Beruflich Dokumente
Kultur Dokumente
AD NUMBER
AD903969
distribution
FROM
Distribution authorized to U.S. Gov't.
agencies only; test
and evaluation; 13 Oct
1972. Other requests shall be referred to
U.S. Army, Attn: SMUP-L3300 Frankford
Arsenal, Philadelphia PA 19137.
AUTHORITY
FA,
9 Oct 1974
DATA SYSTEMS
Litton
FINAL REPORT
SUPPLEMENT TO PARAMETRIC STUDY OF
SYSTEM
SWEAPON(AFAADS)
GAYMEI, ANALYSIS
01
4P
ra
15 September 1972
Submitted to:
I
I
Prepared by:
Data Systems Division
Litton Systems, Inc.
8000 Woodley Avenue
Van Nuys,California 91409
TD 20871
-'
Fi:~rtj n
U
rf
L
ACKNOWLEDGEMENTS
Litton is happy to acknowledge the excellent guidance provided on this contract by Mr. Walter Ryba,
contract technical supervisor, and the substantive con-
I
I
i
j
II
I
t
I
:
j44
FOREWORD
"
I
I
I
I
!
I
I
I
"
vfvI'
I
I
TABLE OF CONTENTS
Section
Page
:2.I
2.2
CONCLUSIONS
2-1
2.1.1
2.1.2
General
Target Acquisition
2-1
2.1
2.1.3
Target Tracking
2-1
2.1.4
2.1.5
2.1.6
2-1
2-2
2-2
2.1.7
Development Considerations
2-2
2.3
2.2.3
2.2.4
2.2.5
2.2.6
2.2.7
2.2.8
2.3
2-4
2.4
2.5
2-5
2-5
2.2.9
Recommended Programs
2-5
3-1
3.1.2
3.1.3
3.1.4
3.1.5
3.1.6
3.1.7
3.1.8
3.1.9
Mask Angles
Visibility Angle
Stochastic Models of Terrain
Terrain Variations Along a Line Transect
Non-Gaussian Probability Density Functions
Stationary Processes in a Plane
Approximations by the Concept of Bandwidth
Level Crossing Methods
3.1
3-3
3-4
3-5
3.9
3-11
3-I 1
3-14
3-15
Night Operations
3-15
3.3.1
3.3.2
Visual Detection
Radar Detection
3-18
3-24
3.3.3
Acoustic Detection
3-27
3.2.2
3.2.3
3.2.4
SENSORS
4.1.3
3-15
3-18
3-18
1.18
3.4
TARGET IDENTIFICATION
TRACKING
4.1
RADAR
4.1.1
Radar Glint Analysis
4.1.2
Effect of Glint on Prediction Errors
3.1
3-I
3.1.1
13.2.1
3.3
2-3
2-3
2.2.2
3.2
I-;
2-1
INTRODUCTION
SUMMARY
I
2
~vii
3-29
4-1
4-1
4-1
4-5
4-7
1.,
Section
4-8
4.9
4-9
4-9
4-1 I
4-17
4-20
4-20
4-21
4-22
4-26
4-26
4-27
4-27
4-27
5-I
5-'
5-2
5-3
54
5-6
Acceleration Filter
5.1.5
Correction of Coefficients for Non-Unit Sampling Interval
5.1.6
5.1.7
Modification of Optimization Function for Non-Unit Interval
Effect of Varying the Sampling Interval
5.1.8
Comparing Filters for Different N
5.1.9
Example of Coefficient Variation with X
5.1,.10
Lag Corrections
5.1.11
Position Filter Lag to Constant Velocity
5.1,12
Velocity Filter Lag to Constant Acceleration
5.1.13
5.1I.14
Position Filter Lag to Constant Acceleration
Prediction A!gorithm
5. I.15
Coefficient for Least Squares Filters
5.1.16
Analog Filter
5.1.17
Generalization
5,1.18
RECURSIVE FILTERS
5.2
Transient Errors
5.2.1
Noise Variance Demerit
5.2.2
Settling Time Demerit
5.2.3
Effective Weighting by Aperiodic Filter
5.2.4
5.2.5
Recursive Filter for Acceleration
RECURSIVE FILTERS WITH TIME VARYING COEFFICIENTS
5.3
PREDICTION
OF FIRE
CONTROL SYSTEMS
6.1
CATEGORIES
6.2
PREDICTION USING
CORRECTIONS
FROM PROJECTILE TRACKINr-
5-8
5-10
5-10
5-10
5-10
5-11
5-1 I
5.11
5.12
5-12
5-12
5-13
5-14
5-14
5-15
5-17
5-17
-17
5.18
5-18
5-18
6-1
6-1
II
I
TABLE OF CONTENTS (Continued)
SSection
Page
6.2.1
6.2.2
6.2.3
6.2.4
Historicai PerspectivL
Automatic Projectile Tracking
Error Sensing with Radar
Error Sensing with an Imaging Tracker
6-2
6-2
6-3
6.3
6-2.6
64
6-5
6-10
6.2.5
6.2.7
6.2.8
S6.2.9
6.2.10
6.21
Simulation Module for Projectile Miss Measurement and Correction
PREDICTION USING TERRAIN INFORMATION
6.3.1
Lateral Aircraft Maneuvers
6.4
UNAIDED TRACKER CONTROL: 'GUNNER'S DLLIGHT'
6.5
THE SNAP-SHOOT GUNSIGHT: SYNTHETIC TRACER CONTROL WITHOUT
TRACERS
6.6
RATE BY TIME PREDICTION
6.7
COURSE AND SPEED SIGHTS
6.8
BARRAGE FIRE
6.8.1
The Remagen Defense
6.8.2
Model of Defense
6.8.3
Effectiveness Computation for Remagen Defense
6.8.4
Conclusions
6.9
IMPROVED BALLISTIC MODULE FOR SIMULATION
6.9.1
Siacci Approximation
6.9.2
Computing Power Law Approximations
THREAT CHARACTERISTICS
7.1
GENERAL CHARACTERISTICS OF THREAT
7.1.1
Numbers of Types of Aircraft
7.1.2
Air to Surface Missile Characteristics: The New Ball Game
7.1.3
Implications for Local Air Defense
7.2
ATTACK AIRCRAFT TACTICS
7.2.1
Delivery Maneuvers
7.2.2
Ordnance Selection
7.2.3
Relative Delivery Accuracy
i7.2.4
Nuclear Weapons Delivery Modes
7.2.5
Flak-Suppression Techniques
7.2.6
General Discussion of Air to Surface Delivery Tactics
7.2.7
Set Up Table for a Bombing Run
7.3
BOMBING ACCURACY AND MUNITIONS EFFECTIVENESS IN THE OLD
BALL GAME
7.3.1
Performance Summaries
7.3.2
SAAB Bomb Sight Characteristics
7.4
AIRCRAFT VULNERABILITY
7.4.1
Interpretation of Combat Loss a-id Damage Statistics
6.3
ix
6-4
6-10
6-10
6-12
6-13
6-14
6-1S
6-16
6-17
6-17
6-18
6-19
6.19
6.20
6-22
6-24
6-25
6.25
7-1
7-1
7-1
7-1
7-4
7-4
7-4
7-5
7-5
7-6
7-6
7-6
7-8
7-8
7-8
7.9
7-12
7-13
Page
7.4.2
Inferences from Aircraft Damage and Loss Reports
7.4.3
Inferences from Antiaircraft Gun Engagement Records
7.4.4
Development of Approximate Vulnerability Functions
7.4.5
Computer Target Vulnerability Module
7.5
SIMULATION MODULE FOR TERMINAL EFFECTIVENESS COMPUTATION
"7.5.1
Determination of Projected Area of an Ellipsoid
7.5.2
Coordinate Systems and Rotation Matrices
7.5.3
Relative Direction of Approach of Projectile to Target
7.5.4
Random Round to Round Dispersion in Angle and Muzzle Velocity
7.5.5
Angular Dispersion
7.5.6
Muzzle Velocity Dispersion
7.5.7
Angular and Muzzle Velocity Biases (Systematic Errors)
7.5.8
Discussion of the 'Diffuse Target' Representation
7.5.9
Computation of Single Shot Kill Probability
7.5.10
Simplified Exprissions for Tracking Module
WEAPON CHARACTERISTICS
8.1
SCALING RELATIONSHIPS
8.2
MUZZLE VELOCITY RELATIONSHIPS
8.3
METHODS OF OBTAINING VERY HIGH MUZZLE VELOCITY USING SUBCALIBER PROJECTILES
8.4
OTHER METHODS OF OBTAINING INCREASED MUZZLE VELOCITY
8.5
WEIGHT PARAMETERS OF SELF-PROPELLED AIR DEFENSE SYSTEMS
8.6
EXTERIOR BALLISTIC TRADEOFFS
INTERACTION OF ATTACK AND DEFENSE OPTIONS
9.1
DEFENSE VERSUS DIVE BOMBING TYPES OF ATTACK
9.1.1
Payoff Functions
9.1.2
Game Theoretic Solutions
9.1.3
Example of Application
9.1.4
Example of Parametric Variations
9.1.5
Effect of Non-Zero Aircraft Value
9.1.6
Effect of Minimum Refease Range
9.1.7
Repeated Attacks
9.1.8
9.2
9.3
9.4
Discussion
7.1S
7.16
7.19
7.24
7-24
7.27
7-29
7.33
7.34
7-34
7.35
7-36
7.36
7-41
7-42
8- I
8.1
8.4
8-5
8-7
8.8
8-9
9-1
9.1
9.2
9-2
9-4
9-6
9.8
9.9
9-9
9-10
9.11
9.13
9-14
9-15
9.17
9.19
9.20
9-20
9.20
9.22
9.25
9-25
9.27
!
4
t11
TABLF. OF CONTENTS (Continued)
Section
IU
r
Page
SIMULATION RESULTS
10.1
INTRODUCTION
10.2
COMMON PARAMETERS AND REFERENCE DATA
10-3
EFFECT OF RATE OF FIRE AND CALIBER
10.4
FOOTPRINTS
10.S
PREDICTION MODES
10'7
10.8
EFFCinOF
10.b
AIRCRAFT'BREAKAWAY' MANEUVER
EFFECT OF MUZZLE VELOCITY
EFFECT OF INCREASED TERMINAL EFFECTIVENESS
10-6
10-6
10-6
10.10
10.11
10-7
I10.9
)10.12
IU
I..I
11.2
1.
11.3
t~
12
10-24
10t25
10-25
GnrlFom.to
S111.6
10-1
10-1
10-1
10.1
10-4
10-5
xi
10-27
10-29
10-29
10-36
10-39
10-42
1I-I
Il-I
I1-I
1111-2
11-3
11P3
11-3
11-4
11-4
11-5
11-5
11.5
11.6
11-7
Ii.8
1112-1
12.1
12-1
12.2
I
I
13
14
is
Page
12.5
AUTOMATIC WEAPONS AND MACHINE GUNS
12.6
RIFLES, CARBINES AND MUSKETS
12.7
FIELD ARTILLERY
12.8
TANKS
12.9
ARMORED PERSONNEL CARRIERS
12.10 ARMORED RECONNAISSANCE SCOUT OR ASSAULT VEHICLES
12.11 TRUCKS
12.12 AMMUNITION COSTS
12.13 SENSORS
12.14 PRODUCIBILITY VERSUS COST
12.15 PERSONNEL COSTS IN MAINTENANCE AND OPERATIONS
ADDITIONAL DATA REQUIREMENTS
TARGET PATH ANALYSIS
13.1
12-3
12-4
12-5
12-5
12.6
12-6
12-6
12-6
12-8
12-9
12-10
13-I
13.1
13.2
13-1
13.2
14.1
14-1
14-I
14.3
15-I
15-1
I5-I
15-I
15-1
15-1
15-2
15-2
15-2
15-2
15-3
15.3
15-3
15-3
15-3
15-4
I
1
Appendix
A
B
A-I
B-I
BIBLIOGRAPHY/REFERENCE LIST
Bibli-I
1
xii
!
LIST OF ILLUSTRATIONS
Figure
3-1
3-2
3.3
3-4
3-5
3-6
3-7
3-8
3-9
3-10
3-11
Page
3-2
3-2
3-3
3-4
3-6
3-7
3-10
3-10
3-1I
3-15
3-12
3-13
3-14
3-15
3-16
3-17
3-18
3-19
3-20
3-21
3.29
3.22
4.1
3,30
4-2
Effect of Radar Frequency on Glint Error at Servo Output as Function of Ratio of Target
3-17
3-21
3-22
3-22
3-25
3-25
3-26
3-27
3-28
3-28
4.2
44
4-7
4.8
4-9
4-10
4-16
4-1I
4-16
4-12
4-13
Prediction Error Standard Deviation of Gyrosight Resulting from Tracking Error with co
mils. Tn = 0.75 sec
4-14
4-3
4-4
4-5
S
4-6
11
xiii
4-6
4.10
4.12
4-13
4.14
4-14
4-15
4-17
4.5
4-18
4-21
Figure
4-15
4-16
4-17
4-18
4.19
5-1
5.2
5-3
6-I
6-2
6-3
64
6-5
6-6
6.7
6.8
6-9
6-10
6-11
6.12
6-13
6-14
6.15
7-1
7-2
7-3
7-4
7.5
7-6
7-7
7.8
7.9
7-10
7.11
7-12
7.13
7-14
7-15
7-16
7.17
7-18
4.24
4-25
4.28
4.29
4-29
5.11
5-13
5-19
6-3
6-3
6-3
Flow Diagram for System Correction Based on Measurements of Projectile Miss Distances
Flow Diagram for Correction Based on Measurements of Projectile Miss Dist-nces Using
Synthetic Trajectory Reference
Flow Diagram for Simulation Module for Projectile Miss Distance Measurement and Correction
Percent Hits versus Range with Machine Guns using Tracer Control: 1926-1929
Machine Gun Fire with Tracer Control: 1926-1929
Contours of Constant Aim Error in Slant Plane with Rate x Time Computation
Section Through Barrage Defense
Full Remagen Automatic Weapon and MG Barrage Over One Defended Point Probability of
Killing Aircraft versus Tons of Projectiles per Barrage
Effectiveness of Defense with 37/40 mm Weapons Alone (240 Tubes)
Effectiveiess of Defense with Cal. 0.50 Machine Guns Alone ( 1100 Barrels)
Barrage Effectiveness versus No. of 37/40 mm Guns Showing Total Wt. of Towed Mounts and
Tons of Ammo (Complete Rounds)
Flak-Suppression Techniques
Hypothetical Variation of Glide Bombing CEP With Duration of Bombing Run
Dispersion of Bomb Delivery With SAAB BT-9 Bombsight
Contours of Constant Dispersion in Meters for SAAB BT9 Bombsight
6-7
xiv
:1
I
6-5
6-11
6-13
6-IS
6-15
6-18
6-21
6-22
6-23
6-23
6-24
7-7
7-9
7-12
7-13
7,17
7-22
7,23
7.25
7-26
7-26
7-27
7.30
7.30
7.30
7.31
7.34
7.37
7.43
LIST OF ILLUSTRATIONS (Continued)
Figure
-1
LXv
8-1
8-2
8-3
8-4
8-5
9-1
9.2
9-3
9-4
9-5
9-6
9-7
9-8
9-9
9-!'0
9-Il
9.- 2
9-13
9-14
9-15
9-16
9-I17
10-1
10-2
10-3
10-4
10-S
10-6
10-7
10-8
10-9
10-10
10-1I
10-12
10-13
10-14
10-15
10-16
10-17
10.18
10-19
10-20
10-21
10-22
Page
Weight af Antiaircraft Gun Systems versus Muzzle Energy
Rate of Fire of Antiaircraft Guns versus Caliber
Projectile Density versus Fraction of Weight in High Explosive
Muzzle Velocity versus Propellant/Projectile Weight Ratio
Typical Sub-Caliber Projectile Designs
Game Theoretic Solution Chart
Simplified Solution Chart For Tactical Ammunition Maximum
Effect of Caliber and Accuracy on Stand-Off Range
Comparison of Conventional and Sub-Caliber High Velocity Rounds
Comparison of Projectiles with 10% and 20% High Explosive Content
Comparison of Caliber, Subcaliber, Muzzle Velocity and Filler Weight
Effect of Value Ratio on Kill Probabilities
Effect of Value Ration on Exchange Ratios
Idealized Defense with Four Fire Units
Idealized Defense with Twelve Fire Units
Probability of No Aircraft in the Defense Zone versus Time from First Arrival
State Transition Flow of Defending Guns
Defense System Activity States with System Saturation
Comparison of Firing Doctrines for Gyrosight Type of Fire Control
Comparison of Firing Doctrines for Fire Control with Second Order Filters
Burst Kill Probability and Optimum Dispersion Ratio
Variation of Optimum Dispersion with Bias
Aircraft Acceleration Program for Jinking Path
One Second Burst Kill Probability Versus Rate of Fire and Caliber
Rate of Fire Relationship from Simulation: Tangential Prediction
Rate of Fire Relationship from Simulation: Linear Prediction
Variation of Survival Probability with Rate of Fire and Target Vulnerability
Footprimt in Ground Plane of Constant Contours of Kill Probability with I-Second Burst
Comparison of Calibers
Comparison of Prediction Modes on Fly-By Path
Comparison of Prediction Modes of Jlinking Path
Comparison of Three Prediction Modes on Jinking Path
Comparison of Prediction Modes Versus Breikaway Maneuver
Effect of Muzzle Velocity on Burst Kill Probabili'y with 20 mm Gun
Effect of Muzzle Velocity on Burst Kill Probability for Three Calibers
Effect of Increasing Terminal Effectiveness
Effect of Aircraft Velocity on Burst Kill Probability with Tangential Prediction
Effect of Aircraft Velocity on Burst Kill Probability with Linear Predic:ion
Aim Errors with Sensor Interrupt on a 500 Meter Fly-By Path
Aim Errors with Sensor Interrupt on a 1000 Meter Fly-By Path
Effect of Sensor Failure on Burst Kill Probabilities
Effect of Selected Sensor Failures on Burst Kill Probabilities
Burst Kill Probabilities in 'Fly-Through' Mode
Effect of Azimuth Bias Error on Burst Kill Probability
8-2
8-3
8-4
8-6
8-7
9-6
9.8
9-9
9-10
9-11
9-12
9-12
9-13
9-13
9-14
9-17
9-18
9-19
9-23
9-23
9-28
9-29
10.4
10-6
10-7
10-8
10-9
10-11
10-16
10-16
10-17
10-17
10-18
10-20
10-21
10-22
10-23
10-24
10-26
10-27
10-28
10-28
10-29
10-31
I
I
Page
10-23
10-24
10-25
10-26
10-27
10-28
10-29
10.32
10-33
10-35
10-36
1040
10,41
10-42
10-10
10-31
10.32
10-33
12-4
!12-2
12.3
12-4
12.5
14-1
10-43
10-46
1047
10-48
12-2
12-15
12-24
12-25
12.27
14.2
A-I
A-2
A-3
A-4
A-2
A-3
A-4
A-5
I
I
I
I
I'
S4
Xvi
SI
LIST OF TABLES
-i
Tables
1-1
*11
Terrain Parameters
Comparison of Probability Density Functions
Comparison of Autocorrelation Functions
C3mparison of A-7 Aircraft Costs
A Sample of Sensors and Weather Parameters That Degride Performance
Glint Parameters
Estimated Angular Response of Aircraft to Turbulence
Summary of Transfer Functions
Sampling Interval Corrections
Effect of Xcn Coefficients for 4-Point Filter
Optimization of Filter with (X/A,3)
Available Observation Times
Average Number of Independent Corrections, N
Remagen Defense Effcctiveness
Assumed Equivalent Remagen Defense
Numerical Strength of Tactical Air Forces in Europe
Comparative Numbers and Types of Somber and Strike Aircraft
Characteristics of Bomber and Strike Aircraft
Smart Munition Costs
Delivery Accuracy of Aircraft Munitions from Unclassified Sources
Delivery Accuracy of Aircraft Munitions in Korea from Unclassified Sources
Munitions Effectiveness in Korea
WW II Aircraft Vulnerability Data
Antiaircraft Effectiveness in European Theater
Antiaircraft Effectiveness in Pacific Theater
Allocation of Kills to Weapon Type
Rounds per Bird by Weapon Type
3-5
3-12
3-16
3-16
3-20
4-3
4-5
4-25
5-10
5-16
5-16
6-4
6-12
6-21
6-22
7-1
7-2
7-3
74
7-10
7-10
7-11
7-18
7-18
7-19
7-19
7-19
VII-18
7.20
7-20
7-24
7.38
7-39
7-39
VII-19
IX-I
IX-2
IX-3
7.40
9-5
9-7
9-7
IX-4
X-I
X-2
X-3
X-4
X-5
X-6
Comparison of Systems
Standard Target Dimensions and Areas
Reference-Point Weapon Characteristics
Burst Kill Probabilities for the Point 500m Before Crossover on a oOOm Fly-By Path
Foot Print Data for 25 mm Gun With Tangential Predictor on Fly-By Path
Foot Print Data For 25 mm Gun With Tangential Predictor on Jinking Path
Foot Print Data For 50 mm Gun With Tangential Predictor on Fly-By Path
9-22
102
10.3
10-5
10-10
10-12
10-13
111-2
3
f
*
111-3
Il-.4
111-5
IV-i
IV-2
IV-3
V-I
V.2
V-3
VI-I
VI-2
VI-3
V14
VII-!
"VII-2
VII-3
VII-4
VIl-5
VII-6
VII-7
VII-8
VII-9
VII-I0
VII-I I
VII-I 2
Si
VII-l 3
VII-14
VII-I 5
VII-16
SIi
1
Page
"1VII-17
xvii
I.I
Tables
X-7
X-8
X.9
X-I0
X-I I
X-1 2
X.1 3
X-14
X-1 5
X-1 6
X-1 7
X-1 8
X-19
X-20
X-21
XIXII-I
XII-2
XII-3
XII-4
XII-5
XII-6
XII-7
XI1-8
XII-9
XIIO-0
XII-. I
XII.-12
XII-l 3
XII-14
XiI-15
XII-16
XlI-17
XII-1 8
XII-19
XII-20
XII-21
XII-22
XII-23
XII-24
XII-25
XII-26
XII-27
XII-28
Page
Foot Print Data For 25 mm Gun With Linear Prediction on Fly-By Path
Foot Print Data For 25 mm Gun With Linear Prediction on Jinking Path
Effect of Muzzle Velocity and Burst Kill Probability
Burst Kill Probabil:y as a Function of Bias Errors
Firing Doctrines Used
Dispersion Values Used
One-Second Burst Kill Probability as a Function of Angular Dispersion
One Second Burst Kill Probability as a Function of Firing Point and Dispersion
Simulation Parameters for Manual Tracking Modes
One Second Burst Kill Probabilities With Simulated Manual Tracking
Major Path Parameters
Burst Kill Probabilities With Linear Predictor and 3600 F/S Muzzle Velocity
Burst Kill Probabilities With Linear Predictor and 5000 F/S Muzzle Velocity
Effect of Prediction Mode on Burst Kill Probability
Jink Effects on Burst Kill Probability
Solutions for Burst Kill Probability
Cost of Antiaircraft Gun Systems
Automatic Weapons Development Program Costs
Cost of Gatling Guns
Cost History of the AN-M2 20-rmm Cannon
Cost of Machine Guns
Cost of Rifles and Carbines
Cost of Rifles, Muskets and Carbines
Field Artillery Costs (Towed Weapons)
Field Artillery Costs (Self Propelled Weapons)
Cost of Civil War Artillery
Cost of U.S. Tanks (Under 50,000 lbs)
Cost of U.S. Tanks (Over 50,000 lbs)
Cost of British World War II Tanks
Cost of Current Foreign Tanks
Cost of Tanks from 1914 to 1957
Cost of Armored Personnel Carriers (APC)
Cost of M113 Derivative Vehicles
Cost of Armored Reconnaissance Scout or Assault Vehicles
Sheridan Cost Build-Up
Cost of Trucks
Cost of Rifle and Machine Gun Cartridges
Cost of Civil War Rifle Cartridges
Cost of 20-mm Ammunition Cartridges
Cost of Howitzer Ammunition
Cost of Gun Ammunition
Cost of Ammunition for Tank Guns
Typical Fuze Costs (Current)
Cost of Sights (Infrared and Image Intensification)
xvlii
xvlli
10-14
10.15
10-19
10-30
10-34
10.37
10-37
10-38
10-38
10-39
10-39
10-44
10.45
10-49
10-50
11-4
12-3
12-4
12-5
12-6
12-7
12.8
12-9
12-10
12-11
12-12
12-13
12-13
12-14
12-14
12-15
12-16
124l6
12-17
12-17
12-18
12-19
12-19
12-20
12-20
12-21
12-21
12-22
12-22
I
U.
3
I
5
II
STables
Page
XII-29
XII.30
XII.31
XII-32
XII-33
XII.34
XII.35
XII-36:
A-i
A-2
12.22
12-23
12-23
12.24
12-25
12.25
12-26
12-26
A-6
A-7
!I
I
I
I
i
!
I
I!
Ixx
-.
SECTION I
INTRODUCTION
During the period of performance., of the work
reported in this contractual effort, antiaircraft guns
IV
1
1.l/l.2
SECTION 2
SUMMARY
fThe
2.1.1 General
a. The principal determinants of potential effectiveness, of predicted fire systems, are the prediction and data smoothing algorithms.
b. There is a strong indication that tangential prediction (i.e., an intermediate position between
linear and full quadratic prediction) provides a
'robust solution' against moderate, intentional
target 'jinking' without suffering unacceptable
degradation against unaccelerated targets.
b. The principal determinants of operational effectiveness, are tracking accuracy, and the ability to
eliminate system errors of boresighting and calibration.
2.1.2 Target Acquliition
a. Visual detection and identification of high speed
2-1
rjm
c. The above considerations suggest that ammunition development for predicted fire weapons is, at
least, as important as gun development and may
have received relatively inadequate funding in
the past.
(3)
Considerations
2.1.7
A development program for a new predicted fire
weapon system should preferably follow the following
guide lines:
2-2
"r
i1'
I
I
I
Since one application will usually dominate a development program, depending upon priorities of the
moment, a burden is imposed on other potential users
to continually fight for significant consideration of
their application in the program. The pay-off for this
effort however, if successful, is greater overall effectiveness of Army weapons within tight development
budgets.
The smaller the number of diverse applications considered, the greater the likelihood of success in a
program of this type. All Army guns have a predicted
fire problem with varying degrees of difficulty. The
ability of weapons to fire on the move is desired for a
widening spectrum of armament; therefore, a dievclopment program integrating the requirements of sensors,
sights, and computers for ground based weapons may
be more feasible than one including helicopter armament as well.
3eitlopm
'I
I
ll
c. Identify those required components which involve new development, because of high payoff
in terms of cost, and allocate the major portion
of the available development budget to those
components.
This method approach is preferable to the alternate
'quick fix' method, of assembling data on available
components and determining the best system than can
be fabricated, by selecting from this limited set of data
and components.
eral of these modes are recommended for future evaluation analytically and by means of the simulation. A
basic model for evaluating 'barrage fire' is developed.
to indicate the very large comparative advantage of
aimed fire.
2.2.3 Parameric Description of Threat and
Defense Systenms
The threat and defense weapon systems characteristics are developed in Sections 7 and 8 with the object
of providing parametric relationships that will be of
assistance in developing trade-offs in system
evaluation.
The threat assessment of the prior report is updated
from unclassified sources. Some effectiveness characfrostuclassifiedtsources.rSomegeouectiveness
ara
teristics of the threat versus ground targets are
Methods are provided for determining the probability density function describing the probability that a
clear geometric line of sight exists, as a fujiction of
terrain type, target altitude, and defense site altitude.
Conditional probabilities of detection, given exposure
are then developed in parametric form for visual
observation as a function of meteorological visibility,
Weather as it affects both visual and infrared sensings
is reviewed, from the viewpoint of the attacker and the
defense.
2-3
turns out to be the most sensitive parameter. Comparing a full caliber ammunition type with a sub-caliber
type fired at very high muzzle velocity, the gain in hit
probability from the increased muzzle velocity slightly
outweighs the loss in terminal effect from the smaller
projectile weight, for the conditions considered.
The concept of 'effective stand-off range' is introduced as a measure of defense effectiveness, all tactical
options considered. This is a range such that if the
attacker released his munitions in dive bombing with
iron bombs, he would have the same probability of
destroying the defended vital area that he would have
using 'game optimal' tactics against the active defense,
where the game optimal tactics depend on the relative
value of the aircraft and the target it is attacking.
shot pattern.
report.
2-4
U
Increasing muzzle velocity yields a substantial improvinent in effectiveness in all cases in which this
parameter was studied. As expected, the gain is most
when the target is 'jinkir,'.
Some rather remarkable possibilities have been discovered for the concise representation and summarization of large numbers of simulation runs in terms of
combinations of parameters. In addition to providing a
simple means for presenting a great deal of simulation
data in concise form, these results lead to parallel
simple analytic models which can replace the need for
simulation in the simpler tactical situations.
unique
advantage
of the
simulation
its ability
to SThe
verify
proposed
sysoem
algorithms
andis to
obtain
effectiveness results against arbitrary and irregular
target paths which would be difficult to handle analytically in a reasonable length of time. The unique disadvantage of this, or any other good simulation, is its
ability to dissimulate. The proposed check-out program
package is intended to constrain the Litton simulation's propensity for creative fabrication. Considering
the much greater complexity of other simulations used
by the military services o weapons evaluation
selection,
the istaruser
r
for cnerpons e
athion
b
and
systems.
ered
view of
unofficial
sources.
However, definitive,
it may be in
helpful
to aitssystems
analyst
in laying
out
the cost
tasks
which
he desires
have performed
by one
of the
many
competent
andto professional
cost
analysis groups within the Army.
2.2.8 Additional Data, Test and Evaluation
Requirements
It is pointed out in Section 13 that in the absence of
a suffint qua
a sufficient quantity of data on actual target paths on
aircraft performing real or simulated attack missions,
additional analytical and simulation effort is of mar-
sof
2lIL
2.5/2-6
3,
SECTION 3
DETECTION AND ACQUISITION
In order to detect a target. an unobstructed line of
remotely.
Figures 3-2 and 3.3. In Figure 3.2, the target approaches at constant a!titude. Once a clear line of sight
exists, it continues to exist, until the target passes over
the observer. As the target recedes, once it is obscured
Si
smoothed (low pass filter) version of the terrain contour. The aircraft as viewed from the observing station
I
Sl
QUANTIFICATION
"-
I
3
.
In order to track a target for antiaircraft fire direction there must be a clear line of sight from the
tracking station to the aircraft. With the exception of
sound. all other sensors (there may be a few unex-
TERRAIN
CHARACTE RISTICS
DETERMINE OR
INFLUENCE
II
A.C APPROACH
PATH ToUA
GROUND
TARGET
IApOAI
TUBLEC
CALGTINTEACRFTION
WATF
AIROFD
ALTITADETEEAATHHFIGHT
jICY
IESNO
II
"9*'LOADINGBTCKGRAUN
!.
CONTOLOBSCUR
STRUCTUR
A/C PATH
IG
3-2
ii
I
los
OS
ArC PATH2NO
ftOBSERvER
/EXPOSURE
EPSR
20871 102
HEIGHT
only within the cone. If n - mask angle and H, aircraft height. then the aircraft is visible at a ground
range from the sensor
(3.1,
R = Htltan m.
A somewhat more sophisticated use can be made o!
mask angle when its distribution function is given, as a
function of horizontal range. Stein has published a
number of these distributions and referenced sources
for others. Stein shows the 'probability' that mask
angle is less than m for range R, as a function of m,
for
specified
values of R.
typical cuives
from(Vol.
his
report
were reproduced
in and
the AFAADS-!
report,
ro P 4.40).
1, p-4stored
Designating
I
F(mlR)-
(3.2)
promaairid
and ?(RH) the probability of having an unobstructed
line of sight to an airplane flying at an altitude difference H, above the sensor altitude
(3.3)
P(Rli) -F(ma I R); tan ma = HtJR
An estimate of the distribution function of line of
sight ranges to a target at given altitude can thus be
ote
Note that the use of a constant mask angle cut-off m,
model defining variations of ground level as a stochastic process. As noted later, elegant solutions of the
stochastic formulation are hard to come hy. and so the
Ca)wood-Schiller solution is something of a inur de
'U
Caywood,
& Co.
inadequate
in some
found studies
of airmask
deangle
conceptSchiller
the simple
ense systems. In their analyses,' they used computer
force.
The method of deriving visibility angle is based on
(3.4)
3-3
VOLUMEVOLUME
MsKEI
MASKD
MAKEDMASKED
SEE
SENSOR
R.
M871-103
Figure 3-4. Boundaries Around Sensor Defined by Mask and Visibility Angle Concepts
and resemble the Torrance. California city boundaries)
then expressing the area as a circle of equal area. This
accounts for the form of the expression, which follows
giving the effective radius (R,) of the equivalent circle.0o
287
1 ,000
1.
Jas
(~)~
(3.6.1)
Reff a Ro 0"21(n + 1.3 m)
where R&is a constant for each terrain
selection. R, isa
given in Table Ill-1. It depends only on Po
Note also that there is a maximum value of visibility
angle associated with each terrain selection. This is
obtained by differentiating Equation (3.6) with respect
to H,, and it is found that the maximum angle occurs
when target height H, - a /0.2 1,i.e., about So.
(3.5)
.n
H
a
H, - target altitude above ;round level vertically
below it (terrain following flight) (meters)
a
H,
(m tesdetails
where
20871-60
.3m) 1/2
085f
In addition to the expressions above for P.., Caywood-Schiller? (R. Rose) have given an expression for
the probability of having a line of sight to RP(R), as
P(R) - min[l,(R.)Ckj
where
C -0.25(re+r,)
- 0.75[;
10(n+4m)J
I -exp
(3.6.2)
5
I
Iogjn
I
I
Koe
00a
20
227
61.> 0.2
ipo-ri, Korea
25
198
Q~ 0.30
2300
Sanyang-ni, Korea
45
0.,39
~1once
1,2,.n
1850
20871-500
limited form
d~d
dimensions.
a~w
-
39
n+vi39%
)a
) + (l/2)J:
D
'j
2
8x
ax.
(3.10)
2 ,..L
aat
xi
mFor
.. x)+V
In the first case, a difference between terrain variations along a line transect and noise amplitude fluctuations with time in a classical 'noise' problem is at
~(
xdt-f(xI
5
3
j~ =n;
X, v1.* ....
fi (xi
(3-7)
20
Wonju, Kra
116
dxifdt
R.0(meters)
3-5
00200
100
10,000
TARGET ALTITUDE
AGL IMETERSI
0
0
EFFECTIVE RADIUS
(METERSM
WETILAR. GERMANYI
3O00
2000
SENSO
2w 71 104
lutely simple.
Uhlenbeck-Ornitein Prwess
w(xb, t) - Oxb-
x(-ao, t
b +Ot
- 0
dx/dt a -. x + v
(3.12)
<y2> - D;<v> -0
aw
B(wx
"+
8WwO
(D/2) a
8x2
(3.13)
w(xo,0)
3-6
I;l
6(x-xo)
I
3
15,.02-
S100
10m000
"AGL
34
TARGET ALTITUDE
METERSI)
10
EVIFECTiE RADIUS
IMETERS)
4000.
1000
2
4,
20871 105
Figure
IIn
Solutions for j8 - 0
When P is very small, the initial evolution of the
process is determined by D. This suggests first solving
for the case of fi - r corresponding to a WienerLevy process (for whit. the variance is v.nbounded).
Lz,,We
4'(3.14)
1/
Offr)l
f~t^(Xt/Xo
iseasyto
1__2__ 1/2__ e
00l .- 2 I
btan an
where
p
2a2
0
(DIP)
2v2 (I p2)
iswellkno
Defining
t
0, the results areTobtained that. setting a - b height of the sensor above grodv
SFor
(e
SAtf
+ A ((lz)-(AzI4)
Efc
where
-M 2a0/D;z
Erfc6y) a 21(iI /2
l(1.0)
I(MO,)
(3.15
= (2Dt)i/ 1aJ
Ay
eX2 dx
Ef(I/z);
d
Erf(y) =I
I - eg2a>/D; 0o
;a 00
0
Erfc(y)
- (
a
);D
;t,0
Solutionsforfi 0 0
..3
P.
(3.17)
D.$j (-X)
(e ost .
B - b/o
(3.18)
sidetermined
by the zeroes of
D$j (-B) - 0
(3.19)
and
D
s the parabolic cylinder function.
Ds(y)
Tables of the zeroes and the partial denvatives with
tI
I
and for .-iall b,xo
case
simnlified
has
given
been
by
traonoics
(3.20)
A comparison of some characteristics of the Gaussian and Laplace distribution functions (also known to
the French as Laplace distributions Types I and 11) is
1,'/2 2
1 (0)
o
2
1(0)
This
aLoge _Do-b/(o)I}
g, ()D[x((a)]
Ro(0
T(0(1/0)as
02
;a
b-x 0
(:3.21)
dxl/dt =x2
dx 2 /dt
(3.22)
(2)!/2/e
(2) 1/20
fl (x 1) + f2 (x 2 ) + v
(3.26)
where
x, - deviation of terrain height about its mean
x2 - terrain slope
If the stochastic process is assumed to be stationary.
probability density functions of height and
then the
slope
are indeperndent of each other.
fl (x 1 )
If
fI
-kxI ; f2 (x 2 )
-cx
(3.27)
X.
fj
A number of studies do indicate that the probability
density f'it',,tion of terrain elevation deviations about
3 the
m
*
then
aw/at = 6 sgn (x) (ow/ax) + (D/2) (a2w/x 2 j
(3.24)
""
(3.25)
(/DD)
!(
L!
-ksgn(x 1) sgn
sign of
(3.28)
3-':
100
34
TARGET ALTITUDE
AGL (METERS|
1'
"EFFECTIVERADIUS
(METERS)
3000
ROUGH TERRAIN
2000
j01
I
2
3
4
SENSOR SITE ALTITUDE IN STANDARD DEVIATIONS ABOVE MEAN
6
271 106
10
1
NOTE.
PROBABILITY
0.1
1 3
0.
IDEGREES)
2W71 107
Figure 3-8. Probability of Not Having an Uninterrupted Line of Sight to all flanges versus Elevation Angle
3-I0
i06.
\\
2000WO
000
4000
"tOORIZONTAL RANGE
50
0?? 108
R IMETERS)
Figure 3-9. Proab;'.ity of Unobstructed Line of Sight Exceeding Range A venrus Elevation Angle of Sight Line
Heine' has generalized Whittle's formulation to a
In a Plane
II
S....SI(x)I
I
I
It is clearly an impor-
3-11
Gaussian
Funhtion Type
ggo0
Rangegol
Mean
,II
Standard deviation
I
Probability density (unction
- 2)
"
(s
e is
(2/1r)"' O
0aA2)I'
3.3 x 90.
1.7 x 10
Lapiscias
; 1' A 2
/' I "/;2
Standard deviation: ov
(O02.o,2)
(o1
*t-t2Oy2/2)
0)
ip,
ly .M x)2
2O
4
2
-.
-I2.
+ (0 -,-/)l
y II1+/222
/2
I2
x I2
Me a n : /4y
eis
Gaussian
Charaiersstic function
0 2/2)
Charactenstic function
e 12':
e-
V.Y
II +
-2''-
For
only
I y I
>
I-1/ff
3/40)o y
Oy
20171-700
3-12
N(b)
p(b)<[ix>
x-
(3.31)
N (b)
<lxi>
y(j)
(3.37)
f(y)dy
(3.38)
-0
o2 Ox2 d((3.32)
2
2=
2 0 (w2) dw
q (x,)
(3.39)
i
2
where 4 (w ) is the Fourier transform of the autocorrelation and vice versa. The time. T., between independent samples is the inverse of the effective bandwidth.
As a result of the above relationships
Weff WOvI/x
h
fq(J)
(3.33)
S-
.-I
y(0)+j L i
(3.36)
wr
S
where
nL
(3.34)
(3.35)
problem.
Then for the line transect case.
3
4
" 1+q(l)+q(l)q(2)*
1
(3.40)
Sline
and for the area problem, the mean clear area about
the site is
3-13
,(
A/Ahex a p(l)+6q(1)p(2)+12q(1)q(2)p(3)+
- I -q(1) +6 [q(1)+q(1)q(2) + .
...
real terrain
]fit
(3.41)
At this point we note simply the expression developed by Cramer-Leadbetter for the expected number
of crossings of the line a + bt (the sighting p.'oblem)
of a stochastic process with normal probability density
function (the method is applicable to other pdfs as
well, and to other curves). The expected number of
crossings in the interval (0,T) is
._/L a lip
q (qlp)
= p/Ahex
Because the crossing rate approach is computationally simple, however, it is worth investigating to use
as a guide for obtaining
possible empirical functions to
data.
rI
-bT\
.-a
X [,
InO)-IW-)lqo+1)
) - 11
2
Then
(3.42)
C
Let
]l
(3.43)
v/2IF
e-x
.-... 1
/2
(X) a2-J
f
L
t2/2
et
dt
(3.45)
where
211aj2
+2 (a2q(j+l) 0
L _q_0_+Ij
p3-14
I,
I-,
AIRCRAFT
27
SALS 3" ,
FLIGHT AITUOI UND101,
M FT
"Msystem
F,oCIEIoE
ovs.
Io
tack.
o001 0
2WO
#W
ASSOLJE
OtVATO'"I
,),,o
LI,
I
SIX
$ROM
-.
no
wcapability
SI
atmosphere.
-l
. 1icos bs
+b
o2
Effective Bandwidth
b un b4c
bli.11
"(iib)
a
bc"ls b -lt 1 s ji/(b-&)
-b
ab
Ratio of
a2 b2
Velocity Vanance to
db
Position Vman'e
A u toco nre:"tio n of
bIsil
Velocity
Gauissant
(.
Laplacian
( 2
b2)lTib~i
)lab)
20871-701
Electonics Costs
A-7A
51,400,000
$210,000
15%
A-79
S 1.440,000
$260,000
18%
A-7E
$2.500,000
$620,000
24%
3
2
20871l-501
(a)
00
02
~25~
..~-"----
"
--
,--.'----
,-
,- I"
Os
,-1
It
SUNRISE
20
10
08
20
60
40
40
14
60soso
16
SUNSET
18
2n
FES
JAN
MAR APR
NOV
OEC
00
I
02
N\IS
04
00
so70
<
70
SUNRISE
08
50
on
60
iSO
40
'2
;0
14
tso
SUNSET
20
00
JAN
FEB
MAR APR
NOV
DEC
."OU71110
Figure 3 -11. Joint celing and visibility probabilities for Sinuiju, North Korea, calculated as a function of
month and hour (smooth dashed curves show times of sunrise and sunset). (a) Ceiling a 10,000 h and
visibility a 7 mi. (b) Ceiling z3,500 ft and visibility > 5 mi.
3-17
III
the ground in the Mekong river valley region is obscured by ground fog in the morning with 50.75%
probability the year round, but the afternoons are
relatively clear over 90% of the time.
Radar surveillance sets may have the future disadvantage of attracting radar-homing missiles. Increasingusage of distributed acoustic sensors on the battlefield
may provide an alternate, or supplemental low cost
means of providing surveillance information for local
air defense.
The relative effect of weather on attacker and defender depends on the sensors with which each is
equipped. How much all-weather capability to provide
for the defense depends on intelligence estimates of
probable all weather capabilities of potential enemies.
Specific effects on the defender's sensors are discussed
in later paragraphs.
tion.
Weather
Huschke,D lists a number of defrom and
Table 111-5,
scriptors
or weather
sensors subject to degradation
For detection at low light levels, target and background brightness must be considered separately, but
for daylight sighting these combine ini a single parameter describing the target contrast against its
background.
i
i
U
u
1an
3-16
/2
Ct2
1.7581/2
190
+
19
C= Co e'
(3.46)
where
R -
where C, is contrast in percent, the numerical coefficients are given for foveal vision, # is the angle of the
target off the center of the fovea (central part of the
retina) in dtgrees, and a is the angle subtended by the
target (represented as an equivalent circle) at the eye in
minutes. A similar expression with different coefficients
applies for other retinal regions.
RIVm ;where,
3.44-3.912
(3.49)
V,. -
[For
ity g is
g a f(C/Ct)
I
(3.47)
T"coefficient
The "intrinsic" contrast of a target against its background is defined as
Co
(Bo- B)/Bs
(3.48)
'
B. Bb -
B.
section
over a number of glimpses, each of which may
be directed
at a different part of the sky as the observer
I
I
C. C, -
Define
p(r) 3-19
or,..
Table 111-5. A Sample of Sensors and Weather Parameters That Degrade Perfoirmanc.
Sensor
LLLTV
TV-Guidance
IR
Laser
Eye
Radar
1. Prec~pitatioflSS
2. Visibli~ty fog
3. Clouds
4. HumidityS. Temperature
6. Wind
7. Ice
8. Turbulenoc
9. Suangi
P(R)
I - p(rj)J
J
1 - P(rJ)I
[IL0
020871-502
3.1
I
-
f te sanrat isU,
su byinintgra.
Appoxiateth
Appoxiatethesumby
n Iteal.If he canrat 15V,
Log~~~~~~
rI(dlrrwtin
f (r~l
og. (t/dr drability
uI -p
(35)
r
P ()~
lasp
3.20
!I
'0
AROB
OF MET VIS
ExCEEINQ RANGE V
PROS OF VISIBILITY ExCE-DING v METERS
I;'
S1PROS
It
PROS OF CE LINIG
EXCEEDING H ME"ERS
r0
S1000
2000
3000
4000
8000I
CEILING H IMETERSI
20871.602
where the sma;ler k values are the asymptotes for very
small tar et subtended angles, and the large values are
I
1
3-21
I-.
LA
I.
PROBABILITY
OF MET VISIBILITY
EXCEEDING RANGE
06--
SUMMER
WINTER
15
i0
20
20871 112
I>
10
40.000 fT
04' -
0I
06
~-
--
0-
--
200
2MII71 113
, All Seasons
3-22I
AL
.,.
--
I
Then for line scan
p
k4 80 jOs ; 0s-
(3.54)
P(R;-
P = k5 0o2 / n ;l n
e*(R
(3.55)
(Cfb)
m (C a2/19)
"K/R
R/V
K2 R/V
(3.56)
U
SI
(3.59)
(3.58)
Svary
R'a e'K4R/VmdR
Experimental Data
C - a0o 1 /2 + boo
o0
f1
e"K3 /V
2900 'R
1
5C1n, f(R 1 )
180-3
scan, f(R)
9.0 R I
35 R1 2
R/I100
;eRI/l4
R 1 1'4
(3.60)
3-23
3.I':1
(3.61)
eAR
I .e"E(R)
E(R) " K
f R R2
e"'RdR
KRWI-'
(e
P(R; R.)
Forlag2
eXR/
E(R)
.]E(R;Re)
(365
P(R;Re)
where
E(R;Re)
(vIV)
fR(XR ) [I-2()R)-1
JRRc
p(r) dr
/ R
fR
+ 6(XR)2
f(r) dr
(3.66)
66)
(3
(3.67)
(3.63)
Now assume that we have data on cumulative detection probability on paths on which the target was
initially exposed at very long ranges, - essentially
infinite. It follows from the above expressions that
)
(3,3)Q(R;
(3.68)
Q(R;RI)Q-R;-
and so, from the sinfle Q(R; co) curve we car. obtain
the cumulative detection probability subsequent to any
initial exposure range.
We should also expect that changes in target speed
will affect the cumulative detection probability as
(3.69)
Q2 (R;-o) - [Q,(R;)] (V! 1V2 )
where the data was taken at V, and it is desired to
The curves are based on the 45-90* scan data, ignoring the maximum of Figure 3-17. Hence, they slightly
over-estimate the cumulative detection probability as
this probability approaches unity. However, they do
show a marked reduction in detection probability ver&us range under atmospheric conditions less ideal than
those of the desert experiments,
Using hindsight, we note that the very simple approxreation to g(R)
aI
0.5""(3.64)
as
(3.70)
by unobstructed
Figure 3-15.
In the experimental data represented
to an
the targets were initially exposed
geometric line of sight at very long, essentially infinite
range. We now indicate how this data can be adjusted
to represent initial target exposure at shorter ranges, as
may be obtained from the terrain models of Section
3.1. This is a simple computation as a result of the
mmm
_r(R) - I - Q(R)
*0 f(Re)dRe
'
f
a I - Q(R, *OeRI
mmIm
I(R)
1
+ )R Ei(-XR)
mmmllm mmmmm
3m.24
o .0.4
11020
45. 90O SCAN
010,
050
450,
ISO. 3~00
900SCAN
SCAN
IdE 004
01000)
-
LOG I0 P)
SCAN
S010..
S~0.0'
SI
________-_____
002So00
1000
10.000
RANGE (Y$S)4
00'
So00
100
2000
-- -- -10.000
S00
RANGE 1YDS)
07
20671-1144
5
1
I.3-25
]..--
LI
,,
- -
' -
'
'I
1,
50 -
40
II
30
I i
45.0gQ SCAN
20
(10) d
.0.
N
I
03.0
1000
10X00
RANOE (YDSI
Sfrom
L
i
dD/dt u -V cos n
(3,71])
and these radars have a blind zone about path midpoint which is narrow for fast targets and wide for
slow targets. The blind zones for a 30 m/s minimum,
window are sketched in Figure 3-20 for a 100 knot
target and a 400 knot target. The lower velocity target
might correspond to a helicopter, however a sepaate3
analysis
required
to estimatefrom
whether
doppler would
signalsbemight
be extracted
the usefui
return
of current
radial velocity
minimum
(Dhe
pulse doppler
surveillance
radars ranges
from coherent
30 m/s
of change of"
targt
isprovides
sant ang
3-26
10
06
II
01
006
qiR)
'|
001
0005
~~0.00!1I
1.0
30
20
40
s0
(AR)
LI
~2M71
117
S|
bardmay
the
hicang
theairraftcanbe
at
3-27
10.
EXCEPTIrw'AL
LY
CLEAR
CUMULATIVE
PA013ABILITY
OF DETECTION
CLEAR
05-LIGHT
HAZEI
METEOROLOGICAL
VISIBILITY (YARDSI
LI ZE
FOG
10.9m
660
20
low~
3000
4000
500
w000
2071 l~e
Figure 3-19. Cumulative Visual Detection Probability versus Target Range and Meteorological Visibility
sound of an aircraft (in these experiments a 8-52) with
TARGET
.G4T
TARGET
DIRECTION
(3.72)
RADAR
Agf 7wDoIV;
F
3es
Figure 3-20. Do
If
I
i0
pler
Radar Blind Zes for 30 rn/sVp
Velocity Minimum
(3.73)
3.28S
3.28
; "'"'
| 'i" I'
Ii
;
"i'
'
'
-i
mT d
...
3fs
i
t
the
eye appear
have more
some information
potential. Modern
datahuman
processing
allows tomuch
to be
extracted from the radar return than is utilized in
detection and tracking. It is possible that advanced
data processing of infra-red signals and laser returns
can
provide identification clues. The same may be
true also
or acoustic
signals.
Whether any of these non-cooperative identification
techniques
can be made
at a high
enough
confidence
level to
to yield
allow identification
shooting is unknown.
However the importance of identification to both the
derense and to friendly air suggests non-cooperative
identification as a useful area for research.
400
30.000
AURAL DETECTION
OP SENS;;NG
IOTAL
DISTANCE
20,000,-
'1
VISUAL DETECTION
so
160
110
VELOCITY (KNOTS)
3-29
20671 120
TIME AUDIBLE
TERRAIN TYPE
& ROLLING
FOR C14-47
POmINS
HELICOPTER
)NTOUR
FLYING
100...
TIME ISEC)
TIME VISIBLE
So-
NI
10.............
so
200
100
SPEED 4KNOTS)
30
400
bOO
M
U71 121
Figure 3-22. Comparison of Visual and Aural Sensing Times versus Helicopters
I3
II
i
It
ii
SECTION 4
TRACKING
*
pect but was very close to 1/2 the target dimension in each aspect.
Additional computations given in the reference,
glint order
error
the computed
of processing
thea result
show
system (second
sensor servo
simulated
through
with 0.707 damping ratio) in terms of the servo out.
put. Standard deviation of glint error after processing
by the servo was determined as averaged over 3600
target rotation, for a range of values of servo bandwidth and constant angular velocity of target rotation.
Only rotation in ajimuth at zero elevation angle was
studied.
For additional details, one may refe: to the referenced report.
4.1 RADAR
Errors of a tracking radar may be categorized and
estimated in a number of categories. The principal
error sources of a monopulse tracking radar at short
ranges are probably I) those associated with 'glint',
2) those resulting from servo lags at higher angular
tracking derivatives, and 3) errors resulting from multipath returns at low elevations. In this section we
consider a simple model of errors resulting from
' glint', i.e., phase interactions among returns from
separate parts of the target, so that the apparent target
direction, measured from the phase front of the return
signal, may often lie off the target.
"
*I
1
-8/0
R4I)
-/002
e
9
(4.1)
where
the target.,
mal distribution.
b. The mean of the distribution varied with aspect,
Themeanaboutn gtouthen
but
.1
tariet dimenaspio t.
but only to about 0.10 of the target dimension.
c. The standard deviation varied slightly with as-
O +o(
e
00,I02
4-1
'characteristic these
angle'areofroughly
each glint
component;
the
angular intervals across which two
samples may be considered
independent.
7-
"
- - --
-.
. .-
"
-.
8.0
TAROGT B
6.0
4.0
2.0
0
20
1.5
1.0
0.6
Own 10/RAO)
=71-122A
Figure 4-1. Reciprocal of Variance at Servo Output for Experimental Data versus Ratio Target Angular Velocity
(/sc) I Servet Bandwidth (red/sei)
Two components were chosen bezause initial computations indicated that a satisfactory representation
could not be obtained with a single component.
r
0) d
Ow
"0/0J(4.2)
has the Fourier transform, on the interval
(O)-)
dw/I
212
;0
p
"'a/
(4.5)
0
(42
-
(og2 /fr)
cThe
wa)esa021
(4.6)
o0 dil
21 -necessary
dI + (coo) 2 ]
a12
11 +2
w b + 0 wTb) 2
(4.4)
2 . 0.50
(4.7)
2t + (w0 Icwn)
i(0sg)2 =in
2 1I + 2t (W01/Wn + (Ioll(wn)]]
4.1.1.1 osaling
The parameters of Table IV- I are for the model and
microwave frequency used in the referenced experiment. To convert to other frequencies and target sizes,
A + n(w2/wn)
2 [I1+ 2t (w02/wn)+(w02/wn) 2J
(4.8)
O -
1.
PmAturbulence.
0.29
0.20
001
0.0930
0.0740
SFighter:
*(0
10.3
A brief survey has been made of the angular velocities caused by air turbulence and normal flight wander
about a mean pathP'
Dunn and Howard' show iome graphs of aircraft
yaw data for a fighter and a bomber attempting to fly
a straight course in clear medium-turbulence atmosphere, with the following rough indications.
Bomber:
4.70
a .5rms
0.33
0 25
data
)2
0.38
Spuametris
0.25
)2
in 17 sec
angular ve!ocity I1/secc
rms
angular
3. I/sec
2087 -103A
4.3
Maximum
excursion, (snread),
50 in three 20 sec traces
(01a/
aircraft.
'characteristic angle' of glint for model
(Table TV- I ).
(I
wavelength in centimeters of radar.
full scale aircraft dimension perpendicular
to sight line (for example, fuselage length
side on) in meters.
3
002
cta
'characteristic angle' of glint for full scale
(t.Xl)
'A
Same
O.A
(4.9)
velocity
1.5
NOTE: For the three traces, rms angular velocity increased roughly as the max'mum spread in
angle.
I
6.0
I,
4.0
0a (METI9S)
3.0
S.BAND
2.0
"
K 'DAND
1.01
-I,
r-
Ip
1.0
0.6
0lwnr
101RAD)--s-,.
1.6
2.0
X471.123A:
Figure 4-2. Effect of Radar Frequency on Glint EGro at Servo Output as Function of Ratio of Target Angular
Velocity f/sec) to Servo Bandwidth (red/eoc)
from the 0.2 Hz assumed to perhaps 0.8 for short
periods of time.
I
Table IV-2. Estimated Angular Response of Aircraft to Turbulence
Standard Deviations
Yaw
Wing
Angie
Anpta
Velocity
Angie
696
30,000
1.3
1.60/1
3.1
5.501
52
318
4,000
4.80
5.70/1
6.2'
11.10Is
87
700
35.000
1.10
1.30/1
2.60
4.60s
Weight
OIbs)
Loading
IbltI2
12,600
So
28,000
12S.000
Velocity
0f/10
208713-04
To show the effect assume a simple prediction algorithm including a velocity term, but not an acceleration
term. The transfer function is
W(s)
Anlar
Velocity
Altitude
(ft)
NACA
Aiicraft
Sdevoted
Roll
I + s(t + nTs)
I
,,
(I + &Ts)n
(4.10)
where
As the radar frequiency is increased, the glint bandwidth increases, and so, although the glint variance is
assumed to remain constant, the variance of tracking
error after processing by a low-pass servo decreases as
shown in Figure 4-2. However, the increased high
frequency content of the noise can result in increased
prediction error because of the differentiating process
in prediction. The increase, if any, depends on the
bandwidth or the velocity filter and the acceleration
filter, if acceleration prediction is employed,
time of flight,
T, -
na
Consider a single glint component with characteristic time To. Then the variance of prediction error is
obtained from
4I
I III
r1
-+/f
-
l..
2t2.
(dP
J+, +2J
(4.1 i)
is essentially unsmoothed.
tprs
P
(4.12)
n = 1;(of)2
Pl
2;(O /)2
P 1
(l +)
'
,A
ATurnn a
(5 +4+2)/2
setting
(4.14)
(T 5 A)(121) > 0.1 for example
It then appears that the shorter the wavelength, he
shorter the smoothing time can be for the same prediction variance, and this is of course desirable for many
other reasons.
3.0-I
40
N*1.0
j7
20
N- 2.0
io/T,- 1 0
if"
~INCREASING
(4.13)
h
.02.0
3!0
,.0
.0
Figure 4-3. Effect of Glint Bandwidth on Prediction Variance for Two Simple Prediction-Filter Algorithms
4-6
.I
U
0 o, 2
Angular
bandwidth
(indians)
0.30
0.001 (V/L)rad
0,10 0./L)rad
Based on the extremely limited glint data it is assumed that glint error can be considered independently
in elevation and azimuth. More accurately, the 'azimuth' computation of error is developed as 'lateral'
error across the line of sight.
,variance
(4.16)
0.70
tile veloci:y).
Smaller
component(2)
Variance
s,
Fraction of
Larger
component(1)
(4.17)
+T
Ow is taken as 0.03 rad/sec - 1.72isec for all computations. To convert azimuth rate to traverse rate
nTA 0lAdAofdt)cos e. I radians/second
(4.18)
(4.19)
The total glint variance as angular velocity approaches zero is asymptotic to (L/2) where L - total
(4.20)
4.1.3.1 Servo Approximation
The transfer function of the servo i approximated
as
H(s)
(I + sT)
I
Ai
Iquate
(4.15)
(L/A
4-7
I'_.t
(,4.21)
for simplicity; this approximation is considered adein view of the uncertainty in the glint characteristics. In addition. T is taken as I/B where B is the
servo bandwidth and is approximated as
B 2ff
1/2 rad/sec where K. a
2
(4.22)
where M 1
xo+l)
aO+l) x)+b(j+l)u(j+lO
aol) = e"S(+l)A
w/B
b(J+l)
Fhe two glint components are considered independent and since the servo is linear, the variances add, to
result in
02- (/2)2
performed according to
(4.23)
14(Ml/2)
(I + M02
1/2)
S+7+0.3
(0+M1 )
["(M
.+0.3
++
21
2j
(1
(4.24)
Aj 1/2
o(i+l) (I -e2So+l)A
(4.26)
(4.28)
4-
I
i
I
I
Target
velocity:
300 meters/second
Target path:
analyze sets of tracking data taken with human operators in the process of tracking real aircraft. However.
the
data
not available
in theinextent
hence
this was
analysis
is more limited
scope. anticipated,
h
4.2.1 Desirable Control Dynamics
Additional experimental data was located on laboratory experiments of human trackings with a variety of
control dynamics. These experiments are summarized
below. The findings support the conclusions and recommendations of the AFAADS-I report.
The second set showed that given the position clement of tracking, the addition of as many as four
integrators in series was helpful when the input has
principally low frequency rather than high frequency
content, and that the integration 'aid' became inferior
only when most of the input had frequency component
so high, that both position and assisted tracking per.
formed poorly.
3lThe
I
I
S4.1.5
i
L. 1
Conclusions
The number of parametric variations in the simulalion runs are too few to draw detailed conclusions. In
particular. as shown in Figure 4-4. with 3 mils angular
round to round dispersion, there is not much room for
additional improvement by reducing tracking error.
The 'zero tracking error' asymptotes shown in Figure
4-4 were hand computed for the assumed target, and 3
mils dispersion.
The analysis indicates the importance of smoothing
time, and the interactions among smoothing time,
prediction mode, and the tracking noise autocovariance. These interrelationships and their effect on burst
kill probability can, of course, be examined by means
of the simulation, at the expense of additional runs.
"4.2 HUMAN OPERATOR
In the AFAADS-I Report. an extensive survey of the
literature on human performance in the tracking function was made. It was hoped in the present contract to
e0/ie
Z K/s
(4.29)
,"4-9
4-
(T s+l) 2
8ol1c
(K/s)
(4.30)
T2 s+1
and the Bode plots are shown in Figures ,.7 and 4-8.
For a single axis control against a mix of sine waves,
the two control laws yielded the same tracking error
averages.
It was suggested that this may have been too simple
a task for the operator; he was then given two axis
tracking of a CRT spot displaced by sine wave mixes
in both coordinates.
I
BURST KILL
PROBABILITY
3 1
0.6
SMETERS
BEFORE MIDPOINT
I,
56.0
RADAR WAVE LENGTH (CM)
10.0
-3-
20B71-126
Figure 4-4. Effect of Radar Wavelength and Number of Rounds Fired on One-Second Burst Kill Probability
from Simulation Position and Velocity Smoothing Only
The 'optimum' control function then yielded about
0.60 the mean absolute errors in each coordinate compared with that obtained with the rate laws.
control function
4-10
II
S00/0c
ocbeen
""with
(4.31)
=o/0c
(1"/400)
Sfor
Ithe
A proper analysis of human error in tracking aircraft targets would require a large batch or tracking
data showing error versus time, with concomitant
records of relevant parameters such as target speed,
heading, range, and the trackings angles and derivatives all as a function of time. From these one might
hope to develop a useful 'model' of the human operator which could be used in simulations and analysis.
Unless taken with the above objective in mind, most
tracking error summaries tend to list only mean tracking error (bias) on each path and the standard deviation about the mean.
An interesting question is whether from such abstracted summaries estimates can be made of the operator bandwidth, or equivalently the approximate correlation time of the error. The following paragraphs
indicate that inferences of this type can sometimes be
made.
i-:
10
BEFORE MIDPOINT
I
0.S
BVjRST KCILL
BILITY
PR tOBA
16 ROUNDS PIERBURST
SMOOTHING
poSITON AND VFLOCITY
1000
0
RADAR
prcSthe
T
WAVE LENG H(CM)
M871,12 6
Positionl and
surat Kill Probability froM Simulationl
onOneSScond
Wvl~f~t~
.3
ofRadr
45.
ffec- f s '
Figue
2
a een t V n,60cnitey S'moothing Only
ue
Fi.Ef~
*
2(l -Cos Ox 21 e
samples
variance of the mean across many
spectral density is Of
power
psrmeohaeses.is
the form
2
(2
am~
44
WM
O(2)
dw
k
(W~Tn)
(4.32)1
eeki sn t uiit(4.34)
kew
Con stant
the
is
#
and
ftval,
inte
m ea urementnoise.
a.' is the variance
where T . is thedensity
of the
spectral
PM admaecmee d*
power
of the mean.Thexoeta
r osdrdao
h aoel~ praltO
term u
Approximate the bracketetd
4-12
I212
I..
BURST KILL
PROBABILITY
0.5
DISTANCE
BEFORE
MICrOINT
SooM
300 M
0
0
5.0
10.0
20871.127
Figure 4-6. Effect of Radar Wavelength on One-Second Burst Kill Probability from Simulation Filter with
Partial Acceleration Corection
quate in view of the limited data to which they will be
applied. There is no problem (but more algebra) in
evaluating the integrals exactly.
(Om/Qo)
-1/2
(435)
elevation coordinates.
The curves of Figure 4-12 are consistent with Equa-
I2Tn
T. - 0.75 seconds
This is not unreasonable for a human operator
employing a rate control. The corresponding bandwidth is about 0.2 Hz, and this is also rensonable, since
,4
1-
-0
2.0
0
AMPLITUDE
RATIO IDS)
OPTIMUM"
a....OPT
.20
"-RATE'
T A AC
0.1
l0.01
1.
F
",
ING
100.0
10.0
AREORAOISECI--.-
20871.128
I,
1
"OPTIMUM"
iI
CONTROL PHASIE
I'
"RATE TRACKING'
0.01
0.1
1.0
10.0
100.0
FRSO IRAD/SEC)
20171.12
4-14
I1
4
CONTROL SYSTEM GAIN
IoI
102
FOURTH ORDER
CONTROL
10
.1.
ZERO ORDER CONTROL
WPOSITION TRACKING)
00`1
1.0
0.1
FREO
100.0
10.0
(RAO/SEC}
20671 .130 A
+so(l)t
+
W(s) "
(4.36)
I +S tP
where s - d/dt,
time of flight, and a - a
constant (here taken as 0.20).
The variance of tracking error, and probably T. as
well will vary with the target path parameters. if we
"assume, however, that the derived parameters of tracking error can be used as average values, we can cornpute the autocovariance of prediction error resulting
from tracking error. Without reproducing the algebra.
ZERO OROER
-,-
100
AVERAcu EPRIRORIRELATIVEI
/CONTROL
60
FOURTH ORDER
I:
20671-131
"
10
II
I
"
I'
0
10
20
30
40
W2071.132
4-16
ELEVAIO
Ii
00
/10IMUTH
-M
010
400
SIRECORD
LENGTH ISC)
operator.
4.2.3 Human Operator Simulation in the
Traoking Function
This section examines the considerations in providing a more explicit representation of the hur' an oea
tor transfer function in the Litton simulation, with the
object of eventually replacing the module now used for
this function, as described in AFAADS-.I.
Output/input isgiven by
y(s)/x(s) - Y(s)/ [1 +Y(s)]
(4.37)
(438)
ASYMPTOTE
20-
STOQ EVIATION
OF PREDICTION ERROR
IMILS}
101
Oi
100
30
2.0
40
20871134
Figure 4-13. Prediction Error Standard Deviation of Gyrosight Resulting from Tracking Error with or.a4.5 mils.
T.a 0.75 sec
Y(s) a (K/s) e0.20s = (K/s)e sTm
(4.39)
where T
(s,/2)
whence
s 21, (2
+q)
-)l)/(l
(4.41)
I/K
(4.40)
t2 a (1/8) [(2tTm)2.1]
f
from
which
for stability
s
T/Tm
(4.42)
> 1.0
2T/Tm - 3.0
*1
I
I
zAz, ,
l)
1.0
(4.43)
i
r2 + 2r-I
TM
z (1.r)/2
.00 of the
1.0.
IY12 - (,T)-2
(4.48)
(445)
wc = T" 1
(4.47)
rV -1.22
and for stability they must lie within the unit circle z
1.0.
Thsclearly
fails if r
This
cwe
2T
(4,44)
+ rz +r -0
zU'+ "z&J
0;r = 0.414
(4.49)
Expanding
Y
(wT)f
41
(4.50)
If ;z 2 1 _ 1. r - 1.00; and since we worked this case for
1.0 for the continuous case.
If we let V = 2, we have
3 2
-z + rz +r
Iz
2, wehavethe
3
(4.51)
00phse lg
(n
(1T2) - iTm
the 90* phase lag comes from the integration, the next
results from the constant delay. Setting iA -vr,
(4.46)+term
c m -/2
Trrm a 2/n
2T/T, - 4/4 -
1.27 20871-605
2081-0
which may be compared with 1.22 for the discrete simulation. This is rather good; in fact it Is better than siirulating this Y by the analog approximation gien earlier.
ZlZ 2 Z3 ' -r
S20871604
4.2.3.1 Conclusion
If the transfer function of the servo loop is
(K/s)esTm
11
(4.52)
It
i
fired during the brief interval that lead is approximately correct at each fly-through.
Whether this is a preferable operational solution
compared with the far simpler course and speed sight
is unlikely to be settled by experiment or analysis, but
the availability of simple low cost laser range finders
should relegate range estimation to a strictly back-up
mode.
II
I
3
400D
3000
RECTIONl
"D--"I)
TARGET
METERS
I:
"O
2000
tI!
1000
S01
II
t
4000
2t0o
OBSERAVERA
1
1
1000
20671 13S
(moetrs)
Absolute Range Estimation Error
Figure 4-14. Footprint of Mean
human response dynamics. The complete systern should feel like good 'aided tracking'.
(3)
a. Human OperatorAssistance
I
V:
(I)
I
(2)
c. Tracking Extrapolation Through Sensor Interrupts. Drive the sensors and provide extrapolated
tracking data in all three coordinates during
intervals of target obscuration, ECM interference
A
with sensor operation, doppler radar blind zones,
and other causes of intermittent sensor inputs.
d. Fail Safe. If the regenerative element fails the
system should retain the tracking capability. The
Then
Zl a Ye
z2 a Yr y
Y = Y, e,
e. -
z,-
dy/dt
Y.(s) e
transfer function of the drive servo, It
may vary with time if, for example,
the loop gain is made a function of
slant range or some other parameter.
transfer function of the regenerative
unit. It will vary with time beause of
the changing geometry and other
causes.
Y,(s,t) -
Y,(s,t) -
Y, -
m -
Y~e
Yc -
z4 -
drive error
Z-ZB
24 -
m -
4;r-
velocity vector
z,-
(4.53)
unit
Yem
YpYc
'; YpYc a Ye
(435)
4-22
Ii
ii
f
'U
y a eSYS
M(z*)YS
y/z = Ys/ (I + s YS)
(4.56)
(4.58)
y/z
UfS
(4.57)
jS v(t-s)W(s) ds
y(t)
(4.59)
in
y/x
= system output/input
e/x
m system error/input
(4.60)
Hence
y/m - system output/control input or: system dynamics as they appear to a human operator if
there is one.
Z2/sx
z(t)
W
measure of how rapidly the solution will degrade if the target is lost.
R(t)"
f s R(t-s)y(t-s)W(s) ds
If we expand
R(t-s) = R(t)- s dR(t)/ds +
T4
to provide 'feed
I) automatic tracking,
2) manual tracking.
In the case of automatic tracking. the servo bandwidth will normally be much wider than the bandwidth of the filters in the regenerative module. Then in
continuous system operation, the transient response of
y/x will be essentially that of the servo. The regenerated rates will display transient response dete:mined
entirely by the filter smoothing time. The servo
and regenerative module components can be designed
separately with only a minimum interface in the lag
correction function.
The best filters for the regenerative unit will have
similar requirements to those of the prediction module,
4-23
Il
oTS ys
(4.62)
(4.61)
ERROR
_NSOR/
_:HUMAN TARGET
-
_ ERROR
PROCESSOR
YpIS),
INCLUDING
OPERATOR
~ S
OPEN IFJ
~~LOST1
DRIVE
"
SERVO
YS
REGEN
Vr
Z2
II
(5.
72%71-136A
I2 (
I+
T 2 l)
,)
(4.67)
where c. and c, are chosen to match the 'ideal' functions described above. Then Equation 4.15 can be
solved for Y, as a series in s" to obtain the required
YAYc
y/n, 1!4'r
(4.66)
(4T66)
(4,65)
4-24
I
Table IV-3. Summary of Transfer Functions
Expression
Function
System output/input
I+ Y
System error/input
System output/control
input
Irates
e/x
Ye
'Yr
y/x
(sI-Y)
I + Ys [+
r
YeY I+
Ye
YY
zz/sx
s I
Yf
.
-
s - Yr
Ire
s rY e
Rcenerated rates/target
Vc
c
I + Ys(iYr
I
y/m
- Yr
YS (s + Ye Yr )s(+Ye'Y
20871-SOSA
.
REGENERATION
SCMUTN
INVARIANTS
COMPUTED
RECONSTITUTED
INVAR:ANTS
RATES
TGT IS
ti
207.11737
4-25
smoothing and recursive smoothing with parameters depending on target path characteristics.
4,3.4 Dlsousslon
Data flow is shown in Figure 4-17.
4.3.4.1 Servo Lag Computation
There is no change from AFAADS-I, except that
range lag is omitted on the grounds that I) lag is a
minor problem with ranging systems as compared with
angle tracking systems and 2) as shown in AFAADS-I
small range errors have a negligible effect on prediction errors.
")I /(KaA)
incorporated
acterlsttcs:
LCA(t)
Ar(t)/Kv + Ar(t)/Ka
..
(4.69)
vided.
c. The major improvement is the provision of a
capability to have the solution switch to regenerated data when tracking loss is simulated, run on
regenerated data until tracking is assumed to be
resumed, and switch back to tracking data incurring only the transients resulting from the difference between regenerated data and tracking
data at the point of resumed normal tracking.
SXr(t -
t ) + VXA
-
Y1(t)
I
I
(t
+
- A-+
AFAADS-I, namely
VZA
(4.70)
.1
4-26
I.
R .A
r
Dr r
rr
yVx
XV
NV
Rz
ZR,
+ x
RrLr =Y> Yy
+record
DAI
P:RTAr
(4.1)
First note that when the target is deliberately performing an evasive maneuver, even very low acceleration maneuvers, the deviation from a predictable path
is so large that it will dominate whatever contribution
-2krAr
f) ()2Z
(4.72)
(to
I-
..
r
!JI
in
since.
notIn
noto include flght roughness
roughnes.....
eosis
.s
.....the simulation,
aatyss
it w conluddtht
fightrouh
sw
"'
LL
SNOISE
-
,+
-TARGET
INTERRUPT
IIL
SMOOTH,
LAGC
VxV
-*PATHORPO
V
Z
0,
CORRECTION
fr
CTOP
Or'
0'o
Zro
REGENERATE
POSITION
NI;
!j
20871.1390 4
Figure 4-17. Radar Tracking, Noise, Smoothing and Regeneration Flow Diagram
can be neglected (although the following method allows their conclusion) and that deviations in a lateral
horizontal plane and in a vertical plane can each be
described by a variance and an autoorrelation consist
ing of a damped sinusoid,
These deviations are computed as a stochastic series
in a manner similar to that now used to generate
tracking noise, but with a different form of autocorrelation (not one-step Markov).
This
"' i gives
i e the
h right
i h answer.
n wr
4-25
WHITE
..
VOICE"
UNIT VARIANCE
ZERO MEAN
GAUUSIAN OR POISSON
SICOMPUTE
FLIGHT
SERIES
i-DEVIATIONS - ICOORDINATES.
'
3-AXESi
IN AIM
POINT By
PREDICT
"PRI:SENT
_COMPUTATION
HT
I+I,
~PATH
NOTE.
ERROR IN
AIM POINT
INCLUDING
FULL COnRECTION
FOR FLT
ROUGHNESS
20671-139A
FLT ROUGHNESS
DE VIATION-.-
"
IAD
-- '---'""l - -- to
----
TO
MEAN F L7"PATH
SU TRACT
FROM PREDICTION ERROR
OR ADD TO FLT PATH
TRACKING ERROR
AT I + tl
'
2071.;40A
I
42 4
S~4-29/4-30
SECTION S
"DATA SMOOTHING
The filters used in the AFAADSI simulation were
optimized for maximum variance reduction of white
"noise with a constant smoothing interval, i.e., memory
time.
tI
- Th
prk-m nalsis xamnesthe onsg,_,nce in
The
present analysis examines the conseou-ences in
terms of filter design of imposing an additional con"straint on the optimization, namely the integrated
square variance of the error during the initial
mean
settling of the filter. An attempt is made to compromise between this transient penalty function and the
reduction of variance or error resulting from tracking
error in the steady state.
The effect is to change the filter coefficients to place
greater weight on the most recent data. The fixed
memory discrete filter then resembles closely the recurv filters.
sive filters generally known as a A tentative conclusion is that the finite memory
filters have no fundamental advantage over the a u algorithms when settling time
is important. Since the
*-
::..
3--
error
with time,
Sun dispersion,
and the distribution
of
optimization,
consideration
of correlation
of prediction
rn[,,(t)l 2 dt
1?
where h makes the dimensions conformable, and expresses the relative importance to be applied to D,
versus D,.
For this analysis we take m - 0. The effe,'t of
making m vcry large would be to se-emphasize the
contributions of the oldest filter points, and is roughly
similar to setting h very small.
We consider the optimization of J for each derivative class of filter separately. It appears that if we
a best ,.ompromse prediction unit as a whole,
the derived filter coefcients will depen d explicitly on
time of flight. This ib contrary to the case of simple
variance reductian, where tlhe optimum prediction unit
ls-II
ll
lllllI
I",
'
-- "
...
x'-
..
n
,cj=0
n
T ic= 0
0
y(N)-
A2
j2 c.
= 2.0
(5.7)
J
J
(5.3)
0
where x(N) is input at data point N and y(N) is the
developed output.
This expression can be conveniently written in terms
of the z-transform as
n
y(z) - x(z) . cj zJ
(5.4)
0
I
1.0
(5.5)
i1
(5.6)
Ds a
-5s.2
(*(k)l 2
(5.8)
Given a sum
(5.9)
Ds =a2 I
s2 =
n
T=0
If(nT)] 2
(5.13)
jfact
U
2nj
An increased penalty for slow settling can be imposed by basing D, on k'e(k). The corresponding
z-transforms are (for E(k) - ke(k))
Im
--
JF@)F(z)z
(f.14)
dz
We wish to obtain
E(z)
i~ajz~j
El(z) =
=;
m 2;
E2(z)
jZ'J
1
2ajz'j
le ( k )] 2
D,
m I, E t z
( 5.15)
nuO
(5.10)
n-I
e(z) =
az
ajz'
(5.16)
~kdz(S.11) Ds"
akz
0; k >-1
Ds-
- ; k -i
0; k <-I
'dz
Z
(f5.17)
ajzf F
21rj
*the
J"- I
(5.18)
0 =2
(5.12)0
5.1.2 Noise Varlanee Dmwit
n
Y=(5.19)
Dn
imo
%
(S.2)
J a DS + X,*1
-o
a,
C-C01
a2
-1-c
1*
0 -C
where A is the relative value o noise nwdactiam compared with sctling time- w, am obtis a best cmmpromise filter iy choosing th c,so minimis J.
5.1.3 Position Filter
Apply a step function at t w 0. It has the
z-transform
(521
Oj
an.-I
E+]I
Ck
n-I
Ck
ck =
-Ck
k-Ok
CJZ'
+n
(5.24)
(5.22)
j-0
' ;n-1
no.o
I,
j-O
n(
Then
Dn Z-
(225
n
e(z)
Cj
C.z.
ao+
(5.23)
I
IS
MM
ii i I
We wish to minimize
We have
(5.26)
Slao
Ds + .Dn
(5.32)
whence
' -an-.I
.c-
(c)
[i
00
I 1 0.
(5.27)
a/ac 2
vc
a/ac 3
AT = TR
AT cT
Now
n-l
ID,
aia cn
D
t. I, 2
- aj:a
2087 1606
Ds-
CTRRTC
ATA
(5.28)
n
.0
+ XUUT + X 11 C. U . 0
0[RRT
so that
+UUT+X-1RRTIC.Uw0
1.0; UT
I o+CTu
... I1
1 n1terms
(5.33)
(5.34)
(5.29)
(5.35)
since
D
Dn
C2+ cTc
(5.30)
substituting for co 2
Dn
(5.31)
35-
andan
[II+ UUT]I-
(5.36)
(.0
(5...)
Then
C - 11 + UUT] -I U
C
U JuuTU (I + n)-I
U.
velocity is therefore
E
-
e(z)
1) (z- I) 0
Dsga xz-
SU I. nj(1 *n )
UDesignate
n+e
I
e()
zJ
(5.41)
X(
II
a2
(5.42)
In view of the constraints and the limit conditions
for the z-transform we know that the numerator of
'
nator. Hence
~n
o
(x)
n .1.0
*i
" -I.02
(5.38)
sao+alx+'+a
Xn.lX'
-X2
(5.43)
ax
0
Apply a unit step in velocity at j , 0. This has the
(5,44)
trans orm
5-
tIt
ix+I
SEquating
coefficients
Now since
at
n-I
SDs
'~
aj2
8 a 1-c0
a 240-I- cwe
can write
1+cTcc
a2 - I 1 -3c,-2c1 -c2
a3
i " 1-
JO.k)ck's-E(kJk
an-I
wee,
ko
"n-2n
Now
Dn =
I -F,(n-l-k)ck
|0
j-o0;
cj2
and the velocity filter has the constraints, for unit sampling
interval
(5.45)
Cn
;
M MT;
where
"-"n
c.O
j
Wj
We obtain, therefore
~iCj
Ii
.. ':
0,
=l0
n(5.49)
aj
kmj
and ao
(5.46)
(k-j)ck
1,0
O; UT 1I
co+cl+CTU
2 100
IA
I1 ".
(5.50)
-3210
:L4
....3 2 1
(5.51)
(5.48)
CT M
""Note t
with
DDT+IDtU] [I+U]f+l
Then as before
express
5.7
[I
(5.52)
Ds + XDn
(5.53)
"-
-( (I I-x)(0l/2) 2n cjxj
the operator
Apply.(.x)2
.4
LV*
+aI
o aoa
+
Uet c
set rc J -0
Performing
coefimcients
and obtain
C
12D+UIX
IB+XN]'
+.nIxn+(5.59)
the
long
division
and
equating
(5.54)
I -(1/2)co
a2
I,
(1/2)(4Co4.C,)
a4
5- 5
e.
(5 ,6
0)
so that in general
aj = I
-k)2 ck j 1
(1/2)
(5.61)
K.k
l(j
(I/2) z (z + I)
S~(z
- 03n
(z-, )
(556)
(I/2) 1
(kj)
ck
j~ l
k-j
t -(I/2)
(5.57)
(5.62)
(1/2) Cr
an'1
[a2
z-2
an.1]
0/2030
S1
00
9/243.1n][
16
9 4 1
Now
j2cj
2.0
,(I/2)CTE
n-I
(5.63)
D22
(5.58)
I + ( - Co)2 + ATA
Because of the constraints we can write the z*transform of the Illter transient error as a polynominal in
and
34
p.H
AT
(5.64)
n
n
Substituting into D
Dn
T
2
C0 2 +C12 +C 2 + C c
CT 16S+12D+2U] +(1/4)CTFC
(5.65)
+ IS + 2U + 3D] IS + 2U + 3DI T + 41
Ds
c2 .
They are
co+c
XDnn=
+C
1 +c 2
c I+ 2c 2 +CT[2U+DI
UT - [1.
" 0
DT -
CI +4c 2 +CT[S+4D+4U]
i,
n.]
11 2 3--.]
J
ST - (1 4 9 16 ...Ia
(5.66)
'l
From these,
c)
U I -(I/2)CTIS+DI
c!
a .2+CTIS*2D]
SC2
1 - (I/2)CTS + 2U + 3D]
(5.69)
The result is
(5.67)
5.70)
Substituting into D(
T
TTEETC
= i +(l/4)CT[IS+DIIS+DIT +EET]C
I+(1/4) cTGC
(5.71)
C - 4F'113S+6D+U]
which is the least squares solution for zero settling
time demerit.
(5.68)
II
1A
_I
.
,,,
5.9
Position
Sampling Interval
Veoct
Velocity ,o,(
piing interval and let b1 be the corresponding coeffiThen for the three
for awesampling
cients
followingA.correspondence.
have the interval
filter types
Position Filter:
Velocity Filter:
bjy
Acceleration
(5.72)
Velocity
Velocity
Dal /A12
Ds
W5t
d~il
= ().I15)
lj/l,
Q/nD
s l + (IAS)Dn]
(5.73)
(D. l/Dn iv
_ 1/4 Urn
5celeration
(X/AS)"
(DsI/Dni)a
Filter
Dal
)[D1
{o
3
I
U
(5.74)
Position
(J 8 IX)
&X/,[s
,(
Do
.)1/5
cjvl
Acceleration Filter: b. = c.
it =j2
(JIp)J a (444%D
5 + (V.A)Dn1l
Acceleration
Dnl/A4
~DtlImemory
20871-506
ml
mmm;nmmmlli~l
ipm
'mmmmmmmlm
plmi
R',mmmmmill
S'
4 POINT FILTER
S POUNT F ILTER
I
1.0
S10
POINT FILTER0
1.1)3
201171-41A
Sshown
in Figure 5-2. For the two point filter, no
optimization is possible and its two points lie at the
AZ/(Z- 1)2
S~points.
I1
I
I
(5.75)
e(z) - Az/lz
1)2
c(z
(5.76)
0
and the steady-state error when the filter has settled is
E(w) -
5.11
im (z - l)E(z)
z. 1.0
- Urn
Zc z
(5.77)
(T ) 2
P/a
polynominal
(&'/2)
j-
, j2 b~.
(.4
A~no+...Snlz'(nl)"](5.79)
C(9
wee .
,Xp
whence
n
T
(580)
(5.80)
Tp/V
where a -a
Now
vo
z(Z+(
(z- 1)30
(12)
(5.81)
tpTv/
+
-(A
n
bjzJ
Z(z
12)
(5.87)
(5.82)
(5.83)
jul
lime of
x. -
v. -
smoothed velocity
and if
x v -
smoothed
tion
smoothed
o
0
pv
(
)
(5.)
p
par of the smoothed position computai-on by
(bpj)c o bpj + Tp/v bvj
5-12
0iLI
prediction.
-( /)2z
1+
(5.86)
z) 2
a 2i + Yo Tp/a +a Tp/a
+ v Tpa +i(Tv/aTp/v + Tp/a 2 )
T,&U z -
xo
The steps are identical with those followed in determining the lag of the position filter to velocity, but the
algetra is more lengthy and will be abbreviated here.
Apply a step acceleration of unit mapnitude. It
generates the following z-transform in position input
(112) .2
(5.85)
(5-90)
nlgh
I'
'I
1 .0
'I
i6
WN
10POINT
0POINT
*-%-
\
.1
~OPOINT
2 POINT
SMOOTHING TIME
-~20071.142A
Position Filter:
Tp
n A/2 a T1/2
2
T2,
/12)n(2n+)
1A
+
Tp/
(Note that this term is negative)
Velocity Filter
PI
Tv/a
5-13
n A/2
Ts!/2
(5.94)
W*(s)
s"=
I + As + (A /2)s
+T
T +T2
(T95
(59)
20871"607
as given in AFAADS-I (Vol. 11, p. 4-3).
5.1.17 Analog Filter
function
+
W(s)
sT2) a d/dt
(1/2)"
(5.91)
I
+ T2 )(TIT 2 )
D =(T
(5.92)
T12 +-2TT 2 + T2 2
5.1.18 Gonwallation
Instead of beginning with a specified
filter structure.
such as Equation (5.91), the problem can be formulated in more general terms, with the weighting funcfilter determined as an
best compromiseproblem
tion of Ittheappears
can be solved by
that the
output.
the Marshafl-Yovits'nmethods (and no doubt others as
well), and the approach is sketched below.
(5.93)
0.707
Hence, as one varies A, the shape of the filter response to a transient is unchanged, but the time scale
depends on X. If one desired to vary the filter value
parameter J with slant range, for example, one would
make A a function of range, but preserve the relationship, Equation (5,94). This is the same conclusion
reached by Benedict and Bordner.
is minimized when
I5
I
5-14
2rrJi
(5.96)
00
Ix~w) 12dw
(S.97)
1.0
where x (jw) is the Fourier transform of x(t).
If for xQt) we take the transient error of a filter in
response to a step input, zero for t<0, we can utilize
this integral.
Consider a velocity filter, with transfer function
Y(s). For small s this must reduce to s so~ we can write,
'n terms of a new function F(s).
(5.98)
Y(s) a s F(s)
Apply a step function in velocity (transform I /s) as
a
ramp function in position (transform I/s?) to the
filter. The transform of the error is
Fs) sF~)
1 / -(I I S)
I/s.(~/~)
1 Fs)
sF~s)
5.9)
(99)
71
Ds 1/21tr
W)2
dw
(5.100)
We wish to minimize
a'.
lFj)2
x(-jcw)xOw~) dw
0
lf(27r)J
I.
(5.102)
0I270
f~
+ , 2NMw') I F(J-)1
s+ NXDn
2
- OD
S0
0.5
0.707
1.0
2.0
3.0
4.0
5.0
1.0
0.58
0.53
0.48
0.41
0.38
0.36
0.35
0.30
-1.0
-0.25
-0.18
-0.12
-0.03
0.01
0.03
0.04
0.10
c2
0.0
-0.25
-0.23
-0.21
-0.17
-0.15
-0.14
-0.13
-0.10
c3
0.0
-0.08
-0.12
-0.15
-0.21
-0.24
-0.25
-0.26
-0.30
Dn
2.0
0.48
0.37
0.3 1
0.24
0.22
0.21
0.21
0.20
1.0
1.18
1.23
1.29
1.39
1.45
1.47
1.49
1.58
c0
20871-507
L/A3)
j/0,/3
0/A
0.5
0.707
1.0
2.0
3.0
4.0
5.0
8.0
1.79
1.68
1.60
1.48
1.46
1.46
1.47
1.59
20871-508
These algorithms were designed for use with trackhence incorporate a one-point prediction which allows them to track through missed
data points. They can therefore be used in regenerative
A.l:,')ri:hm designs. The coefficient shaping is optimum
for irmcking; A has not been verified that the same
e
would he optimum for prediction to many
tumes the sample interval as is required in predicted
fire. Conversely, there is no a priori reason to doubt
their applicability.
smoothed velocity at j
I
1
*,hile-scan radars,
X (J)
a- Wconstants
Let
x1)-
x(j)
0(i)
Xp~jl
X (3)+ &aix(3.)-Xp(3)]
(.1)+( WA) Ix(3)-x(j)]
10
!3:
.'
5t
sampling interval
and y. we obtain
5-16
a* ({ " )z'
(2-oaP) +z'
x(z) = x(z)
I -z
Velocity:
2
[v(z)/X(z)
ev(z) =2
(l -o)
z(z-I +o)
v(z)A
x(2)--
~Ma'
v~~~=(2
(lz1z2z2o-)(-)
0- -1
'I(2. a + z-2 (
+(
Z2) .'(z(-o)0( )
(5.108)
Lim (z-l)ev(z) = 0
z 1za +(0)]
z-1
z 2. z(2 - a - )+ (1 - )
subject to the stability requirement, so that the steadystate velocity error for constant velocity input is zero.
(5.2.2 Noise Variance Demerit
From Jury's tables6 , for white noise of unit variance
(5.106)
Position
Dnp =
(4.2a3-
Velocity
Dnv
A2
202
()
(51l0)
72 a(4-2a.-0)
ix(z)
a(Z_1)
(z.l)2
20871.608
Position e (z)
P
[i
_
(z.I) 2
z [zk'+ (P-o)]
z2. z(2 cr. P)+(l - ot)
E
j=0
20871-609
Az(l .o)
(2. z+(I
The error at
eP(
e c)
oa
at t,
mT.is
is error
obtained
as -+as
or)
[ep 0)]
(5.107)
- Lim (z -1)ep(z)
Z-1
C0
le(t), 2
dt
20871-610
5-17
a - 3/4; 8 - 5/4; a a 1/7 and the velocity coefficients may be compared with those in Figure 5-2 for
the finite memory filter.
cp&)
(S. I*I )
j-0
(5.12a0
(5.113)
filters for tracking systems, and the maximum prediction required is only one sampling interval. This allows
the tracker to carry through missed data points. For
the anti-aircraft problem much greater prediction intervals are required. and the problem and the optimizations need to be reviewed from that point of view.
DW
a 2 (2 -a)i 2 (I -a)
aj3(4-2a-0)
(5.1
a2
Ti2s-a)
(5.114)
This corresponds to a slightly oscillatory transient
response of the filter. There has not been time in the
present sttidy to confirm the optimum or to investigate
= (2-a) + 2(1-a)1/ 2
z
=
I
'p2 . oT02 [1 + 2(tp/1.t) + 2(tp/TS) 2 1
a (j+l)ai + (j-al)ja"
(5.118)
(5.117)
These have been plotted in Figure 5-3 for A -
I;
5-18
or,
(o
a 2112.
1:
II
!
1.0
VELOCITY
I
05
U.
z
01
0
I1"
*,oI
2W71.143
SAMPLE INTERVAL
where L - lead angle, o - angular velocity, and 1 time of flight. The smoothing time constant is at , and
a -
*!
atp dXL/dt
This is not a riew concept. The well known gyrosights compute lead according to the expression
t
at~
p dLdt ++ L
(5.119)
XL
(dXo/dt) t a
1i13
(5.120)
The solution of
XL
y(t)
(dXo/dt)tpIl + a dtp/dt["
*(t,
(5.127)
to) y(O)
(5.121)
This error was first discovered in dynamic tester
runs with the M-7. Even against World War il targets
it was of significa.nt magnitude on the receding leg of"
".he target path.
(t, 0) = L(t)
(5.128)
X,
(5.122)
Vx tp
It is the solution of
t
d/dt [O(t, to)] = A(t) 0(t, t,) 0(t,, to) =
(5.129)
differential equation as
dx/dt - A(t) x(t) + B(t) u(t)
(
(5.123)1
+ L(t)
x(t) = L(t) x(O)
0(to' tl)(tt
0(t 1 , to) =
0(5.124)
0(tI. to)
(5.130)
o")LI(to)
j
Then
0 (to, t1 )
(5.125)
2 ) = *(to, t 2 )
(5,126)
(5.131)
5-20
--
"-P -MM-M
I
If we designate ) t) - B(t)u(t) . - the system input,
!et t, - 0, and let S , t-s
.t
x(t) = 0(t.O)x(O) +j
YJo = Y(O
y)
yO
y(t.S)O(t, t-S) dS
ft x(s) ds
a+tJ
(5.136)
(5.132)
The operational concept was t'-at after target acquisition, f(t) would be allowed to vary as a + t. a could be
very small, and the initial transient would decay as [I
+ (t/al'. The weighting function over t is very close to
a uniform ameraging. which is optimum for white
noise.
To display the solutici for f(t) - kD, consider a
direct incoming target fl),g down the line of sigh,, at
velocity V i.) that dD/dt - -V. Change the variable of
integration to D.
Consider a target acquired at D. and tracked to D,.
Since K has the dimensions of velocity, define K - V,
y(t) -
filter output
x(t) -
variable to be smoothed
x(t)
(5.134)
(5.135)
ds
fto[flu)I'
[fA
+e"oif(s)-
z - D/D
x(z) d(zVs/V);
(5.137)
t Ifs)]'
josj~u)]
ds
du x(s)
,,, ds
du2.0
St5.2
1/5-22f
I
I
g
i
SECTION 6
PREDICTION
This section represents the analyses of a number of
prediction algorithms. These range from a sophisticated system using corrections based on actual measurements of projectile miss distance to barrage fire.
7The types analyzed by no means exhaust the lia of
possible algorithms, nor are the analyses carried to
completion in all cases. It had been hoped to parallel
the analyses with computer simulations but this has not
been possible within the resources available in the
present effort.
6.1 CATEGORIES OF FIRE CONTROL SYSTEMS
In the following paragraphs fire control systems for
predicted fire weapons are cateorized, from the simplat to the most sophisticatedl list is incomplete.
and since there are so many descriptors needed to
completely describe a fire control system, the emphasis
is on the method of prediction. Examples of each
category are given. The categories are:
a. Barrage Fire
(1) Almost all World War !1 heavy gun fire control systems, including the M-4 and M-9,
M-10 series (U.S.).
o
(2) Vigilante fire control system.
h. Complete Solutions with Corrections for Target
Acceleration
(1) U.S. M-33 system for heavy antiaircraft guns.
(
(2) Current Superfledermaus (correction for target acceleration in dive, or 3-D curvature
options).
i. Solutions Using Corrections Based On Projectile
Miss Distance Measurements
(I) US. Navy Phalanx.
j. Solutions Using Internally Generated 'Synthetic
Tracers"
(1) US. Air Force 'Snap-Shoot' Concept.
Although not a prediction function, the provision of
regenerative tracking is considered to be essential for
high performance short r., ,e weapons. Regenerative
tracking has been used in US. Navy fire control systems for about S0 years, beginning with tte Mk I
computer. It was employed in the US. Army Vigilante
system, and is currently used in the Superfledermaus.
(2) Course and Speed Sight, M-38. A sight operator sets estimated target speed and aligns a
miniature target indicator with the real target
heading and dive angle.
e. Rate by Time with Estimated Range
e
(I) The original Kerrison predictor.
) Fife
control ocurrent
British
la)irelcntielon
cro
ntish Falon.
Facon. AgU.
These
are multiplied by time of flight obtained from
estimated range to obtain leadsi
technology to the predicted fire problem is the poten"I Of tracking the projectiles, measuring their miss
distances at the target, and applying appropriate correcons to the fire control system.
degree this kind of observation has always
beta possible by visual observation of tracer rounds or
by visual observation of the burst of t'-ne fuzed projecdtiles.
Thebyusefulness
the method
in thethepast
has been
limited
the man'sofinability
to sense
moment
at
which a tracer reaches target range, or even with
stereoscopic range finder observation of bursts, by the
6-1
o
The same principle was used with the Computing
Sight M-7 (Weissight) for which there were two inputs, aircraft heading, and speed. Heading was adjusted unti! the tracer stream 'crossed' the target, and
speed 2djustments were then made depending on
whether the tracers passed across the face of the target
or were obscured by it.
These systems worked well in proving ground tests.
Whether it was possible to use tracer adjusted fire in
combat, especially with many other weapons firing at
the same target, is unknown.
Tracer control of fire of machine guns and automatic weapons has been extensively used in the past.
The usefulness of tracer observation has been a subject
of continuous controversy, which has some basis in the
fact that the gunner's depth perception at normal
target ranges is nil, and so he cannot tell when the
bullets are at about target range. On the other hand,
tracer observation can assist the gunner in getting the
trajectories into the slant plane defined by tie gun and
target heading, even if tracers are of little assistance in
setting lead within that plane.
the correction based on measurements of the trajectories relative to the target must be used to corret a base
prediction algorithm, because of the time of flight lag.
Because the errors are referenced to the target, the
system has the unique capability of being able to
correct for constant angular errors of the system (such
as boresight and orientation) and for slowly varying
crrors such as may result from imperfections in the
base prediction algorithm, ballistics, non-standard
muzzle velocity and meteorological conditions, etc.
Since the measured error includes the random round
piobable
projectiles,
it seems
dispersion
to
averages
of
mustofbethebased
on weighted
thatround
corrections
o
mcasured miss distances or angles. The weighting
the
s.f
ang
tor
funcs
fs
be
should probably
y ba funcion of tine an some or the
p
engagement parameters such as range.
If it were not for round to rot"' dispersion, there
would be little gain in measuring error more often
I
I
6-2
op~-
'1 1&4
S08
Figure 6-1
2071 146A
components.
S'
~6.2.4
"
-matic
i
2720, 15A
4TV
t
I
su.ple-
mented
indications however,
of when adjustments
to observe.
Without by
thecomputer
range measurement,
can be made to get the prediction into the right 'slant
plane', but not to correct within that plane. But even
this partial correction may be helpful.
6..
ro
esn
ihRdrdlt=-g
or IR. or
velocities can
be asangular
high asvelocity
1600 mils
per second.
The
observed
vertical
component
caused
by gravity drop is approximately
D = slant range
I63
1
shows approximate observation times for two sensor
fields and two angular velocities of tracking.
If one only has the angular miss picture (Figure 6-I)
the best one can do is apply corrections to try to make
the relative bullet trace projected in the sight plane
pass through the target. This will put the bullet in the
right slant plane, but it will not necessarily hit the
target. For example, it will not correct for a systematic
error in bullet velocity.
possible.
b. Provide some compensation if possible for the
varying errors which persist for longer than time
of flight.
How well it does this depends on how latrge the
constant and slowly varying errors are compared with
the rapidly varying and random round to round errors.
300
Trj,kin
1410,
41.3sev
3(0 c
30
An,,lJr \'o,,,', ol
940,
0,03 ,.
0.3 wc
20871-509
6-4
ii
TAGE
TARGE
FiRE
miss
II
GUN
CONTROL
MEASURED
DISTAt[
ERRORS
2W871147A
Figure 6-4. Top Level Flow Diagram of System Using Corrections Based on Measurements of Projectile Miss
Distance
b. The refer. ce estimate contains the error compooy target maneuver and approximanents cation er .s of the basic prediction algorithm.
Hence these can be included or excluded from
the correction as desired.
c. The reference estimate can be used by itself as a
correction device for simple prediction algorithms (for example a course and speed sight),
X,(t) n(t) -
sor)
X,-(t) - sensed target position at time t
H(sp) - data smoothing and prediction
transfer function; s - d/dt, t time of flight
X,(t) - prediction of target position at t +
t
, made at tim
p i
t
round fired at t - o
Its application to the present concept follows a suggestion by Dr. Daniel Willard that one need not wait time
E(t) -
E,(t) -
S6-5
correction C(t) are shown in Figure 6-%, one in advance of the filter and prediction algorithms, and one
at their output. The second mode is preferable because
or discrete implementation. However, since the projectile measurements are discrete events, and the time of
flight lag is most easily handled in the discrete formuR
lation, the analysis is done for 3 discrete system.
= X.(t)
E (t)
G(s)l]
E(s) = Xt(s)
+ w(t)
n(s)
+eO's H(s)J
+easG(s
e'-s
t1 + e"as G(s)]
(6.7)
(6.3)
e-03H - I
E =X
I + " HG]
*va
C C's HiG
+n
e"'sH
.'r+problem
5
ase
+ e_
& SHGJ
The first term in Equation (6.7) defines the prediction function. The system must correctly predict for a
polynomial of assigned degree, and must have a satis-
(6.4)
(6.5)
The third term defines the ability of the miss distance processing element to average out round to
round dispersion, and other errors in miss distance
measurement, and must show a satisfactory reduction
of this error source.
E~t)
Em(t)
CQt)
(+n(t)]
m H(sa) [Xt(t+n
m Xp(t) + C(t) + w(t)
C)
t)correction
a X
aE(t) + Va(t)
-G.t)Era(t)
(6.6)
6-6
S-..-
..
" "-
SYSTEM BIASES.
SLOWLY vAAYiNG
EARORS. FAI5.
ETALIGNMENTS.
SENSOR
ERRtORS
ETC
r~(t
AASOTIG Y
I'f
TAGTADPEITO
PAHAGRTM
MESRDMSSISAC
FLGH
GUN AYIN
*-I
Toidetrmie6-heFlueof
Diagram
for late
Crrectio
useated
syselatios
6(67
-Iz
oz(.8
O-G
CQ)
COj-I) + G Em
(6.10)
G= K
from which
z2(l a)
(6-16)
(z. ,)(z.a)
G(z)
GI(Z)
(6.11)
"E((z))
GI(Z) Em(z)
z-
a)Z
l)(z
-a)
(617)
(6.13)
Ero) v Em(J)
(6.12)
*z
(1
(6.14)
4white'
2(l - a)
(I-K) [2(l +a)+K(l-a)j
point. Then
C(z)
-Em(z)
2
K (2-I)(Z-a)
(6.18)
but they serve to
These aarepedestrian
not optimum
values,
indicate
approach
to symem design as a
preliminary
future analyses using KalmanBuoy methodstoofpossible
optimization.
(6.15)
c
As noted earlier, we should probably prefer to make
K a function of the index of the miss measurement.
K(.), so that the amount of correction applied is large
on the first measurement, and decreases with time. One
may thus have rapid reduction of w. yet maintain
system stability.
"
misses
caused by target maneuvers
misses velocity
caused
by system
if and
a constant
prediction misorientations.
algorithm is usedHence
in the basic
predictor,
and
target
maneuver,
and system
dynamics, but
will
not include
the errors
in boresighting
sensor
and the gun,
or those resulting fromtheballistic
'
Ea(t)
(6.19)
AE a Em -Ea
C - -G A E
(6.20)
We now have
Em(s) a (Xp
C + w) e's.
Xt + v
6-9
(6.21)
z
a good compromise reduction may be achieved. However by periodic calibration f.rings using 'canned'
internally generated target paths and several standard
courses, the component errors can be identified and
conected.
The general topic of projectile tracking in conjunc-
tion with 'canned' courses for calibration firing deserves detailed analysis beyond that which can be
given here.
(6.22)
Consider the term involving w(s). A constant initial
value w, will be eliminated with time if G(l) - I.
A possible algorithm for G(z) is obtained by computing CO) from
CO)
(6.23)
G(z)
G(l) - 1.0
(6.24)
2.
S2+
c.w2
2w 2
2ja-
Although the maximum number of projectile observations is desirab?^ to smooth out the effect of ammunition dispersion, the first observation is not available
until time of flight after firing the first round, and the
effect of making a correction based on that observation
is not seen for an additional time of flight.
(6.25)
T
N Efdt/tp
(6.27)
61
ET
IIt
MOHN
AA
TAGTADPEITOIpl
LORTM
PAH+
PAVORDESA
GU
PNOSITIO
RA
U
COPT
TIEO1RESTM
AI
SYNTHETIC0
8-8.
Figure
ow Diagram fRAJErrCTiORYae
nMaueet
MLAURE
PRJCTL
itne
fPoetieMs
Refrenc
Trajctor
MSS ISTNC
PRCSSNC
sn
ytei
p
mean trajectory to the target. This is not exactly what
would be seen from a ground sensor but the difference
is negligible for present purposes. These errors are
SI
ir
nd"'k
11
4".
.""
2)
_27
'19
20N71-stu
The round to round ammunition dispersion is simulated based on three independent, normal, unit variance sequences. The sampling interval can be as small
as that of the simulation to simulate a high data rate.
or wider sample intervals can be used to determine the
effect of data rate.
The angular miss caused by angular dispersion at
each sample is obtained by multiplying two of the
random sequences by the vertical and lateral standard
deviations now programmed as inputs to the simulation. The simulation now has a module which resolves
muzzle velocity dispersion into lateral and vertical
components at the target. These components are extracted and both are multiplied by the third random
sequence to give vertical and lateral miss of the projectile caused by muzzle velocity dispersion. These are
converted to angular measure.
J
..
j
i
I~~
~~
NGULAR
NO MUZZIAN
Ve
VLOCITY
DISPERSION
STD DEVIATION
VLCY
ANGULAR AND
]MUZZLE VELOCITY
I BIAS COMPONENTS
MODE IIIONLY
COMPONENTS
MRANDOM"SERVE
+
COMPONENTS
RANCESMMULTIPLY
MISs
OF MISS
SAIM
ERROR
COMPONENTS
MODE 11 ONLY
PROCESSING
COMPONENTS
F ROMALOIH
SIMULATION
ICORRECTION
CORRECTION
TO GUN
ORDERS
TO MUZZLE
VELOC17Y
20871.150
Figure 6-7. Flow Diagram for Simulation Module for Projectile Misn Distance Measurement and Correction
I
I
S
*
H.
where T, is a characteristic response time of the aircraft and pilot and may be about 0.50 sec.
Solving this equation for an initial value H(o),
H
H(t) a -(o)e
s
+
+ Ht (t-s)] e
/T
c d(s/T)
(6.30)
H(t)
/
/ .tITc
H(o)e/C + Hokl.e
'+f
t
H, (t-s)e
.srrc
d(siTC)
C
(6.3i)
Define
(6.29)
Tc dH/dt + H = Ho + Ht (t)
&H
1H-
aircraft
attd bv h efi
aim
t altitude above the terrain
ena
mean at
K,.-
2
an(~
H(t)
~.Tc
aircraft
terrain. flight paths. with terrain following over known
d(s/T)
mile would be required than against near sonic aircraft. With each square would be associated a i,,ean
terrain height and this would be stored in memory as a
triad with the coordinates of the center of the square.
a target is acquired its ground track is recorded
+H (t+as)e-sJ/
Ho + Ht
Tc
S"cofH
H
(6.33)
H(t + a) = H(t)- (I .-
d(S/T)
.v/T
tHe
c d(v/Tcc)
v/T
. Ht (t+v) ,- cd(V/Tc)
S(6.3,4)
The algorithm can be p-ogrammed and demonstrated by simulation. This would be interesting, but
not as definitive as testing it against actual recorded
LI
U
i
100
compared with present altitude, and the required acceleration in each case compared. If the no-turn case
indicated a very high acceleration requirement, but one
of the turn options did not, the prediction would be
biased to the turn with lowest acceleration requirement.
64 UNAIDED TRACKER CONTROL: 'GUNNER'S
DELIGHT'
The question of whether a gunner, using a machine
gun without a fire control system, does better wi:h or
without tracer ammunition is controversial, and is
unlikely to be resolved in the near future. Single
caliber 0.50 machine guns using only tracer control
shot down many aircraft in World War I1, and the
Quad 0.50 mount with a fixed reticle sight had an
excellent record.
The time of flight delay between observation and
correction makes it unlikely that simple tracer control
would be much assistance against crossing targets,
however the possibility exists that tracer observation
did help the sight setters of the course and speed sights
(M7). and the angle setters of the Stifflkey Stick and
Peca Sights to get the lead into the right slant plane.
This capability was demonstrated in proving ground
firings.
Since fire with a fixed reticle type sight with or
without tracer is always a back-up mode of operation
for antiaircraft guns, we note, without attempting to
resolve the tracer controversy, some early results with
tracer fire by machine guns against slow targets some
forty-five years ago.' The results may be compared
against modem Firings against helicopters.
1 2 11000)2
A
0
%HITS
- CALo030
- CAL 0 50
o 0
;o
607o
1000
00
20'00
20
RANGE ,YOS---0-A
0o
P I(A1000)2
O-CAL0
0
30
CALO0O
.2
V
1.0
071
,00
o
SPEED iMPHI
2u871 152
Ii
The straight line approximation in Figure 6-9 corresponds to an estimate of equivalent standard eviation
22
22.5 (V/100) mils ;where V a velocit /in mph.
(6.35)
This must have some minimum value as V becomes
r1R
Ut
I
,
t
*
i
Sexpands
(6.39)
and E is in angular measure, (radians). To convert to linear
miss distance
Em = (V/Va)
(XDm/D) (I+2a)
(6.40)
rithms,
andangular
some velocities
foreign systems,
it is believed,
use
smoothed
simply multiplied
by time
of Rlight for an approximate computation.
sap
1+ aStp
(6.36)
I
and for simplicity assume a constant velocity projectile.
slant range
-
I.
Dm
midpoint range
X = distance to midpoint
E,EGoe~tp.
(6.37)
X * distance from
midpoint
Dm
crossing range
(6.41)
0o tp
2 (a + 0.5)
6-17
"A"I
(v/va) t(Ev/V)Dm + Ec X1
I
I
001
TARGET
-020
100 METERS
100 METERS
50 METTERS
S0
EUTE
RS
0A
GUN
1,0METERS
07
5
2W71 153
Figure 6-10. Contours of Constant Aim Error in Slant Plane with Rate x Time Computation
It is doubtful that course can be estimated to better
than about 10 or speed to better than about 15%.
Then FI, works out to about 0.05 D where D is slant
range, for v/v. - 0.30. Surprisingly, this is about the
same order of magnitude as the lag error of the rate by
time predictor over most of the firing region, and if
one allows for some ability to correct course by tracer
observation on direct incoming targets at least, the two
systems are competitive, but neither is very good. The
value for E,, corresponds to about a 50 mil bias over
the whole firing pass.
During World War 11. NDRC personnel ran experiDurng Worlde Wrmine
the
ponneuan
menus to determine the abiliy
ability of human osexpers
observers tto
the
Unfortunately
heading.archives,
andhistorical
target
estimate
if, indeed
deepspeed
in the
rprsare
reportsarey
e deepl
inst
hetremely
they still exist.
I
5
i
-"
I
I
best day inflicted 59% kills on attacking aircraft, including a number of jets.
6.8.1 The Remegen Defense
The Allied advance had succeeded in capturing the
Remagen railroad bridge intact; this bridge was essential to rapid continuation of the advance, arid a total of
64 90-mm guns, 24 37-mm and 216 40-mam automatic
weapons and 1100 Caliber 0.50 machine guns was
emplaced to protect the bridge and assocated Bailey.
Treadway and Pontoon bridges,
in the referenced
itself,
assigned bridge,
(6.42)
. X2+ y2
(6.43)
(
pIN
trajectory intercepts
RmHa
a=-a
r tan
H
I 0 ; r2
Rm - H tan 0,
(.)vNA
(6.44)
RO
(6.45)
N/I(pTRmr)
(6.50)
Rm
hits, or
where A - target presented area to compute
vulnerable area to compute kills. o. - round to round
bullet dispersion and
+ y2
Iq
(6.52)
I
l
+ 2s02)
V-N
(4[r) hits.
VRm
(6.47)
2 - (X -x)
1
I
t' NA
In the horizontal plane containing the target, consider that projectile intersecting the plane at X,Y. when
the target is at x, flying along the x axis. The probability that the target is hit is
A
(6.51)
receive
(6.48)
rounds per
all tracered
spectacular,
aircraft.
This expression can be better understooe' y comparing it with a much simpler case in
,ch fire is
uniformly distributed over a disc of rad,'
RP,/2. The
target
of
the disc in ,/V seconds
drn crosses
hc a diameter
,'
onsaefr..Tetre
The target
2
N A/(ir
P./V R.rounds
during awhich
covers
fraction
14) of are
the firt..,
disc, hence
it will
Pk a I-e' E
(vlV)
A+2f2o
(6.46)
E-
RE
(6.49)
Recognizing the rapid decay of the exponential in
Equation (6.47) with s. and integrating successively
fx y'z
I
3
6-20
i ,
I
Table VI-3. Remagon Defense Effectiveness
Tjrgcs
Ik,,,tria d
Probable%
', Confirmed as
Destroyed
Total Claims
1945
%tjrkIi
1I?
I4
17
9(1
12
59-'
II29
36
48';
12
84'
31
39
371
13
97
31
13
44
32':
14 IlaO
go'.
13
18
6.2%
14 nigh t
14"=
4.3%
19.20
20
I2
20,21
15
SI E-262
''67 Jets
207I27-51 I
A,
DEFENDED TAROET
I.
6.21
i::AI
2007
'
T
Weaoon
Rite of
life
Probability of Kill
Given a Hit
Number of Weapons
-!
CAl 0.50
500 rpm
0.04
1100
37/40 mm
120 rpm
0.60
240
20871-512
I
270
IMP GET
24 2 SiEC
DtSTAOYiP4G
AIRCRAF
PROS Of
IOSEC
SEC
r ES~t121
b
SEC
SECC
SEC
100 MHTARGET
I SECI
10
12
14
toNs OF PROIECTi.ES
PER BARRAGE
1a
s
20
22
20071 156
Figure 6-12. Full Remagen Automatic Weapon and MG Barrage Over One Defended Point Probability of
Killing Aircraft versus Tons of Projectiles per Barrage
Figure 6-15 shows the effectiveness of the automatic
weapons component of the defense if the number of
guns were varied. Also shown are the tons or ammunition (complete rounds) per barrage activation, and tons
Advantages:
S6-22
o010
080
050
040
0
24 2 SEC
~030
..
C0
0 SE..
10 SECC
0.....C
E
ZSEC
2O~~
RDUNDS/KILL
500
Figure 6-13. Etfeciveness of Defense with 37/40 mm Weapons Alone (240 Tubes)
24.2 SEC
20SE
040
~~~~!
SEC0.
0,0000.00).000
40.0
400,000
60,0
ROUN0I/KILL
Figure 6-14, Effectiveness of Defense with Cal. 0.50 Machine Guns Alone (1100 Barrels)
6-23
-,6
0.68
07
PROS OF KILL
0I-;OMHTAI1.111
04
03
011
0
100
200
200
400
50 600 W
700
OW
O0W 1000
NO OFGuNS$
0
1&5
TONS OF GUNS
20
11
30
2000
8
s0
IONS OF AMMO
2101071
'88
Figure 6-16. 8arrage Effectiveness versus No. of 37/40 mm Gums Showing Total Wt. of Tow~d Mounts endI
Tons of Amnmo (Complete Rounds)
Disadvantage:
.-
KV4
K,a&
constants
(6.53)
6-24
(6.56)
It is desired to determine K and a to fit three points
along the trajectory, which is assumed to be a straight
line. Let (V ,D,), (V ,D12 ) be the remaining velocity and
slant range 1pairs at 2these
points.
Define xlI
(6.54)
(6.55)
Vo2-a -v2-a
(2-a) K D
(6.57)
2.0
C am wi/i
w, D
iV V. -
R Logexl - Logex 2
a
22x1.0+
2-
Lge2 2
(6.58)
(6.59)
In most cases 1.0 < a < 2.0 and the two values
obtained above will bracket the correct value. The
above method can be used to develop an iterative
program for computer solution. Once 'a' is obtained,
'K' is easily computed. As an example consider a
6-25
and 1.66.
3
I
I
I
I
I
I
I
I
I
I
SECTION 7
THREAT CHARACTERISTICS
7.1 GENERAL CHARACTERISTICS OF THREAT
NIf
Categories
Tactical aircraft in operational
-erieahbombers
flshtetlpound attack
Southern Europe
NATO
Waruw
Pact
(of
which
USSR)
NATO
Waruw
Pact
ISO
t,130
230
1.400
200
1,100
450
30
IS0
(of
which
USSR)
30
so
intrceptors
..
300
2.100
1,100
275
900
450
recennalssance
..
400
400
300
125
100
40
20871-513A
7-1
Table VII-2. Comparative Numbers and Types of Somber and Strike Aircraft
United States
Number
Type
Category
Long-range
bombers
B-52 C-F
B-52 C/H
Medium-rtnie
Fe-I ItI
Soviet Union
ISO
210
70
Type
Number
Mya-4 Bison
Tu-20 Bear
40
t00
Tu-16 Badger
500
bombers
'a
Strike aircraft
(land-based)
F-1OSD
F4
F- I I A/E
A-7D
(1.200)
TU-22 Blinder
Yak-28 Brewer
Su-7 Fitter
1
MiG-21
FlshbedJ
(iS00)
I
A4
Strike aircraft
(carrier-based)
(900)
A-6A
A.7A
RA-5C
I
I
NATO Countries
Number
Category
Type
Medium-range bombers
Vulcan (Britain)
Canberra B(0)8 (Britain)
56
24
36
80
Type
Number
11-28 Beagle
Su-7 Fitter
(IS0)
(200)
(60)
(400)
F--104
20871-$14A
pal target would probably be ships. Missiles include Kormoran and Oomat.
e. Missiles homing on laser designated target. The
laser may be laid on target by a ground operator,
by an operator in another aircraft (for example a
light artillery spotting aircraft), or by the launch
aircraft. Whatever the laser location, the target
must remain illuminated during the missile
fliSht. Systems include the Hellfire missile, and
laser
homing heads for iron bombs.
The cost of the more sophisticated options is consid.erable, although the added equipment apparently, still
costs less than the flyaway cost of the delivery aircraft
7-2
LL
'Pave Knife' AVQ-10. This is a 'completely scfcontained pod-mounted weapon de!,very system for
the F-4D'. It combines LLLTV with a laser tracker and
designator for a laser guided bomb, and according to
the reference the cost is S350,000 per system.$
Max.
weapons
load (lb.)
ASM
carried
In
service
11,500
12.500
095
0.95
60,000
75,000
Hound Dog
2x Hound Dog
1955
Max. range
(Statute
miles)
Category
Type
Long-range
bomoet
B-52 C-F
B-52 G/H
Mediumrange
bomber
FB-I 11
3,800
2.2
37,500
A-4
A-6A
A-7A/D
RA-SC
F-104C/G
F-105D
I--I I1AIE
F.4
2.300
3.200
3,400
3.000
1.300
2,100
3,800
1,600
0.9
0.9
0.9
2.0
2.2
2.25
2-"/2.5
2.4
10,000
18,000
15,000
13.500
4,000
16,500
25,000
16,000
Stuike
airc aft
1970
2'x
2x
4x
2x
2x
4x
4x
4x
Butipup
Bullpup
Bulipup
Bullpup
Bullput)
Bullpup
Bullpup
Bullpup
1956
1963
1966
1964
1958
1960
1967
1962
ASM
carried
In
service
AIRCRAFT (USSR)
Category
Long-range
bomber
Type
Tu-20
Bear
Mya-4
Bison
Max. range
(Statute
miles)
Max. speed
(Mach.
no.)
Max.
weapons
load (lb.)
7,800
0.78
40,000
6,050
0.87
20.000
1 x Kangaroo
-
1956
1956
Mediumrange
bomber
Tu-16
Badger
4,000
0.8
20.000
2 x Kett
1955
Strike
aircraft
Tu-22
Blinder
3,000
1.5
12,000
1 x Kitchen
1962
2.500
0-81
6,000
1950
1,750
1,250
1.1
1.7
4,400
4,500
1962
1959
900
2.2
2.000
1970
ASM
carried
In
serv ice
(incl
shortrange
bomber)
11-28
Beagle
Yak-28
Brewer
Su-7 Fitter
MIG.21
Fiohbed
Type
Max. range
(Statute
miles)
Max. speed
(Mach.
no.)
Max.
weapons
load (lb.)
BRITAIN
Vulcan B2
Canberva B(I)8
Buccaneer S2
4,000
3,800
2.000
0.95
0.83
0.95
21,000
8,000
8.000
FRANCI
Mirage IVA
2.000
2.2
8,000
S7-3
7-3I
1 x Blue Steel
1960
1955
1965
1964
20871-5ISA
Cost
$5.800
warning and security against enemy fighter attacks to a degree commensurate with the local
Laser Guidance
situation.
s15.000
TV Guidance Package
b. Suppression of Hostile Air Defense. The vulnerability of close support aircraft to ground fire
makes enemy.capabilities in this field a major
$3.100
___erGu__ne____10
consideration.
When
antiaircraft
weapons should
be possible,
rendered known
ineffective
before
2087 1-516
The reported effectiveness of laser target designations suggests a need to provide the defense with a
detector which can sense laser illumination and determine the line of sight to the illuminator. If it is on the
ground at short range it may be engaged directly by
the air defense weapon; at longer range it can be
attacked by artillery if response time of artillery can be
made acceptable.
These counteractions are not necessarily feasible.
However the problem for the defense is now suffi-
A variety of delivery maneuvers have been developed for the various types of air-to-surface weapons.
The maneuvers were developed after careful consideration of such factors as weapon ballistic and fragmentation characteristics, various types of terrain, weather,
enemy defenses, and accuracy requirements. Each delivery maneuver is designed for a particular set of
circumstances. Generally speaking, the delivery maneuvers are divided into four categories -- level, glide,
dive, and loft. Weapon release angles up to and including 45 degrees are referred to as glide deliveries. Dive
OP'
"r
I . .
. .
'
. .. .
". . . F
-'
I/
I
Squently
fire bomb and the retarded snakeye general iarpose bombs. The release height would rat.. e
from 75 feet for the fire bombs to 200 feet for
the 500-pound snakeye while the release airspeed
may range from 150 to 500 knots depending on
the type aircraft and ordnance limitations. The
level deliveries are generally less accurate in
range than glide deliveries. Also the target must
be well marked or have a vertical profile and the
terrain must be relatively flat. The level delivery
may also be used for smoke laying operations
and for battlefield illumination.
77220rdnsnoo Selactlon
The type, size, and delivery method of the ordnance
is selected based upon the nature of the target and the
desired degree of damage. The total number of aircraft
sorties required is determined by dividing the total
amount of ordnance required by the total each aircraft
can carry. In selecting aircraft ordnance for the attack
of specific targets, the following steps are litilized:
a. The vulnerability of a target to attack is consid-
b. Glide delivery. The standard glide delivery maneuvers include 10. 20, 30. and 45 degree release
angle. However, the 10. 20, and 30 degree release
maneuvers are most frequently used for close air
support as they generally provide the best expected accuracy. Since each of these four release
maneuvers has slightly different purposes and
offer different capabilities, they will be discussed
separately.
(I) Ten.Degree Delivery. This is optimum delivery for fire bombs for accuracy and is freused for the delivery of the retarded
snakeye bombs and strafing. The release/firing height is approximately 350 to 400 feet
for all the above weapons
sidered the most accurate and effective delivery for rockets and free-fall bombs, particularly from high speed jet aircraft. The release/firing heights vary from 2,500 to 3,000
feet depending on the type rocket or bomb
being released. The release airspeeds may be
as hibh as 500 knots. Accuracy results from
the lower release point (lower pullout) not
available in the 45-degree delivery,
I
3
(4)
Forty-fiv Degree Delivery. The 45-degree delivery may be used for the delivery of rockets
and unretarded bombs. However, except for
situations where maximum bomb penetration
is required, this maneuver is seldom used on
7-5
3
3,000 feet above the surface. Specific expected accuracies for the various weapons and delivery combina-
in
are
fires
ar c
t
cose ai
fires atc
Flak-suppression
aircraft.
support
air
close
directed against hostile antiaircraft weapons that can
bear on close air support aircraft during the execution
a. Strafing.
c. Rockets.
d. Free fall (unretarded) bombs.
7.2.4 Nuclear Weapons Delivery Modes
Nuclear weapons delivery can be accomplished by
jet aircraft.
fighter/attack lvel
attack/bombersareor higheraltitude
heavy
The varioustactioms
aireeas,
The various tactics are high altitude level releasen
laydown. loft (low. medium, and high), and high
altitude dive bombing.
c. Laydown Tactic. This tactic is a minimum altitude, horizontal bombing maneuver. The target is
approached at low altitude. The weapon is released, retarded by parachute to provide greater
accuracy and increased distance between the aircraft and target. This tactic provides a high
Delivery Tactics
degree of accuracy.
I
i
d. Loft Tactic. The loft tactic of delivery is accomplished by the delivery plane approaching the
target at a very low altitude, making a definite
I
4
CI
R ECOVERY AREA VARIES WITH TYPE
TECHNIQUES EMPLOYED
TYPE OF ATTACK
USC
TARGET
735
III
'o'3A4
0
COMMUNICATION W!TH STRIKE AIRCRAFT
AN~DAGENCIES PROVI DING F LAK-SUPPRESSION
IFIRES
AT ALL TIMES
SrPC
FAC
4
4
Z
0I
ej
'2I
A~frI%
121045
SC1345
0
$ccIIIAND
~MEANSTOCONTROL
FLAIC-UPPRIESSION
~FIRIES
ENSURE MAXIMUM EFFECT
AIRCRAFT SAFETY.
FLAK4UPPRESSION FIRESB
CEASE WH0ENTHEY ENDANGER
STRIKE AIRCRAFT AT ANY TIME
DURING APPROACH ORt RECOVERY
NOTE
7.7
20571 169
1-5
1-5
Low level bombing is cited as having been dangercus to the bomber in Vietnam. and the British view on
Iive and toss bombing is given as that 'both methods
are too dangerous for use in Europe as any aircraft
that gains any height at all will almost certainly be
shot down. Their policy for a Eurpoean war is therefore to use the lay down method wherever possible."
Computer obtains
solution
I+
2-6
Settle tracking
2-10
4-16
Start pullup +
release
1-2
5-18
targets.
Time Joe"
0
Elapsed Time
0
The accuracy with which iron bombs can be delivters. An effective defense forces the attacker either to
release his munitions at a greater range or to use some
other, and possibly more costly air to surface weapon
such as a missile. In this section some information on
the accuracy of bombing with iron bombs is summarized for later use in defense evaluation. All of the
data is from unclassified open sources, and has not
been compared against classified estimates.
Additional open source data will
be found in
AFAADS-I, Table IV-7.
With regard to release altitudes, according to Aviation Week, 0 USAF 'tactical aircraft were limited to a
minimum of 3,000 ft. altitude in making weapons
delivery runs in South Vietnam. The move was designed to reduce the effectiveness of the smaller caliber
anti-aircraft guns being used by the North Vietnamese.' (in May 1972)
Tables VII-S and VIl-6 summarize CEP estimates
from various sources, and Table VII-7 summarizes
some estimates of effectiveness per pass and per bomb
against vehicles and bridges.
It is clear from the tables that an air defense of only
moderate effectiveness could make tactical air attack
with Korean vintage aircraft munitions and fire control cost-ineffective, The 'smart bombs' now operational make it a whole new ball game.
The tabulated data on CEP and effectiveness of iron
bombs and straftn, should therefore be considered as
obsolescing capabilities of unsophisticated, relatively
low cost tactical air systems. An object of local air
defense is to deny the enemy the use of these low cost
solutions and force him to use more costly options.
iir
I
3
1
<
,I
I
'boo
700-
000
400r
CEP (FT)
300
200
10
20
OURATION
O
OF BOMUINO MUN (SMEC)
,r
20671-100
Figure 7-2. Hypothetical Variation of Glide Bombing CEP With Duration of Bombing Run
"7.3.2 SAAB
*
I*
I
Ii
during the bomb run, but the next level of sophistication, allowing non-visual delivery and evasion during
the attack path would also represent a substantial cost
increment.
Svenska Aerplan AB (SAAB, was building a semiautomatic toss-bomb computer in 1939. A fully automatic BT9D (mechanical) was developed, followed by
an electronic model (BT9E).
7.9
Table Vll-5. Delivery Accuracy of Aircraft Munitions fromn Unclen ied Sources
Combat
Conditions
(feet)
Reference
75O
365
Aviation Week
("5)
moderate defense
540
heavy defense
S75
F-105 vs.
"Current" in Vietnam
250
Training
F-105 at Nellis
125
Competition
A7E I%1
pass
150
I00
2nd pass
ATA,
1itt pass
300
2nd pass
200
500
Shallow Dive
F-105
1
Aviation Week
(7.6)
Unspecified Defense
1100
2000
I
Interavia
(7.7)
"Divubombing
8.26 mirctaft
20871- 517A
~CEPI
Conditions
(fee
I:
Table VII-6. Delivery Accuracy of Aircraft Munitions in Korea from Unclassified Sources
!
i
j
I)
Reference
"U.S. A6 Force in Korea"
Dec 1952
340
Jan 1953
314
1951.52
Level bombing, 7000 ft ait
375
175
caise.
In 1962
advertisement
a agraph
of
dispersion
versusthealtitude
and dive carried
angle for
true air
s of" 250 m/s, and this is reproduced as Figure
Tp a
direct
planerelease
towards
the target and
a pickle
switch.his
Bomb
is automatic.
Withpress
an auxiliary
device, the BT9 is also suitable for firing air-toground rockets.'
'The ST9 functions in principle as an analoS computer. In a dive, it automatically registers altitude,
dive
and other
it de.
rives angie.
the correct
releasevariables,
point. Thefrom
pilotwhich
only has
to
3
5
I.
Conditions
Target
Undesitgnated. 1952-53
Measure
Effectiveness
Reference
Bridgs
Hits/pass
Day 0.09
Night 0.04
USAF in Koreu
Bombing
Vehicles
(laims/pass
0.22-0.76
Strafing
Vehicles
Claims/pass
O.12-0.43
OAO Memo.1952
(7-9)
Vehicles
Clatms/bomb
B-26 1951-1952
Fighter-bombers
!95 1-53
0.197
0.189
0.115
0.110
Brldps
Hits/bomb
SO1lb GP average
0.037
0.038
Bridges
Hits/bomb
500 lb OP average
0.08
0.10
Vehicies
Claims/bomb
S00 Ib GP average
1951.1952
1932.
Fighter.bomber (2 bomb)
B.26 (full load of 500 Ib)
OAO Memos.195 2
(.8, 7.9)
0.103
0.089
Vehicles
Claims/sortle
0.262
0.89
OAO Memo.1953
(7.9)
20871-519
7-l
,1
II
Ii
\\
20
ALTTUOE OF
BOMB RELEASE
II
DISPERSION (MILS)
10
40M
3000M
2000M
I000M1
STRUE AIVISPEEO26OM
10
20
30
40
Go*
OIVE ANGLE
so
20871.11
For this sight dive angles between 30.450 are recomSended as optimum.
I1
I!
range independent of the angle of dive. and the angular dispersion in mils almost independent of range as
well.
These errors are probably estimated exclusive of
than aiming
(hose noted
pilot
in Table
errors
and hence
VII-5are
andconsiderably
V/ii-6.
imrller
7AI
7.4 AIRCRAFT VULNERANILITY
The probability that an aircraft will be lost, given a
hit by a round of antiaircraft fire (or by fragments)
depends so specifically on design details that it is
difficult to make simple generalizations. On the other
hand, one feels intuitively that it should be possible to
develop relatively simple parametric expremions relat7-12
I
6000
II
1V
I
I
4000
II t -..
//
METERS
2000
A0
20
METERS
2071-1.62
-0
Figure 7-4. Contours of Constant Dispersion in Meters for SAAB BT9 Bombsight
ing caliber, for example, to kill probability for a
generic type aircraft and homologous projectiles.
P0
The resulting expressions are cqnsidered to be reasonable, but have not been compared with classified
'testimates.
7-13
L/(L+D)
Then
Probability of
rrnbwiits
return withhits
p = po/H
wher
Probability of loss
(7.2)
(7.3)
(7.4)
eP-
e.E
Average number of
hits on survivors
where
H-
L = I - e"Ep
H
qeE
iE r
=
HS
Some rare reports provide information on the numher of aircraft returning with 1,2,...nm hits. This allows
one to do a better job of including correlations in
hitting and killing in the analysis.
75
(7.5)
Average number of
hits on survivors
wi:h hits
'"
Eq
Whicn a mix of weapons has attacked the ai-craft.the problem becomes more difficult. Some reports identify weapon type causing damage. Even better, some
reports identify the components that cause aircraft
vulnerability.
Note that general reports on aircraft damaged will
vary as to whether the damage was repaired by a patch
(likely not to be reported) or by replacement of a
major component (probably reported because of the
requirement for a new component).
(7.6)
Probability of loss
or damage
L+D
(7.7)
Define;
p z L/(L+D)*
0
I-P0
(7.8)
I
i
f
and
I e"E
Ln(- fpo)
Ll-
Define:
(7.9)
0M
1o1! -(112)(f%)]
p E -
q -
(7.10)
7-14
I
The above expression does use the loss rate only in
the correction term. If we believe that L is well known
we can change the approximation form by using
i
L-
p/q
L a I _e-pE
L)(7.20)
(7_12
P - H- p(Q.- 1)
(7.21)
_S
SH
H
H
I -e'Eq
(q/P)Ln(l
p
. L)
(I - L)qP
(7.1)
1)I+(HII
[(qL)/2p] 2 Log tL - Log
- - 1)
- -(II2XH.- 1
(
/
)2
L
F2.5--f-
B-17 and B-24 data on hits by antiaircraft shell fragments, refetence is made to a paper by Weiss.'
7.4.2 Inferences from Aircraft Damage and Lose
Reports
A declassified report on U.S. aircraft damage and
(7.12)
L
21(7.13)
ro
Hence,
(7.14)
~(7quired,
e"EpI poeqE
e-qE
I .(Eq)/H
(7.15)
(7.16)
Then
t(po/%Xq/p)l/H)
t
i
sinformation
Expanding the left side, for small IE
I1+(I/2)pE a (po/qoXq/pXlII
(there
(ePE-I]/(pE)
(7.19)
0I - L) a e-pE
(I Lo ePE
since
pE
and so
(7.17)
(7.18)
-IT
II
- 1I
II III I
Projectile
German 88 mm
of damage (except patches) given a hit could be approximated by a function of the form
p =
UK 3.7" (94mm)
25 - 30 ft.
damaged
to both
and antiaircraft
d m g d and
a d lost
l s o b t airi too air
ar a d a t a r r f
shows a trend with caliber similar to that obtained
from the Korean data, but the loss/hit ratio is substantially higher." It is believed that at least a part of the
difference is accounted for by multiple hits per aircraft
hit in the WW 11 data.
Other Vulnerability Estimates
Drandli provides a curve of kill probability, given a
15 ft.
(7.22)
-kC8
Lethal Radius
C,
Ca'?
World War 11
Looking back to World War II, we find the following estimates derived principally from German proving ground data and camera records of combat with
regard to B-17 vulnerability, and summarized by Weyl
in 1950%' The estimates are shown in Table VII-8.
The second column in the table was based on experiments by Burgsmueller at the Rechlin establishment on
captured
4-engined bombers. It was
estimated that
from 15-18 oz (420-500grammes)
or Hexog~en-Alumi-
Only secondary sources were available for the present study; there may be more information in the
original after action reports.
For the European theater the reporting categories
were
Category 1: aircraft destroyed
each1/(Icateper kill
rounds types
theaircraft
were all
availableAcross
Also (RPB).
gory
,;he inratio
+
11) was between 0.69 and 0.72. Most of the da:a was
on the Mel09, FW 190, and JU 88. An exception was
the Me 262 jet for which the ratio was 0.08 suggesting
that its speed was such that it did not fall close enough
to the battery for a kill to be confirmed.
Table VII-9 summarizes results by weapon, over all
target types." The principal point of interest in this
table is that fewer targets with observable hits escaped
7-16
I
3
I
j.
I
I100
<
I
I
0
- LOGE 11-)0
A/C LOST OR
REPLACE MAJOR
COMPONENT
A/C lr'
Pic
LOST
I
010
0.1
/I
A/C LOST
I7
A
AIR TO AIR, WWII, PACIFIC
0.0
10
20
30
40
CALIBER (MM)
7-17
III
t0
o0
~20671.16,3
Caliber of Projectile
Minimum Number of
Hits Required per
Kill (Test Results)
850
".a.
n.a.
50-60
20 mm
20
20
30 mm
55 mm
n.a.
the 40-mm than the 37-am . The rates of fire were not
greatly different, across these weapons. The caliber
0.50 ratio is explainable if one believes that when ontarget, it usually secured multiple hits.
ness in specific actions in the Pacific Theater is summarized in Table VII-13. 5 In the defense of Corregidor
ammunition was severely limited. The troops trained
in peacetime were highly skilled. Fire control em.
ployed the Sperry mechanical M-4 dircetor and the
stereoscopic height finder. The attacking aircraft
bombed from 15,000 ft. Initially the rounds per aircraft destroyed was about 500 and by April it had been
reduced to slightly below 100 rounds per kill.
At Morotai the 90-mm guns were directed by the
SCR-584 radar and the M-9 electrical director.
i
"
Gu,
90 mm
1+11
Ratio I/(1+11)
S9.S
119.5
0.75
235
RPB I
IRPB 1+11
176
20
25
0.80
178
142
40 mm
244
340
0.72
333
239
37 mm
60
105
0.57
590
337
0.50
73
117
0.62
32.360
20.190
90 mm with VT
20871-521A
7-I1g
L"''"""
SI
-''..."'-"
III
,""
..
'
"
S;Number
Gun
I
90 mm
A1
1i
35
Total Number of
Round%Expended
Ratio
RPB
A/kA+B)
0.31
90 mm. +0.SO
0.17
40 mm
6'A
0.31
4.610
118
12,580
325
10
32
0.31
37 mm
0.00
20
0.50
0.00
165,260
0.50
40 mmm+
of Aircraft
Destroyed by Category
12.700
20871-522
'le in Category
I
I+11
90 mm
2.91
3.89
90 mm + VT
0.65
40 mm
37 mm
0.50
Total
Damaged
90 mm
546
288
0.81
40 mm
409
260
7.95
1.96
2.38
11.07
3.42
3.81
37 mm
400
223
46.S44
23,272
15.85
23.01
0.50
20871.-24
20871-523
iI
.7-19
Number of
Aircraft
Destroyed
Aircraft
Action
Weapon
Engaged
Corregidor
Destroyed
Probably
Destroyed
Damaged
Rounds per
Aircraft
De'troyed
52
120
3"119411
Morotal
(1944)
Leyte
(1944)
90 mm
AU wpms
18
;,278
0.50
0.17
0 33
251
90 mm
61
0.49
0.32
0.19
125
40 mm
133
0.48
0.22
0.30
249
0.50"
57
3.845
2087 I-525A
Kamikaze Attacks
No.
Rounds expended
Percent
Total
Per kill
5" .ommon
22.0
8.9
26,302
1,196
17.0
6.9
7,083
417
3" common
5.5*
2.2
4,667
849
0.5
0.2
544
1,088
115.5
46.8
287,556
2,490
40 mm
1.1"
0.5
0.2
2.695
5,390
20 mm
78.5
31.8
645,315
8.2.,1
.50t CA1
5.5
2.2
119,232
21.678
.30 Cal
2.0
0.8
14,381
7.191
"Total%
247.0
100.0
1,107.775
Ohlcans lhi- 2 or mote dilfferent caliber guns appeared to be responsible for the
7-20
4,480
per kill'
iamekill.
20671-S26
I
parameters so that they can be adjusted to correspond
to more exact (and classified) data. The simulation in
which they are to be used, in addition, has the capability of utilizing more detailed vulnerability data when
this is desired.
For purposes of AFAADS analysis we limit our
interest to two classes of 'kills', those immediately
observable (the Pacific theater Category A of mid air
destruction) and those resulting in target crashes/
within observable range of the battery (Pacific theater
Category B). These will be designed 1-kills' and 'K-
-k2E 1/2
2
(7.24)
kills'.
where E
Scaling Relations
it is assumed that the probability of a K-kill by a
nor-explosive projectile can be approximated by the
function
Pi
IekWhe
where
the explosive change,
weight offiller[
W,,. - the
or projectile
k - 6 x 10' (WV)'
e'kCaVb)p.
(7.23)
(7.26)
w(
where
k C -
(7.25)
(7.28)
_(whiw,)lf2
II
(7.29)
where
II
Z
FRAOMINTS
0 >
0.5
211
1.0
0.5
0
FRACTION OF SHELL
WEIGHT IN HE
3.
20871.164A
Figure 7-6. Fraction of shell weight in HE Energy Division Belween Bloat and Fragments
7-22
IoI
II
APPR
II0XIIIAL
FRAGMENT VELOCITY IF/S)
"GERMAN 9$-MM
0.0
1.0
Figure 7-7. Fragment Velocity versus Fraction of Projectile Weight in High Explosive
E L C3ItY (/t7.30)
/2
frgmntwegsen
calibecie,
/2 f/2
1whe/
7.4.4.1 Summary:
The resulting expressions for the probability of a kill
given a hit, arc summarized in the Table VII-1IS. The K
kill contains I kilts as well.
There are a suficiei t number of coefficients in the
above expressions
fohein to permit some adjustmet to good
vulnerability datr. To show what the exrressions look
like the following coefficient values have been chosen
k, -
k3 k2 -
(7.31)sec
f -
1.0
, 1.0
1.0
1(10)"'2 (4000)1'; for C in mm, v in
10%
"p - -k ewhe/2
-
0---
S.,Target
Component Hit
Wing
Fuselage
he
-k I W.
-klWhe-k3Whe11
K-kill.HE
Pk
K-kill. non.l-II
impacting proj.
Pk
PA
" II
Po (
-c
Il. k4Whe
"k wh4 Whe
NIL
20871-527A
The expressions for the HE projectiles in Table VII15 have been programmed into the Litton simulation.
The k, are input parameters which can be chosen to fit
whatever valid vulnerability data is available,
reprentation,
For small caliber rounds vulnerable components
tend to lie within the target fuselage and an ellipsoid
seems to be a reasonable representation.
round to round disp ion in angle and muzzle velocity. The target used in the simulation for AFAADS-I
was a diffuse circle. The present model consists of two
diffuse ellipsoids representing the target fuselage and
wing.
Since the target vulnerable area can now vary significantly with aspect, it becomes desirab!e to account
properly for the relative direction of approach of the
projectile (resulting from the vector sum of target and
projectile velocity) and this effect has been included,
n ding
gravity drop.
Principal emphasis in the AFAADS study effort is
on the characteristics of the fire control system. A good
overall evaluation of system effectiveness is desired and
so an objective is to describe the target vulnerability
characteristics in sufficient detail to provide an acceptable evaluation,
Given the two basic ellipsoids, they must be projected into a plane perpendicular to the direction of
approach of the projectiles. 'Diffuse' targets are then
constructed from the projected ellipses.
The diffuse target representation has several advantages as indicated in AFAADS-i. It is asymptotically
correct for very small and very large kill probabilities.
The kill computation is much simpler. It also allows a
very simple computation of the probability of a hit on
either the wing or the fuselage in the presence of
'shielding' which is indicated to be suAiciently accurate
for present purposes.
On the other hand, the description of target vulnerability to a degree which includes all of the meticulous
detail with which target vulnerability analyses of specific targets is known would result in the simulation
being dominated by the target model rather than the
fire control model, with excessive running time. Our
object therefore is to strike a reasonable compromise
I
J
H
"
I
IWl
'C
UWW-11
-KL
'KILLIED IN AIRPACIFIC
AFROM BRANOLI'S
CURVEA
0.10
GIVE
PRJ
AAIRT
I-KLL
0.01
7
STRIKING
~0.001
II
10
20
30
40 60
'7-2I
100
0st1e
i
1.0
I/
0.9
o.
a
IL
'r
0.4
OESTRUCTION
K.
0.2IL
Vi
0
,,LL.
10% FILLER
20W PIS
TARGET;
20
40
60
go
FIOHTER-SOMSER
100
120
CALIBER IMM)
20871-167A
Figure 7-9. Conditional Kill Probabilities versus Caliber for 10% Filler
'
10
iI
Iz
0.6 ---
4
0
0.:
70% FILLE.R-
0.2
00
20
IS
TAROET: PIGHTER
POMBER
40
60
s0
100
120
CALIBER (MM)
20871.1
Figure 7-10. Conditional Kill Probabilitiel versus Caliber for 20% Filler
7.26
1A
I
i
I
i
TARGET:
FIGHTER-SOMIER
.2
1012
20871.-16A
2
Figure 7-1 1. Conditional Kill Probabilities versus Projectile Weight
X"- U R'
(7.37)
Ellipeold
U'R'H R U a U'M U a 1.0
(7.38)
Where
X'S(.
[x y
(7.34)
(7.35)
1/22
1i/b2
7.27
F|
Examining the elements of the two terms of Equation (7.44) note that they are equal to each o:her and
expressions is simplified by
the ric
of
Manipulation
es
sn isubstitute
partinip
ong
mt
the
o
"minI m2 l m3
M
m3
m32
i 21
M2 2
m31
m32 .m33
..
M2' M3 J
IUa
uaMiUa
Iw1L] l
[M2'
*wM2.U2
M2 M3 "I M2 ') Ua
(7.45)
10
reference.
Begin by determining the orientation of the ellipse
FMMlin
= 1.0 (7.41)
M2
Define:
M3
1jaM
Ua' [MI
MI 2
(7.39)
2 w+wM3 w -
1.0
v - r sin 0 ; u
r cost,
(7.46)
"
(7.42)
(7.43)
M2 M3]
+ [2ferentiate
0
(7.44)
uE
U t1.0o (U7PU
1.0
(7.47)
I7
U
P7-28
tan 2 4* - 2B/(A-C)
Ae
(7.48)
ff
eU'PU
dU
(7.58)
(rl~~~)
(A+B)
22
1(28)2 + (A-C)21 112
If I PI1/2
2
irI1
(7A9)
rirri
(7.50)
a2 b 2 ] I/2
a2 c 2 + r 33 22a2]/
Ae = ir C 13 2 b 2 c2 + r2 3 2222
B21 1/2222
(AC)
Ae=(7
Referring to Equation (7.45)
M
M .M2
3'
M2'3
(7.52)
and so
A"
(7.53)
IP
I
*/
NMI/ IM3 1
(7.54)
and
2 2 2
(a b c )"
(7.55)
-bank
so that:
At a
nabc IM3 1
1/2
A
e
A,
Se,IMI
IHI
(7.57)
I nplane
The same result can be obtained from the integral
defining the equivalent area of a dilfuse target. We
have
7-29
(7.56)
m
,.U
(7.60)
whence,
At
(7.59)
plane.
relative direction of projectile approach in
azimuth.
relative direction of approach of the projectile in elevation, measured from the
horizontal.
heading of the aircraft in the horizontal
relative to the Y axis.
angle between the projections of the target
velocity vector V and the line of sight to
(7.62)
the target in the horizontal plane.
tv
~NOTE:
A IS
(--)
FoGURE
F
IN THIS
fr
ANGLE
//AZIMUTH
-71-170A--
-208-
///
"
730
AI
aV
S~20971-173
Figure 7-15. Gun Coordinate System
,a
(7.63) CoIonIT.
a,-
a r a Ar-H+ff
10'iJ
0
10o,'|o, co
01l
,, H
o ol"I -S.HHo,
HI
Lf00coJLP.-,, 8 Co0aLo
-I
This set ofoperations isabbreviated
I
"
Invening the
-
lteIT
operations
IIt
131 Ixi
(1.681
('7.65)
IXYZ
Sy1TteIimT
1J
caure the w
7.661
11F-!
lu~l. xz
cosr
A. sin A,0boo]
]
0.J0
oo
)
it.,7,69i)
cot/ e 1o
0o/
LO ijigIner coserj [J
(7.i9)
(7.70)
LetA=O,H=ff/2,ar=r/2,89=
or
gU)T
(n/2)J
* IXTlAl,
SThen
IujT
T.Ix
1.,31T
IITIHITIAIIc, + (ff/2)I-IjTIjT47.7)
R n
This derines
Ihfollows
at once that.
""xI
T-
-,
01
sinY
0
0
cosy
sinY J
Ae " ib
- IRITIUI
IRIT
(37.82)
(7,74)
Let A=
(7.7S)
Then
-sin (e
e)
-COS (e
CoS(Cr
0)
(777)
S- H * a -'
The H and A, matrices can be combined ira, it known,
and
IAITIHI"
Ae =
(7.78)
a[c2os2(eCr .0)+b
sin 2 (r.0)] (4
similat ly
- (e)
,ja
n,i 0 si n a
, ,
*
+
, d son ,
mI csa
in|,i
L i,,o,
er . -
a $ r C sII
+, 10iii Oi1
cota 1 siner
-cosarcose
cose
rl
C7.85)
o"C,-""""
$1111)
11
.
'sn
and
' ,ine
n0U,!
W
-si a. Sin) er
-3) a1
0-
%111.j
fiii
IZ Sifn Cf - 15miSin 0 eio
4:02o,,,,
.cw5 ar
Then
.r - ,,
7,84)
O0
Let - 0,
(7.79)
(7.73)
.1,/2)ITIArITIHIIJl9)]p
1,
(7.72)
and
lxl
For reference
(7.81)
')
whereY = (03[ITIRIT
-cosY
IRI
'1
-,. ...
,i,,i
s,, 0 c
liii, 0ooisC~r,,J
*"liJ liii 0
~ ~
"+a 2c2
,,,,
If a
7.80)
C,,',S%. U
nrJ ( Ofli*tst
7-32
P- c
aos2 er]l/ 2
ar
(7.86)
direction)
*1C/
Ae
2
ln]/2
ffa [bsin2 a+a2cos
(7.87)
where
~Cos 42
#.
=cos a, Cos er
(7.88)
2
Sl/b
) + cos2 I/a2)] -v2/C2
V2
(7.89)
Vp cos e cos A. Vy
V2 + V 2 + V2
0rX
N
rZ
2
Ve-
-0
Vry
(7.94)
VH
Let c
VcosesinA-VX
p
-u 2 J(sin
VrX
* 2
V2 rY
oiotlpae
Te
Then
sin Ar - VrX/VrH
Ae
irabIt 3 3 1
Cos Ar
nab Isini
PsinarCoserCos
CO
0 CS
.
cos Cos
a
er
(7.90/V
sin er
ose
a VrHIVr
V
Sand if
,Ae
ii
=0
(7.91)
nab sin er
(7.92)
(7.95)
. RIt
7.5.3 Rolative Direction .4 Approach of
Projectile to Target
U
3
sine, a VrZ/Vr
(7.90)
=0
W
VrY/VrH
'4
Ar
er =
= Ao ar
0e
ao
(7.96)
fVx
a V cos sinA
pY
Vp
Vc
(.
(.3)
V sine - gtp
":Y
Idifference
The components of target velocity in X Y Z are
V,.V,.Vz and using Equation (7.93) the components of
relative projectile velocity are
7-33
Y]
(.A
AI
a-
A.H+tr
200711"4I
)y
nr(l +n2)'6,os-/(I+n2In
(7.97)
thatcan
othe;,
presently
To the
sources
of degree
dispersion
be resolved
intounidentified,
these three
elements they can be handled in this simulation pso.
gram as inputvparameters.
i
esy
And thereion is worked ou
n
et
e
Beinwithuncg
the
target path
and (e
the tangent
lp
ent devatn
proecntile a isnearthe target).
arge
tsrajtor degaroe at othe,
prseintly uidaelnthifed
traoedtory ar we are on
o
sion set y, coordinates
rMVB
lJ 9t
1[X
gt> -<lo l [x,
o0
z>. i
U.-IL
2oo
q
0
-
R.1
.INS
...
fo
..
fJe
..
.. ::
. .. .., . . .,,
hsiLo,
&l-.
..
I ,,T
. ........ .....
02
....
,d..,,t o
"
. _ ....
Co
e...
U+
. ..e,
,
&1.*
ln
l~
04,
where
(7.98)
A . (A-A,)
'21 '2."
C] Tx,
UTU [. gIT
[R
CI ICI2'x 2
T
U
if <UIu]T>
. .... A
. 2,, 3132"r
C ;I2'
(7.100)
U 2 UVO'IYV
uw1711wThe effect of an error in muzzle velocity point
is to cause
ata
projectile to arrive at the predicted
~the
2
PuvGt f
V2
Pu~wrvav~
uwwv
th
different
time from that computed in the predictor. We
consider
this as a perturbation about the target position.
Pvwj'vedw
I
.2
w
jtio.
(7.101)
If St, -
F-,
then
OR1
2
0
. Puv 'u0v
'v2
(7.106)
(7.102)
The subscript y indicates that the error is olserved
along the y body axis of the target.
CIT _=[eT(A]
T [ [At]
A (r(For
[C]
[er +(ff/2)]
(7.103)
7-3
,,
,,
, ,,
,,I,,
, . ...
. l J
va
Dp/tpV -
(7.107)
'm2"
va
L'u
.. U
where
2P
1
I
.JUVR"
"rE 2
00
TV
(7.15
Then
Logetp Z LogeDp. LOge(Vo.aDp)
and only the four elements in the upper left corner are
retained.
(7.108)
(7.109)
Hence,
E
Divide both side,. by
Vtv)
-*
,
EyNDp f
(voB/V 0) (V/Vp)i0
Vp
[MuBMvBMwB]
(7.112)
(VIVp) 2 D)2
(7.117)
[OBvO)[R]
The Me. term which is along the direction of relative approach of the target is dropped and the remaining two terms are added to the u v miss distances
obtained from the prediction module of the simulation. The resulting, sums g, into the computation of
single shot probability.
The simulation is also programmed to acc:ept constant angular bias errors in azimuth and elevation.
(0vo/Vo)
(7.116)
This miss distance is rotated into the u,v,,,v coordinate system by the operation
(7.111)
WE2-
(7.110)
o.
-tVh (V/va 2 )
ation
U
Before presenting the single shot kill algorithms, we
digress briefly to discuss the diffuse target representation which was presented without discussion in
krAADS-1. We consider the basis for the diuse
target model, the approximation invoived in the simplest method of accounting for wing/target area overxlap, and the 'shoebox' representation versus the
(7.113)
3
d
Jf
L'
I
't t
dEquation
(7.121) as
Pk =
-C t (x, y) f (x,y) dx dy
Pa
f(
)Jf t, (x, y)
where
Av -f, '0 tv (x, y) dx d.
(7.125)
(7.120)
(x, y) f (x. y) dx dy
(7.121)
~t
Mx Y)
--
_____
\4jij
OPAQUE TARGE r
DOIFFUSE TARGET
--
I7.37
f(0, 0) Av
(7.124)
-0"i
dyd
elsewhere
(7.123)
(7.119)
where
tnecessarily
t(x~y)
tv (x, y) f (x, y) dx dy
=!f
=
Ofre
(7.122)
2071.17
i
can be compared with an ellipsoid.
n
If the target is viewcd at an angle 9 from head on,
the presented area of the ellipsoid is
02 ]
22Th
(7.129)
Ae 3 fb
7b[ae
As = 4b
3ta,
sin 0 + bs Cos 0
(7.130)
(7.126)
(A /At)2
P2 A/[A +(21ru2)(.27
P2A(.17
+ -
sin 2W
X '=b/a
tan 8 - A
where
(As/Ae)max = (2)%
I.1.l4...
(7.131)
(7.128)
f0, 0) Av
Pk
.12
computation,
Pl=!.e~"A/2
First consider hit probabilities. Both wing and fuselate are represented by diffuse ellipsoids, with proba-
P)
P2
dent (of course their values depend on u,v) the probability that a round passing u,v hits neither if (I-t)(l-t,)
and the probability of hitting either or both is
0
04.0
.0488
.0473
11.I
.095
.090
1.0
.632
2.1
.864
S1.0
ta 0 tf + tw- tf4w
.500
'7.132)
.667
i-I
20871-528A
7.38
I
I
-.
(7.133)
Ao = (4A/ir)tanzl A
(7.136)
AVL + 1.0 the ellipses become circles and forjt I A. - A.
Ao = ui(! +U)
(7.137)
tion is
the other; the top one is the fuselage, the bottom is the
Aa = 2A - Ao
(7.134)
t . er2/*2
and if we integrate
2
Se
r /a2
2r2/a 2
-lage
(7.135)
presented area.
In fwing
In fact, if we complete the probability of hitting
either
of the two saucers for this case, we find that the
.5 ratio for small p decreases rapidly with increasing p
This is shown in Table VII- I7.
Stacked Saucers
A/(2ro2 )
0
Pexacl
0
Stacked Ellipses
Pdiffux
AI/A
exact
diffuse model
0.2
.18
.24
0.5
.39
.47
0.1
1.87
1.91
.67
0.2
1.75
1.84
1.67
1.0
1.51
1.0
10.
.63
1.0
0.5
1.0
1.50
2OR71I-530
20871-529
7.39
The diffuse wing is considered an acceptable approximately, for the case where the wing is invulnerable
and has a shielding effect.
(7.139)
11
-I
Opaque
Low Wing
Opaque
Mid Win#
"Diffuse"
WingI
1.0
1.0
1.0
1.0
0.1
1.0
0.87
0.93
0.91
0.2
1.0
0.75
0.87
0.84
0.3
1.0
0.51
0.63
0.67
20871-531
7-40
F,
,I
L,
L_
s - [
This is consistent with the objectives of system evaluation, but is not very informative if one has the
objective of reducing the vulnerability of friendly
aircraft to enemy fire.
By expanding the target vulnerability module in the
Litton simulation as described above, and inserting the
fire control system parameters to correspond to those
of enemy weapons, it should be possible to use the
simulation as a tool in vulnerability analysis.
Iin this case the characteristics of the antiaircraft
defense system would be held constant, and the effects
of changes in aircraft design and protection to reduce
friendly aircraft and helicopter vulnerability could be
M.
2T
(7.14S)
74)
-M
(7.144)
assessed.IIM23
r.
(7.143)
11% -()MaT[j*S]'l Ma
r
2
, [RTIIHI[R]
(7.146)
0 2/b2
(7.147)
7
2/c(
2]
02
Is=
POUG
(7.14!)
PiuiS]
o JThe
M.'
for reference
s1 - u2o2(l
p2 )
tematic error.
IMaIT a [MuMv]
1/o2
Is_"I'P
prediction
.pl(ouOv)
i/o2
jIMupfvp] + (MIIuB
J/(Vu0
(7.142)
muzzle velocity
(7.148)
"7-41
II
-.-
..
--
--
...-.
I
earlier, three probabilities
are computed,
then
combined,
(7.149)
[Vf + VW)-
Automatic tracking with radar, gated imaging sensors, and probably manual tracking with visual observation all depend on the aircraft dimensions projected
in a plane normal to the line of sight. Since these vary
with the aircraft aspect, it was felt desirable to modify
the tracking module of the simulation accordingly. In
the AFAADS-I simulation, tracking errors in azimuth,
elevation, and slant range were generated in a manner
independent of target aspect, and this was felt to be
incompatible in the improved simulation with the
improved target representation.
The input parameters to the tracking error module
were chosen to be the dimensions of the rectangular
parallelopiped enclosing the projected target image,
with one side of the rectangle horizontal, and a second
side vertical. Two sets of algorithms were developed,
one using the target ellipsoids, and a simpler set,
developed by J. Jatczynski using simply the orthogonal.I
vectors comprised of the target span, fuselage length,
and fuselage height. The latter set was considered
PC Pf + Pw" Pfw
Note that pfw * Pf Pw in general.
(7.151)
it
1
II
7-42
TAGE
fr
PROJECTILE
VULNERABILITY
ELLIPSES
AI
BIA
21
JIJ
ava
MLOM_
SECTION 8
WEAPON CHARACTERISTICS
The most obvious descriptors or an antiaircraft gun
installation are
Number of guns on mount
Caliber
Muzzle Velocity
Rate of Fire
All-up Weight
In the accompanying Effectiveness Volume. tables
are given of the characteristics of modern antiaircraft
gun systems. In this section some of the possible tradeoffs in choosing weapon parameters are discussed.
It is considerably beyond the scope of the present
"effort to consider weapon design in depth, hence the
discussion is limited to a level consistent with the
sketching of parametric ranges of interest in effective-
One would not expect simple overall scaling relationships to hold well, in estimating all-up weight, because
of at least three major determinants of this value in
addition to the weapon. These are:
ness evaluations.
IAlthough
*
*
*
SThese
9-1i
Ii
100.000
00
TRACKED MOUNTS
$0.000
I/10
00
TOLE_ MOUNTS
1.0
1.
2.0
X|0
2.5
20711t
As noted in the vu!nerability section. a high fractional content of projcctile weight devoted to high
explosive is desirable for terminal effort. Since highh
explosive has a much lower density than steel, as the
HE content is increased, the projectile weight decreases. An additional limit appears when the shell
walls become too thin to withstand the firing stresses.
wp- (CI/1O)
where
w, -
f 8-2
I
-,I
VULCAN (AIR)
GATLING GUNS
VIGILANTE
II
I
I
SI
MESS
820
AM 20"2
1000
"S
OERLIICON
i
a.
o-0
SDOFORS
'100-
10
20
CALIBER (MM)
s0
100
200
-- am-
;.
20671.178A
8-3
caliber in millimeters
proportional
or a projectile
ballistic
constant
shape are isheld
if caliber and
Wa,/C'
and coefficient
to The
n w p/wp
(8.2)
I
401
0I
0I
_J
'9
II
10
LI
0.10
0.20
0.30
HE WEI50T/POJEC.TILE wEIoGT
20671-179A
,,:
II
v2o
Ul
(1 p/)(fgVo2
;k2
= 2.0
.'m
8,000 f/s
n-v/
S(8.8)
(8.3)
w2
p /c +wIh6lv2/g
kw
Wcr
wp + k3Wch
or
(8.5)
CPpwv+k
W/
=1
6kg
68
0.45(10)6 ft
(8.6)
8.3 METHODS OF OBTAINING VERY HIGH MUZZLE VELOCITY USING SUB-CALIBER PROJECTILES
(8.7)
(8.9)
)
(V
8-5
The simplest method of obtaining a very high muzzle velocity is to fire a light projectile. The possible
ain is indicated in Figure 8-4. There is a limit.
owever, to the amount by which one can lighten a
conventional projectile of full gun caliber, and still
retain a good ballistic shape.
.]
6000
/
S40o
4000
AIRCRAFT AND
ANTIAIRCRAFT
AUTOMATIC
WEAPONS
94/
/-
SMALL ARMS
EXPANSION RATIO 6 0
>I
300D/
2000
1000,
1000
0
001
0.02
0.0o
0.1
0.2
0.5
1.0
2.0
Ij
so
10.0
21171.180
h
Very high muzzle velocities can be obtained by firing
sub-caliber projectiles. Two designs for tank gins are
shown in Figure 8-5, one for a spin stabilized aind one
for a fin stabilized round! A sabot is required in each
case; the pieces of the sabot fall in friendly areas and
are a nuisance even when they do not cause friendly
casualities. However this seems to be the only proven
U.S. antiaircraft projectiles approach the % filler content of the Swiss Oerlikon 35-mm projectile.
The muzzle velocity attainable is also increased by
increasing the maximum pressure ., the gun and there
8-6i
I
current method of attaining very high velocities with
gun fired projectiles,
The velocities obtained, however, can be quite high,
in spite of the tact that the sabot is essentially dead
weight, and its acveleration is another cause of lost
Making
a muzzle
charge to
complete
I-Am
figure 8-5. Typic;i Sub-Caliber Projectile Derigns
8 -7
I
with design feasibility, have not been accepted for
operational use.
Periodically, unguided rocket launchers are re-examined as substitutes for guns. Very high burnt velocities
can be attained, but angular and velocity dispersions
have never approached gun accuracies. The current
contender from this type of design is the ThomsonCSF Javelot, which associates a sophisticated fire control system with a multitube launcher firing 40-sm
rockets.
3
-rer
3
U
I
U
(8.11)
where:
W - maximum vehicle weight (Ibs)
L - wheel base (length of track in contact with grout.d) (inches)
.
edges (inches)
e. Rocket propelled spin stabilized, gun fired projectiles. one of which was tested in 280-nim size.
f. Bundles of iub caliber fin stabilized projectiles
fired as a single round. This approach has been
revived recently.'
OfI
9-a
antiaircraft system into the available space and remaining weight margin,
Some current surface to air short range missile
systems employ separate fire control and launch vehicles, but it is doubtful that this expedient will be
accepted for future gun systems. Armor protection
severely from the available armament
Ssubtracts
weight
budget. It is remarkable that technological improvements in radar, servo drives and power sources, and
fire control have made it posible to build tracked,
armored antiaircraft vehicles with both surveillance
and tracking radars in the 25-35 ton all-up weight
range.
(8.12)
(8.13)
D(v, v.)
k 2 (w
2
C
(8.14)
(814)
I"
-~ 'i lI
4-~
I
P
I
i
SECTION 9
MNTERATION OF ATTACK AND DEFENSE OPTIONS
The object of a local defense is to raise the cost to an
attaok of dwsroying a defended target to a level at
which he feels that it is unprofitble to attack. In the
absence of a defense, the attacker can use relatively
inexpensive munitions and aircraft fire control. In the
presence of a light to moderate defense, the attacker
can use the same weapons, but release them at greater
ranges from which they have less effectiveness. This
requires a greater imiber of pases to destroy the
defended tswt. and at some level of defense the
cumulative losses become unacceptable to the attacker,
and his only alternative to abandoning the attack is to
employ more expensive stand-off weapons, call in
ECM support or use other, comparatively costly means.
For a given tactical situation, with the simpler systems, such as conventional dive bombing, both the
attacker and defense have tactical options with regard
to bomb release range, and fire distribution doctrine of
the defense. These in'errelationships are examined in a
game theoretic context, in the one attacker, one defender situation.
The more general situation of a number of defend"ing fire units subjected to repeated attack waves by
In attempting to develop preferred defense performance characteristic-, the object is to find a set of
characteristics which performs well over a wide range
of likely tactical parameters. It is believed that the
contents of this section develop a general analytical
structure within which this objective Lan be pursued.
9-I
Define:
Kw - probability that the bridge is destroyed in one
pass
K. p. -
qd
-p
'value' of the aircraft
V- 'value' of the bridge
-
P/Vd = Kd II -(VaKa/VdKd)l
P=VdKd - VaKa
-a
[Va + VdPdJ
P2/Vd
-VV
(9.3)
9.)munitions
9-2
t
p.(t)
d'Va
V (K&/Kd)
(9.2)
P2
(9.8)
Next consider the relationships if the bridge is subjected to repeated attacks until it is destroyed. The
attack/defense tactics are assumed identical in each
attack pass.
The probability that the bridge is destroyed in txactly 'n passes is as follows:
Nd mi!Kd
I -(VaKa/VdKd)1
(97)
(9.1)
n
Kdn a Kd 1; (1-Kd)
1
I-K.
Q,
V.
I;
U
I
p.(t) -
r, -
N -
maximum number
of rounds which
can be fired by the defender
w(t)
I
P(f, v) =1
o
K. -
P(J. v) -
v) 10(,
define:
b(t)
(9.9)
w (t)
J V(tOp,(t)dt
w (d)
(9.10)
vin
(9.15)
Define:
1
r'I,
(9.14)
(9.11)
Ij
Next define
At)
(O ,V) e
A'(t)
(t)Pa(
in
notation,
(9.12)
(Q,P) df (0
fo
9-3
93
w (t) dt
NO
(9.16)
This is the ammunition requirement to fire the optimum rate of fire schedule over the whole path. If the
available ammunition N <N.,
define:
where:
b6 D. -
fT
velocity.
It would be preferable to use Equation (9.18) in its
complete form for comparative evaluations, but the
simpler form Equation (9.19) allows simple analytic
solutions to demonstrate the kind of results one can
obtain from the game solution. The fact that p, can
exceed unity at very short ranges does not impair the
solution.
fc
(9.17)
w (t) dt =N
Id =
SD
_Av
=A ,+ 2 1-,
o 2 D2+ 2 fr go02
[1 +2 0p/T s) + 2(t p/T s)2]I
Of
(9.18)
a-, -
a,. -
where:
bD
22
iD
t, w time of flight
rr
(9.21)
Swhere
"
weeincludes
A. aircraft vulnerable area
range
D- - slant
saragrange
T, =
(9.20)
Avd + 21rOd 2D2
SPa
A -Nbo/Dm2
om
B a Vm2bo/(4v2N)
K = (Dd/Dm
(9.22)
(9.19)
atake oldhvetesaeprbbliyo
I~
zo
IV
II
-
V,
VI
,;-t
g 02
I'9
combination of his survival probability and distribution of release ranges against the specified defense.
Sz.
40
0.10
SFigufe
..
2bo|/V
o-i. Clam.Teorti
9-6
''-"
Solution OwiN,
?U71 IS7T
System
F-quivalcntr
Angular
Number of
Armament/
%I.\imumKate
of I-ire prr
lire Unit
Iire Units
I ire Unit
(rpm)
I X 20 mm Gatling
3000
1OOO
10
2000
2000
350
4000
3000
300
4000
650
250
3900
X 35 mm
It2
1)
I X 37 mm Galling
2 X 40 nim Bofors
Rounds of Ammo
per l-ire Unit
Mi.,mum
R~ange
~i'per'.on
I mul,%
4onii)4metcr\,)
20871-536
Dd
Target Vulnerable
System
A
B
C
D
2
Area (meters )
With Specified
(meters)
With Unlurnitd
Ammunition
Ammunition
Comments
1300
2410
2590
920
1300
2600
3440
920
1-8
4.0
4.5
4.8
Sand
.
S~9-7
20871-537A
cations of the above comparisons. other than as illustrations of a method for evaluating the interaction
between offense and defense options.
0.01
10.
1
YMAV
M
.
4 1!02VI
001
010
.ji
4.0
To obanaruhieVfth
haeulmtdrt ffiure9-.
feto'nnzr
Shmplatter
Solutico
Char
Tcicammuntineadmu
l~
c.nd
praobabilas
theateiohe the aincrasesortheaitheegun's
the otwc
ground taetwlbedsrgyeti
cageron
9.8
xcfathegni
vnange.ex
agtwill
bwedestproaedhing untyThesegae
04000-
LOCUS OF OPTIMA
WITH INCREASING VALUES
~OF WTo
SM
4
020
40
60
CALIBER
MI-
so
1oo
120
IMM)
2071-189
S
S
S9-9
tt
3600 FFS5
4 MILLS
2000
3000 F1f
4
0
20
406
o100
I.
MI
120I;
9 .1.8 Dlscuaslon
In addition to the conmtraints already noted in the
above derivations, one additional extension may be
noted, namely the possibility of using a mix of defense
weapons to provide a defense in depth. Current doctrine envisions the use of surface to air missiles for
area
defense,
including
Hercules
long range
and high
altitude
coverage.
Hawkforforvery
medium
zone
*
3
9-10
Ik
2I
14000i
//
\\
EFFECT ONLY)
I1
'!U
130
020
40
60
80
t00
t 20
S(.FALISER [M
Sl
208)1li.l-0
I
I9
;ianalysis
*
SI
iewsic t
agt oria
iedt
eeo
h neatos
with more accurately defined probability runcions thar. it has been possible to provide in the present
AND
ULTPLE
IIrCKR
I
*
-I
nieakl n
tion of fire. defense deployment, coordination of enemy attacks, fire unit reload time and time to change
"
)Iwo
.I
20% FILLER
2000
I LLER
S10%
20
40
60
80
100
CALIBER IMMI
Filler Weight
Subcaliber. Muzzle Velocity and
Figure 9-6. Comparison of Caliber,
0.
120
07
207
0.7
--
0.4
/PROBABIILITY
9L.
0.3 -
TRE
THAT DEFE.NDED
ETOE
0..
0.1 -
1.0
2.0
VALUr. RATIO
4.0
3.0
IVAIVOI
!
20871.193
9-121
ile
S10
KA VA
KO VD
5
05
KA
NUMERICAL EXCHANGE RATIO
KO
10
20
30
4.0
5.0
2-'S71-194
MAX EFFECTIVE
RANGE
OF WEAPON
2W71 195
9-13
Dsimilar
SReferring
Units
the attack.
In the case of each event noted, one must make
air defense systems have the capability of tracking several attackers simultaneously in the surveilne foriat This greatly shortens the time
vequired for a newly freed fire unit to acquire
and open fire on a new target, based on information and assignment from the commander at the
surveillance unit. Hence early kill recognition
allows more targets to be engaged in a heavy
attack.
The fire/reload cycle of the fire units is approximated by assuming a statistical model of the process,
described in greater detail in the effectiveness report,
Briefly, if a fire unit fires at a rate Y., and has N
rounds of ammunition on board, it is assumed that the
probability that it exhausts its ammunition in a short
time interval dt is v dt, where v - vo/N. Similarly if
it is in a reload state, with a capability of loading p.
rounds per minute, the probability that reloading is
completed in dt is assumed to be p dt. where p - p.1
N.
ove
We first consider the implications of the
simple model on the number of aircraft in ihe defense
volume as a function of time, ignoring observed destructions for the moment.
Define:
P(;h,t) = probability that there are m targets within
the defense firing zone at time t
Then:
4l
t
dt
Loading
(9.23)
- (X+ Mm)P(m.t)
XdP(,t) + UP(l
9-15
L li
,t)
(9.24)
rJ
P(1,0) . 1.0
P(m,O) = 0; m ; 1.0
(9.25)
et
Tf(o)=
and obtain:
P(0O P= (I -e *t] [
P(mt)
()eM)(l -e
e "M (
(9.26)
(9.32)
(9.28)
(rn-i))where
MMNO
(1 e"A)
e )
\t')
IMmnl
+ e .',7
0,)
(will
(9.29)
a.)=
e (V-)
m!Assigned
Ready, uncommitted
target, attempting to acquire
(9.30)
Firing
Unavailable, being reloaded
Unavailable, malfunction, under repair
Each Target:
obtain:
States Relative to Defense
m e"
+ OVA) (I - e"t)
(9.31)
Undetected
Detected, Unassigned
Detected, Assigned to Fire Unit
Engaged
States Relative to Attack Phase
aircraft.
Approach
on board
On
AttackReleased,
Pass, weapons
Weapons
Evading
In Range, Surviving
Destroyed
above.
constitutes a formidable obstacle to ccntirued development of the stochastic model, even in the simplest
formulation.
Noting however, t
that h
value expression,
(9.31) is easily obtained from a simple differential equation using expected values only, the analysis is simplified to a deterministic formulation in
terms of differential equations, and this turns out to be
tractable.
9-16
1.2
I
1
1.0
I
1/
0.4
0.2
.
M~T
2Qe71 197A
Figure 9 - 11. Probability of No Aircraft in the Defense Zone versus Time from First Arrival
02.4 Simple Deterministic Moel
First consider the case where all guns attempt to fire
as long as there is at least one target within the
defensive volume. The fact that a target is being
engaged already does not prevent a newly available fire
unit from attempting to acquire it and fire on it.
Targets arc chosen at random.
p -
I& -
Ik, -
k2 m -
Define:
i
,
n -
n, -
N, -
.9-17
target at time t
at time t
N -
,n the defense
p
GUNS
/4GUNS
GUNS
FIIGAVAILABLE
BEING
REOAE
2W071.198A
rnk =klnf+P
(k 2 /14
(935)
A gun firing leaves that state if it exhausts its ammunition, it its target leaves the defense zone, or if its
target is observed to be destroyed. It enters the firing
state when it acquires a target.
;r = unf - Pr
I/p.
ri = X-;a-m. kInf
(9.33)
Y =
defense volume
3000 rpm - rate of fire of one fire unit
(9.34)
since the mean firing time is p'. Since targets leave the
system at an average rate jAm. the average rate of
obtaining both observed and de~ayed kills is
approximately:
N. -
10 seconds, -
9-18
*1
I
I:
12
06
08
LOADING
ATTEMPTING
04
-RE
TO .OONCQUIRE
0.2
I
0
20
30
40
so
TIME iSEC)
2W871 .199A
Ar = Unf-pnr
if = a(m-nf) - (u + + k 1 ) nf
nf = a(N-nf-nf) (v
= -
N.nr m
.J + kl)nf ; N-nr m
klnf
k2")A
mk
nf
(9.36)
I +P6 e
"00
S...
...;...
- .........
. . ;.. ...--..
i i . . . i "" "I I ' ' 'I
II
:I
9.2.6 Conclusions
The relative simplicity of the deterministic model
allows a moderate amount of additional detail to be
incorpoiated before it becomes unwieldy. By its use a
rapid preliminary investigation can be made of the
relative effects of acquisition time, reload time, rate of
x.(t) -
0",2
p -
',] 2
f(x.Ix0 ) =
op [21r.(l.p2)] "2
[Xp - (Opod),ol 2
exp2
(9.37)
2 op2 (l.p2)
Since in reality the operator is tracking in two
dimensions, it would be more realistic to write the
corresponding expression for a two dimensional probability density function, and this might be worth doing
in a complete analysis, since the results could then be
expressed in terms of burst kill probability. However,
as will be seen the ont-dimensional case is quite
illuminating.
Next we may recogrnize that even in a one-dimensional problem, the operator could not respond quickly
enough to fire when x6 is exactly zero. Hence it would
be more realistic to integrate f(Qx x.) over a pdf of x.
(small compared with oa,). Likewise in the two dimcnsional case one would establish a ring within which the
operator whould fire since he would never have both
coordinate tracking errors zero at the same time.
To show the maximum possible gain, now,'ver, in
the one-dimensional case we assume that it 4s possible
to fire when x. - 0, and so we need only determine
a 9 2(I-p') and compare it with a,2 to observe the im-
described.
9-20
Lami
I.
-1
'11
I
The prediction variance p2
pr
(9.38)
Ro = 0a 2 e .utTc
(9.43)
up
(9.44)
which we abbreviate:
and we shall determine the effect oil varying T, on the
system performance with and without choice of firing
instant.
Ithe
2X
(9.45)
following:
linear prediction function can be described by
Any
We are interested in the variance resulting with the
designated fire doctrine
ss2 = ap2 (I #P)
xo (t.u)A(
xP ()
dua)
(9.46)
and
(9.39)
= 002 [p.,X21
(9.47)
As examples, we choose I) a predictor transfer function identical with that used in a gyroscopic lead
computing sight, such as Vulcan. This function has the
characteristic that it does not attenuate high frequency
noise. For comparison, 2) a predictor transfer function with second order smoothing is assumed. This
does attenuate high frequency noise.
Then:
Rp
<xp(t)X0 (t)>
xo(t)xo(t.u)A(u) du >
S-0function
j
Ro(u)A(u) du
(9.
(9.40)
I
For the covariance given by Equation (9.38) R, is
therefore simply proportional to the Laplace transform
of A(u),
(9.41)
(9.42)
p
p = 0.2 A(l /T,)
The Laplace transform of Equation (9.39) is:
xp(s) = xo(s) A(s)
9-21
OF,
Sy,tcm
Gyrosight Type
I + ( 1
T"an',ler I mtilion
'p
;a0
+Sa
I sa~p(
(tp + 2T5)
+T
"TrickingI tror
~Autosjinflce
at1
'S
tp
I)
o02c*u/T/
a02e'u/Tc
V j,,in,:,,of!
P fctho
I
Irror
2 11 +"Nill + afal 2
0
1+ X
1 + x I" + 4a 4+
a0
a )II
p
Cojueof
-W
Prcdlictiont ind
Tri.cking Error
R
Itt
+ I
II 1__13__2_1_+q0_+_2._/a_
0.2~
0+
+2a)
0 +I
)
at /!T
Note
20871-$38A
]
]
Wt a
With a system that carried enough ammunition to
fire continuously over the whole available firing time
there is no anticipated advantage in firing only at
selected intervals. For very high rate of fire systems,
and in general for systems with relatively long reload
times, there could be a substantial overall payoff, however, in firing fewer rounds and firing those when the
tracking is very good.
i
.3
i
1<
9-22
II
I,
.
-.
.:..
: .
..
.............
_-.
6.0
024
40
II
NOIS
2087, 20
1.0
0s
0I030
FIRE AT7RANDO
2.4.0006
g
Figure
20071 201
9-15. Comparison of Firing Doctrines for Fire Control with Second Order Filters
9.23
ones,
3
-
computer which points the gun, not at the attacker, but at the point in space where he will be,
provided he continues on the same course, when
the bullet gets there. That is. the computer does
this if the gunner tracks smoothly and on target,
and frames accurately. In addition to these tasks,
which are difficult, the gunner is supposed to
squeeze the trigger.
squeeze manytrguer. ta.
For many gunners this last responsibility is
just too much. They find it impossible to perform
all three tasks (tracking. framing, and triggering)
properly with only two hands and one brain.
Consequently, as the prcject psychologists discovered, the gunners develop a regular rhythm of
triggering which has nothing to do do with the
accuracy of tracking and framing. They squeeme
the trigger, and let it go. squeeze it and let it go,
in bursts of fire which are just as likely to come
when they are off target as on: just as likely when
they are framing poorly as when they are framing properly. The gunners are not to blame. The
task is too difficult for them. It may not be too
difficult for the engineers who designed the
sights. But it was for the average GI who became
a B-29 gunner.
In order to simplify the task to a level where
the ordinary gunner could master it. the psychologists working on this problem did two things:
This determination is already made automatically in some current systems, and the operator is
given a signal to indicate that he can open fire
when he desires. Th. determination of compuiar
settling could be made from a ,nesurement of
elapsed time from target lock-on, or from the
fact that a smoothed measure of target ac.eleration has dropped below a threshold. A determination of system readiness is fundamental to all
systems, whether it is done automaticady, or by
the operator.
b. Phase 11; Threat Evaluation. The following threat
characteristics are observed by the operator, in-
92
r'11
9.24
levels.
Laydown Attack
b)
Strafing Attack
c) Dive/Glide Bomb Attack
d) Release of Standoff Weapons
(2)
h vet
theo tgunhutrs
prfteshotgun
interestin
approach,
even
own bias in favor
fao of
though a simple analysis of the parameters of their
problem indicates that it does not scale at all io the
antiaircraft gun problem.
II
Z () Ze
?Z [(xbx.) 2 + (Yb'y)2]
0 ( Z)
xb
where
yb
( 2 1r)Ee
=/
'A
-OI/E
Eo = na 2 /(a2 2 o2),
f(T) dT
*E.
E 0T
(9.48)
dxbdyb
*E oe *X
T
.XT f(T) dt
(9.56)
(9.49)
vulnerable area
a 2I:=target
=
fut:
(act
)/s-]~
=1"
(9,50)
(9 .51)
A!
T) e"*XT
-XT
f(T) dT 0
(9.57)
()VEo) a(1)fax- .o
(9.58)
(9.59)
dispersion
OZ2 = incremental variance of artificial dispersion
X -2b 2 /(a 2 +202)
Hence:
/
ac~i~la Eo
whence:
Eo( Xe a C
(9.60)
z C .0
(9.55)
aOZ)/aZ.
(Z) 1fe
d(e'w)
(9.61)
(9.52)
()/
.o
(ILlW)w/ewod"
(9.62)
"0(T)
(9.53)
9.26
..
Loge
- 01
(9.63)
indicated, they were associated with burst kill probabilities that were very low, even when optimum dispersion was used.
., - 3.56...
3.56
(9.64)
or
or
(na2)(2
(9.65)
Ab = 2 TrOb
(9.66)
2o-2=[at ob(n/1.78)1/2).a2
(9.67)
(9.68)
deviation of bias, the results of Figure 9-17 are obtained. As expected from the previous section, the
optimum dispersion increases slowly with bias, and is
only about 6.5 mils when the standard deviation of
bias dropped
is 20 mils.
This4%.
occurs when burst kill probability
has
below
expanded with n.
This discussion is not intended to leave the impression that one should not attempt to determine and
employ the optimum dispersion pattern. Rather, the
conclusion is that this is an expedient of limited payoff.
The proper approach is to attempt to minimize prediction errors by a good choice of prediction and smoothing algorithms, to minimize system boresight and
ii
1!
14
ILI
10
'
02
1.0
2.0
3.0
4.0
NA VIAb -'20871
5.0
.202A
10.0
IL
100
ROUND BURST
IISO METER
OPTIMUM OISPERSION>
8.
0.5-b
IN
BUS
-'
ILPRBBLT
I
3
SECTION 10
SIMULATION RESULTS
10.1 INTRODUCTION
It
The simulation is now relatively complex, as compared with the simpler version used in AFA.ADS-l. but
is considerably more realistic, and allows a much
lare vridetaby oftactialisi
e
and
defen
ystm pramch
tlargero variety
tactical and defense system parame-
The price paid for the added capability has been the
difficulty in identifying and eliminating all programming errors in the basic simulation, and those introduced by new operators inputting parameters for new
series of runs.
in Figure 10-1.
c. Dive bombing paths. These paths are described
in Section 10. 16.
In all cases the data filters used had a finite memory
time of 1.8 seconds. The coefficients were chosen as a
compromise between smoothing and rapid settling;
of a real fire control system. which will operate approximately correctly, even thou3h some omponents
S
rxiael
crecl,
ve ho~ One
smecmpnnts
may be slightly
out of adjustment.
danger signal
noted in simulation operation has been occasional
results which show very large dispersions to maximize
I?
The
thesudemonsterated
y
the
wi
reader in this sense- 1) They demonstrate he wide
The
2)
range of capability available on the simulation.
results are probably correct, or very close to correct,
and 3) They can be completely validated by the check-
"
Areas (square meters)
Fuselage
Length
Height
Width
15.0
2.3
1.4
Top
Front
Side
16.49
2.53
27.10
Wing
Span
Chord
Thickness
10.0
6.0
0.36
Top
Front
Side
47.12
2.83
1.10
The entries are the probability of kill with a onesecond burst. These values are plotted in Figure !0-2.
for tangential prediction.
The regular form of the curves led M. Ginsberg to
suggest that a simple scaling relationship could be
developed to describe them. It was found that if the
target survival probability to n rounds was plotted
against the target survival probability to n/2 rounds,
all of the corresponding data pairs for all four calibers
and the three sets of firing points fell on the same
curve for a specified prediction mode. The corresponding plots are shown in Figures 10-3 and 10-4 for the
two prediction modes. The deviations of individual
points from a mean curve is well within the scatter
resulting from the Monte Carlo operation of the
simulation.
2
'Tb
-
e
e*E/
0012)
O(n)
[Q(n/2)] 2
2 Loge Q(n/2)
LogeO(n)
(10.3)
1
j
0(n) B'I(A)"IBI
Q(n/2) B'I(A/2)"l/B
18
(/2)l/
1
0(n)
(10.4)
Define:
If 20- >> aW,as will usually
be the case, B will be
essentially independent of a2.
na 2
2
202
(.
2ub2
a2 B+(10.2)
202=
"I/B
(10.5)
n -
1-
where
o"Z
10-2
I
fCaliber
~~~~Table
X-2. Reference-Point
WeaponCharacteristics__________
20 mm
25 mm
37 mm
50 mm
6000
3840
3000
1920
100
64
50
32
600
500
400
300
Velocity (ft/icc)
3600
3600
3600
3600
2720
3403
5036
6806
1
I
Muzzle
(rpm)
Number of Rounds in
I-sec burst
Number of Rounds on Mount
(meters)
Time of Flilst
4.7
5.9
8.7
11.8
10
jProbability of Kill.
10
0.17
0.24
10
0.44
t0
0.65
Given a Hiat
(C)on
Fuselage
Maximum Kills pci Second
15.5
*17.1
21.8
20.8
I
IIMaximum
0.025
0.042
0.150
2.7
7.5
0.380
*2.5
12.2
The difference among weapons with the linear predictor may result from the fact that the larger weapons
also have some effect against the target wing. In additic'n, the larger caliber weapons have a small time of
flight advantage, which improves their relative
effectiveness.
Thsexrintleutsfo
tesmlaoned
theseiegei mecnta
if
results fromrdn the neeudationtlea
Moto
sinfCanto inferewhcs regardsinglte nedformother
the cost of simulation operation.
For the noise characteristics generated by the radar
model and linear prediction the fact that target sur-
10-
10
II
L
A0ALACELE ATLON
SVERTICAL
wm
ACCELERATION
IACCRDIA
LEATIN
0I
20
TIME (SEC1 ,.
30
35
-*.I
25
CROSS Ovr R
iking Path
20871204
As a demonstration, footprint data have been generated for families of pass courses of two types:
I) constant altitude, constant velocity paths. and
2) 'jinking' paths in which the aircraft weaves vertically and horizontally about a mean, level, straight
line.
j
g
I
the normalization was based on fuselage vulnerability
Table X-3 Burst Kill Probabilities for the Point 500i
Before
o Croisover on a 600 m Fly-By Path
Tangential Pediction
Linear Predictin
Ulibcr
20
Chbcr
2
7
.78
.86
Rc 1 Iir
rde37
S(Rournds/Se)
128
2
Ii
[
Ii
.94
50
50
20
20
25
37
.96
96
.97
.96
.99
.99
1.0
1
only.
1.0
.0
96
72
.81
92.
95
95
.98
10
80
.67
.77
.90
.95
.92
.97
1.0
1.0
64
.62
.72
.87
.93
.89
.95
.99
1.0
48
54
.6S
.82
.90
.82
.90
.98
.99
32
.42
53
.72
.84
.70
.81
.94
.98
16
26
.34
.52
16
47
.59
78
.88
20871.541
At this range and the low target altitude, the fuselage presents a much larger area than the wing, hence
10-5
L7 --
'
'
'
"
IIIII'rrl
10
09
0o
MIN
t~3
"to
036
0
o
16
32
96
80
64
48
12
128
20871
2
205
Figure 10-2. One Second Burst Kill Probability Versus Rate of Fire and Caliber
10. EFFECT OF AIRCRAFT 'BREAKAWAY'
MANEUVER
small.
1
i
The effect of increasing muzzle velocity was determined by a series of simulation runs against the jink-
10-6
p
.2I
-D
' .----,--"
l
,-
I
SLOPE 050
SLOPE 1.0
0100
0.01
20 MM
37 MM
0)
50 MM
0.10
BI
10
QiN}
2o17o.2o6A
_10-
rI
I
|
SLOPE 0510
0.10
S20 MM
25 MM
S37 MM
so0MM1
LI
0.01
o.1
1o0
O(N)
Linear Prediction
Figure 10-4, Rate of Fire Relationship from Simulation:
20671-207A
I OD
;2
f0.607
TANGENTIAL PREOICTION
a-
0.20
|-2
25,MM
.5
7MM
[
010-
II
0.06LINEAR PREDICTION
ii
2M
W3MM,.
so
'
100
1so
200
Figure 10-6. Variation of Survival Probability with Rate of Fire and Target Vulnerability
I10.9
250
I
Table X-4. Foot Print Data for 25 mm Gun With Tangential Predictor on Fly-By Path
4m)
0
Po-t (m)
-Fie
Crossing Range
3000
2500
2000
I SO
1200
900
600
300
-300
037
082
124
198
277
361
469
00
O00
00
034
084
122
195
272
300
5837
185
037
362
362
200
038
Owq-
134
216
292
422
702
483
352
428
300
042
107
152
247
333
488
834
717
580
524
400
123
172
280
377
560
853
?91
703
602
500
05
138
192
310
415
617
845
816
681
579
600
056
152
211
337
448
638
826
800
681
5S4
700
061
166
227
358
473
635
795
760
672
521
800
066
177
241
375
477
620
755
703
643
516
3000
076
194
260
381
446
568
650
573
S45
381
120(
056
204
268
339
399
490
533
436
435
306
1600
095
204
224
255
282
329
331
237
242
176
2000
089
157
165
176
ISO
199
190
126
111
086
2400
065
112
114
1I3
III
104
096
065
045
026
1400
20N71-542
1I
I4
I0.I0I
I
0
25000
I!
2W
0
1000
7MM
GUN
FLY BY PATH
-7-
TANGENTIAL
S.PREDICTION
3
GV0 POI
ON
GUN POSITION
SM
10
50
002So30
20871 209
Burst
Figure 10-6. Footprint in Ground Plane of Constant Contours of Kill Probability with 1 -Second
I
I
I:
Lj
Table X-5. Foot Print Data For 25 mm Gun With Tangential Predictor on Jinking Path
Fire Point (m)
Crossing Range
Im)
3000
2500
2000
1500
1200
900
600
300
-300
Oil
030
052
124
175
353
474
395
008
116
100
Oil
031
061
142
200
356
501
229
087
261
200
012
034
07S
177
248
408
606
327
270
317
300
012
037
089
211
295
492
663
651
559
418
400
013
040
103
240
335
552
672
688
603
440
Soo
013
600
014
700
014
800
044
115
261
365
046
125
275
383
048
133
282
015
049
139
277
1000
018
049
144
1200
013
045
1400
014
1600
013
2000
2400
$79
649
655
625
412
566
618
610
559
377
379
534
579
571
535
352
355
499
533
534
491
360
240
300
405
425
436
405
227
134
197
247
319
323
330
320
173
038
117
157
199
245
243
236
251
135
028
100
122
157
185
180
159
192
110
Oil
015
071
067
093
107
103
069
102
039
010
009
046
029
050
062
065
033
030
"-
010
20871-543
A
"
1
1
1
10-12
I
Table X-6. Foot Print Data For 50 mm Gun With Tangential Predictor on Fly-By Path
Fire Point (m)
Crossing Ran..
im)
300
-300
.713
.454
.103
.48
.694
.880
.69$
.516
.59
3000
2500
2000
1500
1200
900
600
.047
.169
.244
.346
444
.693
.684
100
.059
.148
.230
.383
.517
.598
200
.053
.165
.282
.456
.522
300
.074
.167
.263
.417
.499
.670
.822
.789
.747
.68
400
.086
.198
.341
.481
.550
.750
.857
.888
.835
.755
So0
.080
.209
.291
.526
.571
.721
.885
.916
.823
.68
600
.104
.181
.346
.531
.6"
.804
.897
.896
.846
.735
700
.082
.249
.375
.583
.690
.821
.864
-866
.794
.69
800
.088
.261
.377
.578
.683
.733
.814
.810
.789
.709
1000
.088
.274
.376
.590
.640
.686
.663
.768
.598
.59
1200
.105
.263
.407
.507
.627
.590
.593
.592
.547
.50
1600
.111
.301
.359
.386
.402
.436
.488
.444
.373
.33
2000
.108
.243
.276
.296
.354
.328
.362
.374
.288
.257
2400
.136
.194
.179
252
307
221
.190
.207
.199
.14
1400
20871-544
1I
I
10-13
Table X-7. Foot Print Dota For 25 mm Gun With Linear Prediction on Fly-By Path
,Fire Point (m)
Crssing Range
(m)
3000
2500
2000
1500
1200
900
600
300
-300
02
0o
16
3A
so
59
76
00
00
00
100
02
09
17
37
53
69
83
29
01
42
200
03
10
20
42
$9
78
85
43
26
67
"100
03
1I
23
47
66
85
91
78
86
90
400
03
12
27
52
70
90
95
91
93
92
500
03
13
29
55
"13
92
96
93
94
91
600
04
14
31
56
74
93
95
95
95
87
700
04
14
32
57
74
91
94
94
93
79
800
04
14
32
56
72
89
92
92
90
68
1000
04
13
30
51
63
81
83
82
74
46
1200
02
10
25
40
49
67
66
63
46
19
1600
03
03
09
14
17
25
19
13
03
00
2000
00
00
01
01
02
00
00
00
00
00
2400
00
00
00
00
00
00
00
00
00
00
1400
"
"
--
20871.545
I1
I
I
I
V
10-14
'U
Table X-8. Foot Print Data For 25 mm Gun With Linear Prediction on Jinking Path
Point (in)
______lire
Crossing Range
(M)
3000
2500
2000
1500
1200
t
900
600
300
30
00
06
10
06
10
37
60
49
04
04
100
00
07
II
07
12
43
80
so
14
45
200
00
08
13
08
14
48
84
SI
60
62
300
00
09
15
09
is
47
77
76
85
70
400
00
10
17
09
13
42
67
83
85
62
500
00
12
Is
09
II
33
S6
600
00
13
Is
07
09
24
44
82
77
80
74
55
76
700
00
15
18
06
06
17
30
68
64
36
800
00
16
16
04
04
II
18
56
55
21
1000
00
18
12
02
01
04
04
27
30
09
1200
00
19
08
00
00
01
01
08
09
03
1400
00
18
04
00
00
00
00
o0
01
00
1600
00
00
00
00
00
00
00
01
14
00
00
00
00
00
00
00
00
00
04
00
00
00
00
00
00
00
00
2000
2400
00
00
20871-546
I
I?
10-15
IPI
2.0j
zj
Cc
10
09
NORMAF
IZO
O00 AROSIN
TARNGT
BOYPATH
EIEE
0.CI
30
3018000
20
Io2w3
10-16
FlyBy
at
2007
I
I
09
LINEAR PREDICTION
0I
06--
'A
0.5
~TNGENTIAL
PREDICTION-
300300
"
900
129)0
2S00
I0m)0
1000
"--
3000
20871 21?
O
09
1
I
-0
TANGENTIAL
JINKING PATH
1000 METERS CROSSING RANGE
S03 -
o
300
\L
o,.
0
3o
0ow
1200
to
!2000
2100
3000
10-17
04
03
03
10-
NO BREAKAWAY
02
300
-REAKAWAYf
30000
900
120050
10-11
Z08'1214
20 MM
25 MM
37 MM
I Lit
1
I
Point
(M)
VM= 3600
VM= SOOO
VM - 3600
VM = S000
VM - 3600
3000
00
.00
.00
.00
.00
.00
2500
.16
,22
.18
.20
.29
.24
2000
.03
28
.12
.32
.29
47
1SOO
.00
.13
02
.18
.07
.35
1200
.00
.11
01
.IS
.os
.28
9U0
.02
.21
.04
.24
.11
.39
600
.02
.29
04
.32
.As
.51
300
.21
.57
,27
.58
.48
.72
.23
.53
30
52
.48
.63
300
.06
.25
.09
.24
.24
.33
I20871-547
I
101
r
II
10-19
vv
5000
101
09
081
07
JINKING PATH
05
05
04
03
02
30
0I
0
300
Boo
goo
1200
IS
zo o
2 ,o
o
3 O00
20871 216
!
j
Figure 10-12. Effect of Muzle Velocity on Burst Kill Probability with 20 mm Gun
I
I
!
I
I
I
I
SI.
SIO1-20
L.U
000
x
0-5
-3
0(
mm
0 -5Io0MM
370
,M
AT 3600 FT/SEC
BURST KILL PROBABILITY
Calibers
Probability for Three
Velocity on Burst Kill
Muzzle
of
Effect
Figure 10-1 3.
20871-214,
0I
10-21
.4
30
J
-I
w)
SLOPE 1 3/1.0
0.1020
I-I
I
0
.I
a.
011
1-.102i
25 MM GUN
JINKING
010o
0.20
TARGET
0.301
Effeciveness
Figure 10-14. Effect of Increasiriq-Toprminal
10.22
28127
I]
I 11
10
I
-8
-J
06
TANGENTIAL PREDICTION
I
I
1
10.23
12
10
I0
,.. .
-. --,
.- -.
- .
- --
I.
-.
0.8
2L0
M/S
0.4
02-
Figure 10-16. Effect ot Aircraft Velocity on Burst Kill Probability with Unear Prediction
10.10 OPERATION OF SYSTEM WITH SENSOR
INTERRUPTION
up
r-1+2t
)T
(p+T22
+2(t+T)/T/+2(t+T)2IT
(10.6)
INH
02
'
.-
u
-
Hence, with the 1.8 second filters, the only set used
in these runs, one would not expect the regenerated
solution to hold up well after about 1.5 seconds.
in which it was assumed that rectangular target velocities proportional to real target velocities could be
generated without range data. That is, in theory at any
rate. feasible. System performance was then examined
with a constant range setting.
To perform the experiment, a special modified
model deck was put together. The deck did not use the
time of flight module. Instead, whenever time of flight
was required, a constant value (which was input in the
data deck) was used. Three runs were made with fixed
time of flights corresponding to ranges of 1600, 1400
and 1200 m with the special deck. The system was
tested against an aircraft executing a jinking path with
a minimum ground crossing range of IO00m against
the 25mm gun.
It is remarkable that higher burst kill probabilities
(see Figure 10-21) were generated by the simulation
over an extended path segment with constant range
setting than with continuous range input. The reason
for this mysterious result is unknown.
T
The simulation also has the capability of reproducthe performance of rate x time predictors with
10.23
I.
7,
. . . .. .. . ..
. .
- -.
Si
.'
I1
10
0-
4-
101
20
0-
Ii
I'
TiME
20871 220
TIME
Figure 10- 17. Aim Errors with Senior Interrupt on a 500 Motor Fly-By Path
The resulu Are somewhat mystifying. Elevation bias
is much more serious than azmuth bias, as would be
expected. since the target has on the averag~e, a larger
lateral extent than a vertical dimension. However, a
negative muzzle velocity bias of about 36 ft/sec is
indicated to be beneficial. One would expect muzzle
velocity bias to behave about as azimuth bias near
midpoint, and unfortunately only positive azimuth bias
was run, so it cannot be determined whether one could
get an improvement with negative azimuth bias also.
(10.7)
3
I
1
I1~
a.
.-
...- *--"
..
I
1
15-0
1
I
10
I10I,
10
1.
II
13
14
IS
I
TIMi
Figure 10-18. Aim Errors with Sensor Interrupt on a 1000 Meter Fly-By Path
SE(mils)
- .0.1 Vb(ft/sec)
(10.8)
Hence, a `6 ft/sec muzzle velocity bias should have
about the same effect near midpoint as a 36 mil
azimuth bias. The cosine of elevation angle is close to
unity on these low level paths.
20e71 221
A set of runs was made to demonstrate this capability. A 25 mm gun system with 256 rounds on the
mount was used. Fire rates of 64, 32 and 16 rounds
per second were used against aa aircraft executing a
fly-by with a minimum ground crossing range of
100mm, a constant altitude of 300m, and a constant
velocity of 300m/sec. Ten fire doctrines were used.
They are presented in Table X-I I and in graphical
10-27
101
Ca
1o
06
NO SENSORS LOST
og
04
37 MMGUN
A6METERS
13
LOST
RA0
AD
AAT
LS
ALE
02
0
0.7SNSR
FIRE POINT
iM)
LOST~
20A7, 227
e0i
--
02 1
L
[0
0.7
- ---
ANGLE LOST1
(-,.)RANGE
oI
I ,.
0..
0.6
0.
03
o0
1
OREq
i Pl
O
Wh
# i
,~0.-281
20 97 1 2 2 3~
0.9
07
~----~
0-----
0-*-----
CONTINUOUS RANGE
06
02
0/
02
01
IC
-IFsAS
300
300
600
900
1200
ISO20
POINT (M)
503000
20871 224
Inoted
ofconisistent
hisrou
tofb
t
data, the best value of dispersion
tunsoutioebout 4 to 5 mils. Although not investigated. it wudbe expected that still better results
might be obtained by an elliptical dispersion pattern,
10.14 EFFECT OF MUZZLE VELOCITY
DISPERSION
The lec ofmuzle
eloitydiserson as lso
The
muzlevelcitydiserson
ffet o
as lso
investigated for the target/path situation of the previ.
ous section, and the 23 mm gun. The linear prediction
moewsused. A set of values of muzzle velocity
dispersion was examined in conjunction with two values of an ular dispersion. The results are given in
Table X-Itand the one-second burst kill probabilities
are plotted in Figure 10-26. For this parameter set the
Fot
Uthisset
,-
I SC Burstb
-,,,,, -
Bias Errorss
~-
Muzzle
t- A
Azimuth
(Mils)
Elevation
(miis)
Velocity
.71
.70
75
.66
.71
.48
.98
.94
1.00
.69
.69
.71
.66
.71
.43
.97
.94
1.00
.66
.67
.66
.65
.67
.34
.96
.92
1.00
.61
.64
.61
.61
.60
.23
.95
.89
1.00
.56
.61
.SS
.57
.52
.14
.93
.83
.99
.61
.54
.66
.76
.65
.43
.94
.94
1.00
.45
.36
.49
.73
.48
.31
.82
.89
.90
.27
.20
.29
.38
.25
.15
.58
.74
.88
.12
.09
.14
.33
.08
.06
.31
.43
.60
36
.76
.74
.88
.70
.$5
.36
.99
.8
.61
.61
.76
.46
.25
.13
.97
.63
.99
.35
.37
.43
.20
.04
.06
.77
.2s
84
"4C
(filsec)
-72
0-08
I
3 1
1.00
350
-144
.13
Is
.15
.05
.00
15
.36
.19
-120
.03
03
.03
.01
.00
.21
.Os
.21
.28
-360
.00
.00
.00
.00
.00
.00
.00
209715.48
10-30
,o~o
lIINlUII IIL
L_ IIIIIumul
I ni . i.... nrlm
i S
zI
100
.90
.90
3 SEC BURST FIRED AT CROSSOVER
.80
I.70
t"
AT CROSSOVER
60
300
00
ONE-SECONL) BURSTS
-r
II
.20
.10
"LO
00
is
"20671.226A
Kill Probability
Figure 10-22. Effect of Azimuth Bins Error on Burst
If'
L,
ML I
I10-31
1.01
00
.0
.30
601
K~.40
EOECOSOE
OVER
R
MBEFACROSS
001'L900
.0
.07222
Fiue1-
.10
23
300aio
Sia
EOEro Co SS OVERt
KilPobblt
6~~~~0CMBFOECRS
OE
-'--3
--
SEC BURST
FIRED 900 M
90
BEFORE CFC.SS
OVER TO CsiOSS
OVER
.90
P(KILL) FOR I SEC BURSTS
.?0
S
4
a
,4
AT CROSS OVER
19
.40.I
20
..10
0.0
0
38
72
106
144
ISO
360
1[
20871-227
Figure 10-24. Effect of 1Muzzle Velocity Bias Error on Burst Kill Probability
11
S10-33
Lk
A
*1
Rale of Fire
(Rounds/Sec)
Numb~er
of
Bursts
64
Time*
(Sect
Fire
Distance
from Midpoint
(m)
Continuous/
Noncontinuous
(C oF N)
1800
4
4
Duration
(sc)
64
1200
64
600
64
1800
64
2100
32
10
3000
3600
32
12
32
16
4800
is
16
16
16
4800
16
10
16
16
2400
16
"Relative
Time:
0 = Crossover;
Before Crossover
20871-549
1
i
I
1'
10-34
64 ROUNDS SEC
SEM
}1ROUNDS
16
15
to
to
20
15
TIME (SEC)
(DISTANCE I ROM MIDPOINT (M)
28,
10-35
jpomp
J
DISPERSION VALUES
09
ANGULAR
VELOCITY
[MI 1
SG
30
0S
0i L
300
300
600
900
102000
2500
3000
20871 229
Figure 10-26. Comparison of Burst Kill Probability with Specified Angular and Muzzle Velocity Dispersions
25mm gun against a 300 meter/second target at a
crossing range of 1000 meters. As expected the lags of
the simplistic 'manual' system drastically reduce effec-
10.36
I
Table X-1 2. Dispersion Values Used
Angular ispersion
Litcral
(mils)
vertical
(muig)
Dispersion
2
3
2
3
2
3
0
0
Dispersion Set
No.
it could come 'down the chute' jinking until the moment of weapon release, when it was pointed at the
ground target. For reference Path 2c was a repeat of
2a without jinking. Table X- 17 summarizes these cases.
Muzzle
Velocity
(m/Iec)
20871-550
Number
.2
2.52
16
4s
71
.87
90.87
,87
.96
.97
.97
.58
.89
.96
.97
.97I
.49
.85
.94
.95
.94
.42
.79
.90
.92
.91
.22
.66
.81
.84
.81
.50
.84
.93
.95
.93
.30
.62
.77
.80
.77
.37
.68
.79
.81
.78
tO
.37
.68
.80
.81
.79
20871-531
10-37
I .'
Table X-14. One Second Burst Kill Probebility as a Function of Firing Point and Dipersion
Angular Dispersion .5 Mils
01V
(M/Sec)
(MISeC)
Fire%
Range
(M)
10
is
30
10
IS
30
3000
.03
.03
03
.03
.03
2500
.13
.13
.13
.12
.10
2000
.01
.28
.28
.27
.26
.20
1500
.01
.01
.01
.01
.01
.51
.51
.49
.47
.37
1200
.10
.iI
.II
.11
.10
.65
.64
.62
.58
.46
900
.26
.27
.28
.28
.24
.72
.71
.70
.67
.56
300
.50
.52
.54
.53
.48
.83
.82
.81
.78
.67
.44
.46
.46
.45
.39
.83
.82
.79
.75
61
300
.17
.18
.19
.18
.15
.58
56
.52
.47
.34
20871-S52
SRate
Parameters
Rate
Tracking
,C
5.0
Ks
99999
iL25.0
TR2.0
3.0
X
Regeneration
No
Aided
Tracking
Regenerative
Aided
Tracking
Control
Case
Radar
10.0
10.0
500.0
10.0
90.0
4),C)
25
2.0
3.0
No
251
2.0j
3.0
Yes
.00
2.0
3.0
Yes
20871-553
IU
I
Table X-16. One-Second Burst Kill Probabilities With Simulated Manual Tracking
Rate
Regenerative
Control
Tracking
Aided
-p )L".
Aided
Tracking
Can
Radar
3000
.00
.07
.04
2500
.00
.A
.19
.13
2000
.01
.02
.39
.30
1500
.00
.01
.63
.51
1200
.00
.02
.16
.63
900
600
.02
.03
.12
.83
.81
.10
.S9
.83
300
.06
.19
.86
.82
.12
.20
.30
.74
t10
.37
.60
.46
Fire Point
RIte
(M)
-300
2087.1-554A
Dive
Path
AnSle
Is
45
___
Minimum
Ground Clearance
(M)
1000
2000
1000
2000
1000
Jinkins
Included
yes
Although the individual one-second burst kill probabilities shown in these tables are low, they cumulate to
effective values if the weapons have enough ammunition on board to fire over the whole attack pan.
acceleration algorithm.
c. R
at on level path with constant horizontal
rate of turn. Miss distance should print out zero
with quadratic algorithm.
d. Repeat on straight line dive path through de-
lb
2&
2b
2c
450
300
30P
300
Yes
Yes
yes
No
20871455
10.39
VEOIT
280
11
.O
00
VELQCIVY
300
0o
(M/$SC)
*so
0
I.
@0001
0
100
0
II
O0
0O
10
tio
FCAStINKP46T
Is1
t6
in:20W
SAT
TIME (5EC)
20571.230
5000
4000
000
IMTEFIV
0o.
200
00
11
9V
RAIME
031
(METE20000
Figre1028
Rng
ithelipsida 6)
It.Repat
taget fuelge
lon.
wigbt loean
fseag
ad
in.Ths
Prfie
at
tvel esythadco
pe a adtina
ean
vriictin;
cnbe
orexmpe
evl
h
r-rnnngbustkil
taret
i cangd,ankange nd he eslt
b had
be coputaion
vlidaed
an
roabliie
wthliea
ad
anen
ansompt0
A00mo e s bl
0et0o u co
plfiaio i
hemaycoptaioa
10-41
n eie
lop
os
n0h
300
I,
250
go
200
Sg
10
0g
07122
100
Is
010
TIME (SEC)
the referable
prpri0a,.d42
method is ucioslfrmcnb
probably to use a recur-
Allwin
fw mnorresdua
fothposiblit
ofi esimaed y acominaionof nalsisan
b. Increasing
oweapon muzzle velocity has a high
5000
Iv
4000
RANGE
IMETE n$)
3000
'2000
!0
IMTAS
000
l.
B0
0i
0-
10
-L0
Sc.t sortranestheweapns
'
::
Regarding the simulation results, it appears that
d. No simulation results were obtained indicating
much can be accomplished to develop scaling relations,
singular dispersion& in excess of 5 mils were
which allow a given set of results to be extended to
Sthat
other sets of parameters. This can econonmize computer
Sdesirable.
In those cases where programming
S10-43 running costs. and more important, it allows a large
~~~errors caused faulty prediction, anurdipV
20671.233
against even moderately maneuvering targets. However this interrelates with rate of fire
and caliber and all design tradeoff's must be
in choosing the preferred set of
weapon design parameters.
SIconsidered
20
~TIME (SEC)
2-
Spayoff
is
I'
Table X-18. Burst Kill Probabilities With Linear Predictor and 3600 F/S Muzzle Velocity
2SMM
37MM
Path I
Time
5.0
6.0
.00
O0
Path 2
Path I
Path 2
.00
.00
.00
.00
.00
.00
.A0
.00
.00
.00
.00
.00
.00
.00
7.0
.00
.00
.02
.00
.01
.00
.07
.00
8.0
.01
.00
.00
.00
.4
.00
.02
.00
9.0
.01
-00
.02
.00
.02
.00
.09
.01
10.0
.02
.00
.01
.00
.09
.00
.24
.05
11.0
.00
.02
.10
.00
.00
.03
.17
.00
12.0
.16
.02
.00
.00
.22
.07
.01
.00
13,0
.06
.00
.00
.00
.13
.00
.00
.00
14.0
.00
.00
.00
.00
.00
.00
.0
.00
20871-S56
3
5
Table X-19. Burst Kill Probabilities With Linear Predictor and 5000 F/S Muzzle Velocity
37MM
25MM
Path 2
Path I
Path 2
Path I
a
5.0
.00
.00
.01
.00
.00
.00
.01
.001
C.O
.00
.00
.01
.00
.01
.00
.08
.00
7.0
.01
.00
.06
.00
.08
.00
.08
.00
3.0
.04
.00
.04
.00
.03
.02
.13
.00
9.0
.03
.00
.10
.00
.07
.00
.19
.02
10.0
.10
.00
.25
.00
.12
.03
.44
.03
11.0
.01
.02
.14
.00
.04
.04
.20
.00
12.0
.10
.07
.02
.00
.20
.13
.03
.00
13.0
.12
.00
.00
.00
.16
.00
.00
.00
14.0
.00
00
.oo
.00
.00
.00
.00
.00
Time
1
II
20871-537
10-44
II.
-25
,T
:?7MM SYSTEM
25 MM SYSTEM
""
20
>6
.15
I-
li
:1
,o10
II'
. -0KA
I I
.50
10
1s
2(*71.234
TIME ISI(.,
Figure 10-31
I,
-
20
.I'
10.45
VM - :000 FT/SEC
--
VM - OWFT/SEC
37 MM GUN
.30
-
.3
,.
-'
"
I'|
"
IL 0
.10
""0
I20
I'
A0
TIME
ISEC)
11
20671-235
!0.46
11
IOI
"
I"II'
0.40
030
"V
A
0.20
1:
vo
i(FT/SEC)
CAL
(MM)
sw
25
36W0
37
I
,
0.10
I0.001
10
11
12
13
-Mid
TIME
2C071.236
10 4
10-47
I
Table X-20. Effect of Prediction Mode on Burst Kill Probability
Defense of
Known Point
Time
Tangential
Linear
Quadratic
5.0
.03
.00
.00
.06
.06
6.0
.01
.00
.00
7.0
.01
02
.00
.06
8.0
.05
.00
.01
.03
9.0
.04
.02
.02
.03
10.0
.10
.01
.04
.05
11.0
.02
.10
.01
.03
12.0
.01
.00
.00
.00
13.0
.00
.00
.00
.00
14.0
.00
.00
.00
.00
.
20871,558A
II
!
I
I
!
10-48
Linear
Defense of
T angentialLinearKnown
File
Time
Point
Path
Za
Path
2C
PNth
2a
Path
2c
Path
2a
Path
2C
5.0
.03
.01
.00
.00
.06
.05
6.0
-0t
.05
.00
.00
.06
.07
7.0
.01
.06
.02
.00
.06
10
s.0
.05
.0o
.00
.00
.03
.14
9.0
.04
,08
.02
.00
.03
.12
10.0
.10
.09
o01
.02
.05
.15
31.0
.02
.08
.10
.18
.03
.09
12.0
.01
.00
.00
.01
.00
.00
13.0
.00
.00
.00
.00
.00
.00
14.0
.00
.00
.00
.00
.00
.00
20871-559
1I
-..
10.49/10-50
I
SECTION II
SUPPORTING ANALYTIC TECHNIQUES
I
f
It has been found helpful to have simple approximate analytic expressions for kill probabilities on some
of the simpler target paths employed in simulation
runs. This provides insight as to how the generated kill
probabilities are likely to change with parametric variations, and serves as a gross check on the simulation
program.
Some of these results are developed in this section.
They are presented for unaccelerated target paths, but
can be extended in a straightforward fashion to constant acceleration paths.
properly included.
If range passes through a minimum and then increases, the above integral is divided into two parts. up
to midpoint (D,,) and past midpoint.
Dp2
E = vf
Jp(Dp)IDIp dDp
fDp
(11.3)
,I
(11.4)
(11.5)
Now
tions.
where
On a specific course, suppose that single shot probability has been determined as a function of time. There
is no restriction at this point on whether successive
rounds are or are not correlated. If firing is conducted
at a constant rate v over a time T, the expected number
of hits is
-
Xp2 . D 2_-D2
P"
p m
_ p(t)dt
p
:o
(11.1)
where
vr
projectile velocity at D
Then
* over a
I1-
S.
. .,
(11.6)
;-
;'"
, Il-I
, ,
'
F -t-
I I I
I I I
(11.7)
11202
,
2
2&'A
Dft1
n II pl( : 0V)J p(D )dDlE
2
V 1
02
204D
2)U2,2
DD(a
1.
I1. 12)
Di 2
IDp,/(xpV)l p(Dp)dDp
.Dp
Ppared
[p(Dp)/vJr dD
+fp 2
(11.8)
Dp
a2 < 22
2
D<
(11.13)
a
0
E = (Va2 6)(VD m2)
tan-i xp/Dm
(11.14)
E t 2/)Dp
(11.9)
The angle
e is
a2/(112+202Dp)
(1.0)a
where
Av
V -
1.0 meter
310 meters/sec
H -
200 meters
R. -
i1a2
a,
4 mils
Interms of xp,
E -
Xp(
( o
E
p(xD..)d
0.85 (kill-seconds)
(11.11)
E -
0.8632 (kill.seconds)
11-2
-..
......
....
......-
- ----
.....
--- '1
.J..
..
!I~i
1.1.4 EllipsoMidal Target
11
We can extend the above computation to the case of
an ellipsoidal target, with area A, - ir a! head on, and
- adeo.Thn
_2 2
p IDm
Xp
2
)
a Din2 /Duration
22
I(a
=2
0+P
tp+(Ts/2)12
'a%)sin0cos0+(as+af )0O
2V 02 Dm
can be obtained from 0o' and T, from the filter coefficients, or graphs such as those presented for least
squares finite memory filters in AFAADS-1.
The resulting expression does not integrate easily,
but a workabie expedient is to exoand p,, as a series in
x.. As an example, consider the case of a circular
target. 'circular' round to round dispersion, but tracking error variances which differ in the two coordinates.
Then the denominator of p, can be expanded as a
series in y. where y - x,/D,.. One then obtains,
retaining terms up to
E
E
E - 2M/QI/
dy
A + By + Cy(.21)
2 tan"
(QI/ 2 /B)
(11.22)
(1.18)
Q - 4AC -B 2
'
(11.17)
AwH/Dp
where:
H
9,
(11.23)
IProbability
VE
pit ...
Au
2" 2(OuD)*
2
.
2
...
2 1/2
0p ,* 1
2 +(0,D)" 2 2
s1%
0.09toO.11
0p,"2 1/2
(11.19)
"11-3
S....
.1
Parameter
y
Case
Arbitrary p
0 -I
E"A -)
ures 10.3 and 10-4 suggest that for the system and
noise parameters and algorithms represented in those
results, single shot probability at any point in the
defense volume can be written simply as:
)/,
I--
0'r
lnr.ter
Isa f(Dx).CIXY
(11.24)
where r(D.x) is a term depending on the target posi.
)
1/2
[L L2/2
X2
<<
-07
20871.560
ffe-eo
Y d(eY);E"
nf;n-number
Of rounds
I
If agreement with the simulation is still obtained at
this level, the effect of those target maneuvers which
can be expressed as a constant target acceleration
during the smoothing time and the burst length, can be
handled analytically.
(11.25)
If this computation, f(Dx.x) would be computed from
an expression such as Equation (11,19).
11.4
_
"
assumed. One would expect qualitatively different resuits with an accurate simulation of manual tracking.
-i
2 e-.-
;
2s
1(11.26)
(11.6e)
IA2;A
It
the p
artlett
Furtherand
comments
Bartlett's
paper are
in
discusses
the given
general
Jenkins
Watts." on
Krut'ko0`
Both Mehra' and Jazwinski provide a straightforward method of obtaining the desired result, and it is
their method which is outlined in this paper. Their
method allows the computation to be performed for
time varying systems; for sim licity only linear non-
MJS
()
p(s)=
"I
dX z F X dt + G d3
(I 1.28)
wFs
where X is an n vector. F is an n x n matrix, G is an n
process with:
<do do T > a Q dt
(11.29)
is sampled
If x(t)
quence
can be described as generated by the difference
equations
X(j+I)
(11.30)
~.1
where
(Q+ I j) is obtained
d(u/dt a FO :I(O) a
(11.31)
Solving
andt is C(S).
11
012
-1
w(j+l) a
s)G(i) d
L2
12 a
aI
o (11.32)
00012
021
022
~( 1
.37)
anod
<wO+l) wO+l)T>
f 11 (s)GOs)Q(s)GTls) 4tT(s) ds
e(s)
"0Se'
0
*2
O'2eals-e22
((11.33)
(1
11.2.3 Example
x2
x2 (t)
(11-34)
(I +sTIXI + sT2)
(1134)
[I
Nal
I1
(11.39)
Where xo(t) is loosely specified as 'white noise' with
bandwidth greatly exceeding that of the system. and
specified intensity N2
This expression can be rewritten in the stat. space
form.
Write:
0(1401
A(s)
()Bs
AsB(s)
~B(s)1
Then:
T i t 1 +X I
II x 0
T2 i 2 + x2
LxA
( 11 -35)
x
.,
0[0A
t * I 1A
1 of Ii It
l.x l
,1
I."
. [A~AI0
A Mu
( AA, 0
-114"11
A
"Ak
[Au, 2,.1 IJ2J,
LA,.I Auu,-,4.,
4A
%2.11 [ qA
+
U2
"
III
4A
[00]
'lAB
"A N
'1A 'qAD
tq
111411
(I i.36)
11.6
+ 1)) 2 >
+ 1) w2 (J
- N2a 12 qA
I)> = N2a12
qAB)
xl(J+ 1) = A(A) x 1(J)+wI(J + 1)
II+
) = [Aa)-B~a)]x
1 o)
(11.42)
1B(A1)x2(j)+w2 O+
1)
o 12
(11.43)
1.46)
we obtain:
Since<wlP
<=
l)Wk(j+1+m)>
0o m
O,
0 we
w 2 (j+2)
S genc+at wi as)
ca
ca gA w2 0 + 1)
=+(A'B)
w 1 (j+ I)
c2
el
n2J I
(11.47)
(I 1.44)
j-
e d
A(A)
"J
( + I
vY) = b 2 n (j + 2) + bln(J + 1)
(11.48)
r0
[exponential
A(a) - B()
x20)
their complexity.
11.2.4 Comments
+(11.45)
S|
+is
"varying
j
i1-7
x(k+ 1)
4(k)x(k)+ r(k)w(k)
(11.49)
-0 0(1.0
(11.50)
(11.51)
M u r(k) N(k)T(k)
(11.52)
I
I
11.3.1 Approach
Define:
m anipulate.
lk)
4 /*~ l '
+I}
q l[lk
;k I)
.
El, 5h
11-8
00.
IMlkI
L.
where
(fx(k+i)1
(I11.54)
04)Wkx(k) + Xs(k+l1
i(k+1
0 p[x(k+l)Jx(k)]fJx(k)]q[x(k)] dxk
where
fix(l)
Let:
-:
plx(l)I
Then
qJx(k))
Q(k)
fx(k)qx(k)J dx(k)
(11.59)
0 f~x(k+l)I dx(k+l)
(I 160)
or, alternately
Sis,
given the x(k)
n
n- qlx(k)]
(11.55)
k-I
We need to average this over all x(k) to obtain Q the
average probability that the target survivcs all rounds.
Q(k) - Q(k-l..-
pl
S
,,3 x,,.I
. i.
..
ll pIi ,,l
,=-I'Ipl
end~~
S-,transform
. ),,-i
Ps[X(k)J
1 -q[x(k)j
(11.62)
qx(k)].
.__
flpreferable
(11.61)
ftx(k)IriJx(k)J dx(k)
PIX(i I)l
I II
11.9
A
11.3.3 Comments
The principal interest of this approach is that it joins
the state space system formulation with the value
representing target kill probability in a manfunction
ner which allows the problem to be integrated with the
mathematics.
I
I
II
I
II
II
I
I
14
'I
I1 1
SECTION 12
COST CONSIDERATIONS
This section provides an overview of costs for anti-
are so numerous that they will not be cited individually. They were given in in-process working papers. It
is probable that regardless of the publication or newspaper clipping in which the present writer found the
was the
almost all cases
source in Congressional
original
data, the
hearings.
Army
budget
and associated
A
Without details of where in a production program
costs are noted, or what spares, test unit, or engineering change costs are included, the data cannot be
considered to be accurate. The few indications given of
quantity are particularly unreliable. In spite of these
deficiencies, a reasonable amount of consistency wijl be
12.1 INTRODUCTION
atiarcrat
hve
the defense
place iin thedefnse
have a plae
guns
If antiaircraft Ifgns
arsenal, one of the reasons must be that throughout an
important defense volume they provide effectiveness
equal to that which can be furnished by guided missiles, but at substantially lower cost.
In theory. at any rate, a missile system can do
anything a gun system can do if one is willing to pay
the price. However, guns have certain intrinsic advantages over missiles with regard to reliability, ease of
maintenance, ammunition cost and overall logistic
support costs. These advantages tend to become submerged in the cost of the surveillance and tracking
sensors which are common to both types of systems,
but can be capitalized on by careful design.
Over many years since World War 1I, dollar expenditures for research and development in the United
t a muchlfasertrat tha te
States hav resea
States have increased at a much faster rate than the
engineerin
te phys ocines.
engineering and related physical sciences.
The consequences in terms or the cost of one technical man year in research and development are shown
noted, however, that the starting salaries of new engineering graduates who obtain employment continue to
andgrowth.
might be expected at leaitt to parallel the
increase,
price index
ehas
appeared in the open literature. Because of limited
time, the data presented has been limited to equipment
There is no reason to believe that the rate of inflation will ever become zero regardless of which of the
12-1
600
400/
300
INDEX OF P& D COST
FOR TECHNICAL
MAN YEAR
200
INDEX
100
il
1
1
70
10-
1920
1930
i940
1960
'960
1970
19
9
CALENDAR YEAR
20671.1229
zJ
"I
7/A.
0-mm
Model
Skysweeper
Vulcan XM167
(ib)
20,000
313,000
3,150
52,000
""
2 x 20mm
Rheinmetall
Colt
(dollars)
3,200 (firng)
Cost/Pound
Quantity
90.000
190.000
78
120
31
67.000
1,670
Yewa
($/Ib)
1954
15.50
1968
1968
1972
16-50
25.50
61.50
1972
21.00
for
4,600 (travel)
1974
delivery
20 mm
Vulcan XMI67
26,000
Armament system
XM-741 Vehicle
'Iire Unit
235 mm
OerlikonlCmntraves
PFZ-B
26,000
80,000
97,000
III
1968
209.000
32
1972
56,000
81,250
III
32
1968
1972
153.000
111
1968
15.90
290,000
32
1972
11.00
500-
1972
for
13.80
to
1.100,000
to 1,400,000
600
1974
delivery
t12-3
17.30
Caliber
(million dollars)
Application
25 mm
GAU-/A
lI15
40
(development)
35 mm
GAU-8/A
AX
35
Notes
41)
About halt of the GAU*71A program cost will bc devoted to ammunition development And test. Cascicss
ammunition weighing one half as much as conventional cased cartridges will bc employed. A shoot-out
between GL and Philco-lord is underway.
(2)
GAU-SA will fire aluminum cased ammunition with projectile wcitht about one pound.
Two contractors received firm-fised-price contracts for 3 guns each for a shoot-out at the 0 K Test Range.
They were
GE
59,500.000
$I0.000.000
Philco-lIord
Weight
20-30 mm
ISO Ibs E
Application
ARSV
Quantity
7.000-15,000
:O,O00-1 2,000.
MICV-70
Notes:
I)
Three ontractors to deliver two guns 8.800 rounds each for shoot out.
GE (27 mm)
$1.9 million
$2.1 million
$1.7 million
quantity
much cost reduction is achieved in very large
production.
To broaden the data base. Table XIl-5 summarizes
unit costs on machine guns ranging from World War I
weapons to current designs. The relatively high cost of
the M73 may be associated with a variable rate
capability.
S 4u million
S180 million
"1
11
Ii
Model
6-bbl Vulcan M161
M61A l
IWClsht
(Ibs)
Cost
(dollars)
Minigun Pod
2 x GAU-2B/A guns
058"
Coi.t/Pound
($11b)
18,460
65
1968
700
255
9,900
498
1968
390
10,000
200
1968-72
390
18
96
1458
1968
1500
26.500
117
1968
27,000
351
1968
XM-35 Helicopter
Installation
7.62 mm
Year
265
yearly average
Barrels for M6IAI
Quantity
35/gun
20,400
5.30
18
1966
1863
f
3
12.8
: TANKS
Present practice is to use a tank chassis as a carrier
for a self-propelled antiaircraft fire unit. The more
applications that can be served by a common chassis,
the larger the production run will be, the lower the
unit cost, and the simpler will be the field problem of
11
Comparing the cost of a tank with the cost of a selfpropelled antiaircraft gun system, such as the Oerlikon,
the incremental cost associated with the antiaircraft
sensors and fire control becomes clearly apparent.
12-5
Date
1937
Hispano Suiza offered its Birkilt Type 404 20 mm gun for S3500 each, including mount.
magazine- and tools.
1937
The United States bought otte gun with 2000 rounds of ;mmunttion. tool kit. magazine.
and mount 0fr $8D000
1939
The United States secured production rights to the weapon. and it was produced as the
ANA-2 with the following production cost histoi).
1940 1943
Co--
Manufacturer
Quantiq
First Lot
Last Lot
1 120
458
22.642
Oldsmobile
910
5 10
77.010
Int. flarvester
840
465
24.526
905
565
10.500
Bend,\
Trucks serve as tow vehicles for some towed weapons in an antiaireraft battalion, in addition to being
p;esent in quantity for general tactical mobility support. Throughout the sixties, a fair estimate of the cost
of a truck was about S0.50 to $0.75 per pound of
weight empty. The cost per pound has about doubled
for current conventional wheeled vehicles and a somewhat greater increase has been experienced in the case
of :he articulated Gama Goat design. These costs are
summarized in Table
Xil-20.
12.10
ARMORED RECONNAISSANCE SCOUT
OR
,EONAAUSAT
NCESCOUT
The current M 114 Armored Scout Vehicle is to be
replaced by the XM8O0 for which weight and cost
oojecti;cs have been set in 2n L1u.:crity plan. The:m is a
substantial anticipated cost growth. oY.er the Ml114 but
with greater capability. The M551 Armored Reconnaissance Airborne
Assault
is .hov.n in the
sametabe,
i athogh
ha Vehicle
a itfeentmisson.The
same tahh:, althoe;h it has a different mitioan. The
Sht;idan has the weapcat effectiveneu of a tanl,, hitt
lacks the tank% armor, in order to achieve airtorne
capability. But in spite of sophisticated armprrent it
works out to about the same cost per pound as t
projected XM800. Costs are summarized in T;,.le ,4i.
18.
I
1
Model
Caliber
6.5 mm
Weight
Cost
(ibs)
(S)
Cost/Pound
Quantity
Year
(S1b)
25.5
320
10,000
1914
12.50
"792mm
60.0
1.620
3,000
1917
27-00
7.5 mm
18.3
28
11,000
1931
1.50
(Fyench)
-?
0.30"
M1919A6
0.30"
71.62 mm
M60
32.5
23
188
1945
445
1945
1.700
1.500
500
544
560
577
550
2,857
15,760
3.478
15.031
12,056
850
175
495
670
Navy buy
7.62 mm
2,300
3.400
3.400
M73 coax
0.50"
Unident
0.50"
M85
0.223"
Mk 23 (U. S. Navy)
so 1)
4.000
6.000
882
975
33(7)
1958
1962
1963
1965
74.00
61.00
22.00
24.00
1966
21.00
1967
1968
1970
24.00
25.00
24.00
1972
37.00
1971
5,000
4,700
3e0
1.266
1.050
20
29.00
1969
1970
1971
1945
1950
249
720
5.80
3.10
1970/71
'.970171
IQ72
20871-365
Sroughly $15,000
good.
12.7
Model
Weipht
Jibs)
0 30)
('arbin MI AI. 2
5.20
0 30
9.50
Caliber
Year
Cost/Pound
4S/1b)
35.50
64.00
1945
1950
6.80
12.30
35.00
1945
3.70
41.20
1945
4.30
64.00
94.30
1950
1957
6.70
9.90
1959
1960
1963
1969
15.40
13.20
13.00
13.80
1945
1950
6.90
18.40
Cost
(s)
8.70
133.00
115.00
112.00
120.00
19.40
134.00
358.00
Q0tantity
60.000
120.000
7.62 mm
Rille i14
0.30
Browning Automatic
Rile (BAR) %1918A2
556mm
Rine Ml6
7.4
106.M
600.000
1970
14.00
M16E 1
(Air ForCe buy)
a
Pfogram Average
7.6
107.00
116.00
254.000
65.000
1971
1971
14.00
14.00
128.00
3.200.000
1700
1971
94.00
12.40
20871-566
type of equipment.
12.13 SENSORS
It has not been possible to make an extensilt summary of sensor costs. In this field published or inferred
than in the
to beas widely
likely such
more items
aresimple
costs of
example.
forerror
trucks, in
case
With this caveat lector. we note some pubiished costs
for infrared and image intensifcation sights in Table
12-g
i.
t~I
Table ,I-7.
Caitber
lfinches)
Model
CoMt of RU,
Weight
Cost
(Ibs)
(S)
Cost/Poutid
Quanltt)
Year
(S/b)
6.80
81.652
1862
0.70
10.00
2,181
1862
1.00
0.69
Prusstan
0-69
American
058
9.50
18.20
22,793
1862
1.90
0.58
Sprrolfield
9.90
19.50
6? 1.000
1862
2.00
0 577
Fnfeld
8.91
18.40
428.292
1862
2.05
0.52
8.75
36.00
9.141
1862
4-10
0.52
Spencer !7-%hot)
10.00
37.20
12.471
1862
3.70
0.44
Henr) (12-shot)
9.25
36.50
1.731
1862
4.00
. .
20871-567
A more stable reference may be the number of manhours of effort involved in each case. Manufacturing
S12.14
12-9
,.
Caliber
105mm
Model
Approximate
COst
Weight (Ibs)
(dollars)
Cost/Pound
Year
"
(S/Ib)
M2AI
5000
8,260
1945
1.70
MI01AI
5300
13,670
1950
2.60
MI
12,800
15,640
1945
1.20
Mi .4AI
12,700
26,410
1950
2.10
MI102
XM204
--
|5mm
XM198
203 mm(8")
M1Is
280 mm
1
3
30.000
170,000
437,000
1957
2.60
20811-568
th
iclud ing
st ru te, including
we
n
ppor aa weapon
n u support
and
structure,
allal the
personnel in the weapon's slice of the Army from the
a
o
High reliability and ease of maintenance are obvious
objectives, but the reduction in number of men to
maintain a given effectiveness level is an Army-wide
problem.
I1
12-10
li
Im
Approximate
Cost
Weight (Ibs)
(dollas)
M2A2
52.000
50,400
M4
40,000
Caliber
Model
Weapon
105 mm
M7
M37
($11b)
1945
1.00
1959
2.00
1963
2.30
1959
1.20
115.150
1970/71
2.10
128.100
1970171
2.30
1967
1.70
1960
2.30
103.000
275
I4S,000
M108
155 mm
M41
SO.O00
Ml
44,000
TI96
62.000
M109
Year
53.000
M52
T1 95
CostlPound
Quantity
203 mm(8")
M43
54,000
M2
MI I0
500
83.000
56.000
41,415
Vehicle
only
155 mm
M40
M2
83.000
Unident
175 mm
M107
140.000
64.000
146.000
115
41.000
Vehicle
20871-569
I
1
12-11
i..
I-
__ j
.,
--
I
Ii
Caliber
(inches)
Moel
Gun Weight
(Ibs)
Cost
(S)
Quantity
Gun Cost/lb
($fib)
Carrialge
Cost ($)
Quantity
312
15511
34S
686
S 3.67
4.62
6-pounder
854
400
152
0.45
12ipounder
1757
570
1127
0.32
j"
4.62
12.poundtr
788
385
73
0.49
5.82
24-poundef
1318
425
5e
0.32
440
25
875
689
10
0.78
600
440
25
0.73
6-poundef
1
I
(bronze)
3.67
12-pounder
(bronze)
2.S6
6.pounder
Wisfd (steel)
3.0
10-pounder
Wiard (stel)
790
720
20
0.91
3.0
10-pounder
890
205
587
0.23
S3.67
20-pounder
1750
39%
338
0.22
4.20
30,pounder
4200
557
392
0.13
6.40
IO0-pounder
9700
1304
237
0.13
670
248
8.00
200,poundei
16,300
2244
90
0.14
794
88
10.00
300-poundet
26,500
49371
40
0.19
1656
32
I.
2087t-570
I:
I'
Ii
=
i|
12-12
U!
MOIS
Model
M24
38,000
T 41
26.000 ()
SM
41
46,000
Cost
(S)
Year
Cost/Pound
($/tb)
39.640
1945
1.00
126,030
19S0
4.90
94,420
1959
2.05
20871-571
Model
Weiiht
fIbs)
Cost
(S)
M4
67.300
M26
Year
Cost/Pound
($/|b)
54.836
1945
0.80
86,000
81,324
1945
0.9S
M46
98.000
197,427
1950
2.05
M47
8,000
240.000
1952
2.70
M49
100.000
120,000
19S6
1.20
133,000
1957/59
1.33
109,142
1959
1.24
115,035
1959
1.IS
1959
1.30
1963
L.s0
M48Ai,2
U
3
XM60
102,000
138,000
M60
108,000
162.400
Quantity
(600)
180
203.780
507
1967
1.90
M60A i
225,330
300
1970
2.10
M60AIE2
273,000
1970
2.45
$00,000
(Gst)
1967
6.20
MBT-70
880,000
(est)
1970
11.00
XM$03
620,000
(sot)
1970
7.80
1971
8.00
MBT
$0.000
3000
649,000
20U71.572
12-13
i I I I
Sl ~
Model
Infantry Mk I
Cruiser Mk I
Matilda
Cruiser Mk 11
Cruiser Mk III
Cruiser Mk IV
Covenanter
Crusader
Valentine
Churchill
Cromwell
Weight
(tons)
Basic Cost
(L)
Yewr of
I rst Order
Cost/Pound
(Suib)
1I
13
25.75
14
14.25
14.75
15.55
17.53
IS.6
38.5
28
6,000
12,710
18,000
12,950
12,000
13.800
12,000
13.700
14.900
11,150
10.000
1937
1937
1938
1935
1938
1939
1939
1939
1939
1940
1042
550
970
700
930
840
940
755
780
960
290
360
1.35
2.40
1.70
2.25
2.05
2.10
1.85
1.75
2.10
0.60
0.70
NOTES:
(I)
(2)
A great deal of rework was required on the Churchill because of defective manufacture. This cost
is not included.
(3)
The basic cost quoted is based on contracts, and probably does not include "stores of free imu-'" (GrE).
"20871-573
Cost/Pound
(Slb)
Model
Country
(Ibs)
(S)
L.opard
Germany
88.704
232,000
1967
2.60
252.000
328,000
1970
1970
2.75
3.70
Quantity
3588
You
uI
AMX.30
Frncs
79,000
312.000
1967/70
4.00
Chleftain
England
117,000
308.000
1967
2.65
244.800
1970
2.10
und,,
1970
3.50 (?)
AE S
STRV 103
1weden
(S.Tank)
83,000
350.000
a
_________
___________
12-1,4
201171-374
-o
~Cost=.
Weigeht
(tons)
Model
Country
Renault
land
F'rance
28
7
England
37
IMkIV
oe
Year
Cot/Pound
5.000
,2 0
24350
10,950
094
1914
0.43
0.78
86.400
1920
1.17
0.48
I .T. 17
Mk VIII
18.000
1.3
"7.350
35.300
500
2,400
1928
0.92
Cardcn-Lloyd
England
Mk I
England
II
i,500
7,400
1935
0.34
Renault R35
Irance
II
2.200
10,800
1935
0.50
Vl
Chulchill
England
38.5
15,000
60,450
1941/44
0.78
M4
U.S.
34
10,300
4i.500
1944
0.61
16.600
67.000
1.13
,wM41
U. S.
26
27,200
76,200
1953
1.47
M48
U.S.
49
47.000
132.000
1957
1.35
Centurion
England
50
40,000
112,500
1957
1.12
20871-575
I0-
MT 70Q
0.
lo 1
PS_.f"o
-"o
00
10
~~~0
,I
'I1910
010
1930
192
940
low0
IWtO
12-15
""I
'970
1900
Model
Cost
Quantity
M3
19,000
12.294
M75
36,000
100,000
Yeaw
1740
75,000
41,000
38,000
M59
1800
M113
20,000
M113A 1
0.67
1952
2.80
3954
2.10
1958
1.10
1959
0.90
600
1959
0.80
31,000
600
1959
1.54
28,000
1963
1.40
20-22,000
1969-72
1.10
1969-72
1.5
1971
1.00
22-25,000
M113
1944
29,602
35.000
T113
Cost/Pound
19,955
1125
40,000 Averge
3000
1
1
2.00
(all costs)
Total Ptopam
$120 Million
XM733
35-38000
100,000
(MICV-70)
2500-
1972
2.70
SOO E
Goal
(MechaniW
InfanUy Combat Vehicle)
20871-576A
Table XII-17. Coat of M113 Derivauve Vehicle3
Vehicle
M546:
6T Catlo Caries
XM730
Cost
Quatty
Year
25,000
(basi)
120
1971
29.200
116
1971
chaparral carrier
12..6
";'
Ix
20171-S77A
12-16
''l
1 ,
m'
!W !
,.
----
mr
:'
-I
"'
"I
Model
Weight
M114
15,000
I(ARSV)
XMI1OO
Cost
Quantity
38,000
17.000
116,200 Objective
SOOO E
33,500
(ARAAV)
213.600
______
1662
_________
______
Cost/Pound
1971
2.50
Est. as
of 1972
175.000 Possible
Sheridan
Year
6.90-10.20
1970
6.40
_____________
20871-s578
SMajor
Unit Cost
Component
Vehicle
80A400
8.800
Engine
Tiransmtion
"Gunlauncher
9,000
12,300
Fire control
12.500
25.200
Machine gun
4,400
Searchlight
2,500
Communications
Night vion sight
2,700
2,400
1,600
45,200
2,300
Averae unit price
(I)
213.600
12-.17
L- , ,1
11Ir....
""
- J!
..
-I
iI
Type
1/4 Ton
4x4
Model
Empty
Cost
Jeep
2.400
1,057
1944
0.45
M38AI
2,800
2.000
1953
0.72
MIS)
2,250
3,300
1959
1-46
M15IA1,2
2,400
3,180
1968
1.32
3.242
1969
1.34
3,476
1970
1,44
3,504
1971
1.46
1966-67
0.67
1966-67
0.85
314 Ton
4x4r
M37
1 1/4 Ton
M715
5,700
2 1/2 Ton
4.720
XM.70S
S Cost/Pound
Year
3.400
3,790
5.500
Quantity
5,400
18,000
1970
1.00
8.337
3,452
1971
1.50
M35
13.500
3,352
1959
0.25
M109
I1,000
6,375
1953
0.42
6,650
1955
0.44
7.131
1957
0.48
8,363
M54
S Ton
6x6
1 114 Ton
40.000
M561
6,550
66Gam
1/2 Ton
3.600
19S8
0.67
17,434
S17.196
1969
1970
0.85
17,381
1971
0.85
1970
2.05
1971
1.80
1959
2.40
13,281
ol
15,274
11,722
M274
900
3
1
0.56
13,200
XM 809
3
1
2,150
2.000
0.85
Mach Mule
20871-S30
Ii
12-16
i II 1 I /
I1
k"i
::
|-r
....
=I
i"
::::- 'K"
!
I
Table XiI-21. Cost of Rifle end Machine Gun Cartridges
Caliber
Model
5.56 mm
Weighl
Cost
rtains/lbs
All types
Ball
0.30'
55/.0079
1971
0.060
1971
7.60
0.11
1959
1S2/.0217
0.09
1959
Tracer (Carbine)
107/.0153
0.06
1971
3-90
Bail (Carbine)
I10/.0157
0.07
1971
4.45
0.IL
1959
BaU
IS0/0214
0.087
1971
7.62 mm
3.70
4.05
Linked
0.10
1971
NATO
0.095
1970
0.10
1972
0.325
1957
0.28
1959
0.30
1970
0.38
1971
3.80
0.39
1972
3.90
Ball
70010.10
i"
0.060
S/lb
BaU M2
Tracer
0.50'
Cost/Pound
Year
3.25
3.00
I20871-581
Weapon
Proj Wt
Sains (Ib)
Cost
$)
Quantity
(millions)
Cost/Pound
($/Ib)
0.58
550(.078)
.015
46.4
0.19
0.52
475
068)
.021
16.3
0.31
0.52
385 (.055)
.024
58.2
0.43
216 (.031)
.023
4.6
0.74
S0.44
1I
20871-582
4I21
,,
U.l
,
i!"'
[-
"Weapon
Hispano Suiza 404
Weight
(Ob)
Cost
15)
0.28
All Types
Vulcan IUSAF)
MI 39tHispanoSui.-4820)
Cost/Pound
($/Ibl
1917
7.20
4.00
1937
14.40
1939
8.50
1.30
1970
5.90
3.58
1971
7.20
1.83
1972
8.30
59.5
2.38
VulcanlCobra
Year
2.00
Inert (BIall)
HE
Quantity
thoutands)
0.22
Training
0.22
0.93
1972
4.20
Combat
0.22
2.11
1972
9.60
3.40
1970
TPT M4206
0.25
3.90
1971
7.10
1970
28.40
5.90
1972
23.60
207l1-583
3
3
Type
Cost
(Ib)
(S)
33
26.00
1951
21.50
1972
1972
1972
HE
HE w/o fuze
155 mm
Projectile Weight
Model
M1
33
Cost/Pound
Year
HE with fuze
Illuminating with fuze
M314A3
28.00
53.00
WP
M60
53.00
1972
HE
HE
M107
40.00
30.50
1970-72
Illuminating
M485A2
75.00
1971-72
ICM
XM483
310.00
95
95
1953
(S/Ib"
0.80
0.85
0.42
0.32
1971
Charge
Green Bag
M3AI
13.50
197C-72
White Slag
M4A2
29.50
1970-72
60.00
1970-72
HE
M1306
""CM
M404
200
265.00
1970
Green Bag
M1
20.00
1970-72
Bag
M2
29.00
1970-72
White
0.30
2087-5814
12-20
___
Model
Projectile
Weight (Ib)
Cost
(S)
Year
Cost/Pound
($/Ib)
1.40
Caliber
Type
90 mm
HE (Fixed)
175 mm
Unident.
HE
w/o fuze
M437A2
Charge
M86A2
Navy
RAP
385.00
168
Navy
RAP
350.00
1970
5-38
15354
M71
23.4
33.00
1953-57
94.00
68.00
1960
1970-72
78.00
1970-72
20871-585
Caliber
76 mm
9c, rr.,n
105 mm
152 mm
Model
Projectile
Weight (Ib)
Cost
(S)
Year
HE
21.30
1959
WP
23.90
1959
TP-T
19.20
1959
Blank
8.30
1959
HE
32.20
1959
WP
39.50
1959
TP-T
25.90
1959
HEAT
94.00
1959
Cost/Pound
(S/lb)
3.30
IIEAT-T
M436A2
21
75.00
1970-72
HEP-T
M393A3
24
53.00
1971-72
2.20
APDS
M393A2
13
155.00
1970-72
11.60
TP.T
M490
49.00
1970-72
TP-T
XM411 E4
106.00
1971-72
20871-586
12-21
Model
Application
Cost (dollars)
All types
Hand grrenade,
0.50
Practice
MT (Mechanical Time)
MTSQ (Mechanical
VADS Ammo
3.00 ()i
MS24
81 mm Mortar
8.00
M572
Howitzers, Guns
5.50
MS65
Howitzers
16.00
M564
Howitzers
1700
MS 14AIE I
Howitzers, Guns
45.00
J
J
Time. Superquiek)
Proximity
Note. The World War II Proximity (VT) Fuze for Antiaircraft Guns had an initial cost of $732.00 each in 1942 at the
beg innin of production. By 1945 the cost per ruze was down to $18.00. The total program cost was $ 1.01
billion, and over 22 million fuzes were produced.
20871-587A
Type
Cost
Quantity
ANIPAS-5
Binoculars
350
6300
AN/PAS-6
Rifle Sight
2000
AN!TVS-2
AN/TVS-4
NODMR
I
1
Year
1967
380
3200
1970
1000
1748
1971
j
20871-588
Price
Quantity
AN/VPS-2
26,000
228
25,000
36.000
60
267
35,000
48
Dale
to 1968
I
2087 1-359
12-22
Application
Model
AN/GVS-I
Development Models
ANIGVS-3
Tripod Mount
AN/VVS-1
M5S5I Sheridan
Quantity
110,000
10
53,000
200
Year
Remarks
S,600 E
S00
1972
6,100 F
1000
1973
1967
prototype
462.000
S11,000
243
1969
14.600
176
1969
3.300
300
1970
47,429
176
1971
20871-590
I
I
Cost
Model
SCO
AN/SPS-39A
AN/SPS-40
AN/SPS.48
AN/SPS-S2 (FRESCAN)
AN/SPS-5 8
18,000
1,000,000
190,000
500,000
4.000,000
610.000 (4 proto)
1945
1963
1965
1972
1972
1970
293,000 (6 prod)
1971
20871-591
12.23
'
p i
L Il1
l '
1[
['
el
e !
,~
"
.1
Table X11-32. Cost Of Transportable Radars for Air Surveillance and Control
T) pe
Cost
Year
Unident
384.000
1945
Unident.
2,500.000
1963
2.700,000
1972
AN/TPS-32
AN/TPS-34
For MTDS
100,000
,I
1969
AN/TPS-43
For 40 IL
1.030.000
1970
AN/MPQ-43
(HIPAR) for
1,200,000
1967-70
Nike Hercules
AN/MPQ-49
(FAAR)
7S.000
1971
20e, I-59'
SI
Is
20-
z
20
10
l0
116
.I
210
so
100
20871.1231
12-24
N
Al
I
I
330'Le
so
<
-i
0 0
00
S2o
z0
4
I10
0:
z0
B.
5i
10
20
50
100
200
WT(LRB?
Figure 12-4.
20871.1232
i
Table XII-33. Cost of Army IFF Equipment
Cost
Model
Application
AN/TPX-46
(5)
Quantity
100,000
109
Set
AN/TPX-50
21,500
23
22.500
156
20871-593
Type
AN/AAD-4
SAN/AAS-24
Cost ($)
USAF (FLIR)
550,000
Army
5S0,000
220871-525
12-25
-l
I
I
Cost
Number of
Machining
Weight
Components
Operations
(Ib)
Cost/Pound
($S/b)
Wcrpon
Oerlikon
321
1400
250
3000
136
10.50
"Arsltn
60-70
300
119
900
121
2.50
20871 -95A
'
Manufacturing
Weapon
Man-Hours
MkI
MkV
Bofors 40 ,nm AA Gun
I I Icaly)
5 (late)
12
2420 (teally)
1300 (late)
Field Artillery
Anti-tank
25 pdr
2 pdr
6pdr
17 pdr
308S
2683 (prewar)
1293
2726(1942)
20871-596
1 -
I
I
I
I
z.26
t
K
I
-000
I
7000
5000
ENO OF WAR
4000
I(
S:EPARATION
PAYMENTS
cc
3000
wZ
z
W
1000
--
WW~ 11DRAFT
9~001
I
"I
I,
w4
1920
1930
1940
I91
1900
1970
low0
YEAR
' ''i
20671-1233
'
Figure 12-5. Average Pay and Allowances per Man Versus Time
12-27/12-28
[ II
' 41
SECTION 13
ADDITIONAL DATA REQUIREMENTS
There are two principal determinants of the effec-
(I)
(V2
Normal operation.
I
c. Laser
(1) Rg
counter the always prevalent general opinion that prefire weapons are inherently inaccurate.
In fact, even since World War I, the accuracy of
odicted
( aeonly,
predicted fire antiaircraft weapons has been renmarkable. It has. moreover increased steadily with time in
the case of those systems which have been well designed and engineered,
It is believed that the air defense effort now and i'i
the future would be well served by about a one manyear effort devoted to recovering the test data on past
predicted fire systems from the archives and presenting
a concise summary, showing what was achieved.
If the files of test results at the Air Defense Board
and School have not been purged, they would constitute a gold mine of historical data. Frankford undoubtedly has a great deal of information. The Navy
has excellent data, if it can be made available,
A simple set of curves of predicted fire accuracy vs.
time with a long time t-ase would do a great deal to
b. VTOL aircraft at very low velocities (to deterwine the limitations of pulse-doppler radars).
c. Air to surface missiles (some information on the
ability of radar to track very small targets can be
obtained from counterbattery radar data).
I
I
I
I
"I i
Ii
13-2
4L
SnIi
1W--
--
--
I
I
SECTION 14
TEST AND EVALUATION REQUIREMENTS
It is too late to begin a test program when the first
prototype has been completed. The following brief
outline sketche% some of the considerations to be included in test planning. The plan should be completed
at least in preliminary form when system development
is initiated.
SYSTEM PERFORMANCE
The principal shortcoming of almost all military
field experimentation programs is the limited analysis
of data which is carried out. The usual cause of this
analysis.
partial remedy is to provide instrumentation and
associated data computational capability that will analyze the experimental data in real time, with a printout and summary available after each experimental
run. This approach is limited by the foresight of the
experiment planner in deciding beforehand what data
should be taken and how it should be analysed. However this is an easier task in the case of predicted fire
air defense systems than in most of the more complex
field experiments with troop units and mixed weapons.
The following paragraphs outline a system for real
time assessment of antiaircraft gun systems. The Army
already has under development a 'Vulcan target selector' for which details are not available at the time of
writing, but which it is believed will perform some or
others
all of the functions to be described, and perhaps
rsesiedant erial s othere
asll o wthe
as well. However the present material is offered to
illustrate a concept.
the effectiveness
of a
is to indetermine
The problem
predicted
fire system
a field experiment
in real tm
time.
!I
.q
f~~
activate a development program with the specific object of producing an operational module to be attached
to any predicted fire system for this purpose.
a 'synthetic
the idea
Section
As described
at a
orders
in of
memory
is toin store
gun o0,
trajectory'
a ranging
10/second.
From
sampling rate of about
device time of flight to target present position can be
obtained continuously, gun orders time of flight previous can be retrieved, stripped of superelevation and
other ballistic corrections other than time of flight, and
used to generate a spot on a tracking display, or reticle
showing where the prediction function would have put
the center of aim with respect to the target.
* =
*,
-
gun quadrant
gun superel
"levation.
on.
d. t - , time of fligrn, ,
e. D - slant range to target,
f. t - time.
g
"
"what
The computational process is outlined in Figure 14I. Symbols used are as follows:
a. A. - gun azimuthb.
c.
AN
ALOG
T
ro
DIGITAL
CON
VERSION
COMPUTE
TIME OF
FLIGtiT
STORE
TO
ftoo
It)O
FROM STORED
DATA READ
A9 It
t10
0-1
,
DETERMINE
C OM P UT E
001 If!
MIOM
oil II
AND
10
C O MPU T E
E it t.
DIGIT
*(IJi
f
C0NALRON
if)CONVERSIN
IGL
ANAL OG
F!RING LASER
DMIMUU
A S E R HIT S
StSMPLE RANDOM
POSITION
NUM8FRS
DTERMINGANDI
TGT
PREDICTED AZIMUTH
AMPREDUCTE
AZIMUTH
fit)TI
OT
TifTARGET
AZIMUTH
TIME
SHIIETED B
TIM
ZIMUTU
S17 314ABy
14-2
__.___ -....-
..
be time varying as non-stationary processes, determination of the best way to develop summary descriptions requires careful preliminary analysis.
it is, or course, vital to record the target path data, as
determined from the tracking data corrected for recorded tracking errors, or from a separate tracking
unit of higher precision. This data also should be
reduced and statistically analysed in real time for
including in the engagement summary printout.
If the above objectives are realized, the Army will
have not only an effective device for assessing the
performance of predicted fire systems, but a means for
systematically building a base of" information which
can be used to establish requirements for, and to
design, improved predicted fire systems.
14.3 ANALYSIS OF COMBAT DATA
Quantitative analysis of the combat effectiveness of
weapon systems is an art which almost vanishes in
peace time, and is revived only after a conflict has
been under way for many months. In any war in which
air defense weapons are employed, however, early
measurement of enemy attack tactics will allow a
major improvement in defense weapon system doctrine
and effectiveness.
It is suggested that all except the most primitive of
air defense systems should incorporate, or have means
for attaching, a recording device that will store the
target paths of enemy aircraft as tracked by the defense system. With modern technology, such a device
can be simple, small, and inexpensive. In World War
!I. only the British recorded target tracks but the great
demand in the US. antiaircraft design community for
tracings of the tracks recorded by the British indicated
the recognition of the value of this vital data class in
that era.
,
L
.4,i""iii
14-3/14.4
--
11I
IIIII!
I-
1
SECTION 15
RECOMMENDED PROGRAMS
1
e
15.1 GENERAL
The most important short term objective in short
range air defense should be the construction of a
prototype gun system costing less than $500,000 per
fire unit, in a self-propelled version, with excellent fire
control and growth capability. This statem,;nt is the
opinion of the writer, based on the survey of air
defense requirements under the present and related
study efeort
ucluding
appear to be
Sspecial
b. Design the fire control system so that the prediction algorithms can be easily changed as threat
Riding Projectile&
I
tactics change.
The Army should avoid being locked into a relatively inflexible design concept such as Vulcan.
tage of conventional predicted fire systems, while securing a single shot kill probability at least equivalent
to that of a guided missile by providing projectile in-
Fm
S-m
be placed in proper perspective with respect to conventional systems on an overall effectiveness basis, in
addition indicating possible sources of major cost
differences, and components likely to require considerable developmental effort and/or incurring significant
technical uncertainties.
1
5
are:
concepts. However to fully exploit this capability a check-out package is essential to minimize programming errors, and increase the confidence associated with results. In addition, some module modifications would allow the simulation to support the other
recommended analyses. These are listed below
'-
systems
3
*
. ....
..
..
...
...
There are two possibly useful results of this installation, (I) since errors measured with respect to the
synthetic trajectory arc somewhat similar to those
obtained by projectile trackin;, the brassboard would
allow some experimentation with correction algorithms
prior to the availability of projectile tracking sensors,
with which the synthetic trajectory module would ultimately be combined, and (2) it is a means for field
observation of Vulcan system errors, with a potential
their reduction even with the present fire control.
mended
to supportcomponent
the general predicted
fire effort.
development
is recomThe following
In
a completefeedback
fire control
system,
such
a modulecontrol.
would
to the
human
operator's
regenerative
effective,
solution to
fire control
oto for
o light
ih
o cost
otslto
ofr
efcie low"
antiaircraft automatic weapons. Note that the system
does not have the characteristics of a disturbed reticle
sight, and the adjustment of the input parameters can
be made at a rate that will not throw the tracker off
target.
18.2 General Purpse Regeneraive Tracking
1odle
Module
i Duster, or any other
can antiaircraft
be attached to
Vukan, mount
which
available
travking
to provide a
15-3
WNW-
S.
.. .--
~.
..
..
__-
---
-,--
:1
14
ii
SKV
!V
f!
SV
I.
15-4
_L
"III
I I II
I I-r
, m,-,,m A- :
APPENDIX A
AIRCRAFT AND HELICOPTER LOSSES TO GROUND FIRE IN VIETNAM
.
*
I
[
iI
"1A-
ii-
kg
SUMMARY
Figure A-4 shows the build up of the North Vietnamese air defense system, except for Migs, which
never totalled more than about 75 aircraft in this
period. Almost all U.S. losses to enemy action have
been to ground fire, with a total of 50 aircraft lost to
Migs, through 1971.
I
I
10.000
I;i
8000
0
-
~TOTAL
a
6000
~CAUSES
k-
FIXED WING
AND HELICOPTEA LOSSES
IN INDOCHINA TO ALLI
I...
Iti
TOTAL LOSSES
ACTION
2000
1961
1062 }1963
It
964061
1966
11967 1008
in Vietana
j19609
1070
1971
8000
~~iI
4M0
TOTAL LOST
TO ALL CAUSES
0
wi
&I-
4~2000
LOST TO
1000
ENEMY ACTION
30003
:1
18~~~01
[IFigure
Ii'
182I
1LOOM in Vietnam207-1
A-3
p.-i
30000
"I
1.
TOTAL LOST
TO ALL C.AUSES
uJ
X,
2000
TOTAL LOSSES TO
ENEMY ACTION
1000.TOTAL. LOST
OVER NVN
0.I
1
1962
1963
1964
1966
1966
1967
1968
1969
1970
1971
20871l-317
A
A-4II
.I
10,000
8000
400
U
wzI
z/
Z2
11
4000
200
AAGUNSAND
izI
AUTO WPNS
'
a
ISAM
(
0N
SITES
RADARS
'Joa
100mm
looINTRODUCE0
lei
0!
I/
'I
1941
1962
1063
1WO4
1906
1966
1967
1068
1989
1970
20871-316
Figure
A-5
Air Deftnase
Year
sighted
1965
Ito
1966
990
1967
3500
Mi,%ilc./aircraIt lot[
A'C loit
33
30
mid 1968
55
64
6.
162
Year
Mkgs
lost
Mig%/U. S. lost
NVN Mig
strength
75
1965
3.0
1966
II
29
2 6
70
1967
25
75
30
20 in NVN
SO
50
mid 1968
Curh to 1972 (Jan)
1.1
20-25
40 in Jan 72
20871-635
A-6
,GV
,11.
11
ti*
At
Table A-2. Cumultive U.S. Aircraft Losses in Vietnam since Jan. 1960 as of 5/23/67 (Reference A.2)
Helicopters
Fixed Wing
all types
Total
S65
S71
Lost SVN
to hostile action
over
182
331
513
747
337
1084
646
544
1190
1393
881
2274
Notes:
"'Current" aircraft los rate estimated at S00/year, (1967) 11S Miss, and large numbers of mobile 85, 57.37 mm guns.
constitute dcefenses. Through May 23 the Mig ratio was 71/20 in favor of the U. S.
46 aircraft were lost to SAMS with abo~it 2450 missiles fired.
II%
of losses over NVN were caused by missiles and Miss.
All losses in SVN were caused by small ms and flak.
20871636
Ni
AMA41
A-7/A-
APPENDIX B
TABLE OF DEFINITE INTEGRALS
2
21Ca
irJ
+ w2T 2=
"
dw
w2 T2 )n
CfD
(8.6)
(B.T
Sw2mdw
(1 1(2n - 3)!!
\T)(2n- 2)!-
-(B(/
+w2TI2XI +w2 T2 2 )
w T2))
Sdw
T 2 (2n- 2)!!
(8.7)
T1 +T2
22f where
CI
1)
(2n)!! = (2X4)...(2n)
(B32
Srr~o
(B.8)
CDaT
(aw)
~cos
w"j(I+w2T2XI
-" ~(TI + T2X(TI +T3T + T)
" I
] + w2T2
T -r
Twd
w2 dw
(iJ+w2 Ti 2 XI +w2T2 2X! + w2T3 2 )
2f
o~s_(aw._) d.w
-~a/T(.)
1Iea/TI
dw
+ w2T 2 2)
(B.9)
"T2e'a/T2
T1 2 .T 2 2
(.0
(8.4)
"I
2 ,o -a/
:1~~
a2
~~~ t
IT
w dw
w22T22X + w2T3 2)
jX +
wT
w2 cos (aw) dw
~~(I(+w
2lT+w
2
I2+w2T22)
, (w
2
TIT 2 (TI
-T 2 2 )
T1 +T 2 +T 3
:B-
(8.11)
(B.5)
B-
21cos (aw) dw =I
Wj +WT2
(I +w
)-10
2
11I +(a/)e-a/T
fw
mCoD(w)d
m co- a)d
(z+ W)"
(B.12)
=(.
2T3
dn [zm - 0) eaz%
dz~
CO
1 )m''
*a~j
(B.13)
8-2
BIBLIOGRAPHY/REFERENCE LIST
I
I
Sslon 3
1. Peat. Marwick, Caywood, Schiller & Co., A
Study of the Accuracy of Mask Angle and
Visibility Angle in Representing Terrain Ef.
fects in Analytical Effectiveness Models, October 1965, AD 815, 469.
2. Stein, A., Terrain Mask Angles, Cornell Aeronautical Laboratory. Inc., a paper presented
at the Joint Session of AAAS Section E and
the Geological Society of America, Symposiurni on Quantitative Terrain Studies, Chicago. Illinois, 28 December 1959.
S12.Leadbetter.
gcan
21.
22.
20.
32.
Bertoni. E.A.. Clear Lines of Sight from Aircraft. AF Comabridge Research Laboratories.
AFCRL-67-0435. August 1967. AD 657 801.
50.
J
I
u
Wokoun. W., Detection of Random Low-Al.itude Jet Aircraft by Ground Observers. US.
Army Ordnance Human Engineering Laboratories Tech Memo 7.60. June 1960, AD 238
341.
1L
1gukM.,ealLoPorOpilSyts
35.
54.
I
3
3
722.
Bibli-2
UI
S6.
61.
ber 1968.
I. Mensa, D.L.. Scintillation Characteristicsof
Aircraft Targets, Naval Missile Center, Point
Mugu, California, 26 March 1971. TP-71-13.
2. Control.Display Pilot Factors Program. National Supersonic Transport Program, USAF
Instrument Pilot Instructor School. Randolph
AFB. Texas. December 1963.
: Phillips,
3.
W.H. and Egg1eston, J.M., The Lat.
eral Response of Airplanes to Random Atmo.
spheric
Turbulence,
3954. NASA TR R-74, 1960,
NACA TIN
i iSystems
2.
1962.
1962
.
Simpson. H.R. Performance Measures and
7.
Tracking
between
Comparison
G.G., A Dynamics
5. Frost,'Optimum'
and
Tracking
with a
6
etion
I. Foskett,
R .J., et al, A Review of the Literature
on Use of Tracer Observation as an Antaiarcraft Firing Technique, Technical Report 68I I. Hwn1RRO Division No. 5 (Air Defense),
98AD758.
6.CheroofBoeJH.
Birinham.2
Washington,
'I
Secon5
I. Benedict, T.R., and Bordner, G.W., Synthesis
aa Track.
rc-WieSa
o an
nOtmlSto
While.Scan
Optimal Set of Radar
Smoothing Equations, IRE Transactions on
Automatic Control. Vol. AC-7, No. 4, July
Section 4
8.
3.
13.
15.
16.
17.
18.
6.
I. The Military Balance 1971.1972. The International Institute for Strategic Studies. 18 Adam
Street. London. WC2N 6 AL.
2. Strategic Survey 1971, Ibid.
3. Jane's Weapon Systems 1971-1972, McGrawHill. 1971.
4.
5. Miller. Barry. Lasers Aid Delivery of Weapons. Aviation Week and Space Technology, 3
May 1971.
6.
24.
2s
25.
Ropelewski, Robert R.. A.7E Bombing Accuracy Displayed, AW & Space Tech., 10 May
1971.
10.
II.
12.
Flight
Mazine
Part I 7 December 1950
Difrot Over
Methods
T.. World
Beneckey in
Warof11 Air
In Operational
ReG.
6search Its Practice, Davies, Verhulst edisors,
Pergamon Press, 1958, pp 72-85.
Blbli.4
3
3
3
26. Weyl,
A.R.,
SMagazinei
7. The Republic F-IOSD Weapons System, Interavia, Vol. XV, No. 3, 1960, pp 311-317.
3
I
1
Opera-
I35.
36.
-37.
LJuly
.WiHK.SouonofteFhtrom
r
by the Theory of Games. BRL Technical
Note 23,.June 1949.
2. Karlin, Samuel, Mathematical Methods and
Theory in Games. Programming and Economis
el.IadI.AdsnWse,13,i
particular, Vol. 11. Chapter 8.
Gillman, L., Operations analysis and the theoryofgaes an advertising example, Journal
of the American Statistical Association, 45
54 1-46.
4. Gubler. [.A.. Solution of the Problem of Firing
with Artificial Dispersion for Various Cases, in
Collection of Articles on thte Theory of Firing.
I
omgrstdtr.E
eitTas
RAND Report T- 14. 13 October 1948.
5. Baxter. J.P.. 1ll, Scientists Against Time, MITf
Prss4168
SectIon I1I
M.S.. On the Theoretical Specification
Time-Serles, Jou -nal of the Royal Statistical
SocLiety, B8. ( 1946) pp 2 7-4 1.
2. Jenkins. G.M.. and Watts, DOG.. Spectral
I-Bartlett,
man).