48th AIAA/ASME/SAE/ASEE Joint Propulsion Conference & Exhibit AIAA 2012-4200

30 July - 01 August 2012, Atlanta, Georgia

Closed-Loop Thrust and Pressure Profile Throttling of

a Nitrous-Oxide HTPB Hybrid Rocket Motor

Zachary W. Peterson∗, Shannon D. Eilers∗, and Stephen A. Whitmore†

Utah State University, Logan, UT 84322
Downloaded by "Indian Institute of Technology, Bombay" on March 31, 2025 | http://arc.aiaa.org | DOI: 10.2514/6.2012-4200

Hybrid motors that employ non-toxic, non-explosive components with a liquid oxidizer
and a solid hydrocarbon fuel grain have inherently safe operating characteristics. The
inherent safety of hybrid rocket motors offers the potential to greatly reduce overall oper-
ating costs. Another key advantage of hybrid rocket motors is the potential for in flight
shutdown, restart, and throttle by controlling the pressure drop between the oxidizer tank
and the injector. The proposed research designs, develops, and ground tests a closed-loop
throttle controller for a hybrid rocket motor using nitrous oxide and hydroxyl-terminated
polybutadiene as propellants. The research simultaneously developed closed-loop throttle
algorithms and lab scale motor hardware to evaluate the fidelity of the throttle simulations
and algorithms. Initial open-loop motor tests were performed to better classify system
parameters and to validate motor performance values. Deep-throttle open-loop tests eval-
uated limits of stable thrust that can be achieved on the test hardware. Open loop tests
demonstrated the ability to throttle the motor to less than 10% of maximum thrust with
little reduction in effective specific impulse and acoustical stability. Following the open-
loop development, closed-loop, hardware-in-the-loop tests were performed. The closed loop
controller successfully tracked prescribed step and ramp and with a high degree of fidelity.

Nomenclature

A exponential curve fit scaling coefficient

a1 experimental quadratic curve fit coefficient
a2 experimental linear curve fit coefficient
An nozzle throat area
Av valve port area
Aev effective throttle valve port area
Ainj injector orifice area
b1 experimental quadratic curve fit coefficient
b2 experimental linear curve fit coefficient
c speed of sound
c∗ characteristic velocity
c1 experimental quadratic curve fit coefficient
Cv valve flow coefficient
Cdinj injector discharge coefficient
Cdn nozzle discharge coefficient
Cdv valve discharge coefficient
fc lowpass cutoff frequency
df fuel grain port diameter
∗ Graduate Research Assistant, Mechanical and Aerospace Engineering (MAE) Department, 4130 Old Main Hill, Logan,

Utah, 84322-4130, AIAA Student Member.

† Associate Professor, Mechanical and Aerospace Engineering (MAE) Department, 4130 Old Main Hill, Logan, UT 84322-

4130, AIAA Member.

1 of 42
Copyright © 2012 by Utah State University. Published by the American Institute of Aeronautics and Astronautics, Inc., with permission.
American Institute of Aeronautics and Astronautics
 dp valve port diameter
e controller error
fa acoustical mode frequency
fl hybrid low frequency pole
Ft thrust
Ftf final thrust
Fti initial thrust
Ftn normalized thrust
fvs vortex shedding frequency
h1−3 enthalpies in plumbing feed physics model
Isp specific impulse
K total compressibility factor
KD derivative gain
Downloaded by "Indian Institute of Technology, Bombay" on March 31, 2025 | http://arc.aiaa.org | DOI: 10.2514/6.2012-4200

KI integral gain
KP proportional gain
l Strouhal number characteristic length
L∗ ratio of the chamber volume to the nozzle throat area
Lc combustion chamber length
Lf fuel grain length
ṁi injector mass flow rate
ṁp mass flow rate through valve port
ṁin mass flow rate through valve port in plumbing feed physics model
ṁout mass flow rate through injector in plumbing feed physics model
ṁox oxidizer mass flow rate
ṁtot total mass flow rate through motor
m2 post-throttle plumbing fluid mass in plumbing feed physics model
MB oxidizer bulk modulus
mAh milli-Amp hours
MVT maximum valve travel, see definition in Subsection 3.1.1
dn nozzle throat diameter
NHNE non-homogeneous, non-equilibrium
o/f oxidizer to fuel mass ratio
P̄ average pressure in plumbing
P̄c average chamber pressure
P̄i average injector manifold pressure
∆P pressure drop across orifice
P0 chamber pressure
P1 oxidizer feedline pressure in plumbing feed physics model
P2 post-throttle plumbing pressure in plumbing feed physics model
P0pc pre-combustion chamber pressure
RT product of gas constant and fluid temperature
RT̄ product of the average gas constant and temperature in the combustion chamber
RTc product of the gas constant and combustion temperature at the nozzle entrance
S controller thrust or chamber pressure signal
s differential operator
s1−3 entropies in plumbing feed physics model
Sr Strouhal number
t time
T1 oxidizer feedline temperature in plumbing feed physics model
T2 post-throttle plumbing temperature in plumbing feed physics model
ttot total controller run time
U Strouhal number fluid velocity
Uf fluid velocity in fuel grain
Uv fluid velocity through throttling valve

2 of 42

American Institute of Aeronautics and Astronautics

 V2 post-throttle plumbing volume in plumbing feed physics model
vp plumbing volume
x valve position
x0 exponential curve fit offset coefficient
X1 oxidizer feedline fluid quality in plumbing feed physics model
X2 post-throttle plumbing fluid quality in plumbing feed physics model
Xox oxidizer quality
y general frequency response
βs liquid compressibility factor
ηc combustion efficiency
γ ratio of specific heats
γox oxidizer ratio of specific heats
ωn natural frequency
Downloaded by "Indian Institute of Technology, Bombay" on March 31, 2025 | http://arc.aiaa.org | DOI: 10.2514/6.2012-4200

ρ density
ρ2 post-throttle plumbing fluid density in plumbing feed physics model
ρg average density of gas in plumbing
ρl average density of liquid in plumbing
ρox oxidizer density
τ time constant
τ1 delay transfer function time constant
τ2 delay transfer function time constant
τc combustion chamber fill time
τp plumbing fill time

I. Introduction

During the past 50 years conventional launch systems have been developed to a high state of capability;
however, for a variety of reasons these vehicles have become increasingly expensive to operate. Some of these
reasons include manufacturing and operational complexity, safety and environmental regulations for dealing
with hazardous materials, and the generally large “support army” required for flight preparations. Because
of high launch performance demands on specific impulse (Isp ) and thrust-to-weight ratio, conventional liquid
and solid-propelled rocket stages that employ highly-energetic, explosive, or toxic propellants will likely
remain the systems of choice for large military-class payloads or for human spaceflight. However, there exist
emerging markets, both commercial and government, that are willing to accept a lower system performance
in exchange for reduced operational costs and lower environmental impact.

A. On the Inherent Safety and Operational Flexibility of Hybrid Rocket Systems

A key to commercial spaceflight success is the cost efficient and safe operation of large rocket motors in
civilian environments. Because of the extreme risk and potentially negative environmental impact, large solid-
propelled rocket motors simply cannot be operated at civilian airports. Approximately 70% of catastrophic
space launch failures are attributable to the vehicle’s power plant.1 NASA estimates that the Space Shuttle’s
liquid fueled main engines will fail catastrophically once every 1530 sorties per engine and its solid rocket
boosters will fail catastrophically once every 1,550 sorties per motor.2, 3 While this level of risk is acceptable
for experimental and government-operated vehicles, it is unacceptable for a potential commercial spaceflight
operator.
Hybrid motors that employ non-toxic, non-explosive propellants have the potential to fulfill this market
niche. The inherent safety of hybrid rocket motors greatly reduces the operational risk to a launch vehicle
or any hybrid rocket propelled missile or spacecraft. According to the U.S. Department of Transportation,
hybrid motors can be stored and operated without possibility of explosion or detonation.4 Due to safer and
easier ground handling; transportation, setup, and operating costs can be greatly reduced for hybrid rockets.
Current liquid bi-propellant upper stages feature cryogenic propellant that cannot be stored for any

3 of 42

American Institute of Aeronautics and Astronautics

 extended period of time. The vast majority of in-space propulsion is performed using hydrazine or hydrogen-
peroxide monopropellants (H2 O2 ) or hydrazine-derived hypergolic bi-propellants like mono-methyl hydrazine
(MMH) and nitrogen tetroxide (N2 O4 ).
Unfortunately, all of these propellants are powerful reducing agents that pose serious environmental
concerns. Hydrazine is extremely destructive to living tissues, and is a probable human carcinogen. Exposure
produces a variety of adverse systemic effects including damage to liver, kidneys, nervous system, and red
blood cells.5 In addition to these biological and toxicological impacts, toxic propellants present significant
environmental dangers for the spacecraft and launch vehicle. As chemicals prone to rapidly decompose or
explode when struck, vibrated, or otherwise agitated, both hydrazine and peroxide are among the most
shock-sensitive chemicals listed by the U.S. Department of Transportation.4, 6
Although procedures are in place to allow toxic propellants to be managed safely on tightly controlled
military and NASA-owned flight experiments, the toxicity and explosion potential requires extreme han-
dling precautions. Increasingly, with a growing regulatory burden, infrastructure requirements associated
Downloaded by "Indian Institute of Technology, Bombay" on March 31, 2025 | http://arc.aiaa.org | DOI: 10.2514/6.2012-4200

with toxic propellant transport, storage, servicing, and clean up of accidental releases are becoming cost pro-
hibitive. As space flight operations continue to shift from government–run organizations to private companies
and universities that operate away from government-owned test reservations, servicing payloads requiring
hydrazine as a propellant becomes operationally infeasible. Extreme handling precautions generally do not
favor hydrazine as a propellant for secondary payloads.
A non-toxic, stable propellant alternative is clearly desired. Nitrous oxide (N2 O) is an inexpensive
and readily available propellant that has long been considered as a potential “green” monopropellant for
spacecraft applications. Despite a slight decrease in motor performance, N2 O has several advantages that
make it very competitive as a hybrid rocket oxidizer.
First, N2 O has the clear advantage of being non-toxic to human tissue and is classified as non-explosive
and non-flammable by the U.S. Occupational Safety and Health Administration (OHSA).7 Nitrous oxide
exists as a saturated liquid below its critical temperature of 36.4 C, and studies performed by the USAF
have demonstrated that it is virtually impossible to force a dissociation reaction with N2 O in its liquid form.8
Unlike the highly reactive propellants described in the previous paragraphs, nitrous oxide can be handled
without special precautions beyond those required for any pressurized fluid.
Second, unlike cryogenic liquid oxygen (LOX), N2 O is highly storable and allows rocket systems to
be prepared far in advance of motor use. Long-term propellant storability is a requirement for in-space
propulsion systems. Proposed hydrazine replacements based on aqueous solutions of hydroxyl-ammonium
nitrate (HAN) or hydroxyl-ammonium di-nitrate (ADN) salts are not long-term storable. The high water
content of the HAN/ADN propellants makes them susceptible to freezing during periods of extended cold
soak. These propellants must be temperature conditioned for space applications.
Finally, N2 O at room pressure exists in a saturated-liquid form and has a high vapor pressure exceeding
5000 kPa. This self-pressurizing property can be used to reduce complexity of propellant delivery systems. In
contrast hydrazine and peroxide have vapor pressures less than 1 kPa at room temperature. HAN/ADN and
LOX have essentially zero vapor pressure at storage temperature. It must be noted that the self-pressurizing
property of nitrous oxide typically produces two-phase injector flow, and this property has a significant
influence on the design of the throttling control system discussed in this paper. This effect will be described
in detail later.
While hybrid systems generally deliver lower Isp than conventional bipropellant liquid and lower volu-
metric efficiency than solid rockets of the same thrust level; because the propellant components remain inert
until ignited within the motor chamber, hybrid rockets are inherently safer to transport, load, store, and
operate. This inherent safety greatly reduces ground handling and transport costs, and can potentially lead
to overall reduction in system operating costs. Unlike solid-propelled rockets, where fuel grain flaws and
age-induced cracks present a significant safety issue, hybrid rockets exhibit a relative insusceptibility to grain
flaws.

B. Hybrid Motor Throttleability

Other advantages of hybrid rockets that can potentially overcome the lower performance level include the
ability to be restarted in flight and throttled over a wider range of thrust levels than conventional liquid
bi-propellant systems. Hybrid propelled rocket stages are especially attractive for micro- or nanosatellite-

4 of 42

American Institute of Aeronautics and Astronautics

 scale launch vehicles where high thrust-to-weight ratios are not required. For nano-launcher systems the
ballistic coefficients are significantly lower compared to conventional launch vehicles; consequently, lower-g
acceleration profiles are preferred and lead to better-optimized launch trajectories.
Only a few specialized, very expensive launch vehicles have liquid-propelled upper stages with the ability
for throttle and in-space restarts. Examples include the Lockheed Centaur II upper stage, and the Delta IV
upper stage based on the Pratt and Whitney RL10B engine. These vehicles are almost universally reserved for
launching expensive government-owned reconnaissance, communications, or command & control satellites.
A restartable and throttleable hybrid rocket motor could provide a safer alternative to liquid engines and
a higher efficiency alternative to cold gas and monopropellant thrusters for a secondary-payload satellite,
while still providing multiple-use capabilities for station keeping and orbit transfers. Additionally, an optimal
thrust profile can be implemented such as a ramp up for a rocket assisted take off (RATO) of a UAV, where
high initial thrust could damage the airframe, for minimal drag of a sounding rocket or launch vehicle during
endoatmoshperic flight, or a ramp down for controlled descent of a planetary lander.
Downloaded by "Indian Institute of Technology, Bombay" on March 31, 2025 | http://arc.aiaa.org | DOI: 10.2514/6.2012-4200

While open-loop throttling can be accomplished with few additions to the oxidizer feed system, an open-
loop throttling system cannot adjust for motor variability. In addition to motor variability induced by
inconsistencies in manufacturing processes and operating conditions, hybrid rocket motors are inherently
variable by nature. Hybrid rocket thrust is more closely coupled with oxidizer mass flux than with chamber
pressure. Because mass flux is driven by surface friction along a turbulent boundary layer, hybrid motor
thrust is inherently chaotic. Some motor variability exists even if identical motors are used under identical
conditions. Closed-loop controlled throttling adds slightly more complexity to the throttling system than
does open-loop throttling, but allows the throttling system to compensate for motor variability. Controlled
throttling on a hybrid rocket allows for constant thrust or chamber pressure set points that greatly minimize
motor-to-motor thrust variability.
A high degree of motor-to-motor variability may be acceptable for experimental vehicles but will not
secure FAA certification for non-experimental, commercial spaceflight operations. Additionally, motor-to-
motor variability produces significant thrust asymmetries for clustered hybrid motor configurations, which
represents a significant hazard and currently precludes using multiple hybrid motors for launch vehicles.
Combustion physics require a longer aspect ratio (length to width ratio) for hybrid motors when compared
to their liquid and solid counterparts. Structural components associated with this high aspect ratio make
building a single hybrid motor that can produce sufficient thrust and impulse to achieve orbital velocity
difficult. For clustered motors, a closed-loop throttle controller can significantly reduce demand on launch
vehicle aerodynamic controls and structures.

C. Hybrid Rocket Throttling History

The ability of hybrid rocket motor designs to be throttled has been demonstrated since the mid-1950’s and
open-loop throttling has been implemented on several projects.9 The first published mention of hybrid
rocket throttling was made in 1956.10 In the 1960’s, various organizations in the United States and Europe
developed throttleable hybrid rocket motors for use in sounding rockets, aerial target drones, tactical missiles,
and space launch systems with turndown ratios as high as 10:1.11–16
In the 1970’s and early 1980’s, no significant research was performed on hybrid rockets due to the
operational success of solid and liquid rocket systems developed in the 1960’s. Starting in the mid-1980’s,
the American Rocket Company (AMROC) developed several throttleable hybrid rocket motors in various
sizes and configurations.17–22 In the late 1980’s and early 1990’s NASA investigated throttleable hybrid rocket
launch vehicle boosters with various industry partners.23–25 Lockheed Martin demonstrated a throttleable
hybrid sounding rocket in the late 1990’s.26
More recently, several academic institutions have developed throttleable hybrid rockets. The University of
Arkansas at Little Rock (UALR) developed a lab scale hybrid rocket motor that was used to perform plume
spectroscopy experiments. The oxidizer delivery system could throttle the mass flow rate of the oxidizer
between 18 and 37 g/sec using a Teledyne-Hastings HFC307 mass flow controller.27 Stanford University
developed a custom-made throttling plate for the Peregrine sounding rocket that mates to the injector face
inside the injector manifold.28, 29 The plate is rotated to control the oxidizer mass flow rate between 50-
100% of the nominal value. A series of static test fires performed at Purdue University demonstrated a
throttle-down profile analogous to a powered descent/landing profile and a square wave profile analogous to

5 of 42

American Institute of Aeronautics and Astronautics

 a boost/sustain/boost profile that would be used for a tactical missile flight.30 Throttling with a turndown
ratio of 10:1 was accomplished using a Habonim control valve with a linear flow profile.

II. Research Overview and Summary

This paper summarizes the development of a closed-loop throttle controller for a lab scale nitrous oxide
and hydroxyl-terminated polybutadiene hybrid rocket motor. The development sequence required to design,
test, and verify this closed-loop control system was; 1) development of a semi-analytical closed-loop motor
simulation to tune system gains and predict controller response, 2) experimental classification of system
response parameters using open-loop throttling tests on both a cold-gas N2 O system and a labscale N2 O-
HTPB hybrid rocket motor, 3) experimental evaluation of thrust stability limits while deep throttling the
N2 O-HTPB hybrid rocket motor, 4) static-test demonstration of a closed-loop thrust controller on a cold-
Downloaded by "Indian Institute of Technology, Bombay" on March 31, 2025 | http://arc.aiaa.org | DOI: 10.2514/6.2012-4200

gas N2 O system, 5) static-test demonstration of closed-loop thrust control on the N2 O-HTPB hybrid rocket
motor using sensed thrust feedback, and 6) static-test demonstration of closed-loop chamber pressure control
on the N2 O-HTPB hybrid rocket motor using sensed pre-combustion pressure transducer feedback.
Development of the closed-loop system simulation was necessary to provide an immediate prediction of
system behavior and eliminated extraneous system classification testing. The simulation allowed the system
gains to be quickly tuned by sweeping through a series of gains and observing the simulation response.
System parameters such as time constants of step-input responses, valve opening ranges, and instrument
noise were determined through open-loop testing. An open-loop deep throttle test was performed to evaluate
the limits of stable thrust in the test motors.
The first series of closed-loop throttle tests was performed on a nitrous oxide cold-gas system. Cold-flow
tests were performed first because of their reduced cost and complexity. Multiple tests could be performed
in succession without replacing fuel grains and ignitors, making it easier to compare results from multiple
gain sets to simulator results.
Following cold-flow testing, a thrust-feedback closed-loop throttle controller was demonstrated on a N2 O/
HTPB hybrid rocket motor. Much of the controller setup was the same as for the cold-flow tests. Only the
thrust profile used in the simulator and controller and the gains used for the controller needed to be altered.
Chamber pressure-feedback tests followed the thrust-feedback tests. The same basic form of controller was
used as for the thrust-feedback controllers with the appropriate terms modified for chamber pressure.

III. System Overview

Control laws used in the throttle control system were greatly influenced by the hybrid motor oxidizer
feed system and motor configuration. Much of the infrastructure needed to perform throttle testing already
existed at USU prior to this study.31–33 New piping and instrumentation were added to the existing test
stand to incorporate the throttle control system. All of the other necessary instrumentation and hardware
remained unchanged from previous research projects. This section describes the design and development of
the throttling system hardware and its influence on the control algorithms.

A. Throttle Control System Component Selection and Evaluation

Figure 1 shows a functional block diagram of the throttle control system and the signal flow between each.
An automation controller is used to send a positioning signal to a valve actuator, which repositions the
valve to the command position. The position of the control valve determines the sensed thrust or chamber
pressure, which is sent to the automation controller to finish the signal flow loop in the throttle control
system.
Parameters considered during valve selection were 1) range of control authority, 2) weight, 3) cost, 4)
availability, and 5) linearity of flow profile. Because of cost and schedule limitations, much of the hardware
was not flight weight; however, when convenient and affordable flight weight hardware was used. The valve
needed to provide control authority through typical operational ranges and still be able to reach nominal
thrust levels when fully open. Because the controller could be made to handle non-linear flow profiles, flow
profile linearity was considered but was given low priority. A highly non-linear flow profile creates precision
problems in flow regions where small changes in valve position make large changes in thrust, and delays

6 of 42

American Institute of Aeronautics and Astronautics

 in regions where large changes in valve position are required to modulate thrust. Control laws are more
straightforward and robust if the valve flow profile is linear, but a controller that can handle non-linear
flow profiles should be able to easily handle linear flow profiles. Matching the remote control actuator to
the control valve is a key element of the design process. Parameters considered when choosing an actuator
included 1) weight, 2) torque capacity, 3) power requirement, and 4) actuation speed. The actuator needed
to be able to position the control valve quickly and accurately with minimal bulk and weight. A flight-weight
throttling system requires the valve actuator and power supply to be lightweight and compact.
The following subsections will describe the systems engineering process used to select and size the control
valve and accompanying actuator.

Control
Automation 85 Ω
Downloaded by "Indian Institute of Technology, Bombay" on March 31, 2025 | http://arc.aiaa.org | DOI: 10.2514/6.2012-4200

0-20 mA 0-1.7 VDC Valve

Controller Resistor
Actuator
± 30 mV/

Linkage
0-5 V

Battery

Load Cell/
Control
Pressure Plumbing/Motor Physics
Valve
Transducer

Figure 1: Throttle control system functional block diagram.

1. Control Valve Type Selection

Many valve types were considered during the control valve selection process. Pintle valves are the valve of
choice for most aerospace throttling applications and have long been in use in air-breathing and liquid fuel
rocket engines and have even seen previous use in hybrid rockets.34 However, typically pintle valves are
custom made for an application and were considered cost prohibitive for this initial evaluation project. More
conventional valve designs were considered.
Globe and gate valves have a wide range of control authority and their orifices can be shaped to provide
linear throttling, but linear actuators are required for quick actuation.35 With linear actuators globe and
gate valves are quite large and would never be considered for flight applications. Other valve types that would
require pulse width modulation (PWM) such as solenoid, diaphragm, and pinch valves were also considered.
The nature of a PWM controller can cause rapid, repeated hard starts which increases the probability of
rupturing the motor case. These valve types were held in reserve as options to be considered only if less
risky control valve options prove ineffective. Ball and butterfly valves are cheap, widely available, and due
to their rotary actuation make size- and weight-efficient control valves. Of these types, ball valves are more
common and are available with a wider variety of port geometries for control applications, and were chosen
for the control valve.

2. Control Valve Sizing

Several approaches for reducing the weight of the control valve were examined. Valves made from titanium
or aluminum are much lighter than steel, but also considerably more expensive and have a more limited
selection. Aluminum valves typically have lower operational pressures than were necessary for this project.

7 of 42

American Institute of Aeronautics and Astronautics

 A compromise that allows significant weight reduction at moderate cost was to use a steel valve with the
amount of steel in the body minimized. Criteria used to minimize the amount of steel in a valve include
port geometry that maximizes flow rate and the smallest valve size that could handle the required flow rates.
Preference was given to low-profile versions of the valves that met the other two criteria.
Figure 2a shows how the percentage of the maximum flow coefficient changes with the % MVT of the
valve for several different ball valve port geometries.36, 37 For brevity throughout this document, valve
positions will be defined as a percentage of the maximum travel distance of the valve. For a ball valve, the
maximum valve travel (MVT) is 90◦ and the valve position is reported as 0%-100% MVT, corresponding to
a 0◦ -90◦ rotation. The flow coefficient, Cv , is a constant that relates the geometry to the flow capacity of
a valve. Here it is defined as the number of U.S. gallons of 60◦ F water per minute that will travel through
a valve with a 1 psi pressure drop. For linear control over a wide range, a slotted-port ball valve is ideal
but the lower maximum Cv requires a much larger valve to achieve the required flow rates compared to the
wider port geometries. While Cv is the industry standard to represent the flow rate capacity of a valve, the
Downloaded by "Indian Institute of Technology, Bombay" on March 31, 2025 | http://arc.aiaa.org | DOI: 10.2514/6.2012-4200

effective port area provides a better representation of the linearity of a valve’s flow profile. Figure 2b shows
the effective port area as a function of valve position for a full-port ball valve. Although the flow profile of a
circular-port ball valve is non-linear, the open port provides the most weight efficient flow control ball valve.
Because demonstration of throttling control using a flight-weight valve was desired, a standard circular-port
ball valve was ultimately selected as the control valve.

100 1
30° V-port
90 60° V-port 0.9
90° V-port
80 0.8
Full Port
% Max. Effective Port Area

70 Slotted Port 0.7

60 0.6
% Max. Cv

50 0.5

40 0.4

30 0.3

20 0.2

10 0.1

0 0
0 10 20 30 40 50 60 70 80 90 100 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
% MVT % MVT

(a) (b)

Figure 2: (a) Percent maximum Cv vs. opening percentage for different ball valve port geometries and (b)
Port area percentage vs. % MVT for a circular-port ball valve.

The valve was sized so that the pressure drop through the port was not sufficient to cause cavitation with
the valve fully open. This was mainly done to minimize flow losses and prevent unpredictable flow regimes
during non-throttled tests. During warmer months, the feedline pressure stays approximately 140 kPa above
the saturation pressure of nitrous oxide. To provide a safe margin, a pressure drop no more than 50 kPa
was allowed through the throttling valve when fully open. The port diameter required to maintain a given
pressure drop is given by38 s
4ṁp
dp = √ , (1)
πCdv 2ρ∆P
where ṁp is the mass flow rate through the port, Cdv is the discharge coefficient of the valve, ρ is the fluid
density, and ∆P is the pressure drop across the valve. Assuming a discharge coefficient of 0.8 for the valve,
a 7.8 mm orifice is required. The closest low-profile valve found was a 1/2” valve with a 5/16” (7.9 mm)
orifice sold by McMaster-Carra . The actual discharge coefficient was closer to 0.7 and the pressure drop was
around 70 kPa, but was still within the acceptable limit.
a McMaster-Carr Item# 45395K105, http://www.mcmaster.com/#catalog/118/433/=i4e9p2

8 of 42

American Institute of Aeronautics and Astronautics

 3. Control Valve Actuator
The actuator chosen for the control valve was an Invenscience LC Torxis Servo i00600b rotary servomotor
with a built-in Pololu jrk 21v3c position controller. A servo was chosen as the actuator because they see
typical use in RC aircraft and robotics applications that require mobility; hence they are small, light, fast,
and are designed to run on battery power. Most commercial control valve actuators are designed for longevity
in varying conditions. They tend to be large, relatively slow, and require more power input than could be
provided by a battery.
The servo is capable of producing 1600 oz-in (11.3 N-m) of continuous torque at 12 VDC and 3200 oz-in
(22.6 N-m) of peak torque with a peak current draw of 3 A. The peak torque output is approximately four
times the static opening torque of the valve at the oxidizer feed pressure, which insures that valve damping
will not hinder the speed of the controller. A custom mounting bracket and linkage were designed to attach
the servo to the valve. The servo can be controlled by either a PWM or 0-5 VDC analog input command
Downloaded by "Indian Institute of Technology, Bombay" on March 31, 2025 | http://arc.aiaa.org | DOI: 10.2514/6.2012-4200

signal. The 0-5 VDC analog servo input corresponds to a 0◦ -270◦ full range of servo rotation.

Control
Automation 85 Ω
0-20 mA 0-1.7 VDC Valve
Controller Resistor
Actuator
± 30 mV/

Linkage
0-5 V

Battery

Load Cell/
Control
Pressure Plumbing/Motor Physics
Valve
Transducer

Figure 3: Throttle valve assembly.

For convenience, power is provided by a 12-volt lead acid battery for static testing. However, even at
peak draw the servo only requires 36 Watts of power. Controlling thrust for one minute of powered flight at
peak current draw would require a battery with rated to 50 mAh. It could be powered by a lithium polymer
battery that weighs only a fraction of a kilogram. Weighing in at about 1 kg, the valve and servo assembly
is potentially flight weight. Figure 3 shows the throttle valve assembly as it fits into the functional block
diagram in Fig. 1. Figure 4 shows the relation between the throttle valve assembly and the rest of the test
cart system piping.
b http://www.invenscience.com/index files/Page923.htm
c http://www.pololu.com/catalog/product/1392

9 of 42

American Institute of Aeronautics and Astronautics

 B. Mobile Nitrous oxide Supply and Testing Resource (MoNSTeR) Cart
All of the equipment required for hybrid motor testing could not be permanently stored in the test cell
where static motor testing takes place because it is a shared resource and partially open to the environment.
Everything required for static motor tests was built onto a cart that could be removed from the test cell and
stored elsewhere. The Mobile Nitrous oxide Supply and Testing Resource (MoNSTeR) cart was custom built
for hybrid rocket research at Utah State University (USU). Figure 4 shows the piping and instrumentation
diagram (P&ID) for the hybrid motor test arrangement with the throttling valve highlighted. The MoNSTeR
cart features an oxidizer delivery system, a modular thrust balance platform, and all of the associated system
piping, instrumentation, and hardware required for hybrid rocket testing.
Downloaded by "Indian Institute of Technology, Bombay" on March 31, 2025 | http://arc.aiaa.org | DOI: 10.2514/6.2012-4200

4 Load Cell
5 7 Pressure Gauge
Pressure Transducer
Thermocouple
2 2 2 2 6 Burst Disk/ Safety Valve
Pneumatic Valve
Ball Valve
7
3
8 12
7
9
11 To Small Scale
Thrust Stand
1 1 1 10
1 N2O “K” Size Cylinder 7 Flex/Thin Tubing
2 He “K” Size Cylinder 8 Venturi Flow Meter
3 N2O Feed Manifold 9 98 mm Motor
4 He Feed Manifold 10 CO2 “K” Size Cylinder
2200 psi 800 psi 400 psi
5 He Regulator 11 Throttle Valve
6 Run Tank 12 Throttle Valve Actuator

Figure 4: MoNSTeR cart P&ID.

To allow sufficient mass flow rates with minimal line losses, a predetermined mass of N2 O oxidizer,
nominally 500 g/s of burn time, was delivered to a closely coupled “run tank” from a series of “K” sized
industrial pressure cylinders. Nitrous oxide vapor pressure is highly dependent on temperature. Because the
test cell is open to the environment, the vapor pressure of the nitrous oxide in the MoNSTeR cart’s run tank
can be below saturation pressure if not controlled. Thus, the run tank was pressurized by gaseous helium
(He) to insure a constant feed pressure during the entire length of the burn. The He “top pressure” was set
by a manual regulator and was maintained near 5650 kPa for throttling tests.
The top pressure keeps the N2 O above saturation pressure for the entire run and insures a single-phase
liquid flow through the throttle valve for any expected ambient temperatures. The design motor chamber
pressure was 2760 kPa. A pneumatic run valve upstream of the throttle valve was triggered by an electronic

®
solenoid valve, and was automatically controlled by the instrumentation software.
Oxidizer mass flow was sensed by two vertical Omegadyne LCCD-100 (445 N) load cells mounted on
the run tank, and by an inline Venturi flow meter mounted in the oxidizer feed-line just ahead of the run

10 of 42

American Institute of Aeronautics and Astronautics

 ®
valve. Differential Venturi flow meter pressure was measured using twin Omegadyne PX409-1.0KA5V
®
®
(0-6900 kPa) absolute pressure transducers. Axial load was sensed by an Omegadyne LCCD-500 (2225 N)

®
load cell and chamber pressure was sensed using an Omegadyne PX409-1.0KA5V (0-6900 kPa) absolute
pressure transducer mounted to the motor cap. An Omegadyne Type-K thermocouple was mounted at
the aft-end of the motor case to sense motor case temperature and thermal soak-back following the end of
the burn. All instrumentation was excited using a 10 VDC power source. The output response for the load
cells is 3 mV/Volt, and 0-5VDC for the pressure transducers.
The motors were mounted in a thrust balance on the MoNSTeR cart. A motor mounting bracket in the
thrust balance is supported on the sides by five ball and clevis joint linkages, two in the vertical and three
in the horizontal direction. Motion is constrained in the vertical and horizontal directions and rotations
are constrained about all three principal axes by the linkages. The axial load cell is attached between the
fore end of the motor mounting bracket and a rigid thrust beam using ball and clevis joints on either side.
Linkages and the axial load cell were aligned to within 0.2◦ of the principal axes using precision squares and
Downloaded by "Indian Institute of Technology, Bombay" on March 31, 2025 | http://arc.aiaa.org | DOI: 10.2514/6.2012-4200

inclinometers.

C. Data Acquisition and Automation

®
Two National Instruments data acquisition (DAQ) systems and control devices manage motor fire, control
throttling, and log test data. An NI-compact DAQ 9174d (cDAQ) 4-slot bus controller with multiple analog
input (16-bit), analog output, digital output, and thermocouple modules (24-bit) manage the majority of
the measurements and valve control. A National Instruments Compact Fieldpoint 2020e (cFP) automation
controller is used to manage the throttling valve controllers. An analog input module on the cFP allows
input measurements of thrust and chamber pressure for control loop feedback and to receive communication
signals from the cDAQ. Operators and experimenters are remotely located in a secure control room separated
from the test area. Communications to the test stand are managed by an operator-controlled computer via

®
universal serial bus (USB) using amplified extension cables. General control and measurement functions are
controlled by a LabVIEW virtual instrument (VI) hosted on the control computer. A separate throttle
valve controller VI was run on the cFP. Controller data was stored locally on the cFP and retrieved following
the test. Figure 5 presents a flowchart showing the communication signals between the instrumentation and
control devices.

Motor Load Cell/

Pressure Transducer
Located in Control Room
Test Computer

Located in Test Cell

cDAQ cFP Motor USB Signal
Digital Signal
Analog Signal
Physical Connection
Oxidizer Feed
System Valves/ Throttling Valve
Instrumentation

Figure 5: Instrumentation and control data flowchart.

The cFP automation controller was used to position the servo. The servo has a 270◦ total rotation range,
but only 90◦ was required to actuate the control valve. The 0-20 mA signal output of the 12-bit cFP analog
output module was converted to a 0-1.7 VDC input (5/3 of the 0-5 VDC command input for the 270◦ total
range) for slightly over 90◦ of control authority over the valve with a 0.022◦ resolution. The current to
voltage conversion of the analog output module control signal was accomplished by placing an 85 Ω resistor
in series with the signal wire. Position feedback of the servo to the cDAQ was provided by an internal rotary
d http://sine.ni.com/nips/cds/view/p/lang/en/nid/207535
e http://sine.ni.com/nips/cds/view/p/lang/en/nid/11572

11 of 42

American Institute of Aeronautics and Astronautics

 potentiometer. An Isolation amplifier was inserted between the load cell signal split and the cFP to eliminate
spurious noise.

D. Test Motor
For these tests a commercially available Cesaroni 98-mm solid-rocket motor was modified by replacing the
original motor cap with a custom-designed motor cap with a single port oxidizer injector. A threaded pressure
transducer port was also installed in the motor injector cap to allow for chamber pressure measurements.
The stock nozzle holder was replaced by a custom nozzle holder with a nozzle that had a larger throat
diameter than the stock nozzle holder could support. To reduce run-to-run variability due to nozzle erosion,
nozzles fabricated from a single piece of high-density graphite replaced the original manufacturer-supplied
phenolic nozzle.
The nozzle has a 4.2:1 expansion ratio and has a design throat diameter of 1.7 cm. Two Estes “mini
Downloaded by "Indian Institute of Technology, Bombay" on March 31, 2025 | http://arc.aiaa.org | DOI: 10.2514/6.2012-4200

A” class 10-gram solid rocket motors were inserted into the injector cap as ignitors. Electronic matches
burned by a 12 volt DC signal ignited these small motors. The ignitors were replaced after each test firing.
Additional advantages provided by this configuration are a ready-made flight-weight motor and the ability
to rapidly reload between motor tests.
®
®
HTPB fuel grains were cast using the commercially available Sartomer Poly bd R-45M polybutadiene
resin and PAPI 94 MDI curative. Sartomer R-45M has a polymerization factor of approximately 50 and
a molecular weight of 2800 kg/kg-mol.39 PAPI 94 is a polymethylene polyphenylisocyanate produced by
®
Dow Plastics Inc.40 The formulation contains methylene diphenylene diisocyanate (MDI) in proprietary
proportions. The curative has an average molecular weight of 290 kg/kg-mol. The nitrogen, carbon, oxygen
(N-C-O) bonds in the MDI react with the hydroxyl (OH) terminations in the polybutadiene resin to cure
the fuel grain.
For these tests carbon black was added to the mixture to insure opaqueness and prevent radiative heating
of the fuel grain and motor case liners. HTPB/MDI/carbon black mass proportions were set at 87%/ 12.5%/
0.5%, respectively. Past experience has determined that these proportions assure adequate fuel grain cure
and material hardness. The resin and curative were mixed in a commercial paint mixer that was sealed and
fitted so that the fuel mixture could be placed under a vacuum during the mixing process. A commercial
H-VAC vacuum pump was used to remove gas bubbles created in the fuel grain during the mixing process.

®
The de-gassed mixture was cast in cardboard sleeves with a 2.67 cm OD polyvinyl chloride (PVC) pipe used
as a mandrel. Before casting, the mandrel was coated with Ease Release 400 mold release agent to insure
proper release after the fuel grain cured.
Each fuel grain was approximately 57.15 cm in length, 8.26 cm in diameter, the initial fuel port diameter
is 2.67 cm, and post combustion chambers are 5.66 cm in diameter and 1.27 cm deep. The mean density
of the HTPB fuel grains used for these tests was approximately 966 kg/m3 , and the cast fuel grains had a
mean mass of 2.50 kg. Fig. 6 presents a schematic of the test motor.

IV. Closed-Loop Throttle Controller Development

Prior to the throttling experiments, the motor configuration described in Section III was fired multiple
times and was found to produce a mean thrust of approximately 786 N with a run-to-run standard deviation
of approximately 30 N.33 Figure 7 compares nine motor burn thrust time histories. Even though the motors
were constructed identically and the fire control process was computer automated, the run-to-run thrust
profiles vary substantially. This variability is inherent to N2 O-HTPB hybrid motor designs and can be
caused by a number of factors including oxidizer temperature, ambient pressure, air bubbles in the fuel grain,
improper mixing of fuel grain constituents, and nozzle erosion. Motor variability is a primary motivation for
the closed-loop throttling methods presented in this paper.

12 of 42

American Institute of Aeronautics and Astronautics

 Retaining Post-Combustion Retaining
Ring Pre-Combustion Chamber
Graphite Ring
Chamber Graphite
Motor O-ring Insert Insert O-ring
Cap

Igniter

8.26 cm
5.66 cm
Port
2.67 cm

Igniter Port
Cap Fuel Grain Nozzle
Downloaded by "Indian Institute of Technology, Bombay" on March 31, 2025 | http://arc.aiaa.org | DOI: 10.2514/6.2012-4200

Casting Motor Port

Injector Tube Case Fuel

2.54 cm
Phenolic Grain

1.27 cm
Liner
2.54 cm

57.15 cm

Figure 6: Test motor schematic.

1000

900 Motor
Shutoff
800

700

600 Piping Filling Steady State

Thrust (N)

Dynamics
500 Transient

400

300

200
First Igniter
100

0
0 2 4 6 8 10 12 14
Time (s)

Figure 7: Thrust time histories for previous HTPB motor variability burns.

A. Evaluation of Chamber Pressure and Thrust as Throttle Control Feedback Measurements

The two most obvious measurement feedback options for a hybrid rocket throttle controller are chamber
pressure and thrust. Because both feedback methods are viable options, both were evaluated as part of this

13 of 42

American Institute of Aeronautics and Astronautics

 project. Either option may be preferable under the conditions in which it is being implemented.
Thrust-feedback is convenient for static testing because a load cell can be attached at the front of the
motor. However, in-flight the thrust is measured using the acceleration and rotation rates of the vehicle.
Coupling the controller with the vehicle dynamics eliminates the possibility of a stand-alone throttle controller
and can make the controller much more complicated in some situations. Additionally, load cell based thrust
data tend to be noisy and may feed back unwanted test stand structural harmonics to the control system.
The sensed thrust must be significantly filtered and can complicate the overall control implementation. While
chamber pressure feedback does not provide direct control over thrust, chamber pressure is the strongest
driver for motor mass flow and in the absence of o/f shift and nozzle erosion there is a mostly linear
relationship.
It should be noted that the motor cap pressure transducer does not measure the true chamber pressure,
but rather the pre-combustion chamber pressure, Popc . References to “chamber pressure” in discussions of
controller signals and measured values are actually Popc . For a large, circular port the chamber pressure is
Downloaded by "Indian Institute of Technology, Bombay" on March 31, 2025 | http://arc.aiaa.org | DOI: 10.2514/6.2012-4200

close in value to Popc , and Popc was consistently used for all controller parameters.

B. Control Law Development

Controllers used during this project were either integral (I) or proportional-integral (PI) controllers. Due to
the high noise-to-signal ratio during cold-flow testing, a proportional gain could not be used without causing
servo jitter. PI controllers were used for both hot-flow feedback mechanisms. Figure 8 shows the general
structure of the closed-loop throttling control law.

Command Non-linear (P)I System

+−
Profile Cancellation Gains Dynamics

Non-linear Lowpass
Feedback
Cancellation Filter

Figure 8: General form of control loop used for closed-loop throttling.

Although the same control structure was used for all tests, three different operational gain ranges were
evaluated for the cold-flow controller and the hot-flow thrust-feedback and chamber pressure-feedback con-
trollers. Motor thrust and chamber pressure require a non-linear controller to be controlled directly due to
the non-linear relationship between the valve position and its effective port area. In order to simplify the
controller, non-linear cancellation was used to control the thrust or chamber pressure indirectly using the
linearly controlled valve position. The relationship between thrust or chamber pressure and valve position
was determined experimentally through open-loop testing. This relationship allows valve position to be
parametrized as a function of thrust or chamber pressure inputs. Thrust and chamber pressure command
and feedback signals were subsequently converted to an estimated valve position and the control loop was
closed around the error in valve position. By closing the loop around valve position error, the non-linearity
between the valve position and effective port area was separated from the controller and effectively canceled.
Large deviations in motor performance from the motors used during open-loop testing could potentially
cause instabilities in the controller through the non-linear cancellation, but variances in performance of the
test motor configuration were not large enough to cause instability.
In the controllers, especially the cold-flow controller, feedback noise was a significant factor in controller
stability. A lowpass filter was used to remove higher frequency noise from the feedback signal. Potential
high-frequency instabilities are discussed in Subsection IV.C. All controllers were run at a loop frequency of
100 Hz. The loop frequency was limited by the processing power of the cFP. Steps taken to reduce processing
load of the cFP included limiting data recording to when the control loop was running, building a data array

14 of 42

American Institute of Aeronautics and Astronautics

 and recording data at completion of the test instead of recording real-time, and limiting the number of global
variables.

1. Cold-Flow Controller
Preliminary closed-loop throttle tests were performed using a top-pressured nitrous oxide cold-gas system.
Cold-flow tests were performed first because of their reduced cost and risk. Multiple tests could also be
performed in succession without replacing fuel grains and ignitors, making it easier to compare results from
multiple gain sets to simulator results. The cold gas system used for cold-flow testing consisted of all of the
same MoNSTeR Cart feedline piping used for hybrid motor testing and the motor cap and injector from the
98 mm test motor. Helium top-pressurized nitrous oxide was evacuated through the 3.8 mm orifice in the
injector to ambient conditions.
During testing the motor top cap was mounted to the MoNSTeR Cart thrust stand. The 2225 N axial
Downloaded by "Indian Institute of Technology, Bombay" on March 31, 2025 | http://arc.aiaa.org | DOI: 10.2514/6.2012-4200

load cell was intended to read much higher thrust and the noise level of the signal was a significant portion
of the maximum thrust that the cold gas system could produce. To minimize the effect of the signal noise
on controller performance, an integral controller was used. Only step input command profiles were used in
the cold-flow controller. Because of the high noise-to-signal ratio, ramp input response would be difficult to
discern.
A commanded thrust profile was pre-programmed and used as a tracking signal for the controller. To
synchronize the controller VI with the test computer VI, a thrust trigger was built into the controller. When
the controller VI was set to “active” mode in the test VI, the throttle valve was programmed to open to
an initial position and remain there while the thrust level was below the threshold. At the end of the test
the same threshold value was used to signal the controller to stop. A heavily filtered thrust signal was used
to avoid controller initiation due to noise spikes. Thrust levels were low enough during cold-flow testing
that controller overshoot could cause the thrust level to dip below the threshold and turn off the controller
mid-test. To avoid terminating the controller mid-test, the thrust trigger was set just above the normal noise
level and the lower command setting of the controller was high enough to leave a buffer zone. The highest
noise level before test initiation was around 5 N, so the thrust trigger was set to 10 N. Lower command
thrust levels were set no lower than 30 N to ensure that the threshold value was not reached during the test
due to controller overshoot.
Logarithmic functions of the form
 
Ft
x = − ln − + 1 τ + x0 (2)
A

were used in the non-linear cancellation, where x is the valve position in %MVT, Ft is the thrust, and the
other variables are experimentally derived curve fit coefficients. Equation (2) is the inverse of Eq. (22). To
ensure that values returned from the functions were real and finite, an upper thrust limit was set to a value
just smaller than A. If the measured thrust exceeds the limit, the signal is reset to the limit value.

2. Thrust-Feedback Hot-Flow Controller

While the thrust-feedback controller was less sensitive to noise than the cold-flow controller, load cell signal
noise was still an issue. This load cell noise precluded the inclusion of a derivative term in the control law.
Both a step input command thrust profile analogous to a boost/sustain/boost flight profile of a tactical
missile and a step-down/ramp-up flight profile analogous to an optimum altitude profile of a sounding rocket
were used for thrust-feedback demonstration.
Because both the thrust-feedback controller and the cold flow controller used the same load cell input,
much of the same setup was used. The simulator and controller VI required no re-programming, only changes
to a few parameter values. Curve fit coefficients used in the non-linear cancellation were changed to match
the higher thrust of the hot-flow system; and, consequently, the upper thrust limit was changed. Reaching
the threshold thrust value and terminating the controller mid-test was not as much of a concern as it was
for the cold flow controller, so the thrust trigger was set to 20 N to provide a greater buffer from noise spikes
in the load cell signal.

15 of 42

American Institute of Aeronautics and Astronautics

 3. Chamber Pressure-Feedback Hot-Flow Controller
For the chamber pressure-feedback controller an a priori conversion between thrust and chamber pressure was
made to loosely provide thrust control. Thrust and chamber pressure have a slightly quadratic relationship in
the test motor. The experimentally derived equation that was used to match the chamber pressure command
profiles to command profiles used for the thrust-feedback controller,
p
−4a1 c1 + 4a1 Ft + b21 − b1
P0 = , (3)
2a1
is the inverse of the quadratic equation described in Eq. (17).
The signal to noise ratio of the pressure transducer was much lower than for the load cell. A derivative
gain could be used for the chamber pressure-feedback controller. However, a PI controller was sufficient to
Downloaded by "Indian Institute of Technology, Bombay" on March 31, 2025 | http://arc.aiaa.org | DOI: 10.2514/6.2012-4200

demonstrate the controller and required fewer changes to the controller VI. The same thrust trigger was
used to initiate the controller as in the thrust-feedback controller VI. The simulator also required slight
modification to model the chamber pressure-feedback controller.

C. Hot-Flow System Latencies and Sources

Table 1 presents calculated and measured latencies due to servo speed and filling times of the system piping
and combustion chamber. These were considered the governing delays in the system and smaller delays such
as the oxidizer vaporization delay and the boundary layer combustion delay were assumed negligible. Table
1 also presents the sources expected to dominate the throttle controller. A second order Butterworth lowpass
filter was used in the controller to prevent unwanted modes feeding into the controller as instrument noise.
The following subsections describe these latencies and sources in detail.

Table 1: Hot-flow system latencies and sources.

Latency/Source Approximate Maximum Latency/ Lowest Frequency

servomotor 0.8 s
system piping response 0.14 s
filling dynamics 0.01 s
hybrid low frequency non-acoustic mode 50 Hz
test stand resonant mode 30 Hz
acoustics 1/L mode 350 Hz
Strouhal vortex shedding 350 Hz

1. Servomotor
Delays in the servo response were due to the maximum servo speed and the settling time of the built-in
positioning controller. As provided, the factory preset control gains of the internal PID servo position
controller produced a sluggish response that was insufficient for throttling. To speed up valve actuation
times, the gains were modified so that the internal position controller was critically damped. Figure 9
compares the servo response time for a 10% MVT change in position using the original and the modified
gains. Note that the servo response with the original gains was measured using a lower resolution external
potentiometer so the apparent steps in the response are measurement artifacts and do not reflect the actual
response.
With the modified gains, the time constant for valve repositioning was 0.2 seconds and maximum travel
speed was 60% MVT per second. For the range of valve positions required for this project, the maximum
delay due to the servo was estimated to be 0.8 seconds.

16 of 42

American Institute of Aeronautics and Astronautics

 12

10

8
% MVT

6
Downloaded by "Indian Institute of Technology, Bombay" on March 31, 2025 | http://arc.aiaa.org | DOI: 10.2514/6.2012-4200

2
Original Gains
Modified Gains
0
0 0.2 0.4 0.6 0.8 1 1.2
Time (s)

Figure 9: Servo response time with original and modified gains.

2. Pneumatic and Hydraulic Response of the System Piping

The system piping fill delay is calculated by41


vp ρg P̄i − P̄c
τp = K , (4)
ṁi 2P̄i
where vp is the system piping volume, ρg is the density of the gas phase of the fluid in the system piping,
ṁi is the mass flow rate through the injector, P̄i is the average injector manifold pressure, P̄c is the average
combustion chamber pressure, and the total compressibility parameter, K, is


Xox /γox + βs P̄ (ρg /ρl ) − P̄ ∂Xox /∂ P̄ s
K= 2 , (5)
[Xox + ρg /ρl ]

where Xox is the quality of the oxidizer, γox is the oxidizer ratio of specific heats, ρl is the liquid phase
density of the oxidizer, and P̄ is the average pressure in the system piping. Because a single port injector
was used for this project, there was no injector manifold and the same values were used for P̄ and P̄i . The
liquid compressibility factor, βs , is the inverse of the bulk modulus of the oxidizer, defined as42
P2 − P1
MB = , (6)
1 − ρ1 /ρ2

which describes the difference in the amount that a substance compresses between two different pressures, P1
and P2 ; ρ1 and ρ2 are the densities corresponding to those pressures. The bulk modulus has a nearly linear
relationship with temperature. Using temperatures and pressures in the ranges expected in the oxidizer feed
line, a linear function of temperature was developed to approximate MB .
Two-phase thermodynamic properties were determined using a Helmholtz model.43–45 Measured temper-
atures, pressures, and mass flow rates were assumed to be representative of the average values in the system
piping. The calculated values of τp ranged between 0.02 and 0.14 seconds for the full throttling range.

17 of 42

American Institute of Aeronautics and Astronautics

 3. Combustion Chamber Filling Dynamics
The combustion chamber fill time is given by41

fc (γ) RTc L∗
τc = , (7)
Cdn RT̄ c∗
where Cdn is the nozzle discharge coefficient, RTc is the product of the gas constant and combustion tem-
perature at the nozzle entrance, RT̄ is the product of the average gas constant and temperature in the
combustion chamber, L∗ is the ratio of the chamber volume to the nozzle throat area, c∗ is the characteristic
velocity, and
 (γ+1)/(γ−1)
ηc γ + 1
fc (γ) = (8)
γ 2
Downloaded by "Indian Institute of Technology, Bombay" on March 31, 2025 | http://arc.aiaa.org | DOI: 10.2514/6.2012-4200

where ηc is the combustion efficiency and γ is the ratio of specific heats of the combustion products. Com-
bustion properties c∗ , Tc , T̄ , γ, and the average and nozzle molecular weights were calculated using NASA’s
Chemical Equilibrium with Applications (CEA) program.46, 47 The nozzle discharge coefficient was esti-
mated to be 0.98.38 Previous testing showed the combustion efficiency to be about 0.97 for this motor
configuration.33 The calculated combustion chamber fill time was 0.01 seconds.

4. Low Frequency, Non-Acoustical Pressure Oscillations (Hybrid Low Frequency Mode)

The frequency value of the hybrid low frequency mode is calculated using48
 
1 4 ṁox RT̄
fl = 0.234 2 + , (9)
o/f π Lf P0 d2p
where ṁox is the oxidizer mass flow rate, Lf is the fuel grain length, P0 is the chamber pressure, and dp is the
fuel port diameter. Oxidizer mass flow rate and chamber pressure were estimated from experimental data.
The lowest calculated frequencies occur near the end of the burn when the o/f is about 6 and the average
port diameter is about 5 cm. The calculated values of the hybrid low frequency mode occurs between about
50 and 200 Hz.

5. Test Stand Resonance

The test resonant frequency was determined by delivering an impulse to the thrust stand and measuring the
frequency response in the load cell. A Fourier transform of the load cell noise with and without the impulse
is shown in Fig. 10. The peak of the lowest resonant frequency is at approximately 30 Hz. The noise spike
that occurs in both signals in Fig. 10 is unidentified, but is likely radio frequency interference being picked
by the load cell. This noise is removed by the lowpass filter in the controller.

6. Motor Internal Primary Acoustic Mode

Hybrid rocket motor first longitudinal acoustic modes are calculated as48
p
γRT̄
fa = , (10)
2Lc
is the average speed of sound in the chamber and Lc is the combustion chamber length. The lowest acoustical
mode frequency was calculated to be 350 Hz.

7. Vortex Shedding
The Strouhal number is defined as48
fvs l
Sr = , (11)
U
where fvs is a vortex shedding frequency, l is the characteristic length, and U is the fluid velocity. The
vortex shedding frequency is calculated by rearranging Eq. (11) to solve for fvs . In hybrid rocket motors,

18 of 42

American Institute of Aeronautics and Astronautics

 Sr ranges between 0.25 and 0.5. For the throttling valve, the characteristic length was assumed to be the
fully-open port diameter. The fluid velocity is calculated as
ṁox
Uv = , (12)
ρox Av
where ρox is the oxidizer density and Av is the valve port area. At the valve, the lowest vortex shedding
frequency is about 550 Hz. At the aft end of the fuel grain the fluid velocity is calculated as
4 ṁtot RT
Uf = , (13)
π d2p P0

where ṁtot is the total mass flow rate of fluid at the fuel grain exit. The product of the gas constant and
temperature at the fuel grain exit was assumed to be approximately equal to RT̄ and l was assumed to be
Downloaded by "Indian Institute of Technology, Bombay" on March 31, 2025 | http://arc.aiaa.org | DOI: 10.2514/6.2012-4200

the exit diameter of the fuel grain. Based on these parameters Eq. (13) calculates that the lowest vortex
shedding frequency at the fuel grain exit occurs at approximately 350 Hz.

9
LC noise
8 Test Stand Impulse
Resonance
7

6
Y(Thrust) N

4 Unidentified
Noise
3

0
0 50 100 150 200 250 300 350 400 450 500
Frequency (Hz)

Figure 10: Test stand frequency response to applied impulse.

V. End-to-End System Simulation

Because a PID-type controller was used for throttle the controller, an analytical means for predicting
gains would be difficult to produce. Instead, a simulator was created to provide an immediate prediction
of system behavior for gain tuning and to predict controller behavior. Using a simulator to predict the
controller response for various settings allowed for the controller to be set up without extraneous system
classification testing. The throttling system simulator was developed using MATLAB and Simulink. The
Simulink model is shown in Fig. 11. An accompanying driver code was written in MATLAB. Individual
modules within the simulator are highlighted in Fig. 11 and numbered as 1) proportional-integral-derivative
(PID) controller gains, 2) system piping feed physics, 3) servo, and 4) instrument noise. These modules are
described in detail in the following subsections.

19 of 42

American Institute of Aeronautics and Astronautics

 Downloaded by "Indian Institute of Technology, Bombay" on March 31, 2025 | http://arc.aiaa.org | DOI: 10.2514/6.2012-4200

1 2

20 of 42
3

American Institute of Aeronautics and Astronautics

Figure 11: Simulink throttling system model.
4
 A. PID Controller Gain Model
A PID controller was used in the simulator so that the derivative gain, KD , could be modeled.49 As
mentioned in Section IV.B, the test controllers used either proportional and integral gains, KP and KI , or
only KI , but the capacity to model KD was built in so that no changes to the simulator would be required if
a PID controller is used for future projects. A discrete-time trapezoidal integrator was used for the integral
gain to match the integrator used in the test controllers. Figure 12 shows the PID controller used in the
simulator.

1 In1
DerGainOut

2 In2 Fo=20Hz To Workspace1

Kd du/dt
Downloaded by "Indian Institute of Technology, Bombay" on March 31, 2025 | http://arc.aiaa.org | DOI: 10.2514/6.2012-4200

Gain Derivative
2nd-Order
Filter3 1
K Ts (z+1)
Ki Out1
2(z-1)
Gain3 Discrete-Time
Integrator IntegralGainOut

To Workspace3
Kp

Gain4 PropOut

To Workspace
Figure 12: Simulink PID controller.

The PID controller gains were originally modeled based on the LabVIEW PID controller block to pro-
vide better agreement between the simulator and the controller VI.50 Later, the LabVIEW PID block was
abandoned to provide greater flexibility and the gains were converted to standard format, but the LabVIEW
signal flow path remained. Figure 13 shows how the gain model was integrated into the simulator control
loop.

Command 𝐾𝑃 + System
+
Profile − Dynamics
𝐾𝐼 Trapezoidal
Integrator
Non-linear
+
Cancellation −
𝑑
𝐾𝐷
𝑑𝑡
Non-linear Lowpass
Feedback
Cancellation Filter

Figure 13: Simulator control loop with expanded gain section.

21 of 42

American Institute of Aeronautics and Astronautics

 B. Servo Model
Rather than modeling the individual latencies discussed in Subsection IV.C, an empirically derived transfer
function was used to model the total delay in the thrust or chamber pressure response to changes in valve
position. The assumed form of the transfer function is
1
G(s) = , (14)
s (τ1 + τ1 τ2 s + τ2 )

where τ1 and τ2 are time constants acquired during open-loop testing as outlined in Subsection VI.C below.
The maximum servo speed was modeled by saturating the rate of change in the servo position. Figure 14
shows the simulator servo model.
Downloaded by "Indian Institute of Technology, Bombay" on March 31, 2025 | http://arc.aiaa.org | DOI: 10.2514/6.2012-4200

ServoFeedback 1
Out2
To Workspace7

1
1 Gservoloop du/dt
s
In1
LTI System Derivative1 Saturation2 Integrator1 Saturation3

Figure 14: Simulator servo model.

C. Oxidizer Feed Systems Response Model

Figure 15 shows the basic algorithm used to model the system piping feed physics. Two phase oxidizer
flow through the throttle valve and the injector was modeled using the non-homogeneous, non-equilibrium
(NHNE) model of Dyer, et al.51 Nitrous oxide properties were calculated using a Helmholtz equation of
state as presented by Span and Wagner.43–45 Figure 16 shows the thermodynamic properties used for each
oxidizer state in the oxidizer feed system model. Conditions were assumed to be isenthalpic (constant
enthalpy) between states and isentropic (constant entropy) between a state and the downstream orifice.
Before the system piping model, the servo position signal is transformed to effective valve port area using
a table lookup. Effective valve port areas were calculated by solving for the area required for the system
piping model output to match experimental chamber pressure data at various valve positions, assuming a
constant discharge coefficient. Chamber pressure was used to calculate the effective valve area because it
was needed in both the thrust and the chamber pressure-feedback simulators.
Because the oxidizer system is regulated by a top pressurant, oxidizer conditions upstream of the throttle
valve do not vary greatly between tests. Because the conditions remain relatively unchanged, representative
values averaged from experimental data were used for the first state. These average values were used in the
NHNE model to compute the mass flow rate (ṁin ) as a function of P1 , s1 , and P2 through the valve port
and the conditions at State 2. Table 2 presents these constant values, as well as others used in the system
piping model.
The pressure at State 2 was computed using the Helmholtz equation of state and a Newton-Raphson
solver.52 Temperature and density are required to determine state properties using the Helmholtz equation.
Density, ρ2 , was determined by integrating the net mass flow rate into the post-throttle system piping to
find the mass, or ˆ
m2 = ṁin − ṁout , (15)

and dividing it by the internal volume of the post-throttle system piping, V2 . Temperature was determined
using the Newton solver to match the enthalpies of State 1 and State 2. The pressure in the post-throttle
system piping is calculated as the saturation pressure with the density in the system piping and at the
temperature at which the isenthalpic condition is satisfied.

22 of 42

American Institute of Aeronautics and Astronautics

 Simulator In 𝑉2

𝐶𝑑𝑣 𝐴𝑒𝑣 𝑃1 𝑇1 𝑋1

Injector Feedline
Model
Non-HEM Valve
Port Model

𝑃2

𝐶𝑑𝑖𝑛𝑗 𝐴𝑖𝑛𝑗 𝑇2 𝑋2 𝑚𝑖𝑛

Downloaded by "Indian Institute of Technology, Bombay" on March 31, 2025 | http://arc.aiaa.org | DOI: 10.2514/6.2012-4200

+
−
Integrator
Non-HEM Injector
Orifice Model
𝑚2

𝑚𝑜𝑢𝑡

Chamber Pressure Equation 𝑃0 Simulator Out

Figure 15: Oxidizer feed system response model.

Oxidizer Throttling Post-Throttle Combustion

Injector
Feedline Valve Plumbing Chamber

𝑃1 𝑃2
ℎ1 ℎ2 𝐶𝑑𝑣 ℎ3 𝐶𝑑𝑖𝑛𝑗
𝑇2 𝑚2 ℎ1 ℎ1
𝑇1 𝑠1 𝐴𝑒𝑣 𝑠2 𝐴𝑖𝑛𝑗 𝑃0
𝑠1 𝑋2 𝜌2 𝑠2 𝑠3
𝑚𝑖𝑛 𝑚𝑜𝑢𝑡
𝑋1 𝑉2

State 1 State 2 State 3

Figure 16: Thermodynamic properties used in system piping feed physics model.

Conditions computed by the first NHNE model and the post-throttle system piping pressure were used
in a second NHNE model to compute the mass flow rate through the injector (ṁin ) as a function of P2 , s2 ,
and P0 . Chamber pressure has an approximately linear relationship with injector mass flow rate in the test
motor configuration. Experimental data was used to produce the equation,

P0 = a2 ṁout + b2 , (16)

that calculates chamber pressure based on ṁout , where a2 and b2 are experimentally derived curve fit
coefficients described further in Subsection VI.D. For the pressure-feedback controller model, this chamber
pressure value was output to the simulator.

23 of 42

American Institute of Aeronautics and Astronautics

 Table 2: Constants used in oxidizer feed system response model.

Parameter Symbol Value

valve discharge coefficient Cdv 0.7
oxidizer feedline pressure P1 5500 kPa
oxidizer feedline temperature T1 294 K
oxidizer feedline quality X1 1
post-throttle system piping volume V2 6.32×10−5 m3
injector discharge coefficient Cdinj 0.7
injector orifice area Ainj 1.14×10−5 m2
Downloaded by "Indian Institute of Technology, Bombay" on March 31, 2025 | http://arc.aiaa.org | DOI: 10.2514/6.2012-4200

For the thrust-feedback controller, another experimentally derived equation described in Section VI.D
was used to calculate thrust based on the chamber pressure. Thrust had a slightly quadratic relationship
with chamber pressure in the test motor configuration. The equation used to compute thrust is

Ft = a1 P02 + b1 P0 + c1 . (17)

When the full system piping model was placed in the simulator run times became excessive. In order to
decrease simulator run times, data from the full system piping model was used to make lookup table blocks
so that the imbedded MATLAB functions could be bypassed. The lookup table version of the system piping
model is shown in Fig. 17.

AValveOut1

1 To Workspace8
ValveArea

mdotoxOut1

Product To Workspace1
f(u)
1
P22Gtot
s
m0 xo

2-D T(u) Constant8 Integrator

u1

u2

2-D Lookup
Table mdotinjOut1

1 To Workspace
f(u)
tauMotor.s+1
Transfer Fcn Fcn
1
ChamberPressure pPlumbOut1

To Workspace2
f(u) -K-

rhoP2PP 1/Vol

Figure 17: Simulator system piping model with lookup tables.

24 of 42

American Institute of Aeronautics and Astronautics

 D. Measurement Feedback Noise Model
Figure 18 shows the Simulink noise model for the feedback pressure signal. Instrument noise was modeled in
Simulink by matching the dominant frequency modes in the noise. The Simulink Band-Limited White Noise
block defines the noise power input as the height of the power spectral density (PSD) of the white noise.
The amplitude of the frequency modes were matched by plotting the PSD of the actual noise and adjusting
the noise power level in each Band-Limited White Noise block until the amplitudes matched. A Bandpass
Filter block was used to adjust the bandwidth of each mode to match those in the instrument PSD. Figure
19 shows the PSD of the actual and simulated noise for the pressure transducer.

Bandpass

Band-Limited
Downloaded by "Indian Institute of Technology, Bombay" on March 31, 2025 | http://arc.aiaa.org | DOI: 10.2514/6.2012-4200

Bandpass Filter
White Noise1

1
Bandpass
Out1
Band-Limited
Bandpass Filter1
White Noise2

Bandpass

Band-Limited
Bandpass Filter2
White Noise3

Band-Limited
White Noise4
Figure 18: Pressure transducer noise model.

5
Simulated Noise
4.5 Measured Noise

3.5
PSD (kPa2/Hz)

2.5

1.5

0.5

0
0 50 100 150 200 250 300 350 400 450 500
Frequency (Hz)

Figure 19: PSD of simulated and real pressure transducer noise.

25 of 42

American Institute of Aeronautics and Astronautics

 E. Gain Tuning
Gains were tuned in the simulator by sweeping through a wide range of gains and plotting the results. The
most favorable controller response was chosen from the plots. Figure 20 shows the results from a simulator
run using a step command input with a chamber pressure-feedback controller. In the plots, the green trace is
the command input and the blue and red, respectively, are the raw and filtered response. The proportional
gain was varied between 0.2 and 0.5 by steps of 0.1 and the integral gain was varied between 1.5 and 3.5 by
steps of 0.5. A wider range with larger step sizes was used initially to determine a range of effective gains
and then a finer sweep was made through that gain range to pick final gain values. Table 3 presents gain
values used to produce the plots in Fig. 20 with the “best response” gains highlighted.

1 2 3 4
3000
Downloaded by "Indian Institute of Technology, Bombay" on March 31, 2025 | http://arc.aiaa.org | DOI: 10.2514/6.2012-4200

2000
1
1000
0

3000
Chamber Pressure (kPa)

2000
2 1000
0

3000
2000
3 1000
0

3000
2000
4
1000
0
0 5 10 0 5 10 0 5 10 0 5 10
Time (s)

Figure 20: Plots of controller response to a range of gains.

Table 3: Gain values from plots in Fig. 20.

Row\Col. 1 2 3 4
1 KP = 0.2, KI = 2.0 KP = 0.2, KI = 2.5 KP = 0.2, KI = 3.0 KP = 0.2, KI = 3.5
2 KP = 0.3, KI = 2.0 KP = 0.3, KI = 2.5 KP = 0.3, KI = 3.0 KP = 0.3, KI = 3.5
3 KP = 0.4, KI = 2.0 KP = 0.4, KI = 2.5 KP = 0.4, KI = 3.0 KP = 0.4, KI = 3.5
4 KP = 0.5, KI = 2.0 KP = 0.5, KI = 2.5 KP = 0.5, KI = 3.0 KP = 0.5, KI = 3.5

When a set of gains was selected it was checked to verify that it would work if motor performance varied
from the moderate performance parameters that were used for non-linear cancellation in the controller and
for the system piping model in the simulator. Motor variability was modeled by applying a scaling factor to
the effective valve port area that was input into the system piping model. The scaling factor was applied as
a gain following the table lookup for valve effective area, shown in Fig. 11. Mean motor thrust and chamber
pressure 2-σ variance was less than 8% for all tests performed to date on the 98 mm motor configuration. The
variability was modeled by scaling to ±15% of the effective valve port area, which creates a ±8% variance in
the thrust and chamber pressure. Figure 21 shows nominal simulator results compared to results from 85%
and 115% scaling. Comparison of Figures 20 and 21 shows that motor variability has a similar effect on the
controller to varying the integral gain.

26 of 42

American Institute of Aeronautics and Astronautics

 3000

2500

Chamber Pressure (kPa)

2000

1500

1000

500

0
0 2 4 6 8 10 12
Time (s)
Downloaded by "Indian Institute of Technology, Bombay" on March 31, 2025 | http://arc.aiaa.org | DOI: 10.2514/6.2012-4200

(a) Scale factor = 0.85.

3000

2500
Chamber Pressure (kPa)

2000

1500

1000

500

0
0 2 4 6 8 10 12
Time (s)

(b) Scale factor = 1.00.

3000

2500
Chamber Pressure (kPa)

2000

1500

1000

500

0
0 2 4 6 8 10 12
Time (s)

(c) Scale factor = 1.15.

Figure 21: Controller response for three different scaled valve areas.

VI. Open-Loop Throttle Testing

A. Open-Loop Cold-Flow Testing
Open-loop testing was performed to assess system flow dynamics and to establish a baseline for the control
valve effectiveness. Performing a cold-flow-only open-loop throttle test decouples the dynamics of the oxidizer
delivery system from the more complex dynamics associated with the hybrid rocket motor. During cold-flow
testing, nitrous oxide was evacuated through a 3.8 mm orifice into ambient conditions using the apparatus
described in Subsection IV.B. Figure 22 shows the thrust produced during a cold-flow test. During this test

27 of 42

American Institute of Aeronautics and Astronautics

 the throttling valve started at a closed position and was opened 10% MVT every two seconds until it was
completely open. Plotted points show times when the step input signal was changed. Results showed that
most of the valve’s control authority lies within the 30-70% MVT range. As a result, controlled throttling
was performed mostly withing this range.

80

70 90%
70% 80%
60%
60

50 50%
Downloaded by "Indian Institute of Technology, Bombay" on March 31, 2025 | http://arc.aiaa.org | DOI: 10.2514/6.2012-4200

Thrust (N)

40

30
40% Dsd

20

10

0 30%
0% 10% 20%
-10
0 2 4 6 8 10 12 14 16 18 20
Time (s)

Figure 22: Thrust during cold-flow open loop throttle test.

B. Open-Loop Hot-Flow Testing

Three hot-fire open-loop throttle tests were performed. Results for thrust and chamber pressure are shown
in Fig. 23. During the first test the throttle valve was not adjusted and was simply left completely open
throughout the duration of the burn. For the second test the throttling valve started at 90% MVT and closed
in steps of 5% MVT every second. By the end of the burn the valve was in the 50% MVT open position.
The third test was conducted at the lower end of throttleability. The valve began at the 65% MVT position
and was again closed 5% MVT every second until finishing at the 25% MVT position.
Because the 50-90% MVT throttled test was partially in the region shown to have little control authority
by the cold flow tests, the first valve changes had little throttling effect. As the valve travel entered the
region of more control authority, throttling had the effect of leveling off the increase in thrust and chamber
pressure and at the end of the burn distinct steps are finally discernible. The 25-65% MVT test began in the
region of high control authority. Steps are plainly visible throughout most of the burn duration. The steps
do not level off as they do during cold-flow testing because of the gradual increase in thrust and chamber
pressure during the burn.
Also, as mentioned previously, N2 O-HTPB hybrid motors have a fairly high degree of run-to-run vari-
ability. This variability accounts for the initial peak thrust of the fully-open test being lower than the tests
that were partially throttled. The fully-open test demonstrated that the test motors’ thrust and chamber
pressure increase during the burn.

C. Thrust and Chamber Pressure Response Time Constants

The system thrust response time constant was measured to model system latencies in the simulator. Changes
in thrust due to step changes in the throttle valve position were normalized to a 0-1 range using the normal-
izing relationship
Ft (t) − Fti
Ftn (t) = , (18)
Ftf − Fti

28 of 42

American Institute of Aeronautics and Astronautics

 where Fti is the measured thrust before the step input, Ftf is the measured thrust after settling occurs, and
Ft (t) is the measured thrust at any time during the response. Second order non-linear least squares curve
fits of the form  
1 1 
−t/τ2 −t/τ1
y (t) = 2 1 − τ2 e − τ1 e (19)
ωn τ2 − τ1
were applied to the normalized thrust curves to obtain step input thrust response time constants τ1 and
τ2 and the natural frequency ωn from the time response y (t). Depending on the control law used, y (t) is
either the normalized thrust or pressure response of the system. Figure 24 shows typical normalized thrust
response curves for both cold-flow and hot-flow tests with the throttling valve traveling through the 35-50%
MVT range. For the durations tested, cold-flow thrust can be thought of as primarily a function of valve
position and average time constants can be used. During the hot-flow tests thrust drifts over time. However,
time constant accuracy proved to have little effect on the simulator for the changes in valve position and
thrust durations used for this research. The time variance in the hot-flow step response time constants
Downloaded by "Indian Institute of Technology, Bombay" on March 31, 2025 | http://arc.aiaa.org | DOI: 10.2514/6.2012-4200

were ignored and average values τ1 = 0.0221 s and τ2 = 0.0217 s were used for both cold-flow and hot-flow
simulation. Time constants were the same for thrust and chamber pressure, so the same values were used in
both simulator models. The natural frequency was calculated as part of the curve fit, but was not used in
the simulator.

1000 3500

900
3000
800

700 2500

600
2000
Thrust (N)

P0 (kPa)

500
1500
400

300 1000

200
100% MVT 500 100% MVT
100 50-90% MVT 50-90% MVT
25-65% MVT 25-65% MVT
0 0
0 2 4 6 8 10 12 14 0 2 4 6 8 10 12 14
Time (s) Time (s)

(a) (b)

Figure 23: (a) Thrust and (b) chamber pressure during hot flow open loop throttle tests.

1.2 1.2

1 1

0.8 0.8
Normalized Thrust
Normalized Thrust

0.6 0.6

0.4 0.4

0.2 0.2

0 curve 1 0 curve 1
curve 2 curve 2
curve 3 curve 3
-0.2 -0.2
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
t (sec) t (sec)

(a) (b)

Figure 24: Normalized thrust response curves for (a) cold-flow throttling and (b) hot-flow throttling.

29 of 42

American Institute of Aeronautics and Astronautics

 D. Thrust, Chamber Pressure, and Mass Flow Rate Calibrations
The controllers and simulator requires several computationally efficient conversions between thrust, chamber
pressure, and oxidizer mass flow rate. Expressions used for conversions were obtained from curve fits of the
open-loop test data shown in Fig. 23. Thrust and chamber pressure have a slightly quadratic relationship.
A non-linear least squares curve fit of the form

Ft = a1 P02 + b1 P0 + c1 (20)

was used to describe the relationship between thrust (N) and absolute chamber pressure (kPa). The rela-
tionship between chamber pressure and oxidizer mass flow rate is mostly linear. A linear curve fit of the
form
P0 = a2 ṁox + b2 (21)
Downloaded by "Indian Institute of Technology, Bombay" on March 31, 2025 | http://arc.aiaa.org | DOI: 10.2514/6.2012-4200

was used to describe the relationship between absolute chamber pressure (kPa) and mass flow rate (kg/s).
Thrust can be related to mass flow rate by combining both relationships. Table 4 presents the values of
the curve fit coefficients used in Equations (20) and (21). Figure 25 shows the curve fit data compared to
experimental data. The experimental data shown in Fig. 25a was collected from all of the hot-fire open-loop
tests performed. While o/f shift occurred during each motor burn, it was not a significant factor in the
relationship between thrust and chamber pressure.

Table 4: Curve fit coefficients for thrust, chamber pressure, and mass flow rate calibrations.

Coefficient Value
a1 9.21 × 10−6
b1 0.28
c1 -40.15
a2 8602.47
b2 152.02

E. Thrust and Pressure Curve Fits

Curve fit data required for use for non-linear cancellation in the controllers was obtained during open-loop
testing. Figure 26 shows non-linear least squares curve fits of thrust and chamber pressure as functions of
throttle valve position. Exponential response functions of the form

S = Ae−(x+x0 )/τ + A (22)

were used for the curve fits, where S is the controller input signal (thrust or absolute chamber pressure), x
is the valve position in %MVT, and the other variables are exponential curve fit coefficients that represent
scaling (A), initial offset (x0 ), and time constant (τ ). Equation 22 was rearranged into Eq. (2) for use in
the controllers. Values of the curve fit coefficients used for each controller are displayed in Table 5.

F. Servo Deadband Evaluation

A deadband was built into the servo position controller by the manufacturer to prevent overheating of the
components due to high frequency servo jitter. The deadband limits were measured by commanding a
sawtooth wave input and comparing the response of the servo. Figure 27 shows input signal voltage and
servo position response plots that have been normalized by their respective maximum values for a more
convenient visual representation of the deadband (note that the normalized values are used for illustration
only and were not used to compute the deadband). The measured deadband was ± 0.02 volts or ± 1.25% of
the total travel. Effects of the deadband on the throttle controllers included a lack of response fidelity while
at a constant set point, a small step at the beginning of ramp input profiles, and a slight delay in response to
a ramp input. Although the controller was affected by the servo deadband, the deadband was left in place
to avoid damage to internal servo components.

30 of 42

American Institute of Aeronautics and Astronautics

 800 Test Data
Curve Fit

Thrust (N)
600

400

200

0
0 500 1000 1500 2000 2500 3000
Chamber Pressure (kPa)
(a)
Downloaded by "Indian Institute of Technology, Bombay" on March 31, 2025 | http://arc.aiaa.org | DOI: 10.2514/6.2012-4200

4000
Chamber Pressure (kPa)

Test Data
3000 Curve Fit

2000

1000

0
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35
Mass Flow Rate (kg/s)
(b)

Figure 25: Chamber pressure, thrust and mass flow calibrations compared to experimentally measured values.

800

600
Thrust (N)

400

200 Curve Fit

Experimental Data
0
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
% MVT
(a)

3000
Chamber Pressure (kPa)

2000

1000
Curve Fit
Experimental Data
0
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
% MVT
(b)

Figure 26: Experimentally derived curve fits for (a) thrust and (b) chamber pressure as a function of %
MVT.

31 of 42

American Institute of Aeronautics and Astronautics

 Table 5: Exponential curve fit coefficients used in controllers.

Controller A x0 τ
Cold-Flow 75.04 -0.2527 0.2116
Thrust 819.1 -0.2268 0.1586
Chamber Pressure 2781 -0.2185 0.1522
Normalized Voltage/Servo Position

1.02 Voltage
Position
1.01
Downloaded by "Indian Institute of Technology, Bombay" on March 31, 2025 | http://arc.aiaa.org | DOI: 10.2514/6.2012-4200

1
0.99 ĞĂĚďĂŶĚ
0.98
0.97
0.96

-4 -3 -2 -1 0 1 2 3 4 5
Time (s)

Figure 27: Results for a servo deadband test.

G. Deep Throttling Hot-Flow Testing

An open-loop deep throttle test was performed in order to demonstrate the limits of motor throttling.
Combustion remained stable with throttling ratios as high as 66:1. Figure 28 shows a thrust time history
for the deep throttle test with images corresponding to several of the thrust set points during the test. The
numbers displayed in the images are the percentage of peak thrust and a time stamp corresponding to the
time axis of the plot. Combustion became unstable at point (5) in Fig. 28. Figure 29 shows a more detailed
plot of the unstable region with thrust and chamber pressure normalized by the peak values.
As a point of comparison, the Space Shuttle Main Engine (SSME) was normally throttled within a ratio
of 1.67:1, but was demonstrated as high as 5.88:1.53 Deep throttling is defined as having a turndown ratio
higher than 4:1 for a liquid engine. A 2010 deep throttling study of the Common Extensible Cryogenic
Engine (CECE) achieved a turndown ratio of 17.6:1.54 While liquid engines can be throttled to reasonably
high turndown ratios, deep throttling of liquid systems requires a variable geometry injection system and a
variable area nozzle. Proper fuel and oxidizer atomization is critical for stable combustion in liquid rockets.55
Maintaining a sufficiently high pressure drop across the injector for satisfactory atomization sets a practical
lower limit to the depth of throttling that can be achieved. Specialized injectors are required to maintain
combustion stability during deep throttling. In addition to specialized injectors, specialized turbo pumps or
valves are required. Turbo pumps must be designed to avoid cavitation, stalling, or surging and must have
stable rotordynamics and structural dynamics for a wide flow range. Because both the fuel and oxidizer
must be controlled carefully, valves are required to control flow to high degree of accuracy over a wide flow
range. Regeneratively cooled liquid engines may also have insufficient heat transfer at high throttling ratios.
By contrast, the 66:1 turndown ratio demonstrated during this project was accomplished using a standard,
off-the-shelf components - a ball valve to control oxidizer flow and spray nozzle for the injector. In hybrid
rockets, the regression rate of the solid fuel varies nearly linearly with the oxidizer mass flow rate. The result
is a self-compensating effect in the mixture of the fuel and oxidizer that greatly decreases the complexity and
accuracy of the equipment required to achieve stable thrust over a wide range. Because hybrid combustion
is primarily a surface phenomenon, the atomization of the oxidizer through the injector is not as critical
in hybrids as it is in liquid rockets for combustion stability. The solid fuel regression rate is driven by the
oxidizer mass flux in the motor, so there is a self-compensating effect in fuel mixing as the oxidizer mass flow

32 of 42

American Institute of Aeronautics and Astronautics

 rate drops during throttling. By contrast, if the injector pressure ratio in a liquid rocket is insufficient to fully
vaporize the fuel and oxidizer they will not fully mix, leading to non-homogeneous combustion within the
chamber. The conventional lower limit of injector pressure ratio in liquid engines is 1.25 to maintain stable
combustion. Figure 30 shows that the injector pressure ratio actually increased as the hybrid motor was
throttled down during the deep throttle test. The only physical lower limit to hybrid rocket deep throttling
is the ability to maintain positive pressure between the motor chamber and nozzle exit.

(1) (2) (3)

Downloaded by "Indian Institute of Technology, Bombay" on March 31, 2025 | http://arc.aiaa.org | DOI: 10.2514/6.2012-4200

1000
Raw Thrust
900 (1) Filtered Thrust

800
(4)
700

600
Thrust (N)

500

400

300

200
(2)
(5)
100 (3) (5)
(4)
0
0 5 10 15 20 25 30
Time (s)

Figure 28: Thrust time history with video stills from deep throttle test.

The trade-off for the ease and depth of hybrid rocket motor throttling is the loss of efficiency. Motor Isp
decreased significantly below 10% of peak thrust. Several techniques have been developed to counteract this
drop in efficiency such as oxidizer injection near the nozzle and pressure sensitive fuels.15 However, because
this project was considered an initial evaluation and the main focus was on developing a closed-loop throttle
controller, no steps were taken to improve deep throttling efficiency.

VII. Closed-Loop Throttle Testing

The final phase of throttle testing was demonstration of the closed-loop controllers. Cold-flow testing
was performed first. Thrust-feedback hot-flow testing followed cold-flow testing because much of the infras-
tructure was the same and few changes were required to switch. Following satisfactory demonstration of
thrust-feedback hot-flow testing, chamber pressure-feedback hot-flow closed-loop control was demonstrated.
Table 6 presents a summary of successful closed-loop throttling tests.
To provide a means of comparison between tests, controller performance was rated using the integral of
the absolute magnitude of the error (IAE), calculated as49
ˆ ttot
IAE = |e (t)| dt, (23)
0

33 of 42

American Institute of Aeronautics and Astronautics

 where ttot is the total run time of the controller and e is the error, defined as the difference between the
sensed and command thrust signals. IAE accumulates over the duration of the test so only the first ten
seconds, equal to the shortest test duration, of each test were used so that a more direct comparison could
be made. The controller was run on the cFP, but data recorded on the cDAQ was used to calculate IAE to
provide a semi-independent observation.

/ŶƐĞƚŽĨ&ŝŐ͘Ϯϴ
0.35
Raw Thrust
0.3 Filtered Thrust
Raw P0
0.25
Filtered P0
0.2
Normalized Thrust / P0
Downloaded by "Indian Institute of Technology, Bombay" on March 31, 2025 | http://arc.aiaa.org | DOI: 10.2514/6.2012-4200

0.15

0.1

0.05

-0.05
KdžŝĚŝnjĞƌ
^ƚĂďůĞ hŶƐƚĂďůĞ
-0.1 sĂůǀĞ
ŽŵďƵƐƚŝŽŶ ŽŵďƵƐƚŝŽŶ
ůŽƐĞĚ
-0.15
23 24 25 26 27 28 29
Time (s)

Figure 29: Normalized thrust and chamber pressure during unstable combustion in deep throttle test.

3.5

2.5
Injector Pressure Ratio

1.5

0.5
liquid rocket lower limit

0
0 5 10 15 20 25 30 35
Time (s)

Figure 30: Injector pressure ratio during deep throttle test.

34 of 42

American Institute of Aeronautics and Astronautics

 Table 6: Closed-loop throttle test summary.

Test Number Motor Name Command Profile Feedback KP KI fc IAE (N-s)

ColdFlowCL12 n/a step Ft 0 1 5 70.6
ColdFlowCL13 n/a step Ft 0 0.75 5 93.7
ColdFlowCL14 n/a step Ft 0 0.85 5 96.0
HTPBCL1 A Rising Tide step Ft 0.4 2.0 8 355
HTPBCL2 Pink Elephant ramp Ft 0.4 2.0 8 422
PressHTPBCL3 Bunrake ramp P0 0.4 2.5 20 305
PressHTPBCL4 The Boy step P0 0.4 2.5 20 556
PressHTPBCL5 The Mistress step P0 0.3 2.0 20 695
Downloaded by "Indian Institute of Technology, Bombay" on March 31, 2025 | http://arc.aiaa.org | DOI: 10.2514/6.2012-4200

The cutoff frequency, fc , is the value used in the lowpass filter for each controller. For thrust-feedback
tests (cold-flow and hot-flow) the cutoff frequency needed to at least be below 30 Hz to avoid having the test
stand structural harmonic frequency feed through the load cell into the controller. In both the cold-flow and
hot-flow tests, however, the cutoff frequency was set much lower to reduce the noise amplitude. The hybrid
low frequency mode was not observed in the test motor configuration, so resonance was not as much of a
concern for chamber pressure-feedback tests as it was for thrust-feedback tests. As with the thrust-feedback
tests, the cutoff frequency was selected mainly to reduce noise amplitude.

A. Cold-Flow Test Summary

Figure 31 shows thrust time histories for the final three closed-loop cold-flow throttle tests. Cold-flow tests
were done in series of three to four tests with minor adjustments to settings made between tests. Several tests
were performed to “dial in” the cutoff frequency of the lowpass filter and gains used. Cold-flow testing showed
that the gains in the controller produced a slightly less damped response than predicted by the simulator,
but the cutoff frequency used in the simulator accurately reflected the cutoff frequency used in the controller.
The specific reason for the difference between for the difference between damping in the simulator and in the
test motor configuration is not known. Individual parameter values were tweaked in the simulator, but the
damping level seen in the controller tests could not be duplicated. There is a dynamic system response that
the simulator cannot duplicate, whether because it is missing entirely or because the accumulation of errors
from rounding and calculation assumptions impacts the simulator’s ability to reproduce it. Integral gains
were set lower than those predicted to be stable by the simulator, typically in the range of 0.75-1 instead of
the 1.25-1.75 used in the simulator.

B. Hot-Flow Test Summary

Five successful hot-flow closed-loop throttle tests were performed - two with thrust feedback and three with
chamber pressure feedback. The following subsections describe the closed-loop responses of the thrust and
chamber pressure feedback controllers to the prescribed step and ramp profiles.

1. Thrust-Feedback Tests
Figures 32 and 33 show the results from Tests HTPBCL1 and HTPBCL2. HTPBCL1 was a step input
test with a 10 second duration. HTPBCL2 had a ramp input and a 15 second duration. Cold-flow closed-
loop testing provided a reasonable analog to the hot-flow thrust-feedback tests because the same feedback
mechanism was used. Both controllers performed as expected based on simulator and cold-flow test results.
The step profile tests had ringing in the controller response on the second step as Fig. 32 shows. On
the first step the servo was initialized to a position near where it settled during the initial set point. On
the second step the set point transitioned from a control region where thrust changes nearly linearly with
valve position to a region where large changes in valve position produce little change in thrust. The ringing
may be caused by an increasing controller signal during the transition. Changes in thrust from the nominal

35 of 42

American Institute of Aeronautics and Astronautics

 values due to drift over time could also have an effect. The small step at the beginning of the ramp command
shown in Fig. 33 just before the 8 second mark is the most apparent effect of the servo deadband.

100

80

Thrust (N) 60

40

Raw Thrust
20
Downloaded by "Indian Institute of Technology, Bombay" on March 31, 2025 | http://arc.aiaa.org | DOI: 10.2514/6.2012-4200

Filtered Thrust
Command
0
0 2 4 6 8 10 12 14
Time (s)
(a)

100

80

60
Thrust (N)

40

Raw Thrust
20
Filtered Thrust
Command
0
0 2 4 6 8 10 12
Time (s)
(b)

100

80

60
Thrust (N)

40

Raw Thrust
20
Filtered Thrust
Command
0
0 2 4 6 8 10 12
Time (s)
(c)

Figure 31: Thrust time histories for cold-flow closed-loop tests (a) ColdFlowCL12 (b) ColdFlowCL13 and
(c) ColdFlowCL14.

36 of 42

American Institute of Aeronautics and Astronautics

 1000

800

Thrust (N)
600

400
Raw Thrust
200 Filtered Thrust
Command
0
0 2 4 6 8 10 12
Time (s)

Figure 32: Thrust signal compared to command for Test HTPBCL1.

Downloaded by "Indian Institute of Technology, Bombay" on March 31, 2025 | http://arc.aiaa.org | DOI: 10.2514/6.2012-4200

1000

800
Thrust (N)

600

400
Raw Thrust
200 Filtered Thrust
Command
0
0 2 4 6 8 10 12 14 16
Time (s)

Figure 33: Thrust signal compared to command for for Test HTPBCL2.

2. Chamber Pressure-Feedback Tests

Figures 34 through 36 show results from three of the chamber pressure-feedback tests. Two additional
tests were performed with more aggressive gains that proved to be unstable. PressHTPBCL3 was a ramp
input test with a 15 second duration and PressHTPBCL4 and PressHTPBCL5 were both step input tests
with 12 second durations. The gains used for PressHTPBCL3 were predicted by the simulator to produce
approximately the same damping as those used for HTPBCL2. Comparison of Figures 33 and 34 shows that
damping was higher for HTPBCL2. This was mainly due to nozzle erosion effects.

1000

800
Thrust (N)

600

400
Raw Thrust
200 Filtered Thrust
Command
0
0 2 4 6 8 10 12 14 16
Time (s)

Figure 34: Thrust signal compared to command for Test PressHTPBCL3.

The gains used for PressHTPBCL4 were meant to repeat the amount of damping as observed for HTP-
BCL1. Comparison of Figures 32 and 35 shows that this was the case. Gains were lowered for PressHTP-
BCL5 to reduce the amount of ringing in the controller response to the second step. Figures 35 and 36
show that the trade off for decreased ringing in the controller response is an increase in response time. The
balance between oscillations and response time can be adjusted with the gains to meet mission requirements.

37 of 42

American Institute of Aeronautics and Astronautics

 PressHTPBCL5 exhibits an upward shift from the target thrust because of the chamber pressure-feedback
controller’s inability to adjust for nozzle erosion.

1000

800

Thrust (N)
600

400
Raw Thrust
200 Filtered Thrust
Command
0
0 2 4 6 8 10 12 14
Time (s)
Downloaded by "Indian Institute of Technology, Bombay" on March 31, 2025 | http://arc.aiaa.org | DOI: 10.2514/6.2012-4200

Figure 35: Thrust signal compared to command for Test PressHTPBCL4.

1000

800
Thrust (N)

600

400
Raw Thrust
200 Filtered Thrust
Command
0
0 2 4 6 8 10 12 14
Time (s)

Figure 36: Thrust signal compared to command for Test PressHTPBCL5.

3. Simulator Comparison
Simulator results shown in Fig. 21 had the same operating conditions as PressHTPBCL4, results shown in
Fig. 35. As with the cold-flow controller, the actual response was slightly less damped than the simulator
response. This result was typical of both hot-flow controllers. Although the response was less damped, gain
sets predicted to be stable by the simulator were also stable in the controllers. There was no need to reduce
the gains as there was with the cold-flow controller.

4. Servo Deadband Effects

As discussed in Section VI.F, the servo deadband results in a small step in the servo response for ramp
inputs. Figure 37 compares the output voltage signal from the controller to the actual servo response at the
beginning of the ramp input command for the two ramp-profile tests. Comparison of parts (a) and (b) of
the figure shows that the ramp does not begin until the 0.02 volt deadband has been exceeded, at which
point the control valve position quickly changes to match the controller output signal. The subtle dip in
the servo input voltage observed in Fig. 37(a) is due to the controller compensating for the gradual shift in
thrust over the duration of the burn. Figure 37(b) shows that the control valve does not respond to such
subtle changes in controller output due to the deadband. The lag in the servo position response time to a
ramp input voltage signal due to the servo deadband shown in Fig. 27 can be observed in Figures 33 and
34 as the difference in command and feedback signals during the ramp for a given time.

5. Nozzle Erosion Effects

Following each test the nozzle was cleaned thoroughly, being careful to remove all of the soot from the motor
burn while not scouring into the graphite. A bore gauge was then inserted into the nozzle at four orientations

38 of 42

American Institute of Aeronautics and Astronautics

 separated by approximately 45◦ . The bore gauge was measured at each orientation using calipers with a
0.0005 inch resolution. The average of the four measurements was assumed to be the average diameter of
the nozzle throat. Two of the five tests exhibited a significant amount of nozzle erosion. Table 7 shows
the approximate amount of nozzle erosion during each test. In the table, “negligible” means the amount of
nozzle erosion was less than the resolution of the calipers used to measure the nozzle throat diameter after
each test.

0.68
HTPBCL2
0.66 PressHTPBCL3
Servo Input Voltage

0.64

0.62
Downloaded by "Indian Institute of Technology, Bombay" on March 31, 2025 | http://arc.aiaa.org | DOI: 10.2514/6.2012-4200

0.6

0.58

6 6.2 6.4 6.6 6.8 7 7.2 7.4 7.6 7.8 8

Time (s)
(a)

38
HTPBCL2
PressHTPBCL3
37
% MVT

36

35

34
6 6.2 6.4 6.6 6.8 7 7.2 7.4 7.6 7.8 8
Time (s)
(b)

Figure 37: Servo input voltage and position at beginning of ramp up command.

Table 7: Nozzle erosion during hot-flow throttle tests.

Test ∆dn (mm) ∆An (%)

HTPBCL1 negligible negligible
HTPBCL2 0.10 1.2
PressHTPBCL3 negligible negligible
PressHTPBCL4 negligible negligible
PressHTPBCL5 0.05 0.6

The thrust-feedback controller used in HTPBCL2 was able compensate for the eroded nozzle because
the thrust was controlled directly. Figure 31 shows that the thrust from HTPBCL2 was approximately the
same as the tests without nozzle erosion, while Fig. 38 shows that the control valve was opened substantially
wider to produce the same thrust. The chamber pressure-feedback test, PressHTPBCL5, on the other hand,
was unable to compensate for the change in the nominal relationship between thrust and chamber pressure
using the a priori conversion discussed in Subsection IV.B, even though the chamber pressure remained at
nominal levels. Because thrust cannot be measured directly in-flight, this problem could be encountered
with any in-flight feedback mechanism. However, shifts from nominal controller operating conditions can be

39 of 42

American Institute of Aeronautics and Astronautics

 avoided by using an online conversion capable of compensating for changes. For example, there is a clear
correlation between nozzle erosion and the valve position during the initial set point. A weighting factor
could be applied to the set point based on the deviation from the nominal valve position to adjust for the
larger nozzle diameter. These tests were performed using basic controllers and off-the-shelf equipment. A
more sophisticated throttle control system should be able to easily avoid or compensate for the off-nominal
operating conditions encountered during this project.

90
HTPBCL1
HTPBCL2
80
PressHTPBCL3
PressHTPBCL4
PressHTPBCL5
Downloaded by "Indian Institute of Technology, Bombay" on March 31, 2025 | http://arc.aiaa.org | DOI: 10.2514/6.2012-4200

70
dĞƐƚĚĂƚĂ
ĚĞŵŽŶƐƚƌĂƚŝŶŐ
60 ĞĨĨĞĐƚƐŽĨŶŽnjnjůĞ
% MVT

ĞƌŽƐŝŽŶ

50

40

30

20
0 2 4 6 8 10 12 14 16
Time (s)

Figure 38: Control valve position during hot-flow closed-loop throttling tests.

6. Steady-State Accuracy
Reducing the amount of motor-to-motor variability was the primary motivation for using closed-loop throt-
tling instead of open-loop. The first four seconds of each of the five hot-flow throttle tests presented here
had the same target thrust, 700 N. Steady-state controlled response was evaluated by calculating a mean
and standard deviation of the recorded thrust for all tests in the 2-4 second time interval, avoiding the initial
transients. The standard deviation of the thrust during the initial 700 N set point was 9.97 N for all five
controlled tests. If Test PressHTPBCL5, which used chamber-pressure feedback and had nozzle erosion, is
neglected this amount reduces to 3.34 N, nearly an order of magnitude lower than the standard deviation
of 30.12 N for the uncontrolled motor burns shown in Fig. 7. The mean thrust during the initial 700 N set
point for all five tests was 1.14% higher than the set point as measured on the cDAQ, or 0.53% higher if
PressHTPBCL5 is neglected.

VIII. Conclusion
Hybrid rocket throttling has potential applications both in space and for use in atmospheric flight. Open-
loop throttle testing established physical relationships between the control valve and motor response. Liquid
rocket engines are physically limited in deep throttle by an approximate 1.25 injector pressure ratio. This
limit was not observed in the hybrid rocket test motor configuration. Deep throttle testing showed that the
test motor configuration could produce stable combustion in a 66:1 turndown ratio and that the injector
pressure ratio increased with deeper throttling.
Closed-loop thrust-feedback throttle controllers were developed for a cold-flow nitrous oxide thruster and
for a nitrous oxide/HTPB hybrid rocket motor. A closed-loop chamber pressure-feedback throttle controller
was also developed for the nitrous oxide/HTPB hybrid rocket motor. Closed-loop cold-flow throttle testing
was performed before hot-flow testing began to evaluate the accuracy of the simulator. Slightly smaller

40 of 42

American Institute of Aeronautics and Astronautics

 gains were used in the hot-flow system because damping of the simulator response was observed to be lower
than in the cold-flow system controller. Reduced damping was also observed in the hot-flow controllers, but
reducing gains was not necessary.
Hot-flow closed-loop controller testing showed that thrust could be controlled reasonably accurately even
in the presence of nozzle erosion. The chamber pressure-feedback hot flow controller could not adjust for
nozzle erosion because the conversion between thrust and chamber pressure was prescribed a priori, but still
only deviated from the mean of the controlled thrust by about half of the standard deviation of uncontrolled
thrust using the same motor configuration. For an in-flight system, the impact of off-nominal performance
due to factors such as nozzle erosion can be limited by adding an online conversion between thrust and the
feedback mechanism capable of adjusting the set point. The standard deviation in thrust decreased from
about 30 N for the uncontrolled tests to about 10 N for the controlled tests. Mean thrust at a common set
point for all controlled tests was within 1.14% of the target. For closed-loop throttle tests not affected by
nozzle erosion, the standard deviation was 3.34 N and the mean thrust was within 0.53% of the target.
Downloaded by "Indian Institute of Technology, Bombay" on March 31, 2025 | http://arc.aiaa.org | DOI: 10.2514/6.2012-4200

All throttle controller testing was performed using basic controllers and off-the-shelf equipment. Con-
troller performance can be improved using a more sophisticated throttle control system with features such as
a control valve actuator with reduced or no deadband, a control valve with a more linear flow profile, using
an automation controller capable of running higher loop frequencies, and an online, adaptable conversion
between thrust and the feedback mechanism. Even with all of the hardware limitations in this experiment,
the throttle control system substantially reduced the variability of the test motor configuration as compared
to the uncontrolled tests.

References
1 Chang, I.-S., “Investigation of Space Launch Vehicle Catastrophic Failure,” AIAA Paper 95-3128 , July 1995.
2 Safie, F. M. and Fox, E. P., “A Probabilistic Design Analysis Approach for Launch Systems,” 27th AIAA, SAE, ASME,
and ASEE, Joint Propulsion Conference, Vol. AIAA-1991-3372, 1991.
3 Maggie, G., “Space Shuttle Probabilistic Risk Assessment: Methodology and Application,” International Symposium on

Product Quality and Integrity, Proceedings of the Reliability and Maintainability Symposium, Las Vegas NV , 1996.
4 Anon., “Hazard Analysis of Commercial Space Transportation; Vol. 1: Operations, Vol. 2: Hazards, Vol. 3: Risk Analy-

sis,” U.S. DOT, PB93-199040, Accession No. 00620693 , 1988.

5 Choudhary, G., Hansen, H., Donkin, S., and Kirman, C., “Toxicological Profile for Hydrazines,” US Department of Health

and Human Services Public Health Service Agency for Toxic Substances and Disease Registry (ATSDR). Atlanta GA, 1997,
pp. 1–224.
6 DeSain, J. D., “Green Propulsion: Trends and Perspectives,” Crosslink, http://www.aero.org/publications/crosslink/summer2011/04.html,

[Retrieved 21 March, 2012].

7 Anon., “Occupational Safety and Health Guideline for Nitrous Oxide,” US Department of Labor, Occupational Health

and safety Administration, http://www.osha.gov/SLTC/healthguidelines/nitrousoxide/recognition.html, [Retrieved 28 March

2012].
8 Rhodes, G. W., “Investigation of Decomposition Characteristics of Gaseous and Liquid Nitrous Oxide,” Air Force

Weapons Laboratory, Report AD-784 602, Kirtland AFB, New Mexico, July 1974.
9 Altman, D., “Hybrid Rocket Development History,” 27th AIAA, SAE, ASME, and ASEE, Joint Propulsion Conference,

1991.
10 Moore, G. E. and Berman, K., “A Solid-Liquid Rocket Propellant System,” Jet Propulsion, Vol. 26, No. 11, November

1956, pp. 965–968.

11 Duban, P., “The ”LEX” rocket probe (LEX small rocket probe for in-flight testing of ONERA studies of hybrid propulsion,

discussing design and program),” L’Aeronautique et L’Astronautique, 1968, pp. 47–54.

12 Franklin B. Mead, J. and Bornhorst, B. R., “Certification Tests of a Hybrid Propulsion System for the Sandpiper Target

Missile,” AFRPL-TR- 69 -73 , 1969.

13 Jones, R. A., “Hybrid Propulsion System for an Advanced Rocket-Powered Target Missile,” Quarterly Technical Report,

UTC 2220-QTR2 , 1967.

14 Penn, C. D. and Branigan, J. E., “Preliminary Flight Rating Tests of the HAST Propulsion System,” AFRPL-TR-15-5 ,

1975.
15 Ordahl, D., “Hybrid Propulsion,” Space Aeronautics, Vol. 41, No. 4, April 1964, pp. 108–113.
16 Hamers, J., “Experimental Investigation of Prepackaged Hybrid Propellant Systems,” Final Report AFRPL-TR-67-168 ,

1967.
17 French, J. R., “AMROC Industrial Launch Vehicle: A Low Cost Launch Vehicle,” SAE Technical Paper 871336 , 1987.
18 Kniffen, R., McKinney, B., and Estey, P., “Hybrid Rocket Development at the American Rocket Company,” AIAA-90-

2762 , 1990.
19 Estey, P. N. and Flittie, K. J., “Aquila - The Next Generation Launch Service for Small Satellites,” 14th International

Communication Satellite Systems Conference and Exhibit, Washington DC , March 22-24, 1992.

41 of 42

American Institute of Aeronautics and Astronautics

 20 Flittie, K. J. and Estey, P., “Large-Scale Hybrid Motor Performance and Designs for use in Launch Vehicle Applications,”

JANNAF Propulsion Meeting, Vol. 2, 1993, pp. 37–50.

21 McFarlane, J. S., Kniffen, R. J., and Lichatowich, J., “Design and Testing of AMROC’s 250,000 Pound Thrust Hybrid

Motor,” AIAA/SAE/ASME/ASEE 29th Joint Propulsion Conference and Exhibit, Monterey, CA, June 28-30, 1993.
22 Flittie, K. J., Estey, P. N., and Kniffen, R., “The Aquila Launch Vehicle: A Hybrid Propulsion Space Booster,” Acta

Astronautica, Vol. 28, 1992, pp. 99–110.

23 “Hybrid Propulsion Technology Program Final Report,” NASA-CR-183952 , 1990.
24 Boardman, T. A., Carpenter, R. L., Goldberg, B. E., and Shaeffer, C. W., “Development and testing of 11- and 24-inch

hybrid motors under the joint government/industry IR&D program,” AIAA Paper 93-2552 , 1993.
25 Carpenter, R. L., Boardman, T. A., Claflin, S. E., and Harwell, R. J., “Hybrid propulsion for launch vehicle boosters: A

program status update,” 31st AIAA/ASME/SAE/ASEE Joint Propulsion Conference and Exhibit, San Diego CA, July 10-12,
1995.
26 Arves, J., Gnau, M., Joiner, K., Kearney, D., McNeal, C., and Murbach, M., “Overview of the hybrid sounding rocket

(HYSR) project.” AIAA 2003-5199 , 2003.

27 Wright, A. B., Teague, W., Wright, A. M., and Wilson, E. W., “Instrumentation of UALR Labscale Hybrid Rocket
Downloaded by "Indian Institute of Technology, Bombay" on March 31, 2025 | http://arc.aiaa.org | DOI: 10.2514/6.2012-4200

Motor,” Proc. of SPIE , Vol. 6222, 2006.

28 Dyer, J., Doran, E., Dunn, Z., Lohner, K., Bayart, C., Sadhwani, A., Zilliac, G., Cantwell, P. B., and Karabeyoglu, A.,

“Design and Development of a 100 km Nitrous Oxide/Paraffin Hybrid Rocket Vehicle,” AIAA 2007-5362 , 2007.
29 Doran, E., Dyer, J., Marzona, M. T., Karabeyoglu, A., Zilliac, G., Mosher, R., and Cantwell, B., “Status Update Report

for the Peregrine Sounding Rocket Project: Part III,” AIAA 2009-4840 , 2009.
30 Austin, B., Heister, S., Dambach, E., Wernimont, E., and Meyer, S., “Variable Thrust, Multiple Start Hybrid Motor

Solutions for Missile and Space Applications,” 46th AIAA/SME/SAE/ASEE Joint Propulsion Conference, Nashville TN , July
25-28, 2010.
31 Eilers, S. D., Wilson, M., and Whitmore, S., “Analytical and Experimental Evaluation of Aerodynamic Thrust Vectoring

on an Aerospike Nozzle,” 46th AIAA/ASME/SAE/ASEE Joint Propulsion Conference and Exhibit, Nashville, TN , 2010.
32 Eilers, S. D., WIlson, M. W., Whitmore, S., and Peterson, Z., “Side Force Amplification on an Aerodynamicclly Thrust

Vectored Aerospike Nozzle,” 47th AIAA/ASME/SAE/ASEE Joint Propulsion Conference, 2011.

33 Whitmore, S. A., Peterson, Z. W., and Eilers, S. D., “Analytical and Experimental Comparisons of HTPB and ABS as

Hybrid Rocket Fuels,” 47th AIAA/ASME/SAE/ASEE Joint Propulsion Conference & Exhibit, 2011.
34 Anon., “Liquid Rocket Valve Assemblies,” NASA-SP-8097 , 1973.
35 Anon., Control Valve Handbook , Fisher Controls International LLC, 2005.
36 Anon., “Flow Equations for Sizing Control Valves,” ANSI/ISA-75.01.01 (IEC 60534-2-1 Mod)-2007 , Nov. 2007.
37 “V-Control Ball Valve Selection Guide Technical Bulletin No. 1006,” Tech. rep., Flow-Tek, Sept. 2008.
38 Miller, R. W., Flow Measurement Engineering Handbook, 2nd Ed., McGraw-Hill Publishing Company, 1989.
39 Anon., “Hydroxyl Terminated Polybutadiene Resins and Derivatives,” Sartomer Product Bulletin, March 2006.
40 Anon., “PAPI 94 Product Information,” Dow Plastics Form No. 109-00707-801XQRP , August 2001.
41 Karabeyoglu, A., Stevens, J., and Cantwell, B., “Investigation of Feed System Coupled Low Frequency Combustion

Instabilities in Hybrid Rockets,” AIAA 2007-5366 , 2007.

42 Weast, R. C., editor, CRC Handbook of Chemistry and Physics, 58th Ed., CRC Press, Inc., 1977.
43 Span, R. and Wagner, W., “Equations of State for Technical Applications. I. Simultaneously Optimized Functional Forms

for Nonpolar and Polar Fluids,” International Journal of Thermophysics, Vol. 24, No. 1, 2003, pp. 1–39.
44 Span, R. and Wagner, W., “Equations of State for Technical Applications. II. Results for Nonpolar Fluids,” International

Journal of Thermophysics, Vol. 24, No. 1, 2003, pp. 41–109.

45 Span, R. and Wagner, W., “Equations of State for Technical Applications. III. Results for Polar Fluids,” International

Journal of Thermophysics, Vol. 24, No. 1, 2003, pp. 111–162.

46 Gordon, S. and McBride, B. J., “Computer Program for Calculation of Complex Chemical Equilibrium Compositions

and Applications I. Analysis,” Tech. rep., NASA RP-1311, 1994.

47 McBride, B. J. and Gordon, S., “Computer Program for Calculation of Complex Chemical Equilibrium Compositions

and Applications II. Users Manual and Program Description,” Tech. rep., NASA RP-1311, 1996.
48 Carmicino, C., “Acoustics, Vortex Shedding, and Low-Frequency Dynamics interaction in an Unstable Hybrid Rocket,”

Journal Of Propulsion and Power , Vol. 25 No. 6, 2009, pp. 1322–1335.

49 Dorf, R. C. and Bishop, R. H., Modern Control Systems, 7th Ed., Pearson Education, Inc., 2008.
50 Anon., LabVIEW PID Control Toolkit User Manual, National Instruments Corporation, 2006.
51 Dyer, J., Doran, E., Dunn, Z., Lohner, K., Zilliac, G., and Cantwell, B., “Modeling Feed System Flow Physics for Self

Pressurizing Propellants,” AIAA 2007-5702 , 2007.

52 Chapra, S. C. and Canale, R. P., Numerical Methods for Engineers, 7th Ed., The McGraw-Hill Companies, Inc., 2006.
53 Bradley, M., “Space Shuttle Main Engine Off-Nominal Low Power Level Operation,” 33rd AIAA/ASME/SAE/ASEE

Joint Propulsion Conference & Exhibit, Seattle WA, July 6-9, 1997.
54 Giuliano, V. J., Leonard, T. G., and Lyda, R. T., “CECE: Expanding the Envelope of Deep Throttling Technology

in Liquid Oxygen/Liquid Hydrogen Rocket Engines for NASA Exploration Missions,” 46th AIAA/ASME/SAE/ASEE Joint
Propulsion Conference & Exhibit, July 2010.
55 Betts, E. M. and Frederick, Robert A., J., “A Historical Systems Study of Liquid Rocket Engine Throttling Capabilities,”

46th AIAA/ASME/SAE/ASEE Joint Propulsion Conference & Exhibit, Nashville TN , July 25-28, 2010.

42 of 42

American Institute of Aeronautics and Astronautics