On the Application of .), .

Stereographic Projection to the

Representation of Moving Targets
in Air Traffic Control Systems

Robert G. Mulholland
'.

DOT/FAA/CT-TN85/67

Document is on file at the Technical Center

Library, Atlantic City Airport, N.J. 08405

u.s. Deportmenf of Tronsportoticn

Federal AvIatIon AdminIsllation
Technical Center
Atlantic City Airport, N.J. 08405

, q '­
 Techniclli ~eport Documentation PClge
1. Report No. 3. Rocipiont'. Co'olot No.

DOT/FAA/CT-TN85/67
4. 'Titl. ond Suotit'o S. Ropolt /)0'0
ON THE APPLICATION OF STEREOGRAPHIC PROJECTION TO THE February 1987
REPRESENTATION OF MOVING TARGETS IN AIR TRAFFIC 6. ~ Porlorminll Orgoni ~otion Codo
CONTROL SYSTEMS ACT-130
h;---:--:-":""":'----------------------------!
7. A",horl.)
a. P.r/orminll Orllon; lotion Roport No.
I
Robert G. Mulholland DOT/FAA/CT-TN85/67
9. P.r/orminll Orgonization Name and Addro .. 10. Work Unit No. (TRAISI

Federal Aviation Administration

11. Contract or Grant No.
Technical Center
Atlantic City International Airport "" /M
Ar1~"r;(' r.;rv N,T ORL..OC;
12. Sponsorint A\loncy Namo ond Addro••
U.S. Department of Transportation Technical Note
Federal Aviation Administration 1984 - 1985
Advanced A~tomation Program Office 14. Sponlorint Atoncy Codo
Washington, DC 20590
15. Supplemontary Notos

16. Abstract

In many ~nstances,
a~r traffic control is based on the ~dea that the latitude and
log~tude of an a~rcraft can be represented as a oo~nt ~n a cartes~on plane v~a
the method of s t e r eog r aph i c pro j ec.t i.on , ,In practice, an approximation of t h i s
representat~on ~s obtained through the orocess~n2 of measurements of the alt~tude
of the aircraft and lts slant range az~muth relative to a radar of known
pos~tion. The d~fference between the aooroximation and the -actual representation
of aircraft locat~on ~s viewed as a projection error. In the case of the
Advanced Automation System (AAS), the scheduled replacement for the National
A~rspace System (NAS) des~gn specifications limit the project~on errors to 0.005
naut~cal m~le (nm~). Unfortunately, the dara orocess~ng methods now employed ~n
NAS are insuff~c~ent to meet this requirement.

Project~on errors Ln NAS are examined, and methods for reduc~ng them are
conSLdered. It ~s shown how a simple modifLcatlon of the current NAS methodlogy
can be used to achieve the accuracy Lmposed on the AAS design.

17. Koy Word. 18. Distribution Stot_ _ t

Stereographic Projection Document is on file at the Technical

Air Traffic Control Center Library,Atlantic City International
Airport, New Jersey 08405

19. Socuri'y Clossi/o (01 ,"i. roport) 20. ~ocurity C10... I. (01 .hi. patol 21. No. 01 POll" 22. Prico

Unclassified Unclassified 47
Form DOT F 1700.1 (8-72l Reproduction of completed page authori led
 TABLE OF CONTENTS

Page

EXECUTIVE SUMMARY v

1. INTRODUCTION 1

2. TARGET LOCATION PARAMETERS 2

3. ELEVATION SURFACES 4

4. DEVIATION SURFACES 6

5. COVERAGE REGION 8

6. STEREOGRAPHIC PROJECTION IN THE LOCAL RADAR PLANE 11

7. GROUND RANGE ESTIMATION'WHEN ALTITUDE IS UNKNOWN 13

8. GROUND RANGE ESTIMATION WHEN ALTITUDE IS KNOWN 15

9. STEREOGRAPHIC RELATIONSHIPS 17

10. PROJECTION ERROR IN MAS~ER PLANE 19

11. EFFECT OF MEASUREMENT ERRORS 24

12. CONCLUSIONS 27

REFERENCES 29

APPENDIXES

A - Slant Range Extremum ~n Altitude Window

B .: Error Bound

C - Ell~pso~dal Earth Model

D - Bound on Approximation Error

iii
 EXECUTIVE SUMMARY

Horizontal separation of aircraft under the control of a .single Air Route

Traffic Control Center in the National Airspace System (NAS) is accomplished by
controlling the relative separation of points in a plane that represent actual
aircraft locations. Such a representation is supposed to be the image of the
orthogonal projection of an aircraft onto the mean sea level surface of the
earth under a stereographic~apping. In practice, th~ system implementation of
the mapping is imperfect. Hence, even if the aircraft location is known exactly
in terms of surveillance data, i.e., altitude, slant range, and azimuth, there
is a difference between the system representation of the aircraft and the image
of actual aircraft position under the stereographic mapping. This difference
constitutes a projection error, and the magnitude of the error determines how
well the system representation of an aircraft reflects its location in terms of
latitude and longitude. While the NAS implementation of the stereographic
mapping is simple and computationally efficient, the corresponding projection
errors exceed limits imposed on the design of the Advanced Automation System
(AAS) that is to replace NAS.

The ensuing report examines sources of projection error in NAS. Implementations

of the stereographic mapping other than that employed in NAS are considered.
Both spherical and ellipsoidal models of the mean sea level surface of the earth
are taken into account. An implementation of the stereographic mapping is
disclosed that retains much of the simplicity of the NAS design and yet meets
the accuracy requirements of the AAS specification in the context of an
ellipsoidal earth model.

v
1. INTRODUCTION.

This report is concerned with projection errors in 2-dimensional representations

of airspace employed for air traffic control (ATC) purposes. Some 20 Air Route
Traffic Control Centers (ARTCC's) dispersed across the contiguous United States
provide ATC services to en route aircraft operating under flight plans filed
with the National Airspace System (NAS). The control jurisdiction of a single
ARTCC consists of the combined coverage regions of many radars, and involves
hundreds of thousands of square miles of the surface of the earth. Range and
azimuth data acquired at the radar sites in connection with the location of
airborne targets is transmitted to the ARTCC. Additional altitude information
is supplied to the ARTCC for those aircraft equipped with Mode C transponders, a
mandatory requirement in airspace above 12,000 feet. This surveillance
information is used to create a representation of aircraft positions consisting
of points in a 2-dimensional coordinate system referred to as the system plane
or the ARTCC master plane. Separation of aircraft at the same altitude is
effected by controlling the relative positions of the points.

From a theoretical viewpoint, the system representation of a target should be

the same as the image of the position of the target above the mean sea level
surface of the earth under an explicit mapping of stereographic origin.
However, the system implementation of the mapping is imperfect, and measurements
of target radar coordinates are rarely exact. As a result, there is a
difference between the system representation of the target and the image of
target position under the mapping. This difference constitutes a projection
error. While there are sources of projection error other than inexact
measurements and imperfect implementation of the mapping, they are beyond the
scope of this work.

The objective of this report is to disclose some implementations of the

projection mapping besides the one created by the designers of NAS. In the case
of NAS the implementation is simple and computationally efficient. On the other
hand, the corresponding projection errors exceed limits imposed on the design of
the Advanced Automation System (AAS) that is scheduled to replace NAS in the
next decade. Hence, it is natural t9 ask how one might alter the NAS
implementation without materially destroying the simplicity of the design and
yet achieve an accuracy that will meet the needs of the AAS. This report is a
step in that direction.

The remainder of the report is divided into 12 sections that deal with
projection error in the context of a spherical model of the mean sea level
surface of the earth. Extensions to an ellipsoidal model and details of a
mathematical nature are relegated to four appendixes. The next three sections
review some basic target location parameters and relationships between them that
are used in succeeding developments. The projection error can be expected to
exhibit some dependance upon target altitude and the position of the target
relative to the antenna from which slant range and azimuth are measured. Hence,
some quantitative formulation of what is meant by radar coverage region is
essential to an analysis of the error. Such a formulation is provided in
section 5. The mapping that carries target positions into the master plane can
be regarded as a composition of two maps (references 1 and_2). One of these
stereographically projects airspace into a so-called local radar plane tangent
to the surface of the earth at the latitude and longituqe of the radar site from
 "

l \
R." 0 "IL P I.ATf"oA. IlI\
p\'''''C

FIGURE 1. LOCATION PARAMETERS

3
as S increases from O. Otherwise, it steadily decreases as S increases toward
the critical value

.A
e
<.~,)
= (E + H
R
) ~ '+'
(6)

, ~ !

where it takes on its minimum value

),
,.......,.,.., ( ~,) :. <. E -+ H
R
)
~'Y - E. (7)

In fact, at the critical slant range the w-elevation surface is tangent to the
sphere of radius E + hmin (W) about the center of the earth. As S
continues to increase, the function h(S, W) rises, and it passes through HR as
the slant-range moves through 2s c ( W) .

The relationship

2­

+ (s + <. E· "'R) ~ YJ 1 (8)

can be obtained by completing the square in (3). The solution of (8) for slant
range leads to the functions

and

1. a ,/2.
A ~H,'IJ) "~"E+H) -<'E-+~R·;Z.c:.....a"1l \1-'1 - <E+'"'k' ~lf'. (10)
1,.:

If W is negative and

(11)

then slo (H, W) (shi (H,. W)) represents the smaller (larger) of the .
two possible slant ranges for a target on the w-elevation surface at altitude H.
If either W> a or else W< a and the altitude does not satisfy the constraint
(11), then - slo (H, W) has no physical meaning. On the other hand, when

5
 (16)

about the antenna, and

\, (SiS) ':: <E"i H ) ~$4 - E (17)

c. R

is the altitude of the locus of the points of tangency. Consequently, there are
two possible altitudes for a target on this surface at a slant range S for which

(18)
~ E' of H
R·

These are

lIz.
\, l¢ > ~ -(s
z.
.A 2..
~
...,s" J "'" \,c (~.., (19 )
.l.r

and
1/2.
2.
\, 52-
k
t.¢ ) -= I A
~
(¢>J +'n<sz').
e (20)

When S exceeds E + HR there is only one possible target altitude, and it can
be obtained from (20). Slant ranges of this magnitude are not encountered in
ATC applications.

In the case where S ~ E + HR, the preceding remarks imply the existance of
target locations at slant range S for which the line segment connecting the
target to the antenna is perpendicular to the line segment connecting it to the
center of the earth. These locations correspond to the circular locus of points
defined by the intersection of the sphere of radius S about the antenna and the
,-deviation surface that is tangent to the sphere. The altitude of each of
these tangency points is given by

E (21)

which can be obtained from obvious geometric considerations or the solution of

(16) and (17) for h c(') in terms of 8 mi n (,). Obviously, ,(S,H p (8))
is the maximum deviation angle that can be attributed to a target at slant

7
 f L'Alr;-OF'-
HO.e12.0N
t, .."..,.
.l"aFAe:.tF

CO-JERAGE REGION
9
 o ~ e <'211'. . (30)

As will be seen, this provides a foundation for evaluating the performance of

surveillance techniques based on the method of,stereographic projectioq. For
example, there will be subsequent references to the minimum and maximum
altitudes ~. (S) and H (S) that can be assumed by a target in
1n. ''1llaX • • •
the coverage reg10n at slant range S. It can be eas1ly ver1f1ed that these are
provided by the following algorithms in the practical case where ~max is
positive. The algorithms are written in terms of simple PASCAL assignment
statements (reference 4) with the understanding that the symbolic
representations of mathematical functions on the left (right) side of each
expression are to be treated as real varibles (functions).

Algorithm 1. Maximum Altitude

if (( M >H ) ~
(,M. R

else

Algorithm 2.- Minimum Altitude

if

else
.~ so

6. STEREOGRAPHIC PROJECTION IN THE LOCAL RADAR PLANE.

A target can be stereographically projected onto any plane tangent to the

surface of the earth or, for that matter, tangent to the surface of any sphere
about the center of the earth. The phrase "local radar plane" will be used to
refer to the plane tangent to the surface of the earth directly beneath the

11
 -,/z
'\ (S... H) '= R<. S) \0" \: I")( <. S... \-\)1 (33)

where
\

1. a i l z:
R (S.. ~) =ts - (~- '"' R) 1 (34)

and

.-,( (S H) (35)
J

Thus, target slant range, altitude, and azimuth can be used to compute the
complex representation z(S, H, e ) of the stereographic projection of the target
onto the local radar plane via (31) and (33).

7. GROUND RANGE ESTIMATION WHEN ALTITUDE IS UNKNOWN.

Let ~min(S) ( ~max(S» represent the minimum (maximum) deviation

angles that can be attributed to a target in the coverage region at slant range
S from the antenna. It follows from (32) that

is the arithimetic mean of the maximum and minimum ground ranges that can be
associated with such a target. Moreover, the maximum value of the difference
between rm(s) and the true ground range of the target is

(37)

Thus, when target altitude is known only to the extent that it lies between the
minimum and maximum altitudes Hmin(S) and H x(S) that can be
attr1i b uted to a target 1n
" the coverage region
ma at slant range S, then rm(S) can
be used as an estimator of target ground range, and the absolute value of the
corresponding estimation error cannot exceed Ei:n(S). This estimator is optimal
in the sense that the maximum error associated with any other estimate of ground
range cannot be less than em(S).

13
viewed as an attempt to approximate rm(s). However, in the case where the
slant range exceeds shi (M, ib ), a much closer approximation can be
.
obta1ned max.
from a least squares f1t of a low order po I ynom1a
. If ·
unct10n 0 f 1 / S to
samples of rm(S). In the cases where the slant range falls below Shi(M,
ib max ) ' our experience indicates that rm(S) is essentially a linear function
of S. However, there is a problem with polymomial fitting in the sense that
~m(S) does change with the configuration of the coverage region, and so one
cannot expect the same pair of polynomials to apply to all radar 'si t ea; in NAS.
Thus, one is left with the usual problem of deciding whether the benefit to be
derived from optimality is worth the effort needed to achieve it.

8. GROUND RANGE ESTIMATION WHEN ALTITUDE IS KNOWN.

When both altitude and slant range are known, the ground range can be obtained
from (33). However, there are many ways in which this computation can be
carried out. To the extent that computational speed and consumption of
computational resourses are important, some of these may be desirable and others
not so desirable. In what follows, an approach to the determination of ground
range will be described that is easy to implement in a form that provides quick
results accurate to within 2 m.

There exist efficient high speed techniques (reference 6 and -7) for evaluating
the factor R(S, H) in the formula (33) for ground range. The remaining factor
can be expressed as a polynomial in xes, H) by a straight forward application of
Taylor's theorem with a remainder to the expansion of the reciprocal of the
square root of x about an arbitrary positive number x o• In" other words, the
ground range can be expressed in the form

(42)

where ~ is a number between Xo and xes, H),

r ,..,..
($ H
J )
'1<)
e
(43)

Dk(t) is the k th derivative of (l/x)t evaluated at x = t, and

Thus, if rn(s, H, xo) is used to estimate the ground range then the
corresponding estimation error is given by (44). Unfortunately, ~ is unknown
and so the error cannot be evaluated. On the other hand, it is possible to find
an upper bound on the absolute value of the error that is valid for any pair
(S, H) corresponding to a target location in the coverage region, and, in the
case of NAS, this bound is on the order of a meter when both Xo and n are
assigned the value 1. As a result, the estimator

15
and

(52)

are satisfied throughout NAS. Hence, by using the right sides of constraints
(50)-(52) in place of Smax, ~R, and M in (46)-(49) an upper bound on the
estimation error can be obta1ned that is valid for all radar coverage regions in
NAS. For example,

(53)

and

I. b 0 (54)

From (54) it is concluded that rl(S H 1) is essentially the same as the

ground range over the coverage region of any radar site 1n NAS. On the other
hand, the same conclusion cannot be drawn for roeS, H, 1) from (53).

Although the projection error associated with the estimate (45) is negligible in
the context of a spherical earth model, it does not necessarily follow that the
same will be true in the context of an ellipsiodal model. However, there does
exist an interesting possibility for employing the estimator (45) as a means for
essentially eliminating the projection error when ~he earth is represented by
the reference ellipsoid. This subject has already been touched on in the open
literature (reference 2), and the main features of the idea are provided in
appendix C.

9. STEREOGRAPHIC RELATIONSHIPS.

As already pointed out, an ARTCC is serviced by a multitude of radars, and

control of aircraft in the horizontal sense is effected through stereographic
representations of target locations in a single plane. This center master plane
is tangent to a so-called conformal sphere. The center of the sphere is
collocated with the center of the earth. However, the radius need not be the
same as that of the earth. Target representations in the master plane are
obtained from the stereographic representations of the targets in the local
planes associated with the radars that support the ARTCC. For example, using
appropriate coordinate systems in the local and master planes with origins at
the points of tangency, the master plane representation of a target associated
with the complex number z in a local rad~r plane is given by

(55)

17
10. PROJECTION ERROR IN MASTER PLANE.

Suppose now that ze is the estimate of the stereographic representation of the

target in the local radar plane. For example, it might be the product of
rn(s, H, xo) and the phasor exp (i 6). Also supppose that WN(E x, ze)
is used to estimate the stereographic representation of the target in the master
plane. Then

(E ~)
?C I ri,A - \.I.r (59 )

represents the corresponding projection error. Using (55) and (58), it follows
that

'N \E ) ::.
BI~IZ1[ \E/E J
'" lSI.
tAo\- ) '"D (60)
J
?C .A - J C]

where

... -,
C :=
t 1- .B \.II"'
0 j 1 (61)
,;;

and
I - r. cE IE ~
) vr0
.. B ]A 1
N
'"'

.J):
~
1- (E'/e "1( ) v.rQ
J3 J.o. (62 )

Clearly, (60) cannot be obtained from (59) in the event that either of the
factors Wo*Bz and (E/Ex)wo*Bz e is one. As already indicated, this
is not a problem in the case of the former factor, and, as will be shown, the
same is true in the case of the remaining term. While formula (60) is exact, it
does not provide much insight into the relationship between ze, Ex, and the
projection error. In what follows, consideration will be given to an
approximation that does provide such insight.

In practical situations both C and D are close to unity. Consequently, one is

tempted to conclude that the projection error can be closely approximated by

(63)

Unfortunately, this is not necessarily true. For instance, one can construct
practical examples for which the absolute value of the difference between

19
or 200 nmi so long as the target is within the coverage region of the radar. On
the other hand, the reciprocal of L is typically on the order of thousands of
nmi. Consequently, the second restriction is not a serious impediment to the
validity of the bound. Incidently, (67) together with the fact that the radius
E of the spherical earth model must be between a and b implies that the
magnitude of the factor (E/E x )wo*Bz e is less than 1. Thus, as
indicated before, there is little reason to be concerned with the indeterminate
s i tua t i on in .connectionwith. the expression (60).

When the order of the approximation is 1 the bound is given by

( 68)

When N exceeds 1 it is given by

(69)

where

L 'J' t. \ + r ~'- L\Jl))

(70)
( ,- L 'J ' ) ( ,- i L 'J' )

and P is any number satisfying the inequality

(71)
r II ( L '$ I ) •

The existence of P is guaranteed by the restriction (,67). Moreover, since F is

an increasing function of P it is clear that the same is true'of the bound W,
i.e., the bound is tightest when P is set equal to the left most member of (71).
For future reference it is pointed out that the definition (66) of Limplies
that both U and F increase monotonically wi th I wol and I zl so long as the
constraint (65) on LI zl is satisfied. An example will show how this
monotonicity can be used to evaluate the accuracy of our approximation of the
projection error.

The accuracy of the approximation (64) will be illustrated in the context of the
ErrRouteAutomated Radar Terminal System (EARTS) mosaic model for Alaska
(reference 8). According to the model, there are 15 radars supporting the
ARTCC, the radius Er of the conformal sphere is 3395.7 nmi, the maximum
effective slant range of each radar is no more that 200 nmi, and the distance of
the stereographic representation of each site in the master plan from the point,

21
when the approximation is expressed in m. Consequently, when the approximation
order is greater than 1 the magnitude of the error is at most 7.04 m plus about
one percent of the magnitude of the approxi~ation itself. Clearly, the accuracy
of the approximation (64) is pretty good.

For the most part, one is interested in the magnitude of the projection error
and so it is convenient to have a simple expression for the magnitude of
gN(Ex,ze). Such an expression is easy to derive in the practir~l case I' .
where the phase of the estimate ze is the complement of the target azimuth e •
Then, at least in the context of a spherical earth model, the phases of z and
ze are identical. Since (56) and (57) imply that the phases of A and Bare
always the same, namely, - a , it follows from (63) that the magnitude of the
expression (64) is invariant to azimuth when the approximation order exceeds 1.
On the other hand, when the order of approximation is 1 it is apparent that the
magnitude varies with azimuth and is maximum when the azimuth is v radians out
of phase with the compl~ment of the sum of a and the phase of wOo As a
result,

= ( \ A \ -+ \ B \ \ ~C> \ z 1\ \ c / E. IX ) I J.A.' - \J I \ (76)

and

---~

N= I

(77)

whenever the phase of the estimate of the stereographic representation of the

target in the local radar plane is the complement of target azimuth.

The relationship (77) can be used to illustrate the importance of the

distinction between the radius of the earth and that of the conformal sphere in
the transformation (55) relating the stereographic representations of the target
in the local radar plane and the master plane. In particular, great pains were
taken by the designers of NAS to make the ground range estimate I zel as close
as possible to the true ground range I zl of the target. However, even if the
true ground range and its estimate are the same, a poor choice for Ex in the
approximation (58) can cause projection errors on the order of a nmi. For
example, suppose the radius of the conformal sphere is assigned to Ex ~s is
the current practice in NAS. Then
2­

~
rl E
_"Y +
\ v.i. \
\ \ 'J \
of LtEE (78)
"'r

when the estimate I zel of' ground range is exact. In terms of the Alaskan

23
where

~e (82)
J

A'd = [C s oS ~ c", 4"1 <::..crq. e - ~ (sJ'"'J \) 66 Ai.-.. e

)
(83)

and Cs (CH) is the partial deriviative of rl(S, H~ 1) with respect

to S (H) evaluated at the actual slant range and azimuth of the target. This
expression can be viewed as a first order approximation of the error in the
estimate of the actual stereographic representation of the target in the local
radar plane due to measurement errors. This is not the total projection error.
The total error is given by

(84)

where

(85)

is just the projection error in the case for which the measurements are exact.
To the extent that the first order approximation of the component of the error
due to inexact measurement is valid, the total error in the local plane can be
represented by the sum of ~(S, H, a) and ~ze' Needless to say, the magnitude
of the former is just the absolute value of the difference between the true
ground range of the target and rl (S, H, 1). As already pointed out in
section 8, this is, at most, a few meters for coverage regions of the type
encountered in ATC applications.

The projection error in the master plane can be approximated by substituting

eL for Ze in formula (64). As already demonstrated, the approximation error
associated with this formula is negligible for current ARTCC control
jurisdictions. Also, as pointed out early, it is good design practice to assign
the radius of the earth to Ex regardless of the order N of the polynomial that
is used to map the local plane into the master plane. With this understanding,
the total projection error in the master plane associated with the Nth order
transformation polynomial can be represented by

25
length Aal lies along the line segment connecting the representation of the
target and the antenna in the local radar plane. The result is

_ (.)"''2. /2.) ( 91)

~\"') :. \- f.
, '
It is left to the reader to derive a similar relationship for the master plane
from the relationship (86).

It is emphasized that (91) is based on the assumption that (81) does indeed
qualify as an accurate approximation of the difference between Ze(S+6S,6H+H,e+6~)
and Ze(S, H, e). Clearly, this is not the case when the measurement errors
are large. Consequently, one might do well to at least demonstrate the validity
of the relationship before using it as a design tool.

12. CONCLUSIONS.

There are many ways of going about estimating target ground range in the local
radar plane in the case where altitude is unknown. The approach to ground range
estimation described in this report is optimal in the minimax sense. The
current method employed in NAS LS optimal in the same sense only in the context
of a flat earth model. It is, at best, suboptimal in the context of a
spherical model.

While ground range can be determined exactly when altitude is known, it has been
shown here that computations can be greatly simplified by resorting to an
approximation of the.exact formula that is accurate to within 2 m for any target
in the coverage region of the radar. This, together with a special method for.
calculating the difference between S2 and (H-HR)2, can be used to formulate an
efficient, high speed algorithm for estimating the stereographic representation
of the target in the local plane. Numerical studies (appendix C) indicate the
existence of extensions of the algorithm to the ellipsoidal earth model for
which the projection error can be essentially eliminated under the geometric
limitations imposed by ARTCC control jurisdictions and ATC radar coverage
regions of practical import. One of these extensions involves a ground range
correction factor that limits the ground range error to values less than 5 m.
The corresponding error in phase involves angles that are less than 0.006°.
~hese bounds are. cosiderably less than the slant range and azimuth quantization
'i nte rva l s vof 0.0078 nmi and 0.022° associated with the Mode S s'y s t em that is
scheduled to replace the current operational ATC radar beacon system.
Unfortunately, at ground ranges of 200 nmi from the antenna the phase error can
lead to projection errors on the order of 0.02 nmi which exceeds the 0.005 nmi
limit imposed on projection errors by current AAS requirements. However, by
introducing a phase correction factor in addition to the ground range
correction, it is possible to hold the projection errors in both the local radar
plane and the ARTCC master plane to levels that are well within AAS
specifications. In fact, this can be accomplished in such a manner that the
master plane magnification factor (reference 2) is optimized. Insofar as size
of the control jurisd~ction is concerned, the 2500 x 2500 nmi limitation imposed
by the AAS specification on the surveillance coverage region of the Area Control
Computer Complex does not present an obstacle to successful application of the
range and phase correction fa c t or s-,

27
 REFERENCES

1. Mulholland, R. and D. Stout, Numerical Studies of Conversion and

Transformation in a Surveillance System Employing a Multitude of Radars, Part I
and II, Astia Documents AD A072-085 and 086, National Technical Information
Service, Springfield, VA, April 1979.

2. Mulholland, R. and D. Stout, Stereographic Projection in the National

Airspace System, IEEE Transactions on Aerospace and Electronic Systems,
January 1982.

3. Abramowitz, M.· and 1. Stegin, Handbook of Mathematical Functions, National

Bureau of Standards, Washington DC, 1966.

4. Koffman, E., Problem Solving and Structured Programming in PASCAL, Addison

Wesley, Re~ding, MA, 1981.

5. Saleh, N., A. Smith, B. Sokkappa, The Stereographic Projection in the

National Airspace System: Principles, Approximations, and Errors, MTR-83W67,
The MITRE Corporation, Mc Lean, VA, May 1983.

and Implementation of Efficient

Corrected Slant Ran e, Report No.

7. Stout, D., and~. Mulholland, Approximation of Corrected Slant Range in a

Radar Surveillance System, IEEE Transactions on Aerospace and Electronic
Systems, September 1981.

8. Test Report for EARTS Tracking Improvement, Contract FAA No.

DTFAOl-84-C-00007, NAS Modernization - En Route Automateo Radar Tracking System
with Mosiac Software (EARTS-M), Sperry Corporation, July 1984.

9. Advanced Automation System, System Level Specification, Design Competition

Phase, FAA-ER-130-005E, August 1984.

29
 APPRENDIX A

SLANT RANGE EXTREMUM IN THE ALTITUDE WINDOW

Let J represent the~et of points in the intersection of the beam and the window
comprising all altitudes from 0 to M. Suppose there are targets in J at
altitude H. Since slant range increases with increasing deviation a~gle at
constant altitude it follows that among all such targets the one with the
greatest deviation angle will be farthest from the radar.

Now consider the case where ~min is less than the elevation angle of the
line-of-sight horizon. Clearly, the target at altitude H in J at the greatest
distance from the antenna must lie on the ~o-elevation surface, and the
corresponding slant range is Shi(H,~o). But formula (10) implies this is
an increasing function of altitude. Hence, the greatest distance Sl between
the antenna and a point J must be Shi(M, ~o). In like manner, it can be
shown that Sl . (.1.) .1.
1S shi M, ~min when min ~ ~o.

Finally, since the slant range of a constant altitude target decreases with
decreasing deviation angle, shi(M,l/J) is a decreasing function of elevation
angle, and so shi(M, l/Jo) exceeds shi(M, l/Jmin) if, and only if,
!/Jmin >!/J O• Thus, in general, Sl is given by (25).

A-I
 APPENDIX B

ERROR BOUND

Consider any pair (S, H) corresponding to a target in the coverage region.

Expression (34) implies that R(S, H) cannot exceed the maximum slant range
Smax of the radar. Moreover, as will ·be seen, xes, H) is bounded above by
v and .be.l ow by u , Al so , ' the magnitude (47) of ' the (n+l)thderivative of lWis
a decreasing function of t. Consequently, (46) must be an upper bound for
en (S, H, x o, ~ ) •

In order to establish u as a lower bound for xes, H) one need only note that
Smax is certainly less than the polar radius of the earth in ATC
applications and the latter is not greater than E. Consequently, (35) implies

2­
/X ( S,J ~n ~ I - l S~ / (2. E) ~ ~ ~. (B-1)

The upper bound v can be obtained from (35) and the fact that M 15 the max1mum
target altitude in the coverage range, 1.e.,

I')(. (S 1..1) < (B-2)

J
IV'" •

B-1
 APPENDIX C

ELLIPSOIDAL EARTH MODEL

Under the assumption of an ellipsoidal earth model the altitude of a point above
mean sea level is the distance separating the point from its orthogonal
projection onto the ellipsoidal surface. The geodetic coordinates of a target
at altitude H are specified in terms of the triplet (L, ~,H) where L and A are i ,
the geodetic latitude and the longitude of the target. The latitude measures
the angular deviation from the equatorial plane of the normal to the ellipsoidal
surface at the project~on of the target. Thus, it specifies the location of the
projection to within a circle about the polar axis of the earth. The longitude
represents an angle measured in the plane through the circular locus from the
projection to a fixed half plane defined by the polar axis and a predetermined
point on the surface of the earth. As a result, latitude and longitude uniquely
specify the projection of the target on the surface of the earth and this,
together with altitude, provides the location of the target itself.

If it is imagined that the lengths of the axes of the ellipsoid approach one
another, then L approaches the angular deviation from the equatorial plane of
the line segment connecting the target to the center of the earth. In fact,
points on the ellipsoid are often associated with points on the surface of a
so-called conformal sphere abottt the center of the earth. Specifically, the
point (L, A, 0) on the ellipsoidal surface is equated to the point on the
surface of the sphere at longitude A and latitude VeL) where

. L -.. J6
-z.,. cc-i»

is the mapping defined by the equation

(C-2 )

and ~ is the eccentricity of the ellipsoid (references I and 2). The angle VeL)
is sometimes referred to as the conformal latitude of the target. As will be
seen, this concept bears directly on the definition of the local radar plane in
the context of an ellipsoidal earth model.

Consider a radar with geodetic coordinates (L s , As, HR)' The antenna axis
corresponds to the normal to the ellipsoidal surface at (L s , As' 0), and
rotation is commonly in the direction counter to that dictated by the right hand
rule with respect to the direction along the axis at the radar away from the
ellipsoid. The azimuth of a target is determined by two half planes having the

C-I
 as the range and phase components of the error. If the lengths of the
semi-minor and major axes of the ellipsoid approach the common value E, then
(C-4) reduces to the projection error in the context of the spherical model, the
angle error vanishes, and, as pointed out in section 8, the range error is
negligible for radar coverage regions like those encountered in practical ATC
.app l i ca't Lons , . However, it still r emai.ns to show how the projection error for
such coverage regions can be neutralized in the situation where the mean sea
level surface of the earth is represented by the reference ellipsoid.

If we employ the reference ellipsoid as the earth model and assume that the
coverage region of the radar is constrained by inequalities (50) - (52), then
the range and phase components of the projection error can be represented by
sinusoidal functions of the phase of the true representation z of the target in
the local radar plane with amplitudes that vary with the magnitude of the
representation. In particular, consider the case where z is constrained to lie
on a circle about the point of tangency of the local radar plane, target
altitude is held constant, and the phase of z is increased from 0 to 2v
radians. Empirical data (references I and 2) suggest that both the range and
phase errors oscillate in an almost sinusoidal like manner. The same data also
suggest that the amplitude of the range (phase) error oscillation is a nearly
quadratic (linear) function of ' the radius of the circle. Although the
ampitudes of the oscillations do vary with the geodetic latitude of the radar
site, the radar site altitude, and the altitude of the target, the variation
with target altitude is extremely small over the values assumed by this
parameter in ATC applications. In more explicit terms, empiri~al data suggest
that the approximations

(C-7)

and

(C-8)

are highly accurate representations of the range and phase components of the
projection e~ror wh~re the coefficients ao, aI, a2, bO' and bl' are functions of
the radar s~te lat~tude Ls and the radar site altitude HR.

The coefficients aO through bl' can be determined by applying least squares

methods to samples of the range (phase) errors corresponding to a single target
altitude and a sequence of values of z along the positive half of the imaginary
(real) axis of the local plane coordinate system where arg(z) is rr/2 (0)
radian. The target altitude chosen for this task might be half the maximum
altitude of interest within the coverage region of the antenna. For example,

C-3
 into the master plane. The answer to this question is affirmative. In
particular, 'the reader need only refer to the formula provided in reference 1
for the impact of the projection error in the local plane on the corresponding
error in the master plane. A little thought will lead to the conclusion that
the projection error in the master plane associated with the use of (C-9) and
(C-lO) in the estimation of the local plane representation of the target is
essentially the same as the corresponding error in the local plane under the
. usually. constraints imposed on the geometry, of an ARTCC control jurisdiction.
The recent numerical work performed at the FAA Technical Center substantiates
this conclusion for cases where. the radar site is located at distances up to
1460 nmi from the latitude and longitude marking the so-called master plane
point of tangency.

It is emphasized that the accuracies claimed for the estimates (C-9) and (C-lO)
are based on empirical data rather than a mathematical derivation. Moreover,
the data were collected for selected radar site latitudes in the northern
hemisphere for which the coverage region of the antenna does not include the
north pole or the equator. We do not expect the relationships (C-9) and (C-lO)
to hold for radar sites near a pole for which the coverage region involves
points on the surface of the ellipsoid on both sides of any plane passing
through the polar axis. Likewise, these approximations cannot be expected to be
valid for sites in the neighborhood of the equator where the coverage region
involves portions of the earth'surface in both the northern and southern
hemispheres.

Finally, the projection error in the master plane is dependent upon the -.
..
corresponding error in the local plane as well as the bilinear transformation
used to map the l~tter plane into the former. The transformation is itself a
function of E and another parameter ~r that is sometimes referred to as the
radius of the conformal sphere supporting the master plane (references 1 and 2).
If Er is improperly evaluated, then there is no guarantee that the projection
error in the master plane will be commensurate with rhat in the local plane.
This problem can be avoided by choosing Er to optimize the so-called
magnification factor in accord with the method outlined in reference 2. In
fact, no difficulty whatsoever should be encountered for the control
jurisdiction geometries cited in the AAS specification, and there is good reason
to believe that the approach to the projection problem outlined in this appendix
can be successfully applied to the case where the size of the jurisdiction is
extended well beyond the limits prescribed by the specification for the
surveillance cov~rage region of ~he Area Control Computer Complex.

C-5
 (0-6)

corresponds to (63) and

(0-7)

By expressing x as a function of z through (0-1) and using the fact that E lies
between a and b, it can be verified that I uI is upper bounded by U of (68).

Turning to the case where N exceeds 1, it will be noted that

N
(0-8)
[
\.~ I

where

(0-9)

and

As a result, (0-4) can be written as

(D-11)

where

(D-12)

0-2
Under the constraint (65) the magnitude of v is upper bounded by that of u.
Hence, U of (68) must upper bound I vi. Thus, to establish the validity of the
bound (69) it only remains to verify the inequality

(D-13)
F

Toward achieving the last stated goal we point out that

~ (D-14)
''d I

Moreover, by multiplying the leftmost inequality of the relationship (71)

through by I wo*~ and recalling that the radius of the spherical earth model
cannot exceed a, it is apparent that

(D-15)

Consequently,

..k ~-,
~ lrx \ -I [
= (D-16)
,'" ,~-,
.~
~:o
U-f )
r
'-, .~ "¢ I

From this relationship and the obvious inequality

(D-l7)
I d(')(" IJ' I
the objective can be achieved.

In what follows, we verify (D-13) for the case where P is not one. Verification
of the inequality for the remaining case is left to the reader. When p is other
than one the relations (D-16) and (D-17) imply