single magenta diamond
represents the many paths that are line-of-sight. Since the line-of-sight model is only a
function of distance and time percentage, all points have a single value of 108.05dB. A
second line of brown circles represents all those points that are classified as sub-path
diffracted. These do not extend down to the line-of-sight value, as there is no signal
enhancement being applied to even those that suffer only a small incursion into the first
Fresnel zone. The transhorizon paths have been analysed into three groups; those that are
unchanged by the model changes, those that are changed but have a path angular distance, θ,
of greater than 1 mrad and those changed but with a path angular distance of less than 1 mrad.
The angular blend being proposed in 3M/58 only has an effect within the path angular range
from 0 to 1 mrad approximately. Hence, this latter grouping is indicative of the number of
paths that are experiencing a line-of-sight to ducting/layer reflection blend.

Figure 3 shows an histogram and cumulative probability distribution plot of the same paths
analysed into the same groups. It is immediately apparent that, although those points that are
marginally line-of-sight can undergo a blend down to the line-of-sight value, there are not
very many of them relatively. There are many more transhorizon paths that show large
unintended changes of value brought about by the proposed model changes. Figures 4 and 5
illustrate the equivalent analysis for paths that are identical except for a marginally increased
length of 20.01 km. This shows that although the transhorizon paths with small path angular
distances still undergo a change, the large grouping with large angles is no longer present.
The changed distribution of points near to the line-of-sight value is a consequence of the
ducting/layer reflection model and diffraction models being applied to both x and y axes in
the case of 20.01 km paths.

Annex 2 Page 50 of 68
Removal of discontinuities from ITU-R P.452

These results provide the clearest indication that the removal of the 20 km lower limit switch
from the ducting/layer reflection model is responsible for the changes seen. Simply removing
the switch is not the complete solution for a blend.

Path #22078

FIGURE 2
P.452 model losses pre- and post- model changes proposed within 3M/58. The 25000,
19.99 km length paths have been colour coded to show the transhorizon paths affected by
the changes. The terminal heights are 100m and 5m respectively.
The box shows a path, #22078, selected for a detailed analysis

Annex 2 Page 51 of 68
Removal of discontinuities from ITU-R P.452

Breakdown of model losses; 19.99km paths


1000

800

600
Number

400

200

0
120 140 160 180 200 220 240 260
P.452 Loss (dB)
LOS paths
Sub-path diffracted
Transhorizon pre-changes
Transhorizon post-changes
Transhorizon post-changes < 1mrad

Cumulative distribution
100

80
Percentage

60

40

20

0
120 140 160 180 200 220 240 260
P.452 Loss (dB)
Pre-model change
Post-model change

FIGURE 3
Histogram and cdp showing the distribution of model losses pre- and post-proposed
changes in 3M/58. The 19.99 km paths have been analysed into line-of-sight, sub-path
diffracted and transhorizon types. The first two types are unaffected by the changes

Annex 2 Page 52 of 68
Removal of discontinuities from ITU-R P.452

FIGURE 4
P.452 model losses pre- and post- model changes proposed within 3M/58. The 25000,
20.01 km length paths have been colour coded to show the transhorizon paths
affected by the changes

All of the results presented thus far relate to terminal heights of 100m and 5m for the
transmitter and receiver respectively. Runs have been made using a number of different
terminal heights to look for any other effects. Table 1 shows the number of paths in each of
the groupings used above. Although the numbers of paths in each category vary considerably
across the table, the trends remain fairly stable. There are always many more paths that
undergo an unintended change than those intentionally changed by the blending equations in
3M/58, whatever combination of terminal heights is chosen.

Annex 2 Page 53 of 68
Removal of discontinuities from ITU-R P.452

Breakdown of model losses; 20.01km paths


1000

800

600
Number

400

200

0
120 140 160 180 200 220 240 260
P.452 Loss (dB)
LOS paths
Sub-path diffracted
Transhorizon pre-changes
Transhorizon post-changes
Transhorizon post-changes < 1mrad

Cumulative distribution
100

80
Percentage

60

40

20

0
120 140 160 180 200 220 240 260
P.452 Loss (dB)
Pre-model change
Post-model change

FIGURE 5
Histogram and cdp showing the distribution of model losses pre- and post-proposed
changes in 3M/58. The 20.01 km paths have been analysed into line-of-sight, sub-path
diffracted and transhorizon types. The first two types are unaffected by the changes

Annex 2 Page 54 of 68
Removal of discontinuities from ITU-R P.452

TABLE 1
The number of 19.99 km paths in each category as a function of the terminal heights.
The same random set of 25000 paths for each combination is analysed.
The paths are categorised as line-of-sight (L), sub-path diffracted (D) and
transhorizon (H). A further breakdown of those transhorizon paths that have
changed value is shown as path angular distances above and below 1 mrad
Tx height
10m 20m 30m 50m 100m
Rx height
173 234 284 396 684
1m L
1631 2687 3670 5429 8600
D
23196 22079 21046 19175 15716
T
19197 18003 16948 15334 12460
>1mrad
961 1121 1182 1074 777
<1mrad
180 245 295 409 729
2m L
1919 3085 4189 5984 9112
D
22901 21670 20516 18607 15159
T
18860 17558 16468 14786 11909
>1mrad
1032 1183 1155 1066 767
<1mrad
200 271 357 490 956
5m L
2667 4048 5298 7229 10158
D
22133 20681 19345 17281 13886
T
17996 16517 15294 13503 10780
>1mrad
1138 1256 1183 1050 648
<1mrad
10m L 258 369 491 712 1670
D 3601 5333 6634 8664 10842
21141 19298 17875 15624 12488
T
16904 15136 13846 12036 9532
>1mrad
1280 1269 1218 889 556
<1mrad
20m L 367 604 821 1440 5230
D 5255 7160 8484 9932 8925
19378 17236 16695 13628 10845
T
15152 13252 11920 10331 8121
>1mrad
1316 1168 1012 671 420
<1mrad
50m L 687 1476 2478 6157 12387
D 8626 9896 10167 8122 4056
15687 13628 12355 10721 8557
T
12053 10268 9132 7833 6195
>1mrad
904 705 627 440 218
<1mrad

Annex 2 Page 55 of 68
Removal of discontinuities from ITU-R P.452

3.2 Detailed analysis of a selected path


Having ascertained that the 20 km ducting/layer reflection switch is creating many more
abrupt changes of value than the line-of-sight to ducting/layer reflection transition in any set
of randomly generated paths, a path which has one of the largest differences was selected for
further study. Path #22078 exhibits a 45 dB difference in loss when the ducting/layer
reflection switch is removed from the model and is worthy of further study. The terminal
positions of the selected path are from a latitude of 36.56722 deg and longitude of 33.93616
deg to a latitude of 36.41406 and longitude of 34.0534. Figure 6 shows the path profile and
both terminal horizon positions.

Path profile of point #22078


3000

2500

2000
Height (m)

1500

1000

500

0
0 2 4 6 8 10 12 14 16 18 20
Dist (km)
Ground height
D lt
D lr
-
-
Theta T
-
m
m’
Theta R

FIGURE 6
Path profile of point #22078. The horizon distances and elevation angles are shown
for each terminal. The black dotted and black solid lines are the P.452 slope parameters,
m and m‘ respectively

Annex 2 Page 56 of 68
Removal of discontinuities from ITU-R P.452

The path itself has a large, reasonably uniform gradient with a high degree of site shielding at
the receiver. Figure 7 shows the individual P.452 losses for the ducting/layer reflection and
diffraction models

Ducting Loss vs Time%


95

β0

115

135

155
Loss

175

195

215

235

255

1 .10
3
0.01 0.1 1 10 100
Time %

FIGURE 7
The individual P.452 model losses for the ducting and diffraction models
as a function of time percentage for path #22078. The broken red line is
the enhanced line-of-sight curve for a path of the same length

The 45dB difference between the diffraction and ducting/layer reflection losses at small time
percentage is apparent. This difference is the cause of the large changes seen when
comparing the existing to the proposed P.452 model.

For any practical path it is evident that an infinitesimally small change in path length cannot
result in a 45 dB change in the returned path loss. One or both of the models in use must be at
the extremity of its range of applicability. The ducting/layer reflection model is attempting to
model anomalous atmospheric conditions that result in an enhancement of the radio signal.
Undoubtedly, anomalous conditions can and do occur for shorter distances. The problem
comes in that the ducting/layer reflection model was formulated from fitting to data from the
European COST210 program. The minimum path lengths in the data set were in excess of
50km and the model was not intended be accurate at distances much below this figure. In
practice, a new short-range model is required to represent anomalous conditions on short
Annex 2 Page 57 of 68
Removal of discontinuities from ITU-R P.452

transhorizon paths for small percentages of time, but the necessary measurement data is
unavailable. Conversely, diffraction models have been developed for short distance paths but
primarily they attempt to describe changes in bulk atmospheric radio refractivity lapse rates
and not atmospheric stratification that the ducting/layer reflection model is successfully
addressing, for longer distances at least. In addition, the Deygout diffraction model currently
in use was chosen as the result of minimising the standard deviation between model and data
over a wide range of very many different paths. Being a statistical model, it is inevitable that
accuracy will degrade as the path length gets very short where a range of very path specific
features could be influencing the result.

Given the fact the two models described above are the only contenders at present, one
solution is to take a pragmatic approach until a better solution is found. A blend over a
suitable distance range between the two models is feasible. The next section is concerned
with the development of such a blend.
4 A New Distance Blend

Since the path profile described above is fairly uniform with much of the signal loss coming
from the site shielding at the receiving end, it is a useful path to use to provide an answer to
the following question:
What happens to the individual ducting/layer reflection and diffraction model losses as the
total path length decreases?

This question needs to be answered, as there is the potential for a recovery effect to occur as
the distance increases when using blending. Figure 8 shows the resulting losses from the
ducting/layer reflection and diffraction models as the transmitting terminal is moved towards
the receiver by one profile point at a time. This procedure is stopped at 2 km as the region
giving the site shielding near to the receiver is reached. The evidence from this figure is that
the difference between the two models lessens as the distance becomes smaller mitigating any
recovery effect to some extent.
It is proposed to use a hyperbolic tangent curve to blend the diffraction result into the
ducting/layer reflection result over the range from approximately 5 to 40 km such that it is
exactly half way between the two at the nominal switch position of 20 km. The equations
have been formulated with two parameters, κ and dsw, that control the severity and position of
the switch. These will allow subtle adjustments to be made if needed. The following
equations can be used.
Calculate the interpolation factor, Fk:
  κ⋅ 
Fk = 1.0 − 0.5 ⋅ 1.0 + tanh 3.0 ⋅ (d − d sw )  (1)
  d sw 
where:
d: the great circle path length.
dsw: a fixed parameter determining the distance range of the transition; set to 20.0
κ: a fixed parameter determining the approach slope at the ends of the range; set
to 0.5

Finally the overall modified ducting/layer reflection model loss, Lbda(p), is calculated from:
L ( p ) for Lminba ( p ) > Lbd ( p )
Lbda ( p ) =  bd (2)
 Lminba ( p ) + (Lbd ( p ) − Lminba ( p )) ⋅ Fk for Lminba ( p ) ≤ Lbd ( p )

Annex 2 Page 58 of 68
Removal of discontinuities from ITU-R P.452

where:
Lbd(p) diffraction loss from equation 14 of the recommendation
Lminba(p) modified ducting/layer reflection loss defined in 3M/58.
These equations are inserted into the sequence provided in 3M/58 and the notation is
consistent.
To best illustrate the effect of these equations, a synthetic profile based upon the major
features of path #22078 but of length 50 km was constructed.
Losses vs distance @ 0.001%
95

105

115

125

135
Losses (dB)

145

155

165

175

185

195

205
2 4 6 8 10 12 14 16 18 20
Distance (km)
Difference (dB)

40

20

0
2 4 6 8 10 12 14 16 18 20
Distance (km)

FIGURE 8
Individual model losses for path #22078 as the transmitter is moved toward the receiver.
The thick blue solid line is the ducting/layer reflection loss and the magenta dashed line
the diffraction loss. The black dotted line is the overall model loss which switches
abruptly as 20 km is reached. The equivalent line-of-sight value is shown as red dots.
The lower plot is the difference between the ducting and diffraction models

Annex 2 Page 59 of 68
Removal of discontinuities from ITU-R P.452

Synthetic path profile

2500

2000
Height (m)

1500

1000

500

0
0 5 10 15 20 25 30 35 40 45 50
Dist (km)

FIGURE 9
A synthetic profile modelled upon the real one of Figure 6. The regions in
the immediate vicinities of the terminals have been retained to give similar terminal
horizon positions (dotted lines). The slope of the path is also comparable to the
real path but the overall length has been increased to 50 km

Figure 9 shows the synthetic profile chosen to evaluate the new blending procedure. The
same analysis as before was applied to the path. The length of the path was reduced a profile
point at a time whilst still retaining the terminal features and horizon distances. The overall
characteristics and model losses thus closely matched those of the real path, but without the
‘profile noise’.
Figure 10 clearly reveals the effect of the blend upon the overall P.452 model loss. Although
there is still a recovery effect it is halved in magnitude compared to the step value recovery
that a significant number of random paths undergo with the current implementation method.
In practice, many of these trends will be obscured by the small-scale random variation of
profile heights. Figure 11 shows the effect of successively moving the transmitter towards the
receiver for another path profile drawn from the earlier dataset.
Figure 12 shows the overall effect that the new blend has had on the earlier histograms in
Figures 3 and 5. The results for path distances either side of 20 km are now the same as one
another owing to the distance blend. However the existing P.452 model will place them
either on the ducting/layer reflection values or the diffraction values depending upon the path
length. Paths that are transhorizon, but marginally line-of-sight, are left unaffected by the
careful choice of the order in which the blends are made. The distance blend should be
performed after the ducting result is clamped to the modified notional line-of-sight line but
before the angle blend is made (see 3M/58). In any case, paths that are marginally line-of-
sight will not have large excess diffraction values and will need little blending from the
distance blend proposed here.
Annex 2 Page 60 of 68
Removal of discontinuities from ITU-R P.452

Losses vs distance @ 0.001%


95

105

115

125

135

145
Losses (dB)

155

165

175

185

195

205

215
5 10 15 20 25 30 35 40 45 50
Distance (km)

40
Difference(dB)

20

0
5 10 15 20 25 30 35 40 45 50
Distance (km)

FIGURE 10
Individual model losses for the synthetic path profile of Figure 9 as the transmitter
is moved towards the receiver. The thick blue solid line is the ducting/layer reflection loss
and the magenta dashed line the diffraction loss. The black dotted line is the overall
existing P.452model loss which switches abruptly as 20 km is reached. The red dadot line
shows the proposed blend. The equivalent line-of-sight value is shown as red dots. The
lower plot shows the differences between the ducting and diffraction models

Annex 2 Page 61 of 68
Removal of discontinuities from ITU-R P.452

1400

Height (m)

1200

1000
0 5 10 15 20 25 30 35 40 45
Distance (km)

Losses vs distance @ 0.001%


95

105

115

125

135

145
Losses (dB)

155

165

175

185

195

205

215

225
5 10 15 20 25 30 35 40 45
Distance (km)

FIGURE 11
Individual and overall model losses for the path profile provided in the upper half of the
figure as the transmitter is moved toward the receiver. The lines have the same
meanings as before but the blend is being applied to the overall result

Annex 2 Page 62 of 68
Removal of discontinuities from ITU-R P.452

Breakdown of model losses; 19.99km paths


1000

800

600
Number

400

200

0
120 140 160 180 200 220 240 260
P.452 Loss (dB)

Breakdown of model losses; 20.01km paths


1000

800

600
Number

400

200

0
120 140 160 180 200 220 240 260
P.452 Loss (dB)
LOS paths
Sub-path diffracted
Transhorizon pre-changes
Transhorizon post-changes
Transhorizon post-changes < 1mrad

FIGURE 12
Histograms showing the distribution of model losses produced before and after the
changes proposed in 3M/58 and this document. The paths have been analysed into line-
of-sight, sub-path diffracted and transhorizon types. The transhorizon results have
displaced in one direction for the 19.99 km paths and in the other for the 20.01 km paths

Annex 2 Page 63 of 68
Removal of discontinuities from ITU-R P.452

5 Conclusions

This paper has demonstrated that the rigorous testing of the blending techniques proposed in
the earlier Document 3M/58 has resulted in an unwanted side effect related to the removal of
the 20 km ducting/layer reflection switch. Through the use of 25000 randomly generated
paths over a sufficiently large an area to encompass all feature types, the relative incidence of
marginal line-of-sight transhorizon paths compared to those affected by the 20 km ducting
switch is very low. This is not to say that either transition can be left as they are within the
existing P.452 model, since the impact on affected paths can be dramatic.

A simple solution to allow the 20 km switch to be abolished as originally proposed in 3M/58


has been found. Provided this new blend is integrated into the earlier blends in the
appropriate order, then the P.452 model produces no abrupt changes in overall model answers
with small changes in distance or terminal heights.

It is becoming increasingly apparent that the user community is now exercising the P.452
model over input parameter ranges that were never envisaged by its originators. The
individual models within P.452 were developed to best model the data gained from running
the COST210 experiment. These paths, over 50 km in length, were all taken from NW
Europe and used, in the main, fixed microwave links with relatively tall, uncluttered towers.
Obviously boundary conditions will, and do, exist for the input parameters to each of these
models. It is uncertain, however, where all these boundaries are and some latitude has been
given to regions where one model is taking over from another. The very fact that the
ducting/layer reflection model is exhibiting lower losses on a marginally line-of-sight path
than the line-of-sight model is evidence that one or other of these models is beyond its
operational range. It is also unfortunately true that in the absence of reliable data, it will be
very difficult to devise a new short-range anomalous propagation model to replace the
existing ducting/layer reflection model in the short term. Meanwhile users require the
existing P.452 model to behave rationally to changing input conditions within simulators, etc.
This document should be viewed in the light of a pragmatic approach to remove the worst of
the discontinuities that currently exist within P.452. It is not the intention to endorse the use
of any of the individual models for use outside parameter ranges that were carefully chosen
by the earlier contributors to P.452. Rather, the blending techniques employed should be seen
in terms of mathematical procedures to smoothly join results from well inside the range of a
model’s applicability and no more. By adopting this approach, the door is left open for future
workers to change or replace an individual model or indeed create a new one. Also smooth
transitions have to be closer to the truth than sharp discontinuities, which this document
shows can be currently in excess of 45 dB.

The testing matrix for these changes is so large that independent testing is still a requirement,
as fine-tuning may still be needed.

Appendix 1 takes the additional blend described in this paper and integrates it into the
appendix from 3M/58. As before, there may still be textual changes required in other areas of
the recommendation.

Annex 2 Page 64 of 68
Removal of discontinuities from ITU-R P.452

APPENDIX 1 TO ANNEX 2

Proposed revision of Recommendation ITU-R P.452


The following text describes the changes required to the Recommendation to implement the
new method.
1 In section 2, at the end of the last paragraph, add:
“This overall prediction is made using a blending technique that ensures for any given path
distance and time percentage that the signal enhancement in the equivalent notional Line-of-
sight model is the highest attainable.”
2 Replace Step 5 in section 3.2.1 as follows:

Step 5: Calculation of propagation predictions


Table 4 indicates, for each type of path, the propagation models that are appropriate. The
necessary equations for these individual propagation mechanism predictions are to be found
in the text sections indicated in the table. In order to build an overall prediction, the individual
propagation mechanism predictions must be calculated and combined in the manner shown in
section 4.7. For trans-horizon paths, elements from both the line-of-sight and diffraction
models are re-used within the combining process. Once this has been achieved for each of the
required time percentages, the prediction is complete.

3 In Table 4, under Trans-horizon Models required, remove the bracketed expression
following the words “Ducting/layer reflection”.
4 In Table 5, under Trans-horizon Actions required, change the text to read:


The overall prediction is obtained in two stages:
The modified ducting/layer reflection model loss, L‘ba ( p), is found by
application of the algorithm in section 4.7.1. The overall prediction can
then be obtained by applying the following ancillary algorithm:
Lb( p) = −5 log (10– 0.2Lbs + 10– 0.2Lbd + 10– 0.2L‘ba) + Aht + Ahr dB (8c)
where Lbs ( p) and Lbd ( p): individual predicted basic transmission loss for
p% of time given by the troposcatter and diffraction propagation models
respectively.
NOTE 1 – Where a model has not been proposed for a path (because the
conditions given in Table 4 were not met), the appropriate term should be
omitted from equation (8c).

Annex 2 Page 65 of 68
Removal of discontinuities from ITU-R P.452

5 Add a new section 4.7 as follows:


4.7 The Overall Prediction


Table 5 gives the actions required to build the overall prediction for each classification of path
type. For paths classified as line-of-sight or line-of-sight with sub-path diffraction, no further
pre-processing of the individual model results are required before applying the required action
from the Table.

4.7.1 Trans-horizon paths


In the case of trans-horizon paths, although the line-of-sight model is not a required model,
use is made of the equivalent notional line-of-sight model loss in the combining process. The
overall prediction is based upon calculation of a modified ducting/layer reflection loss,
L‘ba(p), from the following function prior to the application of equation (8c) in Table 5:
L’ba ( p ) = Lbda ( p ) + (Lminb 0 ( p ) − Lbda ( p )) ⋅ F j (28)
where:

L ( p ) for Lminba ( p ) > Lbd ( p )


Lbda ( p ) =  bd (29)
 Lminba ( p ) + (Lbd ( p ) − Lminba ( p )) ⋅ Fk for Lminba ( p ) ≤ Lbd ( p )

Lbd(p): the diffraction loss evaluated at p% time from equation (14):

Fk: interpolation factor which blends the ducting/layer reflection


into the diffraction loss with distance:

  κ⋅ 
Fk = 1.0 − 0.5 ⋅ 1.0 + tanh 3.0 ⋅ (d − d sw )  (30)
  d sw 
where:
d: the great circle path length defined in Table 3
dsw: a fixed parameter determining the distance range of the
transition; set to 20.0
κ: a fixed parameter determining the approach slope at the ends of
the range; set to 0.5

Lminba(p): modified ducting/layer reflection loss:


  L ( p)   L ( p) 
Lminba ( p ) = η ⋅ ln exp ba  + exp b 0   (31)
  η   η 

where:
Lba(p): the ducting/layer reflection loss from equation (16)
Lb0(p): a notional line-of-sight loss for the path evaluated from equation (16).

η = 2.5

Lminb0(p): the notional minimum propagation loss that the modified ducting/layer
reflection loss can attain:

Annex 2 Page 66 of 68
Removal of discontinuities from ITU-R P.452

 Lb 0 ( p ) for p < β 0
Lminb 0 ( p ) = 
 Lbd 50 − (Lbd 50 − Lb 0 β )⋅ Fi
(32)
for p ≥ β 0
where:
Lb0β: the notional line-of-sight loss evaluated at β0% time from
equation (9):
Lb 0 β = Lb 0 ( β 0 %) (33)
Lbd50: the diffraction loss evaluated at 50% time from equation (14):
Lbd 50 = Lbd (50%) (34)
Fi: interpolation factor based on a log-normal distribution of
diffraction loss defined in equation (13c)

Fj: interpolation factor which blends the modified ducting/layer reflection


into the notional line-of-sight loss:
  ξ⋅ 
F j = 1.0 − 0.5 ⋅ 1.0 + tanh 3.0 ⋅ (θ − Θ )  (35)
  Θ 
where:

Θ =0.3
ξ = 0.8
θ: path angular distance defined in Table 7.

5 Renumber the remaining equations accordingly.

__________________

Annex 2 Page 67 of 68
Removal of discontinuities from ITU-R P.452

__________________

Annex 2 Page 68 of 68

You might also like