Sounding Rocket Fin Design to Mitigate Roll Lock-In

C.P.Hoult (Author) Hien Tran (Author)

Rocket Science and Technology Rocket Science and Technology
4363 Motor Ave. 2822 W. Borchard Ave
Culver City, CA 90232 Santa Ana, CA 92704
(310) 839-8959 [email protected]
[email protected]

Abstract— Roll lock-in is a persistent high angle of attack,

nonlinear coning motion sometimes observed in the flight of
reentry vehicles and sounding rockets. For example, it has
occurred during Aerobee sounding rocket and Sidewinder I. INTRODUCTION
missile flights prior to the introduction of rollerons. This paper Roll lock-in, or catastrophic yaw, is the most challenging,
focuses on how fin design effects the probability of lock-in. The non-linear phenomenon in the field of sounding rocket
high angle of attack response is driven by misalignment and/or dynamics. There are several causes for this phenomenon:
offset thrust and drag forces, amplified by during pitch-roll
First, poor design leads to configurational asymmetries.
resonance. High angles of attack engender nonlinear roll
Second, lateral offset of the center of mass relative to the
moments which cause the roll rate to follow the pitch natural
frequency. It is well known that such roll moments can arise
symmetry axis leads to both pitch/yaw perturbing torques, and
when the center of mass is offset from the vehicle symmetry axis. to nonlinear roll moments. Finally, the interaction of fore body
However, this paper explores another source of nonlinear high vortices and tail fins causes nonlinear, high angle of attack roll
angle of attack roll moments, interaction between vorticity shed moments. Since only the last is significantly relevant to fin
from a fore body and tail fins. Both kinds of roll moment have design, it is the only one studied here. Note, however, that the
similar magnitudes. However, lock-in due to center of mass first two have generally valid remedial design prescriptions, e.
offset is, apart from static margin, not affected by fin design. g., roll balancing.

The location and strength of the shed vortex pair are found
Experience has shown that sounding rocket roll lock-in
from wind tunnel data. Roll moments are estimated from strip usually begins in powered flight at the time of yaw-roll
theory assuming incompressible cross flow. Conditions for resonance when a rocket's pitch/yaw natural frequency equals
steady state roll lock-in, and its probability of occurrence, are its roll rate1. During resonance the angle of attack response to
derived from the rigid body moment equations. A technique for body-fixed perturbations is significantly amplified. Re-entry
significantly reducing the probability of roll lock-in by adjusting vehicles are similarly afflicted during yaw-roll resonance2.
the fin exposed semispan and static margin is presented, and Then nonlinear roll moments will compel the roll rate to follow
used to show that, for a typical university sounding rocket, static the yaw natural frequency indefinitely. Such locked-in motion
margins larger than the classical two caliber heuristic rule can is said to be lunar with one side of the rocket always facing the
mitigate this problem. Fin taper ratio was studied, and found to cone axis. The consequence is a prolonged period of excessive
have a relatively minor effect. Little difference between three drag with severe adverse effects on the mission. The body-
and four fins was found. But, more than four fins, at fixed static fixed perturbations driving all this are random. Therefore, for a
margin, can significantly reduce the incidence of roll lock-in. Six given rocket flying a given mission, the occurrence of roll
fins are much better than four, and eight are better still. lock-in is also random.
In the past, most studies of roll lock-in have focused on
TABLE OF CONTENTS nonlinear transient phenomena1,4. But, some previous studies3,4
I. INTRODUCTION ............................................................ 1 have shown that catastrophic yaw is a continuing (quasi-steady
II. NOTATION ................................................................... 2 state) process. Two significant aerodynamic nonlinearities,
III. VORTEX PAIR DESCRIPTION ......................................... 3 Magnus moment and vortex-induced roll moment, are
IV. ANALYSIS OF VORTEX-INDUCED ROLL MOMENT.......... 4 important. Our approach is to first estimate the steady state
coning rate from the yaw moment equation, and, using the
V. LOCKED-IN LUNAR MOTION ........................................ 5
pitch moment equation, the fin panel size to ensure the desired
VI. EXAMPLE ..................................................................... 7
(input) vehicle static margin is attained. Next, the minimum
VII. IMPLICATIONS FOR FIN DESIGN .................................... 8 angle of attack for which steady state lunar roll lock-in is
VIII. RECOMMENDATION .................................................... 10 possible is found from the roll moment equation. The
REFERENCES ........................................................................ 10 probability of no roll lock-in is the objective function for
APPENDIX I: STABILITY DERIVATIVES .................................. 10 optimizing fin design, and is estimated from known body-fixed
APPENDIX II: CENTER OF MASS OFFSET MOMENT................. 12 perturbations15,16 .

1
 II. NOTATION g Indefinite integral in the roll moment induced
Mnemonic_________Definition________________________ by a single fin by a single vortex,
a Distance from a vortex core to a point on a fin, IP Pitch/yaw moment of inertia,
b Exposed semispan of a fin, L Rocket overall length,
CD Fin airfoil drag coefficient for a section normal lT Distance from the nozzle to the rocket center of
to the leading edge, mass,
 2C D M Free stream Mach number,
CD  ,
 2 N Number of fin panels,
C M Pitching moment coefficient slope, Pr Probability on no roll lock-in,

CN Normal force coefficient of the entire vehicle, p, q, r Roll, pitch and yaw rates,

C N Fin span-average airfoil normal force q Dynamic pressure,

coefficient slope without any body R Body radius,
interference,
r Distance from the body centerline to the core
C NF Fin assembly normal force coefficient slope, of a free vortex,
CNN Nose normal force coefficient slope, S Aerodynamic reference area  R 2

CL Roll moment coefficient CL  L qSd , SM Pitch/yaw static margin,

s Distance from the body centerline to the core
C L Single fin roll moment coefficient due to fin
of an image vortex,
cant,
T Thrust force,
C Li Amplitude of the vortex-induced roll moment
coefficient, U Free stream velocity,

C Lp Roll moment coefficient due to fin damping in v Tangential velocity around a vortex,
roll, vN Velocity normal to a fin panel,
C LV Nonlinear roll moment due to interaction x Distance aft of the nose tip,
between nose vorticity and tail fins,
xCG Distance from the nose tip to the Center of
C LO Nonlinear roll moment coefficient due to CG Gravity,
offset,
xCPF Distance from the nose tip to the fin assembly
C Np Yawing moment coefficient due to roll rate, center of pressure,
C Np Yawing moment coefficient due to roll rate and xCPN Distance from the nose tip to the nose center of
angle of attack, pressure,
C Nr Yawing moment coefficient due to yaw rate, xS Distance aft of the nose tip where vortex
separation occurs,
CY Magnus side force coefficient per unit length
based on cross flow velocity, y Distance from the body centerline to a point on
a fin,
c( y ) Local fin chord at spanwise station y,
yV Lateral distance from the body centerline to a
cR Root chord, free vortex core,
cT Tip chord, zV Vertical distance from the body centerline to a
free vortex core,
d Aerodynamic reference length  2R ,
 Body angle of attack,
fi Integral functions,
* Minimum angle of attack for which lock-in can
occur,

2
  local Local angle of attack at a point on a fin,
 Dimensionless exposed semispan  b / R ,
 Fin cant angle,
T Single plane thrust misalignment angle,
 Body roll angle,

V Roll angle attitude of a free vortex,

 Circulation around a free vortex,
 Fin sweep back angle, between root chord and
leading edge,
 Fin taper ratio  cT / cR ,
 Coning rate,
n Pitch/yaw natural frequency, and Figure 1 - Free Vortex Strength

 Single plane center of mass offset, The approximate lateral location yV of a vortex core is

( )o Value of ( ) at low angle of attack. yV

 0.7 (2)
R
III. VORTEX PAIR DESCRIPTION
It will probably come as no surprise to the reader to find
that there is an extensive literature on the vortex wakes in the
lee of slender bodies of revolution5,6,7. These will provide what
we need to know.
Our chosen point of departure is the longitudinal location
xS where the vortex pair separates from the body. While this
is difficult to accurately determine experimentally, it would
appear that the separation point approximately corresponds to
the location where the zero  static pressure on the body is a
minimum. Aft of this point, the longitudinal pressure gradient
is adverse, i.e., pressure increasing as x increases. This is a
strong argument for locating the vortex separation point there.
In Fig. 1 the definitions are:
( x  xS )
Dimensionless Distance Aft of Separation  
R
 Figure 2 - Vertical Location of a Free Vortex
Dimensionless Vortex Strength 
2RU The wind tunnel data6 for the vertical location of a free
vortex is shown in Fig. 2.
Using these definitions, the empirical data in Fig. 1 can be
well represented by zV  1
Here, Dimensionless Vertical Location  .
R
 ( x  x S ) 
1.52

 0.6  0.08  (1)
2RU  R 
The data displayed above cover the angle of attack range
from 10o to 30o, and extend aft from the body nose tip about 10
body diameters. The dashed line in Fig. 2 is a rough fit to the
data that could be used in other analyses. Its equation is
zV ( x  xS )
 1  0.21  (3)
R R
The image vortex is collinear with the centerline and the
external vortex, but with opposite sign. It can be shown8 that if

3
 rs  R 2 (4) The flow around an isolated vortex follows circular
streamlines centered on the vortex8. The tangential velocity v
the boundary condition on the cylinder surface is satisfied
at a radial distance a from the vortex core to a point on a fin is
everywhere.

v  (6)
IV. ANALYSIS OF VORTEX-INDUCED ROLL MOMENT 2a
First, consider the effects of vorticity shed from the fore
body at high angles of attack. In the Reynolds no. range
commonly encountered with sounding rockets at resonance a
pair of vortices are shed into the leeward wake over the
forebody. The vortex pair is initially symmetric about the
angle of attack plane. As the leeward wake moves aft, the
vortex pair grows stronger from additional vorticity shed from
the body. In the neighborhood of the tail fins the vortex pair,
and their image vortices induces a roll moment.
Figure 3 shows the important features of this problem. The
solid dots represent the cores of the shed free vortices. The
open dots represent the cores of the image vortices, located at a
distance s from the centerline8. Each free-image vortex pair
will satisfy the boundary condition of no flow through the
cylindrical body surface. As described above the location and
strength of the shed vortex pair ( r, v ) must be found from Figure 4 - Vortex and Fin in the Cross Flow Plane
wind tunnel experiment. The sign convention used here is to view the rocket from its
The fundamental assumption beneath the induced roll nose tip. A positive  and v follow the fingers of the right
moment calculation is the idea of incompressible cross flow. hand when the thumb points out of the paper. A positive roll
That is, the flow in the cross flow plane may be considered moment also follows this right hand rule.
incompressible so long as the overall hypersonic similarity
parameter H is small compared to unity: Now, look at Fig. 4, and consider the interaction
between the fin at top dead center and the free vortex at ~2
H  M 2  1 sin( )  1 (5) o’clock. The distance between the vortex core and a point on
the fin, a , is
Since for roll lock-in the angle of attack is often less than
about 0.2 radians, this implies the flight Mach number M  5 , a y 2  r 2  2 yr cos(V   ) (7)
or so. Equation (5) is automatically satisfied for subsonic free
stream Mach numbers. Next, the angle  between the tangential velocity and the
normal to the fin panel is
a2  y2  r2
    cos 1 ( ) (8)
2ay
Then, the vortex-induced velocity normal to the fin panel is
vN  v cos , or

( y  r cos(V   ))
vN   , (9)
2 ( y 2  r 2  2 yr cos(V   ))
Assuming a fin planform with straight taper, the local chord
is given by
c( y )  C1  C2 y , where (10)

R R  R 
C1  cR (1  ( ))  cT ( )  cR 1  ( )(1   )  c1cR
b b  b 
and (11)
cR  cT  1
Figure 3 - Roll Geometry Looking Forward C2    cR  c2 c R
b b

4
 With this the rolling moment due to a fin strip in the The variable appears explicitly in eq. (17) because has not
presence of an isolated vortex is yet been evaluated at the limits of integration. The vortex-
induced roll moment for a single fin is
d Induced Moment   qyc ( y )dyC N
vN
U  g First quadrant free vortex  
 
( y  r cos(V   )) qC N     g First quadrant image vortex  
 qy (C1  C 2 y )C N dy 
2U ( y 2  r 2  2 yr cos(V   )) 2U   g Second quadrant free vortex  
 
  g Second quadrant image vortex 
qC N  ( yC1  y 2C2 )( y  r cos(V   ))  
 dy , or
2U ( y 2  2 yr cos(V   )  r 2 ) Here g is the definite integral form of the indefinite
integrals given by eq. (18).
Induced Moment One Fin, One Vortex  
b R  g (C1 , C2 , r ,V , b  R, )  g (C1 , C2 , r ,V , R, ) 
qC N  1  
2U  y  2 yr cos(V   )  r 2
2
(12)

qC N    g (C1 , C2 , s,v , b  R, )  g (C1 , C2 , s,V , R, ) 
2U   g (C1 , C2 , r ,V , b  R, )  g (C1 , C2 , r ,V , R, ) 
R (19)
 yC1r cos(V   )    g (C , C , s, , b  R, )  g (C , C , s, , R, ) 
 2  dy.  1 2 V 1 2 V 
 y (C1  C2 r cos(V   ))  y C2 
3

The signs in the various terms of eq. (19) arise from

To provide a useful result, this must be coded in a computer integration limits and the signs of the various vortices.
program. To this end define some functions:
The roll moment contribution from the second fin panel
dy takes exactly the same form except that the roll angle argument
f0  
y  2 yr cos(V   )  r 2
2  is increased to   2 / N . The roll angle argument for each
1 y
(13) successive fin panel is increased by 2 / N . The moment from
 tan 1 (  cot(V   )) each of N panels takes two function evaluations to implement
r sin(V   ) r sin(V   ) the integration limits times four for the four vortices. For a
ydy four-finned configuration, 32 g function evaluations are
f1  y 2
 2 yr cos(V   )  r 2
needed for each roll angle.
(14)
1 Right-left symmetry is important. Assuming an
 log( y 2  2 yr cos(V   )  r 2 )  r cos(V   ) f 0 airfoil normal force coefficient slope of 4 / radian, Fig. 6 below
2 shows the induced roll moment coefficient for a four-finned
y 2 dy rocket at various angles of attack as found by the methods of
f2  y 2
 2 yr cos(V   )  r 2
this paper. Note that for small angles of attack, the phasing can
change by 180o. Figure 6 clearly shows the sinusoidal
 y  r cos(V   ) log( y 2  2 yr cos(V   )  r 2 ) (15) character of a typical induced moment curve. It follows that
the induced moment can be represented well by a simple sine
 r cos( 2(V   )) f 0
2
wave whose amplitude can be approximated by a single
and calculation at a roll angle of about  / 2 N . The maximum
amplitude of C Li over the entire roll angle range is shown as
y 3dy solid curve in Fig. 7.
f3   y 2  2 yr cos(V   )  r 2
y2 1 V. LOCKED-IN LUNAR MOTION
  2 yr cos(V   )  r 2 ( 2 cos 2 (V   )  ) (16) Next, consider the steady state lunar coning motion of a
2 2
rocket in roll lock-in. In lunar coning motion one side of the
log( y 2  2 yr cos(V   )  r 2 )  r 3 cos( 3(V   )) f 0 
rocket always faces the spin rate vector  as shown in Fig. 5:
We can now write the vortex-induced roll moment for one A further description of this motion is: The rocket is assumed
fin in the presence of one vortex in the form to be both slender and to have pitch-yaw symmetry. It follows
that the roll moment of inertia can always be neglected when it
qC N  appears in combination with the pitch or yaw moment of
 g (C1 , C2 , r,V , y, ) (17)
2U inertia.
with the function g given by The forcing function for this motion is assumed to be a
combination of thrust misalignment and lateral C.G. offset
g  C1r cos(V   ) f1 acting in the pitch plane. Its palliation, mass balancing, is
(18) known to be easy and effective.
 (C1  C2 r cos(V   )) f 2  C2 f 3
Because the motion is assumed to be steady state, the angle
of attack  is assumed to be constant. In other words, the

5
rocket is in a "flat or lunar spin" with one side always facing  R sin R cos  
inward. M Yaw  2qR 3  C Nr  C Np  (24)
 U U 
Then the steady state roll (x-axis) and yaw (z-axis) rates
  I P sin cos 
2
are:
p   cos  These three equations have three unknowns, a ,  and  .
Combining the second and third gives:
and (20)
r   sin 2qR 3C M   T (lT  T  )
 R sin  R cos  
 2qR 3 C Nr  C Np 
 U U 
Assuming the trim angle of attack is small lets us
estimate  :
T (lT  T  )

R 


2qR 3 CM  C Np  C Nr
U 


Solve the third equation for  under the same conditions:

2qR 4

UI p

C Nr  C Np  (25)

2
Note as always the period P of the motion is P  .

Combining eq's. (24) and (25) results in:
Figure 5 - Lunar Locked In Motion T (lT  T  )
 (26)
In body-fixed axes the three equilibrium moment  2qR 5 
equations8 are 2qR C M  2 (C Nr  C Np ) 2 
3

 U Ip 
 R cos  
M Roll  2qR 3 CLp  NCL  F  CLi ( ,  )  0 (21) Equation (26) is important because it can be used to
 U  estimate the probability of not exceeding  . Due to roll
and, assuming thrusting flight, symmetry, the pitch axis in this analysis is defined by the
direction of the vector sum of the two transverse components
M Pitch  2qR 3CM a  T (lT  T  )  I P 2 sin cos  (22) of the perturbing torque. If the statistics of both transverse
components have the same Gaussian distribution, then  has a
The perturbations driving this motion,  T and  , are Rayleigh distribution. Let
random. Some estimates on them can be found in ref's. (15)
and (16). The static pitching moment in eq. (22) above should T 2 var( l T  T  )
var( )  , where
be treated with respect. At sufficiently high angle of attack the 2
 2qR 5 
fins are susceptible to stalling if the Mach number is subsonic. 2 q R C M  2
2 2 6
(C Nr  C Np ) 2  (27)
Now, the derivative C Np appearing in the third ( M Yaw ) of  U Ip 
these equations is tricky. If evaluated at zero  it will be var( l T  T  )  l T var( T )  var( )
2

found to vanish. But, if evaluated at a non-zero  it can have Then, the probability of exceeding  is
significant dynamic effects. It's basically a Magnus torque.
That's because C Np is really CNp as shown in the Appendix.   2 
Pr  1  exp   (28)
 var( ) 
Thus,
R cos  R cos  Next, to find the maximum  for which no lock-in is
CNp  CNp (23)
U U possible, return to the roll equilibrium condition, eq. (21):
Then, the yaw moment equation becomes  R cos  
C Lp  NCL  F  C Li ( ,  )  0
 U 

6
 Now, at low angle of attack (just prior to entering the First, estimate the vortex-induced roll moment as a function
resonant condition), of roll angle and plot it in Fig. 6 below:
pR
NCL  F  C Lp
U
The roll rate just before resonance is just the pitch natural
frequency. Then

 2qR 3 C M
p  n  , and
IP

 R 5 C M
NC L  F  C Lp
IP
With this, the roll equilibrium equation becomes

C Lp   2qR 5 C M 
 R    C Li ( ,  * )  0 (29)
U  IP 

Equations (28) and (29) are the key to our problem. At
small angles of attack, the C Li term is too small to satisfy Figure 6 - Vortex-Induced Roll Moment for Various
eq.(29) for any roll angle. That is, no steady state locked-in Angles of Attack
solution is possible. The smallest angle of attack  * for which Figure 6 shows the induced roll moment vs. roll angle at
a steady state locked-in solution exists is given implicitly by angles of attack = 0.05, 0.1, 0.12 and 0.15 radians for the
example vehicle. First, note that the roll moment vs. roll angle
C Lp   2qR 5 C M  curve goes through a 180° phase change from a small angle
 R    C Li MAX ( * ) (30) attack regime (  6  ) to different regime at larger angles of
U  IP 
 attack. As  increases beyond this transition, the amplitude
of the induced roll moment increases to very large values. This
where C LiMAX ( * ) is the maximum value of C Li obtained by is caused by the changing position of the fin tips relative to the
finding , at a fixed angle of attack, that roll angle  which shed vortex pair. This pattern is generally observed for many
maximizes/minimizes C Li . Since the induced roll moment different fin geometries and flight conditions.
oscillates between positive and negative maximum values, Next, consider how this leads to roll lock-in. Begin by
equation (30) should be interpreted to mean, for a lock-in finding the maximum value of C Li as a function of angle of
solution to exist, that the C Li term has a sign opposite to that attack. This is shown as the solid line hill-valley-mountain
of the C Lp term. In effect, eq. (30) defines a necessary and curve in Fig. 7 below. When the number of fins is increased
sufficient condition for roll lock-in to occur. Unless it is to, say 6, the single humped shape (think dromedary) shown in
satisfied, roll lock-in cannot happen, even though the Fig. 7 often changes to a two humped (think Bactrian camel)
perturbing torque is very large. But, if it is satisfied, then it's form. If lock-in were to occur it would first manifest itself at
only a matter of statistics as shown in eq. (28). these relative maximum  ’s because at other roll angles (same
 ) the necessary induced roll moment would not be as large.
The horizontal line in Fig. 7 corresponds to possible roll lock-
VI. EXAMPLE
in conditions. In regions with a thin, solid horizontal line roll
As an example of this process, consider the following set of lock-in cannot occur (not enough C Li ), but where the heavy
parameters:
dashed line appears, steady state roll lock-in can happen.
U  10000 in / sec q  4 lb/in 2 L  122 in During the run up to pitch-roll resonance the angle of attack
starts small and increases as resonance is approached. Lock-in
W  60 lb XCG  71 in I P  27734 sl  in2 will then be observed where the vertical dot-dashed line
labeled  intersects the other two curves.
b  6 in Rad  3 in  2
  0.3 SM  3.33 cal SM  20 in

C Nn  2 / rad C Na  4 / rad X S  36 in

XCPF  108 in XCPN  39 in N 4

7
 Once we have a model of vortex-induced roll moment and
how it influences roll lock-in, it is appropriate to consider what
can be done to mitigate steady state roll lock-in. The obvious
first thing is to carefully roll-balance and align the structure to
minimize the pitch / yaw perturbing torques. This has been
known for a very long time.
So the next step in our mitigation strategy is to try to hold
the left hand side of eq. (30) above the C Li hump(s) occurring
low angle of attack. In achieving this goal, static margin is
more important than generally realized. There is a long
tradition of designing for two calibers (body diameters) of
static margin. But Fig. 8 shows that, apart from the perturbing
torque statistics, more than two calibers can extend the region
where lock-in is dynamically impossible.
The left hand side of eq. (30) depends significantly on
flight condition, more specifically dynamic pressure as shown
in Fig. 9 below. Larger dynamic pressure acts like larger static
margin by reducing angle of attack. It is expected that the
Figure 7 – Typical Conditions for Roll Lock-in
smaller dynamic pressures at higher altitude will exacerbate
Figure 7 above sketches the approach to solving eq. (30) any lock-in tendencies.
with the horizontal line representing the C Lp term. To this
varying amounts of C Li are added with more as the angle of
attack is increased. Lock-in first becomes possible when the
maximum C Li is just tangent to the horizontal line. Thus,
eq.(30) implicitly defines the minimum angle of attack for
lock-in. Now, since the first (C Lp ) term in eq. (30) does not
depend on roll angle, the boundary between no roll lock-in and
just possible roll lock-in is that C Li be stationary with respect
to roll angle, and that it be large enough to just balance the
C Lp term. Thus, at any specified angle of attack, there is a
maximum / minimum for C Li . This can be found using
numerical methods from the induced roll moment data like that
plotted in Fig. 6 above.
Now add in the remainder of eq. (30) as depicted in Fig. 7.
The left hand side, the C Lp term, appears as a horizontal
straight line. Those parts of this straight line shown as heavy Figure 8 - Typical Effects of Varying Static Margin
dashed lines are regions where roll lock-in is possible, that is,
where eq. (30) is satisfied. Equation (26) establishes a pitch
perturbing torque needed for the lowest angle of attack for roll
lock-in shown by the vertical line in Fig. 7. Thus, no roll lock-
in is possible in the region to the left of this vertical line.
Given this, eq. (28) can be used to estimate the probability that
no roll lock-in can occur.

VII. IMPLICATIONS FOR FIN DESIGN

Begin by observing that there is no guaranteed fin design
that will always prevent roll lock-in. It is always possible to
have very large misalignment torques that, no matter how
statistically rare, will cause roll lock-in. But, it is possible to
design fins that have a very low probability of lock-in. Figure
7 shows that if we can move the horizontal line high enough to
clear the “hill” lock-in will be deferred to very high, and
uncommon, angles of attack.

8
 Figure 9 - Effect of Dynamic Pressure on the First
(C Lp ) Term in Eq. 30)

The remaining choices involve manipulating fin geometry.

Fin design usually is one of the last steps in vehicle design.
The Fin Designer considers most of the key system parameters
identified in the example as givens. His free variables are fin
size (Static Margin), number of fin panels (N ) and fin shape
(  and  ) . The simplest approach is to manipulate the
appearance of Fig. 7 by varying SM , N ,  and  to move the
horizontal line above the camel's hump(s).
By far the most important fin parameter for mitigating lock-
in is the number of fin panels, N . Most sounding rockets have
flown with 4 fins with a small minority using only 3 fins, and
almost none with 5 or more. Our model shows that there is
very little difference between 3 and 4 fins when attempting
implementation of the first part of our strategy. Small tweaks
in taper ratio  or exposed fin span  can bring about very Figure 11 - CL of 4 fins for Various Angles of Attack
similar performance in these two cases. But, increasing the
number of fins to 5 or more leads to dramatic improvements.
Our model shows that with 6 fins major improvement is
relatively easy, and it becomes trivially easy with 8. Note that
mortar bombs are commonly flown with 8 fin panels.
Assuming the same baseline for all fin counts, results are
shown in Fig’s. 10, 11, 12 and 13 for 3, 4, 6 and 8 fin panels.

Figure 12 - CL of 6 fins for Various Angles of Attack

Figure 10 - CL of 3 fins for Various Angles of Attack

Figure 13 - CL of 8 fins for Various Angles of Attack

9
 The effect of increasing the number of fins beyond 4 where the second term in  is the inverse square upwash
dramatically, improves resistance to roll lock-in. around a cylinder8. The third term is the fin cant angle, the
fourth is due to roll damping and the last is the sum of the
VIII. RECOMMENDATION  in ’s induced by the four vortices.
To improve resistance to roll lock-in, consider using static
margins greater than two calibers and more than four fins.

REFERENCES
[1] Nicolaides, J. D., "A History of Ordnance Flight Dynamics", A.I.A.A.
paper 70-533, 1970.
[2] Platus, D. H., "Ballistic Re-entry Vehicle Flight Dynamics", J. Figure 14 - Airfoil Sketch
Guidance, Vol. 5, no. 1, pp 4-16 (1982).
Before analyzing the roll moments, take a break and recall
[3] Ananthkrishnan, N., and Raisinghani, S. C., "Steady and Quasisteady
Resonant Lock-in of Finned Projectiles", Journal of Spacecraft and
that while we are trying to design fins to minimize nonlinear
Rockets, Vol. 29, no. 5, 1992, pp 692-696. roll behavior, any resulting design must have the necessary
[4] Clare, T.A., "Resonance Instability for Finned Configurations Having static margin. This implies that the constraint takes the form of
Nonlinear Aerodynamic Properties", Journal of Spacecraft and Rockets, eq. (33) below10,
Vol. 8, No. 3, 1971, pp 278-283.
R2
[5] Jorgensen, L. H. and Perkins, E. W., “Investigation of Some Wake
Vortex Characteristics of an Ogive-Cylinder at Mach Number 2”,
d Fin Normal Force Slope  qC N c( y)dy(1  ) (32)
N.A.C.A. Report 1371, 1955.
y2
[6] Thomson, K.D. and Morrison, D. F., “The Spacing, Position and The last term accounts for body-fin upwash interference8.
Strength of Vortices in the Wake of Slender Cylindrical Bodies at Large
Incidence”, Journal of Fluid Mechanics, Vol. 50, Issue 04, 1971.
Integrating across the exposed fin span gives the single panel
result,
[7] Spahr, J. R., “Theoretical Prediction of the Effects of Vortex Flows on
the Loading, Forces, and Moments of Slender Aircraft”, N.A.S.A. Single Panel Fin Normal Force Slope 
Technical Report R-101, 1961.
[8] Rauscher, M., “Introduction to Aeronautical Dynamics”, John Wiley &
Sons, New York, 1953. qC N c R 
 
 c1b  c2 b 2  2bR / 2  c1 R 2 b / Rb  R 

  c R 2 logb  R  / R  
[9] Hoult, C. P., “Strip Theory for Setting Fin Angles (revised)”,RST  2 
Memo, 13 September 2009.
[10] Barrowman, J.S. and Barrowman, J.A., “The Theoretical Prediction of For a complete symmetric fin configuration ( N  3) it can
the Center of Pressure”, Master of Science thesis, Catholic University, be shown10 that the
1966.
[11] Babister, A. W., "Aircraft Dynamic Stability and Response", Pergamon Total Fin Assembly Normal Force Coefficient Slope 
Press, London, 1980.
 
 c1b  c 2 b 2  2bR / 2 
Nc R   (33)
[12] Naselius, K. A., "A Wind Tunnel Study of Magnus Effects on a
Differentially Rotating Missile", AFIT M. S. Thesis, 1992. C NF  C N   c1 R 2 b / Rb  R  
[13] Reid, E. G., "Tests of a Rotating Cylinder", N.A.C.A. TN-209. 1924. 2R 2

  c 2 R 2 logb  R  / R 
[14] Jacobson, I. D., "Magnus Characteristics of Arbitrary Rotating Bodies",  
AGARD-AG-171, edited by P. F. Yaggy, Harford House, London,
England, 1973. Keep in mind that the tail fins are the most important
[15] R. L. Ammons and C. P. Hoult, “Standardized Perturbation Values for contributor to vehicle static margin. The static margin, and
Several Sounding Rockets to be Used in the Calculation of Dispersion hence the fin assembly normal force slope, result from
and Structural Loads”, Aerojet-General Corporation memo requirements unconnected to roll lock-in. Therefore eq’s. (33)
8110:M0658:ak, 16 September 1970.
and (37) acts as a constraint on the allowable changes in fin
[16] R. N. Knauber, "Thrust Misalignments of Fixed-Nozzle Solid Rocket
Motors", Journal of Spacecraft and Rockets, November-December
geometry. To apply it, select C N , b, R and  . Solve for c R .
1996, pp794-799.
Then, the roll driving and damping moments of a single fin
[17] Abbott, I. H., von Doenhoff, A. E., and Stivers, L.,Jr., "Summary of
Airfoil Data", N.A.C.A. Report 824, 1945. panel9 are
b R
APPENDIX I: STABILITY DERIVATIVES
Strip theory9 considers a fin panel to have only two
Driving Moment  qC N   yc ( y)dy
R
dimensions; that is, it’s a pure airfoil. However, use of a 2D C C
airfoil lift curve slope can give highly erroneous results so the  qC N  ( 1 ) y 2  ( 2 ) y 3
reader is advised to use the 3D lift curve slope averaged over 2 3
the exposed fin area. Now, consider Fig. 14 below. The local  C1 2 
angle of attack  local on an airfoil strip is ( 2 )(b  2 Rb) 
 qC N   
N  ( C 2 )(b 3  3Rb 2  3R 2 b)
 local   (1  ( R / y) 2 )    R y / U    in (31)  3 
1

10
  c1  where the summation extends over all body elements
( )(( b  R) 2  R 2 ) 
C N c R  2
including those with fins.
C L    (34)
2R 3  c 2 3 

 3( )(( b  R ) 3
 R ) 
C. C Np
The yaw moment due to rolling and angle of attack is tricky
b R because it requires estimation of the Magnus force on the body.
p
Damping Moment  qC N y
2
c( y )dy Reference (13) provides an experimental estimate of the
U R Magnus side force acting on a uniform spinning cylinder:
 C1 3 
( )(b  3Rb 2  3R 2 b) Rp
p 3  CY  0 when  0.5 (39)
 qC N   U sin 
U  C2 
  ( )( b 4
 4b 3
R  6 R 2 2
b  4 R 3
b )  Since this condition is abundantly satisfied for the current
4 Rp
problem (  0.05) , the body contribution to Magnus
 c1  U sin 
( )(( b  R) 3  R 3 ) 
C N c R  3 moment can be neglected. But, the data8,12,13,14 show that there
C Lp    (35) is a significant nonlinearity near zero roll rate. Once the
R 4  c2 4 
 ( 4 )(( b  R)  R )
4 condition in eq. (39) is exceeded, the side force increases more
Rp
or less linearly with .
The stability derivatives needed for the lunar coning U sin 
version of roll lock-in are estimated above. The fin contribution can be found with the aid of Fig. 15
below. Consider a strip of fin of span dy viewed from the left.
A. C M While the basic, non-rolling angle of attack is still  cos 
A simple estimate on the static margin is there is a additional local angle of attack due to roll rate:
xCPN C NN  xCPF C NF p 
py
SM  xCP  xCG   xCG
C NN  C NF U cos 
If we define lift L and drag D the usual way, normal to
the static pitching moment derivative is then
and parallel the relative velocity vector, the x -force X acting
dCM  xCG (C NN  C NF )  xCPN C NN  xCPF C NF (36) on the chordwise strip for small  is:
Finally, the required fin assembly normal force coefficient py
dX  ( cos   )dL  dD
slope is U
 SM  xCPN  xCG
C NF  C NN (37)
SM  xCPF  xCG
This is used in eq. (33).

B. C Nr
The yaw moment due to roll rate can be estimated for a
single fin panel using the strip theory methods of ref. (10).
First, the definition of the yaw moment coefficient:
N
CN 
qSR
Its derivative with respect to pR 2U is defined as C Np .

Now, decompose the body into a sequence of elements,

each of which has a normal force coefficient slope c N . If an
element were located a distance xi  xCG  xi from the
center of mass, then
xi 2
C Nr  2 c N i ( ) (38)
R
Figure 15 - Rolling Fin Geometry

11
 The yawing moment contributed by the strip is so, in incompressible flow. For fins with supersonic leading
edges, the Ackeret result for airfoil wave drag should be used.
dN   y cos  ((local cos    p )dL  dD) The correction for sweep angle is based on infinite swept wing
theory.
Next, recall that body upwash adds to the local lift
distribution, and that C L  C N . Then,
APPENDIX II: CENTER OF MASS OFFSET MOMENT
 R 2
py  A second major source of nonlinear roll moment of offset
 ( (1  2 ) cos   )  center of mass. This is sketched in Fig. 16 below
 y U 
dN  q cos   2  ycdy
 C ( (1  R py
 N 2
) cos   )  C D 
 y U 
Converting to coefficient form, and noting that a
symmetrical airfoil drag has a parabolic relationship with local
angle of attack,
1  2CD 2
CD  CD0   local sin 2 
2  2
cos   2C D
dC N   (C N  sin 2 )
2R 3  2
R2 py
( (1  2 ) cos   ) 2 ycdy
y U

Retaining only those terms contributing to C Np , results in

 (C  CD sin 2 ) cos 2  

dC N   N 
 Figure 16 - Roll moment due to Center of Mass Offset
 R 3  (40)
p R 2 We easily find that the induced roll moment is
 (1  2 )(C1 y  C2 y )dy
2 3

U y 
CLCG  CN  sin(  CG ) (42)
The single panel stability derivative is found after d
integrating from R to b  R , and differentiating: The total nonlinear roll moment is then the sum of that due
to nose vorticity and that due to CG offset, including statistical
 (C  CD sin ) cos  
2 2
CNp   N 

correlations
 R 4 
 C1 C 4 
 ((b  R)  R )  2 ((b  R)  R ) 
3 3 4

 3 4 
 C2 
  R (C1b  ((b  R)  R ))
2 2 2

 2 
The full fin assembly C Np requires adding the terms from
all fin panels. Thus, for N  3 ,

 Nc (C  C D sin 2 ) 
C Np   R N 

 2R 4 
 c1 c2 4  (41)
 ((b  R)  R )  ((b  R)  R ) 
3 3 4

 3 4 
 c2 
  R (c1b  ((b  R)  R ))
2 2 2

 2 
Two comments are in order. Due to root and tip effects,
C N will vary across the fin span. A good first approximation
is to take the full, 3D C N without body upwash, and average
it over the span. Second, the derivative CD can be estimated
from wind tunnel measurements17. Typically, CD  0.2 , or

12