Star Trackers

Celestial attitude determination, star-image processing, measurement modelling, error sources, spacecraft integration, and sensor fusion for precision GNC.

4.1.4 Star Trackers

A star tracker is a precision optical attitude sensor used to determine the orientation of a spacecraft relative to an inertial celestial reference frame. It observes stars within its field of view, determines their directions from the focal-plane image, identifies the observed stars using an onboard star catalogue, and solves for the orientation of the star-tracker frame relative to the inertial frame.

Star trackers are among the most accurate absolute attitude sensors used on spacecraft. They are therefore particularly important for missions requiring precise attitude knowledge, such as Earth observation, astronomy, formation flying, rendezvous and docking, laser communication, precision payload pointing, and deep-space navigation.

The basic star-tracker processing chain can be represented as

\[ \boxed{\text{Celestial Star Field}} \rightarrow \boxed{\text{Optical System}} \rightarrow \boxed{\text{Detector}} \rightarrow \boxed{\text{Star Detection}} \rightarrow \boxed{\text{Centroiding}} \rightarrow \boxed{\text{Star Identification}} \rightarrow \boxed{\text{Attitude Determination}} \rightarrow \boxed{\mathbf q_{ST}} \]

Unlike a gyroscope, the star tracker does not primarily measure angular velocity. It provides an absolute celestial attitude reference. This makes star trackers and gyroscopes highly complementary sensors.

4.1.4.1 Role of Star Trackers in GNC

The attitude of a spacecraft describes the orientation of the spacecraft body frame relative to a chosen reference frame. For many missions this reference is an inertial frame.

Suppose the inertial direction of a catalogue star is

\[ \mathbf s_i^I. \]

The same star is observed by the star tracker as a direction

\[ \mathbf s_i^S \]

expressed in the star-tracker sensor frame \(S\).

Ideally,

\[ \boxed{ \mathbf s_i^S = C_I^S \mathbf s_i^I } \]

where \(C_I^S\) transforms vectors from the inertial frame into the star-tracker frame.

If the mounting transformation between the star-tracker frame and the spacecraft body frame is known,

\[ C_S^B, \]

then the spacecraft attitude can be obtained from

\[ \boxed{ C_I^B = C_S^B C_I^S } \]

for this particular DCM convention.

Therefore, the fundamental star-tracker problem is

\[ \boxed{ \{\mathbf s_i^I\} + \{\mathbf s_i^S\} \rightarrow C_I^S \rightarrow C_I^B } \]

or equivalently,

\[ \boxed{ \text{Known Celestial Directions} + \text{Measured Star Directions} \rightarrow \text{Spacecraft Attitude}. } \]

4.1.4.2 Why Stars Are Useful Attitude References

Stars provide excellent attitude references because they are extremely distant. Their directions therefore behave approximately as fixed directions in an inertial celestial reference frame over the short time scales relevant to spacecraft attitude determination.

For a catalogue star,

\[ \mathbf s_i^I = \begin{bmatrix} s_{ix}\\ s_{iy}\\ s_{iz} \end{bmatrix} \]

is known from astronomical catalogue data.

The spacecraft rotates this inertial direction into the sensor frame:

\[ \boxed{ \mathbf s_i^S = C_I^S(\mathbf q)\mathbf s_i^I. } \]

The star tracker observes several such directions and solves the inverse problem:

\[ \boxed{ \{\mathbf s_i^I,\mathbf s_i^S\} \rightarrow \hat{\mathbf q}. } \]

This provides an absolute attitude reference and prevents the unbounded long-term drift associated with integrating gyroscope measurements alone.

4.1.4.3 Star Tracker vs Gyroscope

A gyroscope measures angular velocity:

\[ \boxed{ \boldsymbol{\omega}_m = \boldsymbol{\omega}_{true} + \mathbf b_g + \mathbf n_g. } \]

The spacecraft attitude must then be propagated using the measured angular velocity. Because gyro bias is integrated with time, attitude errors accumulate.

A star tracker instead provides an absolute attitude measurement. A simplified quaternion measurement model is

\[ \boxed{ \mathbf q_{ST} = \delta\mathbf q_{ST} \otimes \mathbf q_{true} } \]

where \(\delta\mathbf q_{ST}\) represents the star-tracker attitude measurement error under a specified quaternion convention.

Consequently,

\[ \boxed{\text{Gyroscope}} \rightarrow \boxed{\text{High-Rate Relative Attitude Propagation}} \]

while

\[ \boxed{\text{Star Tracker}} \rightarrow \boxed{\text{Accurate Absolute Attitude Correction}}. \]

A spacecraft attitude estimator therefore commonly combines both:

\[ \boxed{\text{Gyroscope + Star Tracker}} \rightarrow \boxed{\text{EKF / MEKF}} \rightarrow \boxed{\text{Attitude + Gyro Bias Estimate}}. \]

4.1.4.4 Main Components of a Star Tracker

A modern autonomous star tracker combines optical, electronic, and software subsystems.

The principal components are

\[ \boxed{\text{Baffle}} \rightarrow \boxed{\text{Optics}} \rightarrow \boxed{\text{Image Detector}} \rightarrow \boxed{\text{Processing Electronics}} \rightarrow \boxed{\text{Star Catalogue}} \rightarrow \boxed{\text{Attitude Algorithm}}. \]

Optical System

The optical system focuses light from stars onto the detector. Important optical parameters include:

Detector

The detector converts incident photons into electrical and eventually digital signals. CMOS and CCD detector technologies are commonly associated with star-sensing systems.

Baffle

The baffle suppresses off-axis illumination from bright objects such as the Sun, Earth, and Moon and reduces stray light reaching the detector.

Processing Electronics

The processing electronics may perform image acquisition, thresholding, star detection, centroiding, catalogue matching, attitude determination, quality assessment, and communication with the spacecraft computer.

Star Catalogue

The onboard star catalogue provides the celestial reference directions needed to identify the stars observed in the image.

4.1.4.5 Field of View

A star tracker observes only a limited region of the celestial sphere at any instant. This region is described by the instrument field of view.

For a detector dimension \(D\) and focal length \(f\), an approximate one-dimensional field of view is

\[ \boxed{ \mathrm{FOV} = 2\tan^{-1} \left( \frac{D}{2f} \right). } \]

For a rectangular detector,

\[ \boxed{ \mathrm{FOV}_x = 2\tan^{-1} \left( \frac{D_x}{2f} \right) } \]

and

\[ \boxed{ \mathrm{FOV}_y = 2\tan^{-1} \left( \frac{D_y}{2f} \right). } \]

A wide field of view generally increases the number of stars available for identification and can improve lost-in-space acquisition robustness.

However, for a fixed detector, increasing the field of view generally reduces the angular scale represented by each pixel.

Thus an important optical trade-off is

\[ \boxed{ \text{Wide FOV} \leftrightarrow \text{Fine Angular Resolution}. } \]

4.1.4.6 Star Magnitude and Limiting Magnitude

Stars have very different apparent brightnesses. Apparent magnitude is a logarithmic measure of brightness.

For two stars,

\[ \boxed{ m_1-m_2 = -2.5 \log_{10} \left( \frac{F_1}{F_2} \right) } \]

where \(F_1\) and \(F_2\) are their observed fluxes.

A star tracker has a limiting magnitude \(m_{\mathrm{lim}}\), which describes approximately the faintest stars that can be reliably detected under specified operating conditions.

The limiting magnitude depends on

A tracker capable of detecting fainter stars generally has a larger candidate star population available inside a given field of view.

4.1.4.7 From Starlight to a Detector Image

Because stars are extremely distant, the light arriving from an individual star can be approximated as a bundle of parallel rays.

The optical system focuses those rays onto the focal plane.

An ideal point source would produce an infinitesimally small point. Real optical systems instead produce a finite intensity distribution known as the point-spread function:

\[ \boxed{ PSF(x,y). } \]

Consequently, a star normally illuminates multiple pixels rather than a single detector element.

The image-formation chain is therefore

\[ \boxed{\text{Star Direction}} \rightarrow \boxed{\text{Optical Projection}} \rightarrow \boxed{\text{PSF}} \rightarrow \boxed{\text{Detector Pixels}}. \]

The spread across several pixels enables sub-pixel centroid estimation, which is one reason star trackers can achieve angular measurement precision much finer than the angular width of a single pixel.

4.1.4.8 Pinhole Camera Model

A useful first-order model of the star-tracker optics is the pinhole camera model.

Consider a star direction expressed in the star-tracker frame:

\[ \mathbf s^S = \begin{bmatrix} s_x\\ s_y\\ s_z \end{bmatrix}. \]

Assuming the optical axis is the \(+z\) direction, the ideal focal-plane coordinates are

\[ \boxed{ x_f = f\frac{s_x}{s_z} } \]

and

\[ \boxed{ y_f = f\frac{s_y}{s_z}. } \]

If the detector pixel pitches are \(p_x\) and \(p_y\), then the corresponding pixel coordinates may be written as

\[ \boxed{ u = u_0+\frac{x_f}{p_x} } \]

and

\[ \boxed{ v = v_0+\frac{y_f}{p_y} } \]

where \((u_0,v_0)\) represents the principal point.

4.1.4.9 Pixel Position to Star Line of Sight

After a star image has been detected and its centroid measured at \((u,v)\), the processing algorithm converts the centroid back into a line-of-sight direction.

The focal-plane coordinates are

\[ x=(u-u_0)p_x \]

and

\[ y=(v-v_0)p_y. \]

An unnormalised direction is

\[ \tilde{\mathbf s}^{S} = \begin{bmatrix} x\\ y\\ f \end{bmatrix}. \]

The corresponding unit vector is

\[ \boxed{ \mathbf s^{S} = \frac{ \tilde{\mathbf s}^{S} }{ \|\tilde{\mathbf s}^{S}\| }. } \]

Thus,

\[ \boxed{ \text{Centroid} \rightarrow \text{Calibrated Pixel Coordinate} \rightarrow \text{Star LOS Vector}. } \]

4.1.4.10 Camera Intrinsic Matrix

The projection can be expressed more compactly using the camera intrinsic matrix

\[ K = \begin{bmatrix} f_x & \gamma & c_x\\ 0 & f_y & c_y\\ 0 & 0 & 1 \end{bmatrix}. \]

For an image point represented in homogeneous coordinates,

\[ \tilde{\mathbf p} = \begin{bmatrix} u\\ v\\ 1 \end{bmatrix}, \]

the ideal projection satisfies

\[ \tilde{\mathbf p} \propto K\mathbf s^S. \]

Conversely,

\[ \boxed{ \mathbf s^S = \frac{ K^{-1}\tilde{\mathbf p} }{ \|K^{-1}\tilde{\mathbf p}\| }. } \]

Errors in the calibrated values of \(f_x,f_y,c_x,c_y\), detector orientation, or pixel geometry therefore translate into star-vector errors and ultimately into attitude errors.

4.1.4.11 Optical Distortion

Real optical systems do not follow the ideal pinhole model exactly. One important source of systematic error is lens distortion.

For radial distortion, define

\[ r^2=x^2+y^2. \]

A commonly used polynomial model is

\[ \boxed{ x_d = x \left( 1+k_1r^2+k_2r^4+k_3r^6+\cdots \right) } \]

and

\[ \boxed{ y_d = y \left( 1+k_1r^2+k_2r^4+k_3r^6+\cdots \right). } \]

Tangential distortion and other optical imperfections may also be included.

A realistic processing chain therefore contains

\[ \boxed{\text{Raw Centroid}} \rightarrow \boxed{\text{Distortion Correction}} \rightarrow \boxed{\text{Calibrated Star Vector}}. \]

4.1.4.12 Star Detection

The digital detector image contains stars, background illumination, detector noise, hot pixels, radiation events, and possibly stray-light features.

The processing software must first identify candidate star regions. A simple detection condition is

\[ I(u,v)>I_{\mathrm{threshold}}. \]

An adaptive threshold may be based on the estimated background:

\[ \boxed{ I_{\mathrm{threshold}} = \mu_{bg} + k\sigma_{bg} } \]

where

Connected pixels exceeding the threshold can then be grouped into candidate objects.

Further checks may reject

4.1.4.13 Centroiding

Once a candidate star has been detected, its location must be estimated more accurately than the raw pixel coordinates.

A common intensity-weighted centroid is

\[ \boxed{ u_c = \frac{ \sum_i I_i u_i }{ \sum_i I_i } } \]

and

\[ \boxed{ v_c = \frac{ \sum_i I_i v_i }{ \sum_i I_i }. } \]

The summation is performed over the pixels associated with the detected star image.

Because the PSF spreads the star signal across multiple pixels, centroiding can estimate the star location to a fraction of a pixel.

More advanced methods may include

4.1.4.14 Centroid Error and Angular Error

Suppose the centroid uncertainty is \(\sigma_c\) pixels and the detector pixel pitch is \(p\).

The approximate focal-plane position uncertainty is

\[ \sigma_x \approx p\sigma_c. \]

For small angles, the corresponding angular uncertainty is approximately

\[ \boxed{ \sigma_\theta \approx \frac{p}{f}\sigma_c. } \]

This relationship demonstrates several important design effects:

However, a longer focal length for a fixed detector also reduces the field of view.

4.1.4.15 Signal-to-Noise Ratio

The quality of star detection and centroiding depends strongly on the signal-to-noise ratio.

A representative detector-level expression is

\[ \boxed{ SNR = \frac{N_s} { \sqrt{ N_s+ N_b+ N_d+ N_r^2 } } } \]

where, depending on the detector model,

Higher SNR generally improves both star-detection probability and centroid precision.

4.1.4.16 Detector Noise Sources

Photon Shot Noise

Photon arrival is a statistical process. For \(N\) detected photons, the standard deviation associated with ideal Poisson shot noise scales approximately as

\[ \boxed{ \sigma_{shot}\approx\sqrt{N}. } \]

Read Noise

Electronic circuitry introduces noise while reading detector pixels. This becomes particularly important for weak star signals.

Dark Current

Thermally generated charge accumulates even without incident starlight. Dark current generally increases with detector temperature.

Fixed-Pattern Noise

Different pixels may have slightly different gain and offset characteristics.

Quantisation Noise

The analogue signal is converted into finite digital levels by an analogue-to-digital converter.

Hot and Dead Pixels

Some pixels may produce unusually high output or fail to respond normally.

Radiation Events

Energetic particles can produce transient bright features that may initially resemble stars.

4.1.4.17 Exposure Time

The exposure time \(T_e\) determines how long the detector integrates incoming light.

A longer exposure generally collects more photons and can improve the detection of faint stars.

However, if the spacecraft rotates during the exposure, the star image moves across the detector.

Therefore,

\[ \boxed{ \text{Longer Exposure} \rightarrow \text{Higher Sensitivity} } \]

but also

\[ \boxed{ \text{Longer Exposure} \rightarrow \text{Greater Motion Blur}. } \]

Exposure time is therefore one of the principal star-tracker design and operational trade-offs.

4.1.4.18 Motion Blur

If the spacecraft angular rate during an exposure is approximately \(\omega\), then the first-order angular displacement during the exposure is

\[ \boxed{ \Delta\theta_{blur} \approx \omega T_e. } \]

For small angles, the corresponding focal-plane displacement is approximately

\[ \boxed{ \Delta x \approx f\omega T_e. } \]

In detector pixels,

\[ \boxed{ N_{blur} \approx \frac{f\omega T_e}{p}. } \]

Excessive blur can cause

4.1.4.19 Maximum Tracking and Acquisition Rate

Star trackers usually specify angular-rate limits.

A simplified relation is

\[ \omega_{\max}T_e < \theta_{\mathrm{allowable\ smear}}. \]

Thus,

\[ \boxed{ \omega_{\max} \approx \frac{ \theta_{\mathrm{allowable\ smear}} }{ T_e }. } \]

Actual limits depend on detector sensitivity, exposure-control algorithms, image processing, star brightness, and field of view.

It is important to distinguish between maximum acquisition rate and maximum tracking rate. A tracker that already knows approximately where the stars should be may continue tracking under conditions where a complete lost-in-space solution would be difficult.

4.1.4.20 Star Catalogue

The onboard star catalogue contains the reference information needed to identify observed stars.

A catalogue entry may contain

\[ \{ ID_i,\alpha_i,\delta_i,m_i,\ldots \} \]

where

The corresponding inertial unit vector can be formed as

\[ \boxed{ \mathbf s_i^I = \begin{bmatrix} \cos\delta_i\cos\alpha_i\\ \cos\delta_i\sin\alpha_i\\ \sin\delta_i \end{bmatrix}. } \]

The onboard catalogue is normally optimized for attitude determination rather than simply containing every known astronomical object.

4.1.4.21 Catalogue Epoch and Apparent Star Direction

For very accurate attitude determination, catalogue vectors cannot always be treated as mathematically immutable.

Depending on the required accuracy, corrections may be considered for

A higher-fidelity reference-generation chain is therefore

\[ \boxed{\text{Catalogue}} \rightarrow \boxed{\text{Epoch / Apparent-Position Corrections}} \rightarrow \boxed{\mathbf s_i^I(t)}. \]

4.1.4.22 Star Identification

After stars have been detected and converted into measured directions, the tracker must determine which catalogue stars produced those observations.

This is the star-identification problem.

The central problem is

\[ \boxed{ \text{Observed Star Pattern} \leftrightarrow \text{Catalogue Star Pattern}. } \]

Useful geometric quantities include

4.1.4.23 Inter-Star Angle

For two measured star directions

\[ \mathbf s_1,\qquad\mathbf s_2, \]

their angular separation is

\[ \boxed{ \theta_{12} = \cos^{-1} \left( \mathbf s_1^T\mathbf s_2 \right). } \]

A rigid spacecraft rotation changes the coordinates of both vectors but does not change the angle between them.

Therefore, inter-star angular separation provides an attitude-independent geometric quantity that can be compared directly with catalogue data.

4.1.4.24 Pattern-Matching Algorithms

A variety of methods can be used for autonomous star identification. Examples include

Although the computational details differ, the objective remains

\[ \boxed{ \{\text{Measured Geometric Features}\} \rightarrow \{\text{Catalogue Correspondences}\}. } \]

Once sufficiently reliable correspondences have been established, the tracker can solve for attitude.

4.1.4.25 Lost-in-Space Mode

When the tracker has no useful prior knowledge of spacecraft attitude, it must determine attitude directly from the current star image.

This operating condition is commonly called lost-in-space.

The processing sequence is

\[ \boxed{\text{Acquire Image}} \rightarrow \boxed{\text{Detect Stars}} \rightarrow \boxed{\text{Calculate Star Pattern}} \rightarrow \boxed{\text{Search Catalogue}} \rightarrow \boxed{\text{Identify Stars}} \rightarrow \boxed{\text{Solve Attitude}}. \]

Lost-in-space processing generally requires more computation than tracking because the catalogue search cannot rely strongly on a prior attitude estimate.

4.1.4.26 Tracking Mode

Once attitude is known, the previous attitude solution can be used to predict which catalogue stars should appear in the next image.

Conceptually,

\[ \boxed{ \mathbf q_{k-1} \rightarrow \text{Predicted Star Locations} \rightarrow \text{Local Image Search} \rightarrow \mathbf q_k. } \]

Tracking mode can therefore be computationally faster and more robust than repeating a complete lost-in-space solution every frame.

A typical operational sequence is

\[ \boxed{\text{Lost-in-Space Acquisition}} \rightarrow \boxed{\text{Initial Attitude}} \rightarrow \boxed{\text{Tracking Mode}}. \]

4.1.4.27 Wahba's Problem

Once observed stars have been associated with catalogue stars, the attitude can be found by solving a vector-observation attitude determination problem.

The classical formulation is Wahba's problem.

Suppose

\[ \mathbf r_i \]

is the known inertial reference vector and

\[ \mathbf b_i \]

is the corresponding measured sensor/body-frame vector.

The attitude matrix \(C\) is selected to minimize

\[ \boxed{ J(C) = \frac{1}{2} \sum_{i=1}^{N} a_i \left\| \mathbf b_i-C\mathbf r_i \right\|^2 } \]

where \(a_i\) represents the weighting assigned to observation \(i\).

The weights may reflect the relative quality or covariance of different star measurements.

4.1.4.28 Attitude Determination Algorithms

Several algorithms can solve the attitude-determination problem. Important examples include

The output is normally represented as

\[ \boxed{ \hat C_I^S } \]

or

\[ \boxed{ \hat{\mathbf q}_{SI}. } \]

4.1.4.29 Davenport's q-Method

Define the attitude profile matrix

\[ \boxed{ B = \sum_i a_i \mathbf b_i \mathbf r_i^T. } \]

Then define

\[ S=B+B^T, \]

\[ \sigma=\operatorname{tr}(B), \]

and

\[ \mathbf z = \begin{bmatrix} B_{23}-B_{32}\\ B_{31}-B_{13}\\ B_{12}-B_{21} \end{bmatrix}. \]

The Davenport matrix is

\[ \boxed{ K = \begin{bmatrix} S-\sigma I & \mathbf z\\ \mathbf z^T & \sigma \end{bmatrix}. } \]

The quaternion corresponding to the optimal attitude is obtained from the eigenvector associated with the maximum eigenvalue of \(K\), subject to the quaternion ordering and attitude convention used in the implementation.

4.1.4.30 QUEST

QUEST stands for Quaternion ESTimator. It is an efficient method for solving Wahba's problem and is strongly associated with spacecraft attitude determination.

Conceptually,

\[ \boxed{ \{ \mathbf b_i, \mathbf r_i, a_i \} \rightarrow \text{QUEST} \rightarrow \hat{\mathbf q}. } \]

QUEST combines several noisy vector observations to obtain an attitude quaternion consistent with the weighted vector measurements.

The star-tracker processing chain therefore becomes

\[ \boxed{\text{Centroids}} \rightarrow \boxed{\text{Star LOS Vectors}} \rightarrow \boxed{\text{Catalogue Matching}} \rightarrow \boxed{\text{QUEST}} \rightarrow \boxed{\mathbf q_{ST}}. \]

4.1.4.31 Number of Stars Required

A single measured star vector does not uniquely determine complete three-axis attitude because rotation about that vector remains unobservable.

Two non-collinear vector observations can theoretically determine full attitude.

Therefore, under ideal geometry,

\[ \boxed{ N_{\mathrm{stars}}\geq2 } \]

can provide a complete attitude solution.

Practical star trackers normally use more than two stars because additional observations improve

4.1.4.32 Importance of Star Geometry

The number of observed stars alone does not completely determine attitude accuracy. Their geometric distribution is also important.

Stars spread over the field of view generally provide stronger attitude information than stars concentrated in a small region.

The attitude covariance therefore depends on

4.1.4.33 Star Tracker Outputs

A star tracker can provide more than a quaternion. Depending on the hardware, outputs may include

For a GNC estimator, a useful abstract output interface is

\[ \boxed{ \{ \mathbf q_{ST}, R_{ST}, t_{ST}, valid_{ST} \}. } \]

4.1.4.34 Ideal Star Tracker Model

At the simplest GNC modelling level, an ideal star tracker produces exactly the true spacecraft attitude:

\[ \boxed{ \mathbf q_{ST} = \mathbf q_{true}. } \]

At vector level,

\[ \boxed{ \mathbf s_{m,i}^S = C_I^S \mathbf s_i^I. } \]

This ideal model assumes

4.1.4.35 Small-Angle Attitude Error Model

For GNC simulation and estimation, star-tracker error is often represented as a small three-axis rotation:

\[ \delta\boldsymbol{\theta} = \begin{bmatrix} \delta\theta_x\\ \delta\theta_y\\ \delta\theta_z \end{bmatrix}. \]

For a scalar-first quaternion and sufficiently small error,

\[ \boxed{ \delta\mathbf q \approx \begin{bmatrix} 1\\ \frac{1}{2}\delta\boldsymbol{\theta} \end{bmatrix}. } \]

A measured quaternion can then be modelled as

\[ \boxed{ \mathbf q_m = \delta\mathbf q \otimes \mathbf q_{true}. } \]

The multiplication order depends on whether the error is defined in the left- or right-multiplicative sense and on the adopted frame convention. The convention must therefore be defined explicitly in both simulation and flight software.

4.1.4.36 Measurement Covariance

The small attitude error can be modelled as a random vector:

\[ \boxed{ \delta\boldsymbol{\theta}_{ST} \sim \mathcal N ( \mathbf 0, R_{ST} ). } \]

A simplified diagonal covariance is

\[ \boxed{ R_{ST} = \begin{bmatrix} \sigma_x^2&0&0\\ 0&\sigma_y^2&0\\ 0&0&\sigma_z^2 \end{bmatrix}. } \]

A more complete model may contain non-zero cross-correlation terms.

The covariance is particularly important when the measurement is fused inside an EKF or MEKF because it determines how strongly the estimator should trust the star tracker relative to its propagated attitude.

4.1.4.37 Cross-Boresight and Roll Accuracy

Star-tracker attitude accuracy is often not identical about all three axes.

The optical axis is called the boresight. Errors perpendicular to the optical axis are commonly described as cross-boresight errors.

Rotation about the optical axis is commonly referred to as roll or twist about the boresight.

A simplified covariance may therefore have the form

\[ R_{ST} = \begin{bmatrix} \sigma_{CB1}^2&0&0\\ 0&\sigma_{CB2}^2&0\\ 0&0&\sigma_{roll}^2 \end{bmatrix}. \]

The roll accuracy may differ from cross-boresight accuracy because roll information depends strongly on the relative geometry of multiple stars across the focal plane.

For precision simulations, using one scalar accuracy value for all three axes can therefore be insufficient.

4.1.4.38 Accuracy, Precision and Attitude Knowledge

Precision

Precision describes the repeatability of measurements.

Accuracy

Accuracy describes closeness to the true attitude.

Attitude Knowledge

Attitude knowledge describes how accurately the complete spacecraft navigation system knows the spacecraft orientation.

Therefore,

\[ \boxed{ \text{Star Tracker Accuracy} \neq \text{Spacecraft Attitude Knowledge}. } \]

Final attitude knowledge can also be affected by

4.1.4.39 Star Tracker Bias

A simplified attitude-error model can be written as

\[ \boxed{ \delta\boldsymbol{\theta}_m = \mathbf b_{ST} + \mathbf n_{ST} } \]

where

Systematic error may arise from

4.1.4.40 Mounting Alignment

The star tracker measures attitude in its own sensor frame \(S\), while the GNC system normally requires attitude in the spacecraft body frame \(B\).

The calibrated sensor-to-body transformation is

\[ C_S^B. \]

Thus,

\[ \boxed{ C_I^B = C_S^B C_I^S. } \]

If the actual mounting transformation differs from the calibrated transformation, a systematic spacecraft attitude error results.

For a small alignment error, a first-order representation is

\[ \boxed{ C_{S,true}^{B} \approx \left( I- [ \delta\boldsymbol{\theta}_{align}\times ] \right) C_{S,cal}^{B} } \]

subject to the selected perturbation convention.

4.1.4.41 Thermoelastic Alignment Error

The mechanical relationship between the star tracker and spacecraft body can change as temperatures vary.

Possible causes include

A simple first-order model is

\[ \boxed{ \delta\boldsymbol{\theta}_{align}(T) = \delta\boldsymbol{\theta}_0 + K_T(T-T_0). } \]

For high-accuracy spacecraft, thermoelastic alignment errors can become comparable to or larger than the intrinsic random measurement noise of the tracker.

4.1.4.42 Timing Error

A star-tracker attitude solution corresponds to the time at which the image was observed. If the measurement is associated with the wrong time, the spacecraft may have rotated between the true observation time and the assumed measurement time.

If the timestamp error is \(\delta t\), then for small rotations

\[ \boxed{ \delta\boldsymbol{\theta}_{time} \approx \boldsymbol{\omega}\delta t. } \]

Therefore, timing accuracy becomes increasingly important as spacecraft angular rate increases.

4.1.4.43 Latency

Latency is the time between image acquisition and availability of the processed attitude solution.

The chain is approximately

\[ \boxed{\text{Exposure}} \rightarrow \boxed{\text{Readout}} \rightarrow \boxed{\text{Image Processing}} \rightarrow \boxed{\text{Star Identification}} \rightarrow \boxed{\text{Attitude Solution}} \rightarrow \boxed{\text{Communication}}. \]

If an image corresponds to time \(t_k\) but is delivered at

\[ t_k+\tau, \]

then \(\tau\) is the measurement latency.

A known latency can often be handled by the navigation architecture, provided the measurement is correctly time tagged.

4.1.4.44 Update Rate

Star trackers normally provide measurements at a finite update rate. If the update period is \(T_{ST}\),

\[ \boxed{ f_{ST} = \frac{1}{T_{ST}}. } \]

Gyroscopes generally provide measurements much more frequently. Therefore, the estimator commonly uses

\[ \boxed{\text{Gyroscope}} \rightarrow \boxed{\text{High-Rate Propagation}} \]

and

\[ \boxed{\text{Star Tracker}} \rightarrow \boxed{\text{Periodic Absolute Correction}}. \]

4.1.4.45 Sun Exclusion Angle

The Sun is enormously brighter than the stars observed by a star tracker. Direct sunlight or strong near-axis sunlight can therefore saturate the detector or generate severe stray light.

The tracker typically requires

\[ \boxed{ \theta_{Sun} > \theta_{Sun,exclusion}. } \]

If the Sun approaches the exclusion region, possible effects include

Sun exclusion is therefore both a sensor specification and a spacecraft attitude-planning constraint.

4.1.4.46 Earth Exclusion Angle

The bright Earth can also produce significant background and stray light, particularly for spacecraft in low Earth orbit.

A tracker may specify a minimum separation from the bright Earth limb:

\[ \boxed{ \theta_{Earth\,limb} > \theta_{Earth,exclusion}. } \]

Earth exclusion geometry should therefore be included when analysing star-tracker availability over an orbit.

4.1.4.47 Moon Exclusion Angle

The Moon is another bright object that can interfere with star detection.

A corresponding constraint may be

\[ \boxed{ \theta_{Moon} > \theta_{Moon,exclusion}. } \]

The required angle depends on the tracker optics, baffle, detector, and image-processing architecture.

4.1.4.48 Stray Light

A useful focal-plane intensity model is

\[ \boxed{ I_{measured} = I_{stars} + I_{background} + I_{stray} + n_{detector}. } \]

Stray-light sources can include

Stray light raises the image background and reduces the contrast between stars and the detector background.

4.1.4.49 Baffle Design

The baffle prevents or attenuates unwanted off-axis light before it reaches the optical system.

Important baffle characteristics include

The conceptual process is

\[ \boxed{\text{Bright Off-Axis Source}} \rightarrow \boxed{\text{Baffle Rejection}} \rightarrow \boxed{\text{Reduced Stray Light}} \rightarrow \boxed{\text{Improved Star Detection}}. \]

4.1.4.50 Earth Albedo and Spacecraft Reflections

Even if the Sun itself is outside the exclusion angle, sunlight reflected from Earth or spacecraft surfaces may enter the star-tracker optics.

Potential reflecting structures include

Consequently, star-tracker accommodation should consider the complete spacecraft optical environment rather than only direct celestial sources.

4.1.4.51 Radiation Effects

The detector and electronics operate in the space-radiation environment.

Possible effects include

Radiation-induced detector events can be rejected using spatial shape, temporal persistence, predicted star position, or catalogue-pattern consistency.

4.1.4.52 False Stars and Non-Stellar Objects

Not every bright object detected in the focal plane is necessarily a catalogue star.

False candidates can include

Reliable attitude determination therefore depends on geometric consistency between multiple detected objects and the catalogue.

4.1.4.53 Blinding and Loss of Tracking

A star tracker may temporarily fail to provide a valid attitude solution.

Possible causes include

A realistic sensor interface therefore requires a validity flag:

\[ \boxed{ valid_{ST} = \begin{cases} 1,&\text{valid attitude solution}\\ 0,&\text{invalid attitude solution}. \end{cases} } \]

4.1.4.54 Acquisition and Reacquisition Time

After power-up or loss of tracking, the tracker may require finite time before a valid attitude solution becomes available.

Relevant specifications can include

These parameters are important for safe-mode and manoeuvre recovery analysis.

4.1.4.55 Quaternion Sign Ambiguity

The two quaternions

\[ \mathbf q \]

and

\[ -\mathbf q \]

represent exactly the same physical orientation.

Therefore,

\[ \boxed{ \mathbf q\equiv-\mathbf q. } \]

A tracker may consequently output an attitude quaternion with the opposite sign from the previous sample even though the physical attitude has changed only slightly.

A common continuity test is

\[ \mathbf q_k^T\mathbf q_{k-1}<0. \]

If this condition occurs, the sign can be changed:

\[ \boxed{ \mathbf q_k \leftarrow -\mathbf q_k. } \]

This prevents artificial sign discontinuities in interpolation, plotting, and some estimator interfaces.

4.1.4.56 Quaternion Measurement Residual

Quaternion components should generally not be treated as four independent Euclidean attitude errors.

Suppose the predicted attitude is

\[ \hat{\mathbf q} \]

and the star-tracker attitude is

\[ \mathbf q_{ST}. \]

An error quaternion can be formed as

\[ \boxed{ \delta\mathbf q = \mathbf q_{ST} \otimes \hat{\mathbf q}^{-1} } \]

for one common error convention.

For a small attitude error,

\[ \delta\mathbf q \approx \begin{bmatrix} 1\\ \frac12\delta\boldsymbol{\theta} \end{bmatrix}. \]

Hence,

\[ \boxed{ \delta\boldsymbol{\theta} \approx 2\delta\mathbf q_v } \]

where \(\delta\mathbf q_v\) is the vector part of the error quaternion.

4.1.4.57 Star Tracker in a Multiplicative EKF

A common spacecraft attitude-estimation architecture uses a nominal quaternion together with a small attitude-error state.

A representative error state is

\[ \boxed{ \delta\mathbf x = \begin{bmatrix} \delta\boldsymbol{\theta}\\ \delta\mathbf b_g \end{bmatrix} } \]

where

The gyro propagates the nominal attitude:

\[ \boxed{ \dot{\hat{\mathbf q}} = \frac12 \Omega \left( \boldsymbol{\omega}_m-\hat{\mathbf b}_g \right) \hat{\mathbf q}. } \]

When a star-tracker measurement arrives, an attitude-error quaternion is formed between the measured and predicted attitudes.

The small-angle residual is then used in the Kalman measurement update. The filter can correct both attitude error and gyro-bias error.

4.1.4.58 Gyroscope + Star Tracker Sensor Fusion

                 Gyroscope
                     |
                     v
              Angular Rate
                     |
                     v
            Bias Compensation
                     |
                     v
          Quaternion Propagation
                     |
                     v
             Predicted Attitude
                     |
                     +----------------------+
                                            |
                                            v
Star Tracker --------------------------> MEKF / EKF
Absolute Attitude                          |
                                            v
                                  Attitude Innovation
                                            |
                             +--------------+--------------+
                             |                             |
                             v                             v
                     Attitude Correction            Gyro Bias Estimate
                             |                             |
                             +--------------+--------------+
                                            |
                                            v
                                  Estimated Attitude
                                            |
                                            v
                                    GNC Controller

The gyroscope provides high-rate rotational information, while the star tracker periodically anchors the estimate to an absolute celestial reference.

4.1.4.59 Measurement Gating

A spacecraft estimator should not automatically accept every star-tracker measurement.

Suppose the attitude innovation is represented by

\[ \mathbf y. \]

If its innovation covariance is

\[ S, \]

then a normalized innovation statistic can be formed:

\[ \boxed{ d^2 = \mathbf y^T S^{-1} \mathbf y. } \]

If \(d^2\) exceeds a selected statistical threshold, the measurement may be rejected or flagged.

A large innovation can indicate

4.1.4.60 Measurement Validity Logic

A more complete validity function can be represented as

\[ \boxed{ valid_{ST} = f( status, N_{stars}, quality, timestamp, R_{ST}, innovation, geometry ). } \]

A star-tracker measurement may be accepted only if

4.1.4.61 Star Tracker Dropout

During a star-tracker outage,

\[ valid_{ST}=0. \]

The attitude estimator can continue using gyroscope propagation:

\[ \boxed{ \mathbf q_k \leftarrow \text{Gyro Propagation}. } \]

However, the uncertainty grows because gyro bias and noise are no longer periodically corrected by an absolute attitude sensor.

Therefore,

\[ \boxed{ \text{Longer Star-Tracker Outage} \rightarrow \text{Larger Attitude Uncertainty}. } \]

4.1.4.62 Multiple Star Trackers

Spacecraft requiring high availability or fault tolerance may carry more than one star tracker.

Advantages include

The architecture becomes

\[ \boxed{\text{ST-1}} + \boxed{\text{ST-2}} + \boxed{\text{Gyroscope}} \rightarrow \boxed{\text{Attitude Estimator}}. \]

4.1.4.63 Multiple Optical Heads

Some star-tracker architectures use multiple optical heads connected to common processing electronics.

Each head has a different mounting transformation:

\[ C_{S_1}^{B}, \qquad C_{S_2}^{B}, \qquad C_{S_3}^{B}. \]

This can improve total sky availability because a bright object that affects one optical head may not affect another.

However, a shared processor or power interface can introduce common-mode failure mechanisms, so multiple heads do not necessarily provide the same redundancy as completely independent tracker units.

4.1.4.64 Relative Alignment Monitoring

If two trackers provide simultaneous attitude solutions, their relative solutions can be compared.

The nominal relative orientation between two sensor frames is

\[ \boxed{ C_{S_1}^{S_2} = C_B^{S_2}C_{S_1}^{B}. } \]

A systematic change in the measured relative orientation may indicate

4.1.4.65 Star Tracker with Sun Sensors

A Sun sensor provides a direction to the Sun, whereas a star tracker can provide complete three-axis celestial attitude.

Sun sensors are nevertheless valuable for

A layered attitude-sensing architecture may therefore use

\[ \boxed{ \text{Gyroscope} + \text{Sun Sensor} + \text{Star Tracker}. } \]

4.1.4.66 Star Tracker with Magnetometers

A magnetometer provides the local geomagnetic-field direction in the spacecraft body frame.

A Sun sensor and magnetometer can provide coarse attitude information during early spacecraft operations.

A representative mode transition is

\[ \boxed{\text{Magnetometer + Sun Sensor}} \rightarrow \boxed{\text{Coarse Attitude}} \rightarrow \boxed{\text{Star Tracker Acquisition}} \rightarrow \boxed{\text{Fine Attitude Knowledge}}. \]

4.1.4.67 Safe-Mode Considerations

A star tracker is not always sufficient as the only safe-mode attitude sensor because it may require

A possible recovery sequence is

\[ \boxed{\text{Safe Mode}} \rightarrow \boxed{\text{Detumbling}} \rightarrow \boxed{\text{Coarse Sun / Magnetic Attitude}} \rightarrow \boxed{\text{Star Tracker Acquisition}} \rightarrow \boxed{\text{Fine Pointing}}. \]

4.1.4.68 Fine Pointing Applications

Star trackers are particularly important for missions requiring fine pointing.

Examples include

The control chain may be represented as

\[ \boxed{\text{Star Tracker}} \rightarrow \boxed{\text{Attitude Estimator}} \rightarrow \boxed{\text{Attitude Error}} \rightarrow \boxed{\text{Controller}} \rightarrow \boxed{\text{Reaction Wheels / CMGs / Thrusters}}. \]

4.1.4.69 Attitude Knowledge, Pointing Error and Stability

Three related but different concepts should be distinguished.

Attitude Knowledge Error

The uncertainty in the estimated spacecraft attitude.

Pointing Error

The difference between the actual spacecraft pointing direction and the commanded pointing direction.

Pointing Stability

The variation of the pointing direction over a specified time interval.

Therefore,

\[ \boxed{ \text{Sensor Accuracy} \rightarrow \text{Attitude Knowledge} \rightarrow \text{Control Performance} \rightarrow \text{Pointing Performance}. } \]

They should not be treated as interchangeable specifications.

4.1.4.70 Jitter and High-Frequency Motion

A star tracker has finite exposure time and finite update rate. It may therefore not fully observe very high-frequency spacecraft motion.

Potential jitter sources include

A tracker can therefore provide excellent low-frequency absolute attitude knowledge while additional sensors may be needed for high-frequency line-of-sight stabilization.

4.1.4.71 Payload Boresight and Structural Alignment

The star tracker measures the orientation of the star-tracker frame. The payload is interested in the orientation of its own boresight.

The complete transformation may be written as

\[ \boxed{ C_I^{Payload} = C_B^{Payload} C_S^B C_I^S. } \]

Therefore, even a highly accurate star tracker cannot guarantee equivalent payload pointing if the star-tracker-to-payload alignment is poorly known.

This leads to an important spacecraft-level principle:

\[ \boxed{ \text{Excellent Sensor Accuracy} \not\Rightarrow \text{Excellent Payload Pointing} } \]

unless the structural and alignment chain is also accurately characterized.

4.1.4.72 Ground Alignment Calibration

Before launch, the orientation of the star tracker relative to the spacecraft reference frame can be measured using precision metrology.

Methods may involve

The result is the calibrated mounting transformation

\[ \boxed{ C_S^B. } \]

The uncertainty of this transformation must be included in the spacecraft attitude knowledge error budget.

4.1.4.73 In-Orbit Alignment Calibration

Launch loads and the orbital thermal environment may change alignment slightly after ground calibration.

In some missions, alignment can be refined in orbit using

Let the alignment parameters be

\[ \mathbf p_{align}. \]

An estimation process can seek

\[ \boxed{ \hat{\mathbf p}_{align} = \arg\min_{\mathbf p_{align}} J(\mathbf p_{align}). } \]

4.1.4.74 Complete Star Tracker Error Budget

The total attitude error may be decomposed conceptually as

\[ \boxed{ \begin{aligned} \delta\boldsymbol{\theta}_{ST} ={}& \delta\boldsymbol{\theta}_{centroid} + \delta\boldsymbol{\theta}_{optics} + \delta\boldsymbol{\theta}_{catalogue} \\ &+ \delta\boldsymbol{\theta}_{alignment} + \delta\boldsymbol{\theta}_{thermal} + \delta\boldsymbol{\theta}_{timing} \\ &+ \delta\boldsymbol{\theta}_{motion} + \delta\boldsymbol{\theta}_{processing} + \delta\boldsymbol{\theta}_{noise}. \end{aligned} } \]

If independent zero-mean contributions can reasonably be assumed, an approximate root-sum-square uncertainty can be written as

\[ \boxed{ \sigma_{total} = \sqrt{ \sigma_1^2+ \sigma_2^2+ \cdots+ \sigma_n^2 }. } \]

For correlated errors, the complete covariance must be propagated rather than simply using root-sum-square addition.

4.1.4.75 High-Fidelity Attitude Measurement Model

At spacecraft GNC level, a useful high-fidelity model is

\[ \boxed{ \mathbf q_{ST}(t_k) = \delta\mathbf q_{err} \otimes \mathbf q_{true}(t_k-\tau) } \]

where \(\tau\) represents measurement timing or latency effects as defined by the simulation.

The total attitude-error quaternion can depend on

\[ \delta\mathbf q_{err} = f( \delta\boldsymbol{\theta}_{noise}, \delta\boldsymbol{\theta}_{align}, \delta\boldsymbol{\theta}_{thermal}, \delta\boldsymbol{\theta}_{optics}, \delta\boldsymbol{\theta}_{timing}, \ldots ). \]

The measurement is supplied to the estimator only when

\[ \boxed{ valid_{ST}=1. } \]

A realistic star-tracker model must therefore represent both measurement error and measurement availability.

4.1.4.76 Vector-Level Star Tracker Model

A more detailed model generates individual star measurements.

For each catalogue star,

\[ \boxed{ \mathbf s_i^I } \]

is transformed according to spacecraft attitude:

\[ \boxed{ \mathbf s_i^S = C_I^S(\mathbf q) \mathbf s_i^I. } \]

The direction is then processed through

\[ \boxed{\text{FOV Selection}} \rightarrow \boxed{\text{Optical Projection}} \rightarrow \boxed{\text{Distortion}} \rightarrow \boxed{\text{Centroid Error}} \rightarrow \boxed{\text{Measured LOS}}. \]

The resulting noisy vectors are passed to an attitude solver such as QUEST.

4.1.4.77 Image-Level Star Tracker Model

The highest-fidelity simulation creates a synthetic focal-plane image. 

Star Catalogue
      |
      v
Reference Star Directions
      |
      v
Spacecraft Attitude
      |
      v
Sensor-Frame Star Directions
      |
      v
Field-of-View Selection
      |
      v
Optical Projection
      |
      v
Lens Distortion
      |
      v
Point-Spread Function
      |
      v
Motion Blur
      |
      v
Photon Statistics
      |
      v
Detector Noise
      |
      v
Background + Stray Light
      |
      v
Synthetic Pixel Image
      |
      v
Star Detection
      |
      v
Centroiding
      |
      v
Star Identification
      |
      v
QUEST / Wahba Solver
      |
      v
Attitude Quaternion

This level is appropriate when developing or validating the star tracker itself. It is usually unnecessary for early spacecraft controller design.

4.1.4.78 Three Levels of Simulation Fidelity

Level 1 — Attitude Measurement Model

Start with truth attitude and add

For example,

\[ \boxed{ q_{ST} = \delta q_{err} \otimes q_{true}. } \]

This level is suitable for most GNC Monte Carlo analysis.

Level 2 — Star Vector Model

Generate visible star directions and solve the attitude using QUEST or another Wahba solver.

This tests the attitude-determination algorithm as well as the estimator.

Level 3 — Image-Level Model

Generate synthetic focal-plane images containing optics, detector noise, PSF, motion blur, stray light, centroiding, and pattern recognition.

This level is appropriate for detailed star-tracker development.

4.1.4.79 Simulink Star Tracker Model

For spacecraft GNC simulation, a practical Simulink architecture is 

True Spacecraft Attitude
          |
          v
Sensor Mounting Transformation
          |
          v
Star Tracker Frame Attitude
          |
          v
Alignment Error
          |
          v
Thermal Alignment Drift
          |
          v
Attitude Measurement Noise
          |
          v
Sample / Update-Rate Logic
          |
          v
Latency / Delay
          |
          v
Sun / Earth / Moon Exclusion Logic
          |
          v
Angular-Rate Validity Logic
          |
          v
Dropout / Availability Logic
          |
          v
Measured Quaternion
          |
          +---------> Measurement Covariance
          |
          +---------> Validity Flag
          |
          +---------> Measurement Timestamp
          |
          v
       MEKF / EKF

Useful Simulink blocks include

4.1.4.80 Star Visibility Model

For vector-level simulation, the model must determine which catalogue stars lie inside the tracker field of view.

Let the star-tracker boresight expressed in the inertial frame be

\[ \mathbf b^I. \]

For catalogue star \(i\),

\[ \mathbf s_i^I, \]

the angular separation from the boresight is

\[ \boxed{ \theta_i = \cos^{-1} \left[ (\mathbf b^I)^T \mathbf s_i^I \right]. } \]

For a circular field of view with half-angle \(\theta_{FOV}\),

\[ \boxed{ \theta_i \leq \theta_{FOV} } \]

is the basic geometric visibility condition.

Further checks should then consider magnitude, occultation, exclusion angles, and detector boundaries.

4.1.4.81 Star Occultation

A star may lie inside the instrument field of view but still be blocked by a celestial body.

For example, in low Earth orbit the Earth can occult a large portion of the celestial sphere.

Therefore,

\[ \boxed{ \text{Inside FOV} \neq \text{Visible Star}. } \]

A realistic star-selection algorithm should evaluate both

4.1.4.82 Probability of Detection

A high-fidelity star model need not treat star detection as perfectly deterministic.

Detection probability may depend on

\[ \boxed{ P_D = f( m, SNR, T_e, \omega, I_{background} ). } \]

A bright star under low-background conditions may have a detection probability close to unity, while a star near the limiting magnitude may be detected only intermittently.

4.1.4.83 False-Alarm Probability

Noise, hot pixels, and radiation events can generate false star candidates.

A detailed model may therefore include

\[ \boxed{ P_{FA} = P(\text{false detection}). } \]

The pattern-recognition algorithm must remain robust when the candidate set contains both real and false objects.

4.1.4.84 Dynamic Exposure Control

Some star trackers can adjust exposure according to the current operating condition.

Conceptually,

\[ \boxed{ T_e = f( \omega, I_{background}, \text{star brightness} ). } \]

At low spacecraft angular rates, a longer exposure can improve sensitivity.

At high angular rates, a shorter exposure reduces star streaking.

Thus adaptive exposure helps balance

\[ \boxed{ \text{Sensitivity} \leftrightarrow \text{Motion Blur}. } \]

4.1.4.85 Rolling-Shutter Effects

Some CMOS detectors use rolling-shutter readout. In this architecture, different detector rows correspond to slightly different observation times.

The effective observation time can be represented as

\[ \boxed{ t=t(v) } \]

where \(v\) is the detector row.

If the spacecraft rotates during readout, the star pattern can become geometrically distorted.

A global-shutter detector exposes the complete image more nearly at the same time and therefore avoids this particular row-dependent temporal distortion.

4.1.4.86 Optical Defocus and PSF Design

It might appear that the ideal star image should always occupy the smallest possible number of pixels. In practice, controlled spreading of the star image can be useful.

A PSF distributed over several pixels provides intensity information that supports sub-pixel centroiding.

However, excessive spreading can

The optical PSF is therefore part of the complete centroiding and accuracy design.

4.1.4.87 Calibration Parameters

A detailed star-tracker calibration model may contain

\[ \boxed{ \mathbf p_{cal} = [ f_x, f_y, c_x, c_y, k_1, k_2, \ldots, \delta\boldsymbol{\theta}_{align} ]^T. } \]

Calibration can estimate

The calibrated line-of-sight vector can be represented as

\[ \boxed{ \mathbf s_{cal} = f_{cal} ( u,v,\mathbf p_{cal} ). } \]

4.1.4.88 Reading a Star Tracker Datasheet

Star-tracker selection requires careful interpretation of the datasheet. The following parameters are particularly important.

Parameter GNC Significance
Attitude accuracy Accuracy of the absolute attitude measurement.
Cross-boresight accuracy Angular accuracy perpendicular to the optical axis.
Roll accuracy Accuracy for rotation around the optical boresight.
Noise-equivalent angle Short-term angular measurement noise.
Field of view Determines sky coverage and available star population.
Limiting magnitude Indicates the faintest usable star population.
Update rate Frequency of attitude solutions.
Acquisition time Time required to establish an initial attitude solution.
Reacquisition time Recovery time following loss of tracking.
Lost-in-space capability Ability to determine attitude without a prior attitude estimate.
Maximum tracking rate Maximum angular rate while maintaining tracking.
Maximum acquisition rate Maximum angular rate for successful acquisition.
Sun exclusion angle Minimum allowable angular separation from the Sun.
Earth exclusion angle Minimum allowable separation from the bright Earth or limb.
Moon exclusion angle Minimum allowable separation from the Moon.
Latency Delay between observation and availability of attitude data.
Timestamp accuracy Determines how accurately the attitude measurement is associated with time.
Number of tracked stars Affects robustness and attitude-solution geometry.
Output format Quaternion, DCM, star vectors, covariance, status, etc.
Electrical interface Compatibility with spacecraft avionics.
Power Spacecraft electrical power requirement.
Mass Spacecraft mass allocation.
Dimensions Mechanical accommodation requirement.
Operating temperature Thermal compatibility during normal operation.
Survival temperature Allowable non-operating temperature range.
Radiation tolerance Suitability for the expected mission environment.
Shock and vibration qualification Compatibility with launch loads.
Mission lifetime Compatibility with required spacecraft lifetime.
Space heritage Evidence of prior operational use in space.

4.1.4.89 Important Datasheet Distinctions

Accuracy vs Resolution

Fine detector resolution does not automatically imply equivalent absolute attitude accuracy.

Update Rate vs Measurement Bandwidth

A high numerical output rate does not necessarily mean that all high-frequency spacecraft motion is measured accurately.

Tracking Rate vs Acquisition Rate

The maximum rate for maintaining an existing star lock may differ significantly from the maximum rate at which a lost-in-space attitude solution can be acquired.

Latency vs Timestamp Accuracy

A known processing delay can often be compensated. An unknown timing error can directly become an attitude error during spacecraft rotation.

Sensor Accuracy vs Spacecraft Attitude Knowledge

The final spacecraft attitude knowledge also includes mounting, thermal, timing, structural, and estimation errors.

4.1.4.90 Star Tracker Selection for a Space Mission

Sensor selection should begin with the mission-level pointing and attitude-knowledge requirements.

A useful requirements flow is

\[ \boxed{\text{Mission / Payload Requirement}} \rightarrow \boxed{\text{Pointing Requirement}} \rightarrow \boxed{\text{Attitude Knowledge Requirement}} \rightarrow \boxed{\text{Sensor Requirement}} \rightarrow \boxed{\text{Star Tracker Selection}}. \]

Important selection questions include:

  1. What absolute attitude knowledge is required?
  2. What pointing stability is required?
  3. What is the spacecraft maximum angular rate?
  4. What is the expected slew rate?
  5. Is lost-in-space acquisition required?
  6. What acquisition time is acceptable?
  7. What update rate is required?
  8. How much latency can the estimator tolerate?
  9. What timestamp accuracy is required?
  10. What Sun exclusion angle can the mission accommodate?
  11. Will Earth frequently approach the FOV?
  12. Is Moon exclusion important?
  13. What star-tracker outage duration can be tolerated?
  14. Is one tracker sufficient?
  15. Is redundant or multi-head architecture required?
  16. What alignment stability is required?
  17. What thermal environment is expected?
  18. What radiation environment is expected?
  19. What mass and power are available?
  20. Is the communication interface compatible?
  21. What environmental qualification is required?
  22. What space heritage is available?

4.1.4.91 Example Star Tracker Error Model

Consider a simplified star tracker having small-angle standard deviations

\[ \sigma_x=10~\text{arcsec}, \]

\[ \sigma_y=10~\text{arcsec}, \]

and

\[ \sigma_z=40~\text{arcsec}. \]

Since

\[ 1^\circ=3600~\text{arcsec}, \]

the equivalent values in degrees are

\[ \sigma_x = \frac{10}{3600} \approx 0.00278^\circ, \]

\[ \sigma_y \approx 0.00278^\circ, \]

and

\[ \sigma_z = \frac{40}{3600} \approx 0.0111^\circ. \]

The simplified covariance is

\[ \boxed{ R_{ST} = \begin{bmatrix} \sigma_x^2&0&0\\ 0&\sigma_y^2&0\\ 0&0&\sigma_z^2 \end{bmatrix}. } \]

For implementation in attitude algorithms, the angular quantities should normally be converted into radians.

4.1.4.92 Monte Carlo Star Tracker Model

Monte Carlo analysis can be used to evaluate the sensitivity of the spacecraft GNC system to star-tracker uncertainties.

Parameters that can be randomized include

One realization can be represented schematically as

\[ \boxed{ q_{ST} = \delta q_{noise} \otimes \delta q_{align} \otimes q_{true}(t-\tau). } \]

A corresponding validity signal is

\[ \boxed{ valid_{ST}(t)\in\{0,1\}. } \]

This allows Monte Carlo analysis to evaluate not only measurement accuracy but also loss and recovery of absolute attitude measurements.

4.1.4.93 Failure Modes

Potential star-tracker failure modes include

A critical distinction is

\[ \boxed{\text{No Measurement}} \]

versus

\[ \boxed{\text{Incorrect Measurement}}. \]

An incorrect measurement can be particularly dangerous because it may appear valid unless additional consistency checks are performed.

4.1.4.94 Fault Detection, Isolation and Recovery

Quaternion Norm Check

A valid attitude quaternion should satisfy

\[ \boxed{ \left| \|\mathbf q_{ST}\|-1 \right| < \epsilon_q. } \]

Innovation Check

The difference from the gyro-propagated attitude should remain within an appropriate uncertainty-dependent bound.

Timestamp Check

Measurements with invalid, repeated, or stale timestamps can be rejected.

Tracker Status Check

Internal hardware status and validity flags should be interpreted by the flight software.

Star-Count Check

A solution based on insufficient or poor-quality stars may be rejected.

Cross-Sensor Check

The tracker solution can be compared with

4.1.4.95 Accuracy vs Availability

A star tracker may provide extremely accurate measurements when valid, yet be unavailable during some orbital or spacecraft attitude conditions.

Therefore, mission analysis must consider both

\[ \boxed{\text{Measurement Accuracy}} \]

and

\[ \boxed{\text{Measurement Availability}}. \]

Availability may be represented statistically by

\[ \boxed{ P(valid_{ST}). } \]

It depends on

4.1.4.96 Spacecraft Placement

Star-tracker placement is a spacecraft-level design problem. The sensor should have a sufficiently clear view of the celestial sphere while remaining mechanically and thermally stable.

Placement should consider

For precision missions, the tracker should ideally be mounted on a stable structure whose alignment relative to the precision payload is well characterized.

4.1.4.97 Optical Contamination

Because a star tracker is an optical sensor, contamination can directly affect its performance.

Possible contamination sources include

Contamination can reduce optical transmission or increase scattering, thereby reducing star-detection margin.

4.1.4.98 Operational Modes

A star tracker may operate through several internal modes:

\[ \boxed{\text{OFF}} \rightarrow \boxed{\text{BOOT}} \rightarrow \boxed{\text{ACQUISITION}} \rightarrow \boxed{\text{TRACKING}}. \]

Additional states may include

The GNC system should use the tracker mode and quality information rather than treating every received quaternion as equally reliable.

4.1.4.99 Initialization Sequence

A representative startup sequence is

\[ \boxed{\text{Power On}} \rightarrow \boxed{\text{Self Test}} \rightarrow \boxed{\text{Detector Initialization}} \rightarrow \boxed{\text{Image Acquisition}} \rightarrow \boxed{\text{Star Detection}} \rightarrow \boxed{\text{Star Identification}} \rightarrow \boxed{\text{Initial Attitude}} \rightarrow \boxed{\text{Tracking}}. \]

The spacecraft operations concept must account for the time required to reach a valid tracking solution.

4.1.4.100 Performance During Spacecraft Slews

During a large spacecraft slew, angular rate may become too high for normal star-tracker operation.

Possible strategies include

A common operational sequence is

\[ \boxed{\text{Fine Pointing}} \rightarrow \boxed{\text{High-Rate Slew}} \rightarrow \boxed{\text{Gyro Propagation}} \rightarrow \boxed{\text{Slew Completion}} \rightarrow \boxed{\text{Star Tracker Reacquisition}} \rightarrow \boxed{\text{Fine Pointing}}. \]

4.1.4.101 Star Tracker and Spacecraft Autonomy

An autonomous star tracker can determine spacecraft attitude without ground intervention.

This capability is particularly valuable following

The autonomy chain is

\[ \boxed{\text{Unknown Attitude}} \rightarrow \boxed{\text{Star Image}} \rightarrow \boxed{\text{Lost-in-Space Identification}} \rightarrow \boxed{\text{Absolute Attitude}} \rightarrow \boxed{\text{Autonomous GNC}}. \]

4.1.4.102 Comparison with Other Attitude Sensors

Sensor Primary Measurement Main Strength Main Limitation
Gyroscope Angular rate High-rate attitude propagation Bias causes long-term drift
Magnetometer Magnetic-field vector Simple absolute reference Magnetic disturbances and moderate/coarse accuracy
Sun Sensor Sun direction Simple celestial reference Sun availability and single-vector limitation
Star Tracker Absolute celestial attitude Very high absolute attitude accuracy Bright-object sensitivity, rate limits and greater complexity
Earth / Horizon Sensor Earth limb or nadir information Useful Earth reference Mission geometry and lower precision for many applications
Fine Guidance Sensor Precision line of sight Very high pointing precision Usually payload or mission specific

4.1.4.103 Star Sensor vs Star Tracker

The terms are sometimes used interchangeably, but a useful conceptual distinction can be made.

A star sensor may primarily detect stars or provide star directions.

A modern autonomous star tracker typically combines

\[ \boxed{\text{Optical Sensor}} + \boxed{\text{Image Processing}} + \boxed{\text{Star Catalogue}} + \boxed{\text{Star Identification}} + \boxed{\text{Attitude Determination}}. \]

It therefore provides a complete attitude solution rather than only raw optical measurements.

4.1.4.104 Complete Star Tracker Signal Flow

Celestial Star Catalogue
          |
          v
Reference Star Directions
          |
          +------------------------------------+
                                               |
                                               v
Spacecraft True Attitude --------------> Rotate Star Vectors
                                               |
                                               v
                                     Star Tracker Frame
                                               |
                                               v
                                      FOV / Occultation
                                               |
                                               v
                                        Optical System
                                               |
                                               v
                                      Focal Plane Image
                                               |
               +-------------------------------+-------------------+
               |                               |                   |
               v                               v                   v
       Optical Distortion                Detector Noise       Stray Light
               |                               |                   |
               +-------------------------------+-------------------+
                                               |
                                               v
                                         Digital Image
                                               |
                                               v
                                         Star Detection
                                               |
                                               v
                                           Centroiding
                                               |
                                               v
                                      LOS Reconstruction
                                               |
                                               v
                                      Star Identification
                                               |
                                               v
                                    Catalogue Correspondence
                                               |
                                               v
                                      Wahba / QUEST Solver
                                               |
                                               v
                                    Sensor-Frame Attitude
                                               |
                                               v
                                    Mounting Calibration
                                               |
                                               v
                                      Body-Frame Attitude
                                               |
                                               v
                              Timing / Quality / Validity Logic
                                               |
                                               v
                                     Star Tracker Output
                                               |
                                               v
                                           MEKF
                                               |
                                               v
                                Estimated Spacecraft Attitude
                                               |
                                               v
                                    Guidance and Control

4.1.4.105 General GNC Measurement Equation

The star tracker fits into the general spacecraft sensor model

\[ \boxed{ \mathbf z_k = h(\mathbf x_k) + \mathbf v_k. } \]

If the spacecraft navigation state is

\[ \mathbf x = \begin{bmatrix} \mathbf r\\ \mathbf v\\ \mathbf q\\ \boldsymbol{\omega}\\ \mathbf b_g\\ \vdots \end{bmatrix}, \]

the star-tracker measurement function depends primarily on spacecraft attitude:

\[ \boxed{ \mathbf z_{ST,k} = h_{ST}(\mathbf q_k) + \mathbf v_{ST,k}. } \]

For an error-state filter, the attitude innovation is normally expressed using a three-component small-angle error rather than treating the four quaternion elements as independent linear states.

4.1.4.106 Major Star Tracker Design Trade-Offs

Field of View vs Angular Resolution

For a fixed detector,

\[ \boxed{ \text{Wide FOV} \leftrightarrow \text{Fine Angular Sampling}. } \]

Exposure Time vs Motion Blur

\[ \boxed{ \text{Long Exposure} \rightarrow \text{More Photons} } \]

but

\[ \boxed{ \text{Long Exposure} \rightarrow \text{More Motion Blur}. } \]

Aperture vs Size and Mass

A larger aperture can collect more light but generally increases optical size and accommodation demands.

Accuracy vs Update Rate

Higher update rates can require shorter exposure and increased processing throughput.

Baffle Performance vs Accommodation

Greater bright-source rejection can require a larger or longer baffle.

Redundancy vs Resources

Additional trackers improve availability and fault tolerance but increase mass, power, cost, thermal load, and avionics interfaces.

4.1.4.107 What Matters Most for Spacecraft GNC?

If the objective is spacecraft GNC design rather than development of the star tracker itself, the most important star-tracker parameters are typically

  1. absolute attitude accuracy,
  2. cross-boresight measurement noise,
  3. roll measurement noise,
  4. mounting alignment error,
  5. thermoelastic alignment drift,
  6. update rate,
  7. latency,
  8. timestamp accuracy,
  9. maximum acquisition rate,
  10. maximum tracking rate,
  11. Sun exclusion angle,
  12. Earth exclusion angle,
  13. Moon exclusion angle,
  14. dropout behaviour,
  15. reacquisition time,
  16. measurement covariance,
  17. validity and quality flags.

These parameters can be modelled without generating a complete synthetic star image.

4.1.4.108 What Matters for Star Tracker Development?

If the objective is to design or validate the star tracker itself, a much more detailed model is required.

Important quantities include

4.1.4.109 Recommended GNC Star Tracker Model

For a spacecraft attitude simulation, a useful model is

\[ \boxed{ \mathbf q_{true} } \]

\[ \downarrow \]

\[ \boxed{\text{Sensor Mounting}} \]

\[ \downarrow \]

\[ \boxed{\text{Fixed Alignment Error}} \]

\[ \downarrow \]

\[ \boxed{\text{Thermal Alignment Error}} \]

\[ \downarrow \]

\[ \boxed{\text{Random Attitude Noise}} \]

\[ \downarrow \]

\[ \boxed{\text{Sampling / Update Rate}} \]

\[ \downarrow \]

\[ \boxed{\text{Latency / Timestamp}} \]

\[ \downarrow \]

\[ \boxed{\text{Sun / Earth / Moon Exclusion}} \]

\[ \downarrow \]

\[ \boxed{\text{Angular-Rate Validity}} \]

\[ \downarrow \]

\[ \boxed{\text{Dropout / Reacquisition}} \]

\[ \downarrow \]

\[ \boxed{ \mathbf q_{ST}, R_{ST}, valid_{ST}, t_{ST} } \]

\[ \downarrow \]

\[ \boxed{\text{MEKF / EKF}}. \]

This model provides enough fidelity to investigate most spacecraft attitude-estimation and control questions while remaining computationally manageable.

4.1.4.110 Example End-to-End Spacecraft Operation

Consider a spacecraft that has just completed a large slew. During the manoeuvre, the star tracker may temporarily be unavailable because the angular rate is too high.

The estimator therefore propagates attitude using

\[ \boldsymbol{\omega}_m-\hat{\mathbf b}_g. \]

After the spacecraft rate decreases, the tracker begins acquisition. It captures a star-field image, detects candidate stars, computes their centroids, converts those centroids into line-of-sight vectors, identifies the stars against its catalogue, and solves Wahba's problem.

It then outputs

\[ \mathbf q_{ST}. \]

The navigation filter forms an attitude residual between \(\mathbf q_{ST}\) and the gyro-propagated attitude.

If the measurement passes validity and innovation checks, the filter updates the attitude and gyro-bias estimates.

The corrected attitude estimate is then supplied to the spacecraft controller.

The complete sequence is

\[ \boxed{\text{Slew}} \rightarrow \boxed{\text{Gyro Propagation}} \rightarrow \boxed{\text{Star Tracker Acquisition}} \rightarrow \boxed{\text{Absolute Attitude}} \rightarrow \boxed{\text{MEKF Correction}} \rightarrow \boxed{\text{Fine Pointing}}. \]

4.1.4.111 Key Equations

Star Vector Transformation

\[ \boxed{ \mathbf s_i^S = C_I^S\mathbf s_i^I } \]

Catalogue Vector from Right Ascension and Declination

\[ \boxed{ \mathbf s_i^I = \begin{bmatrix} \cos\delta_i\cos\alpha_i\\ \cos\delta_i\sin\alpha_i\\ \sin\delta_i \end{bmatrix} } \]

Pinhole Projection

\[ \boxed{ x_f=f\frac{s_x}{s_z}, \qquad y_f=f\frac{s_y}{s_z} } \]

Centroid

\[ \boxed{ u_c = \frac{\sum I_i u_i}{\sum I_i}, \qquad v_c = \frac{\sum I_i v_i}{\sum I_i} } \]

Inter-Star Angle

\[ \boxed{ \theta_{ij} = \cos^{-1} \left( \mathbf s_i^T\mathbf s_j \right) } \]

Wahba's Problem

\[ \boxed{ J(C) = \frac12 \sum_i a_i \| \mathbf b_i-C\mathbf r_i \|^2 } \]

Small-Angle Measurement Error

\[ \boxed{ \delta\mathbf q \approx \begin{bmatrix} 1\\ \frac12\delta\boldsymbol{\theta} \end{bmatrix} } \]

Timing Error

\[ \boxed{ \delta\boldsymbol{\theta}_{time} \approx \boldsymbol{\omega}\delta t } \]

Motion Blur

\[ \boxed{ \Delta\theta_{blur} \approx \omega T_e } \]

Centroid-to-Angular Error

\[ \boxed{ \sigma_\theta \approx \frac{p}{f}\sigma_c } \]

Measurement Covariance

\[ \boxed{ \delta\boldsymbol{\theta}_{ST} \sim \mathcal N(0,R_{ST}) } \]

4.1.4.112 Summary

A star tracker is a precision celestial attitude sensor that determines spacecraft orientation by observing stars and comparing their measured directions with known directions from an onboard star catalogue.

At the most fundamental level,

\[ \boxed{ \mathbf s_i^S = C_I^S \mathbf s_i^I. } \]

The tracker determines the unknown attitude \(C_I^S\) from several such vector correspondences.

Internally, the complete process is

\[ \boxed{\text{Starlight}} \rightarrow \boxed{\text{Optics}} \rightarrow \boxed{\text{Detector}} \rightarrow \boxed{\text{Star Detection}} \rightarrow \boxed{\text{Centroiding}} \rightarrow \boxed{\text{Star Identification}} \rightarrow \boxed{\text{Wahba / QUEST}} \rightarrow \boxed{\text{Attitude}}. \]

The principal error mechanisms include

\[ \boxed{ \begin{aligned} &\text{Centroid Noise}\\ &+\text{Optical Calibration Error}\\ &+\text{Mounting / Boresight Error}\\ &+\text{Thermoelastic Alignment Error}\\ &+\text{Timing Error}\\ &+\text{Motion Blur}\\ &+\text{Detector Noise}\\ &+\text{Catalogue / Processing Error}\\ &+\text{Stray-Light Effects}. \end{aligned} } \]

For spacecraft GNC, the star tracker is normally combined with a gyroscope:

\[ \boxed{\text{Gyroscope}} + \boxed{\text{Star Tracker}} \rightarrow \boxed{\text{MEKF}} \rightarrow \boxed{\text{Accurate Attitude + Gyro Bias Estimate}}. \]

The gyro provides high-rate propagation while the star tracker provides the absolute celestial reference needed to prevent long-term attitude drift.

For system-level simulation, it is also essential to model availability, not only accuracy. Sun/Earth/Moon exclusion, high spacecraft rates, dropout, latency, timestamp error, and reacquisition behaviour can all affect the GNC system even when the nominal star-tracker accuracy is extremely high.

Ultimately, spacecraft attitude knowledge depends on the complete chain:

\[ \boxed{ \text{Celestial Reference} \rightarrow \text{Optics} \rightarrow \text{Image Processing} \rightarrow \text{Attitude Solution} \rightarrow \text{Mounting Calibration} \rightarrow \text{Timing} \rightarrow \text{Sensor Fusion} \rightarrow \text{Spacecraft Attitude Knowledge}. } \]