POINTING ERROR ENGINEERING FRAMEWORK FOR HIGH POINTING
ACCURACY MISSIONS
Massimo Casasco(1), Sohrab Salehi(2), Sven Weikert(3), Jochen Eggert(3),
Marc Hirth(3), Haifeng Su(4), and Thomas Ott (5)
(1)
ESA - ESTEC, Keplerlaan 1, 2201 AZ Noordwijk, The Netherlands
(2)
Rhea System for ESA, ESTEC, Keplerlaan 1, 2201 AZ Noordwijk, The Netherlands,
+31(0)715655126, Sohrab.Salehi@esa.int
(3)
Astos Solutions GmbH, Meitnerstrasse 8, 70569 Stuttgart, Germany
(4)
Institute of Flight Mechanics and Control, University of Stuttgart, Pfaffenwaldring 7a,
70569 Stuttgart, Germany
(5)
Airbus Defence and Space, D-88039 Friedrichshafen, Germany
Abstract: The publication of the ECSS Control Performance Standard and the ESA
Pointing Error Engineering Handbook forms part of the recent standardization effort in
Europe to define a clear pointing error engineering methodology for ESA projects. This
is complemented by the development of the Pointing Error Engineering Tool, a prototype
software, released under the ESA Software Community License and intended to support
dissemination of this pointing error engineering methodology. This paper describes the
mathematical framework and the steps of the methodology for pointing error budgeting
that is implemented in the Pointing Error Engineering Tool and highlights the benefits
that this tool can provide with respect to traditional conservative approaches,
particularly for high pointing accuracy missions. Finally, the perspective activities to be
promoted by ESA in the area of tools for pointing error engineering are outlined.
Keywords: Error budgeting, software tool, high accuracy pointing
1. Introduction
Pointing error engineering covers the engineering process of establishing system pointing error
requirements, their systematic analysis throughout the design process and eventually the
verification of compliance. For technical as well as historical reasons, pointing error engineering
in the European space community has been implemented on the basis of engineering practices
that were often tailored on a case-by-case basis and no standard practice was in place. This has
changed through the initiative of the European Cooperation for Space Standardization (ECSS).
The ECSS Control Performance Standard [1] now provides a solid and exact mathematical basis
for constructing performance error budgets. The Pointing Error Engineering Handbook (PEEH)
is an ESA applicable document [2], containing a step-by-step process with guidelines,
summation rules, recommendations and examples. It provides ESA projects with a clear pointing
error engineering methodology.
ESA PEEH users are supported by the Pointing Error Engineering Tool (PEET). System
engineers and control engineers use it for the pointing requirements allocation activities, during
the early phases of a project, and for the pointing error budget verification activities, in later
1
phases. ESA initiated and coordinated the development of a PEET prototype, carried out by
Astos Solutions GmbH with the support of the Institute of Flight Mechanics and Control of the
University of Stuttgart. This approach is intrinsically capable of minimizing the margins and
uncertainties in pointing budgets and is expected to prove extremely valuable especially for high
pointing accuracy missions, where accurate analytical results obtained using PEET could make
the difference in taking the correct design decisions.
PEET has since been extended for relative position error budgeting and is released under the
ESA Software Community License. Operational software, targeting use in Phases B and C, is
under development for ESA to be released under the same license type as PEET. This will permit
processing complex calculations for high accuracy pointing, frequency domain techniques
introduced in the handbook, provision of traceability and a common platform for exchange of
information between the various entities. ESA is using PEET for the pointing error engineering
analyses of a number of missions, including Euclid [6], MetOp-SG [7], and Proba-3 [8].
The paper provides a detailed summary of the pointing error analysis and evaluation method
according to the PEEH. The implementation and features of the PEET prototype are described.
2. Pointing Error Analysis and Evaluation Methodology
While the ECSS Control Performance Standard E-ST-60-10C provides normative clauses with
clear mathematical elements for control performance analysis in general, which apply to all
disciplines involving control engineering and at different levels, ranging from equipment to
system level, the ESA PEEH embeds the elements of the ECSS standard in a step-by-step
engineering process for the specific case of satellite pointing errors. The four analysis steps for
pointing error analysis and evaluation are described in Sections 2.3-2.6. More details on the
theoretical background of the ESA PEEH are available in [4] and [5].
2.1. Nomenclature and Definitions
A pointing error e can be considered as the response of a system to external or internal physical
phenomena affecting the system’s pointing performance as illustrated in Figure 1.
Figure 1. Pointing error source transfer
Each such physical phenomenon is referred to as a pointing error source (PES) and denoted by
es. A PES is categorized as being either constant in time (time-constant), random in time (timerandom) and/or random in its realization (ensemble-random).
2
A pointing error contributor (PEC), denoted by ec, represents the contribution of one or more
pointing error sources es to the overall pointing error e. A PES becomes a PEC through
undergoing a pointing system transfer, such as a coordinate frame transformations, control
system, or structural transfer function. In order to analyze pointing performance, a pointing
system is broken down into subsystems with individually controlled (active or passive) transfer
properties. The pointing error e is the sum of the different PECs.
e(t)
∆t1
∆t2
∆t3
...
t
instantaneous time
∆t window time
∆ts1
Observation Period 1
∆ts stability time
∆ts2
Observation Period 2
t
Figure 2. Time dependency of pointing errors
Several types of time dependencies of pointing errors can be distinguished. These are graphically
illustrated in Figure 2 for the observation requirements of a pointing system, e.g. a satellite and
its payload [3]. Pointing errors can depend on:
• Instantaneous time t: pointing error at any point in time t during system lifetime or a
defined observation period.
• Window time Δt: pointing error within a time window Δt, where the time window can
occur at any point in time t during system lifetime or a defined observation period.
• Stability time Δts: pointing error describing stability, thus the relative error, among
pointing errors in time-windows of length Δt. The time-windows are separated by a time
difference of length Δts, and can occur at any point in time t during system lifetime or a
defined observation period.
The time-dependent pointing errors are summarized in Table 1.
Table 1. Definition of pointing error indices
Index
AKE
KDE
Name
Absolute Knowledge
Error
Absolute Performance
Error
Mean Knowledge Error
Mean Performance Error
Relative Knowledge
Error
Relative Performance
Error
Knowledge Drift Error
PDE
Performance Drift Error
KRE
Knowledge
Reproducibility Error
Performance
Reproducibility Error
APE
MKE
MPE
RKE
RPE
PRE
Definition
difference between the actual parameter (attitude, geolocation, etc.) and the
known (measured or estimated) parameter in a specified reference frame
difference between the target (commanded) parameter (attitude, geolocation,
etc.) and the actual parameter in a specified reference frame
mean value of the AKE over a specified time interval Δt
mean value of the APE over a specified time interval Δt
difference between the AKE at a given time within the time interval Δt and the
MKE over the same time interval
difference between the APE at a given time within the time interval Δt and the
MPE over the same time interval
difference between MKEs taken over two time intervals separated by a specified
time Δts within a single observation period
difference between MPEs taken over two time intervals separated by a specified
time Δts within a single observation period
difference between MKEs taken over two time intervals separated by a specified
time Δts within different observation periods
difference between MPEs taken over two time intervals separated by a specified
time Δts within different observation periods.
3
The comprehensive set of pointing error indices, categorized as knowledge or performance errors
and depending on instantaneous time, window time and stability time, is formulated in Table 2.
Table 2. Mathematical formulation of pointing error indices
Pointing Error Indices
index
= e P (t )
eAPE(t)
eMPE(t, ∆t)
eMKE(t, ∆t)
eRPE(t, ∆t)
eRKE(t, ∆t)
ePDE(t, ∆t1, ∆t2, ∆ts)
ePRE(t, ∆t1, ∆t2, ∆ts)
eKDE(t, ∆t1, ∆t2, ∆ts)
eKRE(t, ∆t1, ∆t2, ∆ts)
∆ts
instantaneous time
= e K (t )
eAKE(t)
∆t
instantaneous
= e P (t , ∆t )
window time
= e K (t , ∆t )
= e P (t ) − e P (t , ∆t )
= e K (t ) − e K (t , ∆t )
stability time
= eP (t , ∆t1 ) − eP (t + ∆ts , ∆t2 )
= eK (t , ∆t1 ) − eK (t + ∆t s , ∆t2 )
window time
eindex
instantaneous error
stability time
e K (t )
knowledge error signal
e P (t )
performance error signal
time average:
e(t , ∆t ) = e(t )
∆t
1
=
∆t
∫ e(t ) dt
t + ∆t / 2
t − ∆t / 2
The frequency domain classification of time windowed and windowed stability errors allows an
exact evaluation of the metrics defined Table 2. They can be used to determine the contribution
of the PEC signal PSD to the different time windowed and windowed stability errors. For this the
PES PSD needs to be characterized or at least approximated with reasonable assumptions.
The frequency-domain pointing error metrics are specific PSD weighting functions Fmetric. In
order to perform analysis, rational approximations, F~metric , of the weighting functions are given
in [5] and summarized in Table 3 such that Fmetric(ω) ~ |F~metric(jω)|2 and with s=jω. The metrics
can be understood as a function by which the PEC signal power, described by its PSD, is
weighted. The weighting function corresponds to a low pass, a high pass or a combination of
both. As can be seen in Table 3 the weighting functions have the form of a sinc-function. This is
due to the fact that the windowing in the time domain is equivalent to filtering the time signal by
a rectangular function, which has the sinc-function as frequency domain equivalent.
4
Table 3. Pointing error metrics - frequency domain
Pointing Error Metric Weighting Functions Fmetric
FWM with ∆t=0.18s
Windowed Mean (WM)
2(1 − cos(ω∆t ) )
(ω∆t ) 2
-2
Magnitude
FWM (ω , ∆t ) =
0
10
rational approximation:
~
FWM ( s, ∆t ) =
2(s∆t + 6 )
10
-4
10
-6
10
(s∆t )2 + 6(s∆t ) + 12
-8
10 -3
10
FWM
rational approximation
-2
10
2(1 − cos(ω∆t ) )
~
FWV ( s, ∆t ) =
10
(
s∆t s∆t + 12
(s∆t )
2
Magnitude
(ω∆t ) 2
rational approximation:
1
2
10
10
)
10
10
10
+ 6(s∆t ) + 12
0
-2
-4
-6
FWV
rational approximation
-8
10 -3
10
10
-2
10
-1
10
0
10
1
10
2
Frequency [Hz]
FWMS with ∆t=0.18s and ts=60s
Windowed Mean Stability (WMS)
FWMS (ω , ∆t , ∆t s ) = FSTA × FWME
4(1 − cos (ω∆t s )2
=
(ω∆t) 2
Magnitude
10
10
10
10
rational approximation:
~
~
~
FWMS (ω , ∆t , ∆t s ) = FSTA × FWM
0
-2
-4
-6
FWMS
rational approximation
-8
10 -3
10
10
-2
10
-1
10
0
10
1
10
2
Frequency [Hz]
FSTA with ts=60s
Stability (STA)
FSTA (ω , ∆t s ) = 2(1 − cos(ω∆t s ) )
Magnitude
10
rational approximation:
~
FSTA ( s, ∆t s ) =
0
10
FWV with ∆t=0.18s
Windowed Variance (WV)
FWV (ω , ∆t ) = 1 −
-1
10
Frequency [Hz]
2 s∆t s (s∆t s + 6 )
(s∆t s )2 + 6(s∆t s ) + 12
10
10
10
0
-2
-4
-6
FSTA
rational approximation
-8
10 -3
10
10
-2
10
-1
10
0
Frequency [Hz]
5
10
1
10
2
2.2. Pointing Error Engineering Methodology Framework
Figure 3 shows a flow diagram for the pointing error engineering methodology. The awareness
of the whole pointing error engineering cycle is key for pointing error engineering activities from
requirement specification to performance verification: indeed for specification of pointing error
requirements relevant analysis and verification methods have to be identified and vice versa.
Application
Pointing System
AST-1
Application
Requirements
PES
Characterization
Mapping
PES
Characteristics
System Pointing Error
Requirements per Index
System Design
System
Specifications
AST-2
PES Transfer
Analysis
Compliance or
Redefinition Request
Compliance or
Redefinition Request
PEC
AST-3
Error Index
Contribution Analysis
Absolute & TimeWindowed PEC
AST-4
Application
Performance
System Pointing Errors
per Index
System Pointing Error
Evaluation
Mapping
Figure 3. Pointing error engineering methodology workflow
The process starts with mapping the user specified Application Requirements into unambiguous
System Pointing Error Requirements, formulated according to the classification in Table 2. The
compliance of the system pointing error requirements is analyzed by estimating and combining
the different occurring error sources in four analysis steps, AST-1 to AST-4. Note that the
mapping process is not treated in the PEEH because it is application specific.
6
The analysis steps AST-1 to AST-4 from Figure 3 should then be applied to each subsystem of
the pointing system. For complex cases, the pointing system can be broken down into several
subsystems. Then step AST-4 is performed again at pointing system level in order to compile
and evaluate the overall pointing error budget.
This methodology can be implemented via two different main analysis methods:
a) simplified statistical method: analysis with standard deviation, σ, and mean, μ, and their
summation per ECSS pointing error index under the assumption of the central limit
theorem.
b) advanced statistical method: analysis by joint PDF characterization via convolution of
different error probability density functions (PDF), p…(e), and evaluation of level of
confidence for required ECSS pointing error indices.
The currently available PEET prototype implements the simplified statistical method. It is
however foreseen that future releases of PEET will also include the implementation of the
advanced statistical method.
2.3. Step AST-1: Characterization of Pointing Error Source
Selecting the eligible mathematical elements for PES characterization are selected in step AST-1,
depends on the PES error data characteristics as well as the type of available PES error data.
Figure 4 shows the categorization of a PES as time-constant or time-random, and its description
as random variable or random process.
time-constant
nth Pointing Error Source (PES)
identification
time-random
Inputs
nth PES error data eS
eS ← yes
timeconstant
no → eS(t)
yes
PES description:
random variable
Outputs
pSC
σSC
RP-data
available
PES description:
random process
μSC
pSRP
Gss
no
PES description:
random variable
σSRP
σSR
μSR
pSR
Figure 4. Characterization method
Depending on the maturity of the available data, a PES can be described according to its
fundamental properties:
• ensemble-randomness
• time-randomness
• power spectrum
7
In the decision tree, a PES is categorized depending on its fundamental properties. The first
decision is whether a PES is time-constant or time-random. Time-constant PES do not vary
randomly with time, but in their ensemble of realizations. On the other hand, time-random PES
have a magnitude that varies randomly in time. Of course a combination of both is also possible,
meaning that a PES can have time-random and time-constant properties (e.g. a periodic signal
whose amplitude varies over different realizations). A time-constant PES is described as a
random variable in line with the mathematical elements provided in the PEEH. A time-random
PES is ideally described as a stationary random process if sufficient information about the
underlying process is available. This is because describing a PES as stationary random process
has the advantage that exact window time and stability time properties of the PES are captured.
A random pointing error process {ek(t)} is an ensemble of k sampling function realizations that
are random in time t (time-random) and random in its ensemble of realizations (ensemblerandom). The ensemble is the set {…} of all realizations k of the random pointing error ek(t). The
probability properties of a random process are described by the ensemble statistical quantities
(e.g. mean or variance) at fixed values of t, where ek(t) is a random variable over the index k. In
general, the statistical quantities are different at different times t. If the statistical quantities are
equal for all t the random process is said to be stationary. A stationary random process is
described by its PDF p(e). In practice most stationary random processes have a Gaussian PDF
and thus are completely defined by their mean value and covariance respectively [2].
The frequency domain characteristics of a random stationary process are described by means of
its power spectral density (PSD). This becomes important when considering time-windowed
pointing errors because windowing in time domain is equivalent to low-pass-filtering in
frequency domain. This enables mathematically exact analysis of time dependent pointing errors.
The PSD is a powerful formalism to describe random stationary noise processes. The doublesided PSD of ek(t) in [unit2/(rad s-1)] is defined as See(ω), based on which the single-sided PSD is
given as Gee(ω)=2See(ω), with ω being the frequency in [rad s-1]. The single-sided PSD is
commonly also defined in [unit/√(rad s-1)], in which case it is referred to as Pee(ω)= √Gee(ω).
If time series data is not available, [1] provides guidelines for an approximate random variable
description. Examples of application of the decision tree for the random process and random
variable cases are provided in [3].
2.4. Step AST-2: Transfer Analysis
The description of the PES is given with respect to its point of origin. In order to evaluate a
pointing error requirement, the transfer of a PES from its origin to the point of interest needs to
be analyzed in step AST-2 to determine the pointing error contributor (PEC). This can be done
by decomposing the pointing system into subsystems, as exemplified in Figure 1. These PECs
are obtained by a transformation, which depends on the system under evaluation. The transfer
characteristics of each system are tunable to a certain extent and thus can be used to perform
trade-offs with the aim of making pointing errors compliant with their requirement. The input
(PES) and output (PEC) parameters of the transfer analysis are shown in Figure 5.
8
Inputs
pSC
σSC
μSC
pSRP
GSS
transfer analysis
Outputs
pCC
σCC
σSRP
σSR
μSR
pSR
μCR
pCR
transfer analysis
μCC
pCRP
Gee
σCRP
σCR
time-constant
time-random
Figure 5. Transfer analysis
There are various techniques in the frequency domain and the time domain for the system
transformation of time-random PES described as random processes. The analytic methods are
based on linear transformation of statistical properties, whereas the numerical methods rely on
simulations and experimental results. In the case of a time-constant PES, the system
transformation analysis is simply the multiplication of the bias/mean value with the system
steady-state gain.
The frequency domain approach relies on the observation that if the input error signal of a
system, the PES, is known and the system can be represented by a linear time-invariant (LTI)
transfer function H(jω), being stable and strictly proper, the output error signal, the PEC, can be
determined. The variance of a PES described as random processes is related to its PSD as shown
in Eq. 1.
∞
1
2 (�)
= 2� ∫0 ��� (�)��
����
(1)
The PSD Gss of the input error signal es(t) is transformed by the system according to the wellknown relations in Eq. 2 (SISO case) and Eq. 3 (MIMO case).
��� (�) = |�(��)|2 ��� (�)
��� (�) = � ∗ (��)��� (�)�(��)
(2)
(3)
The variance of the output error signal ec(t) is thus computed from its PSD Gee via Eq. 4.
∞
1
2 (�)
= 2� ∫0 ��� (�)��
����
(4)
This transfer can be analyzed by various methods [9][10]. The advantage of this analytical
approach, fully implemented in the PEET prototype, over numerical methods is that it can be
used to tune the system transfer function H based on signal and system norms. Guidelines for
transfer analysis based on simulations and experimental results are provided in [1], but these
approaches are not implemented in the PEET prototype.
9
2.5. Step AST-3: Pointing Error Index Contribution
The contribution of PECs to the pointing error indices is determined in step AST-3. This step can
be skipped for random time-constant PES because it does not depend on time. In addition, AST3 can also be skipped for the analysis of errors, such as APE and AKE, which only depend on the
instantaneous time, and not on the window time or stability time. The contribution analysis is
shown in Figure 6 with input and output parameters.
Inputs
pCC
σCC
μCC
pCRP
σCRP
Gee
σCR
μCR
pCR
error index contribution analysis
Outputs
pBC
σBC
μBC
μindex
σindex
pindex
time-constant
time-random
Figure 6. Pointing error index contribution analysis
Guidelines are provided in [1], for evaluating the pointing error index contribution of a random
variable description of time-random PECs. These take the form of tables that quantify the
contribution for a number of different error probability distribution functions. These tables are
used in the PEET prototype when the random variable description of a PEC is selected.
On the contrary, for PECs described as random processes, an exact evaluation of the contribution
to an error index is possible by evaluating the integral of the PSD associated with the stationary
random-process and applying a suitable spectral weighting function [2][3]. A summary of the
exact expressions and rational approximations for the weighting functions can be found in [2].
The PEET prototype includes algorithms that implement and perform all the calculations needed
to evaluate the error index contributions for PECs described as stationary random processes.
A worked example, that illustrates the application of the process in the frequency domain makes
use of the weighting functions described above, can be found in [3].
A PEC needs to be interpreted with respect to the required statistical property in line with
statistical interpretation guidelines provided in the PEEH. If a PES is described as random
process, the statistics is interpreted for each pointing error index contributor at the end of AST-3.
On the other hand, if a PES is described by a random variable, an equivalent mean and variance
is determined based on the statistical interpretation already in AST-1. In the following a short
summary is given on the statistical interpretation.
The properties of physical phenomena, and thus the pointing errors and their sources, are
described in terms of their probability characteristics. Three statistical interpretations [1] for
describing the property and statistical characteristics are defined: mixed, ensemble, and temporal
10
In the mixed interpretation one considers the probability P greater or equal to a level of
confidence Pc such that the ensemble of pointing error realizations {ek(t)} or e(k,t) is less than a
required error value er in its ensemble of realizations k and in time t. This mathematically
translates into Eq. 5.
����[{|�� (�)|} < �� ] ≥ ��
��
����[|�(�, �)| < �� ] ≥ ��
(5)
In the ensemble interpretation the probability P greater or equal to a level of confidence Pc is
considered, such that a realization k of the ensemble of pointing error realizations {ek(t)} or e(k,t)
is less than a required error value er for all times t. This mathematically translates into Eq. 6.
����[|���� (�)| < �� ] ≥ �� ���ℎ ���� (�) = max� [{�� (�)}] �� max� [�(�, �)]
(6)
����[|���� (�)| < �� ] ≥ �� ���ℎ ���� (�) = max� [{�� (�)}] �� max� [�(�, �)]
(7)
In the temporal interpretation the probability P greater or equal to a level of confidence Pc is
considered, such that the entire ensemble of pointing error realizations {ek(t)} or e(k,t), or just the
worst case realization, with realization index k, is less than a required error value er for a fraction
of time t. This mathematically translates into Eq. 7.
2.6. Step AST-4: Pointing Error Evaluation
Pointing error is evaluated in two steps, as shown in Figure 7.
Inputs
pBC
pN∗pN-1∗…∗p1
σBC
μBC
μindex
σindex
Σ
Σ
Σ
Σ
pindex
pN∗pN-1∗…∗p1
Pc evaluation
Pc evaluation
εindex
B
compilation of total pointing error per index
Outputs
eAPE/AKE
eMPE/MKE
eRPE/RKE
ePDE/KDE
ePRE/KRE
time-constant
time-random
Figure 7. Pointing error evaluation
The time-constant and time-random error contributors are first combined together separately,
taking into account possible correlations between errors. The probability with the applicable
11
confidence level is then computed. The total pointing error is subsequently computed per error
index from both intermediate results.
The rules for summing means and variances of error contributors are the following: the means
are summed linearly, the uncorrelated variances are RSSd, and upper bound estimation is used
for the correlated variances.
The error index is then computed for the applicable confidence level. First the standard deviation
is multiplied by np, where np is a positive scalar such that for a Gaussian distribution the np
confidence level encloses the probability Pc, as specified in the requirement. Then the scaled
standard deviation is summed with the mean value. Finally, the time-constant and time-random
pointing errors are summed per APE, AKE, MPE and MKE indices. Note that the time-constant
error does not contribute to the other pointing error indices.
3. The Pointing Error Engineering Tool (PEET)
The Pointing Error Engineering Tool (PEET) automates the 4-step pointing error engineering
methodology presented in the ESA Pointing Error Engineering Handbook (PEEH). The current
prototype has a special focus on pre-phase A and phase A activities. PEET is released under the
ESA Software Community License.
The architectural structure of PEET is shown in Figure 8.
MATLAB
Java Runtime Environment (JRE)
PEET_GUI
MATLAB_Engine
MATLAB Toolboxes
System_Editor
MATLAB_Block_Classes
MATLAB_Block_Classes
MATLAB_Block_Classes
Tree_View
Java MATLAB Interface
(JMI)
Plotting
Java_Excel_API
MATLAB GUI
Database_Browser
Block_Database
Figure 8. Architectural structure
The graphical user interface is written in Java Swing with a core implemented by MATLAB
classes. The Java GUI runs completely within the virtual machine of the MATLAB installation
12
and does not need any additional Java runtime. It communicates with the MATLAB core through
MATLAB’s Java Matlab Interface. With the exception of the Control System Toolbox, PEET
relies on a standard MATLAB installation. An additional interface is available for import/export
of data with Excel. Data from the MATLAB workspace can also be accessed by PEET.
3.1.Graphical User Interface
The user interface is very similar to Simulink. To build up pointing systems, the user can add
various building blocks and connections between them, using predefined building blocks
applicable to pointing error engineering, provided in the PEET database.
The user first constructs the pointing system by dragging building blocks, such as pointing error
sources (PES) or transfer systems, into the System Editor (see Figure 9). Each type of block has
a mask through which the parameters of the block can be configured. The block is then
connected to other blocks.
Figure 9. System editor and block mask
The user then defines the error indices applicable to the pointing scenario, as well as any
correlation between error sources or correlation between axes of a single error source. PEET
currently supports uncorrelated and fully correlated error sources. The user then runs PEET and
can inspect the results of the error computations in the Tree View (see Figure 10).
The pointing system is shown in a tree-like structure, on the left side of the screen. Selecting a
particular block causes the error information of its input and output signals to show up on a
number of tabs, on the right side of the screen. The error signals are split up into components
(e.g. random variable part, drift part, etc.) to give a better overview about the error contribution
at this point of the pointing system. The final block shown at the bottom of the tree has a special
tab called Pointing error. On this tab, all the information about the final pointing error and the
line-of-sight error is presented to the user.
MATLAB® is a registered trademark of MathWorks.
Excel® is a registered trademark of Microsoft.
13
Figure 10. Tree View
Additionally PEET has a sensitivity analysis capability, which can be used for the identification
of error drivers in the pointing system. In general, the sensitivity value expresses the change of
the components of the final pointing error with respect to a change in some scalar parameter
value. Figure 11 shows an example of the sensitivity analysis. In this case, changing the value of
the parameter’s upper left matrix element will have a large influence on the y-axis of the final
pointing error.
Figure 11. Sensitivity analysis manager
3.2. PEET Core
PEET supports a wide range of pointing error sources which can be either 1-dimensional or 3dimensional. Depending on the available data, each PES can be described by its statistical
properties or by a PSD (see Section 2.3). The time-correlation of the noise, i.e. its coloring, can
be fully taken into account by using the PSD representation. PEET automatically converts PES
signal representations into an eligible format using a structure with 5 signal types (time-constant
random variable, time-random random variable, PSD, drift, and periodic). Each PES signal type
is internally mapped either to a PSD or to a covariance matrix, as the underlying treatment in the
system transfer is different for each signal type.
14
PEET provides tools and models to support the user in defining the system transfer from PES to
contributors. PEET provides parameterized blocks for the most common models. The following
models [11] are currently supported by PEET:
• Pointing Error Sources
• Mapping from 1D signals to 3D signals
• Coordinate Transformations
• Flexible and Rigid Plant Models
• Static Systems (e.g. matrix)
• Dynamics Systems (e.g. transfer function, state-space, zero-pole-gain, etc.)
• Feedback Systems
• PID Controllers
• Gyro-Rate Noise (parametric model based on [12])
• Gyro-Stellar Estimators
• Star Tracker Noise (parametric model for sensor field-of-view and pixel noise)
• Reaction Wheel Microvibrations
The system transfer, depending on the signal type, is automatically performed either via Eq. 3 as
most accurate way to regard the propagation of time-correlation, or in terms of covariance via
covariance propagation. The correlation of signals is fully considered in this computation.
The user can select pointing error indices (among those defined in Table 2) and associate a
statistical interpretation (temporal, ensemble and mixed) to each index according to [1][2]. The
pointing error indices are automatically computed by PEET, using a globally defined level of
confidence.
PEET can be used throughout all design phases by successive modelling refinement. This means
that starting with variances, mean values, and simple transfer models, the system description can
be refined in later design phases by time-series data originating from simulations or by using
PSDs with their cross-spectral densities together with sophisticated transfer models.
The interfaces to Java and analysis algorithms of the PEET core are implemented in an objectoriented manner using MATLAB classes. These classes are suitable for future extension and
maintenance. The resulting concise structure inherently improves code re-use. The dedicated
internal data structures and the corresponding categorization algorithms are suitable for both
GUI-based computations and script-based computations. This enables the user to use batch mode
operations and recursive computations. Algorithms for handling the signal transfer analysis
history are developed, which support correct computation of signal correlations, and enable
extension for user-specified correlation.
The current implementation of PEET assumes the applicability of the central limit theorem, i.e.
non-Gaussian distributions are converted into equivalent Gaussian distributions. The user needs
to be aware of this when evaluating the confidence level, in case of dominating non-Gaussian
pointing error contributors. Furthermore, all PES are assumed to be stationary and systems
transfers are treated as LTI models. Accurate computation of cross-correlation is currently
limited to only full or no correlation of PES.
15
4. Applications
An artificial example of an high accuracy pointing satellite mission with typical pointing error
sources and system transfers that convert these sources into final pointing error contributors is
described in [13][14].
The pointing error engineering methodology and the current prototype PEET tool are already
being used for the error analyses of diverse European satellites.
Euclid, in an orbit around the second Sun-Earth Lagrange point (L2), is an ESA mission to map
the geometry of the dark Universe. Its AOCS is responsible for high pointing accuracy during
science observation. The stringent pointing stability requirement (25 milli-arcsec rms over 500 s,
TBC) require high performance AOCS sensors and a cold gas micro-propulsion system. MetOpSG, in Sun synchronous orbit, forms the space segment of the EUMETSAT Polar System, and
aim to collect consistent long-term remotely sensed data for meteorology and climate
monitoring, forecasting and operational service provision. It is composed of multiple
instruments, on several satellites, with various scanning requirements for pointing and pointing
knowledge. PROBA-3 is a low-cost ESA mission for solar coronagraphy and technology
demonstration for formation flying of spacecraft in a highly elliptic orbit. The two spacecraft will
operate as a single unit, pointing at selectable directions, with a relative position accuracy at the
millimeter level. EDRS is a geostationary constellation of GEO satellites intended to relay user
data between LEO satellites (as well as UAVs in the future) and ground stations. The major
requirements are on pointing knowledge and pointing stability, which is challenging with respect
to a typical telecom satellite.
5. Conclusion
The methodology of the ESA PEEH and the accurate calculations supported in PEET can be
beneficial in reducing the uncertainties in pointing budgets. This in turn can prove essential in
guiding design decisions for high pointing accuracy missions, where comfortable margins with
respect to requirements cannot be allocated.
The successful application results are paving the way for the development of an enhanced
software framework that will develop specialized modules for Earth Observation, Science,
Telecommunication and Navigation missions, which ESA will pursue in the near future. PEET is
expected to become the reference tool for pointing error budgeting activities, processing complex
calculations for high pointing accuracy and providing traceability in a common platform for
exchange of information among the various stakeholders (ESA, industrial primes, industrial
subcontractors, and scientific community).
6. References
[1]
European Cooperation on Space Standardization, “Control Performance Standard ECSSE-ST-60-10C.” ESA-ESTEC Requirements & Standards Division, 2008.
16
[2]
ESA Engineering Standardisation Board, “Pointing Error Engineering Handbook ESSBHB-E-003.” ESA-ESTEC Requirements & Standards Division, 2011.
[3]
Ott, T. Benoit, A. Van den Braembussche, P. and Fichter, W. “ESA Pointing Error
Engineering Handbook.”, 8th International ESA GNC Conference. Karlovy Vary, Czech, 2011.
[4]
Lucke, R.L. Sirlin, S.W. and San Martin, A.M. “New Definition of Pointing Stability: AC
and DC Effects.” The Journal of the Astronautical Sciences, Vol. 40, No. 4, pp. 557-576, 1992.
[5]
Pittelkau, M.E. “Pointing Error Definitions, Metrics, and Algorithms.” American
Astronautical Society, AAS 03-559, p. 901, 2003.
[6]
Laureijs, R.J. Duvet, L. Escudero Sanz, I. Gondoin, P. Lumb, D.H. Oosterbroek, T. and
Saavedra Criado, G. “The Euclid Mission.” Proceedings of SPIE 7731. San Diego, USA , 2010.
[7]
Kangas, V. D'Addio, S. Betto, M. Barre, H. Loiselet, M. and Mason, G. “Metop Second
Generation microwave sounding and microwave imaging missions.” Proceedings of the 2012
EUMETSAT Meteorological Satellite Conference. Sopot, Poland, 2012.
[8]
Tarabini Castellani, L. Rodriguez, G. Llorente, S. Fernandez, J.M. Ruiz, M. Mastreau, A.
Cropp, A. and Santovincenzo, A. “Proba-3 Formation Flying Mission.” 7th International
Workshop on Satellite Constellations and Formation Flying. Lisbon, Portugal, 2013.
[9]
Bayard, D. S. “A State-Space Approach to Computing Spacecraft Pointing Jitter.” AIAA
Journal of Guidance, Control, and Dynamics, Vol. 27, No. 3, pp. 426-433, 2004.
[10] Bayard, D. S. and Neat, G. “Performance Characterization of a Stellar Interferometer.”
IEEE Control Systems Magazine, pp. 90-108, October 2007.
[11] Eggert, J. Hirth, M. Ott, T. and Su, H. “Pointing Error Engineering Tool – PEET Software
User Manual.” Issue 1.7, 26 July 2013. URL: http://peet.estec.esa.int/files/ASTOS-PEET-SUM001_Iss1.7_PEET_Software_User_Manual.pdf.
[12] IEEE-Std-952-1997, Appendix B (R2008) “IEEE Standard Specification Format Guide and
Test Procedure for Single-Axis Interferometric Fiber Optic Gyros.”
[13] Hirth, M. Ott, T. and Su, H. “Pointing Error Engineering Tool - PointingSat Definition.”
Issue
1.3,
8
July
2013,
URL:
http://peet.estec.esa.int/files/ASTOS-PEET-TN001_Iss1.3_PointingSat_Definition.pdf.
[14] Casasco, M. Saavedra Criado, G. Weikert, S. Eggert, J. Hirth, M. Ott, T. and Su, H.
“Pointing Error Budgeting for High Pointing Accuracy Mission using the Pointing Error
Engineering Tool.” AIAA GNC Conference. Boston, Massachusetts, 2013.
.
17