Chapter 10: On-Board Data Handling¶
Key Concepts in This Chapter
-
OBDH functions: data acquisition, command distribution, housekeeping telemetry, and payload data handling
-
Radiation-hardened and radiation-tolerant processor architectures for space
-
Memory technologies and error correction: EDAC, scrubbing, and Triple Modular Redundancy
-
Space data buses: MIL-STD-1553B, SpaceWire, and high-speed interfaces
-
CCSDS standards governing spacecraft data communications
-
Flight software design principles, RTOS selection, and autonomy/FDIR
-
On-board AI processing for next-generation payload autonomy
10.1 Introduction and System Architecture Overview¶
The On-Board Data Handling (OBDH) subsystem is the electronic nervous system of a spacecraft. It acquires measurement data from sensors distributed throughout the satellite, decodes and distributes commands received from the ground, maintains the spacecraft’s health through telemetry generation, manages on-board time, stores payload and housekeeping data, and executes the autonomous on-board software that controls every other subsystem (Wertz et al. 2011).
The OBDH architecture of a modern spacecraft can be understood as a layered hierarchy:
-
Sensor/actuator layer: raw measurements from temperature sensors, current monitors, position encoders, and payload detectors; actuator interfaces to reaction wheels, heaters, and pyrotechnics.
-
Remote Terminal Unit (RTU) layer: data concentrators that aggregate analogue and discrete measurements from the sensor layer and communicate digitally with the central computer via a data bus.
-
On-Board Computer (OBC) layer: central processing unit that executes flight software, performs command decoding, schedules activities, and manages data flow.
-
Mass memory layer: solid-state recorder or mass memory unit that stores payload science data pending downlink.
-
Data bus layer: the communication backbone connecting all elements.
10.2 On-Board Computer Architectures¶
10.2.1 Radiation Environment and Its Effects on Electronics¶
The space radiation environment consists of galactic cosmic rays (GCR), solar energetic particles (SEP), and trapped particles in the Van Allen belts. These produce several classes of radiation-induced failures in microelectronics:
- Single Event Upset (SEU): a heavy ion or proton deposits sufficient charge along its track to flip a memory bit from 0 to 1 or vice versa (a “soft error”). The cross-section for SEU depends on the linear energy transfer (LET) of the particle and the critical charge \(Q_c\) of the cell. The SEU rate is:
where \(\sigma(LET)\) is the cell cross-section as a function of LET and \(\Phi(LET)\) is the differential particle fluence spectrum.
-
Single Event Latchup (SEL): an ion triggers a parasitic thyristor structure in CMOS logic, creating a low-impedance path from supply to ground. SEL is destructive unless power is cycled within milliseconds.
-
Total Ionising Dose (TID): cumulative ionisation in gate oxides shifts threshold voltages and degrades transistor gain. Expressed in Gray or rad(Si), typical TID limits are \(20\)–\(300\,\mathrm{krad(Si)}\) for radiation-hardened devices.
-
Displacement Damage: non-ionising energy loss (NIEL) from protons and neutrons creates crystal lattice defects, degrading minority carrier lifetime in bipolar transistors and CCDs.
10.2.2 Radiation-Hardened Processor Options¶
| Processor | Architecture | Frequency | TID | Notes |
| (MHz) | (krad) | |||
| LEON3-FT (GR712RC) | SPARC V8, dual-core | up to 100 | 300 | European rad-hard standard |
| LEON4-FT (GR740) | SPARC V8, quad-core | up to 250 | 300 | Next-generation European |
| RAD750 | PowerPC 750 | 110–200 | 1000 | Mars and lunar missions |
| Microchip SAMRH71 | ARM Cortex-M7 | up to 100 | 100 | Rad-hard microcontroller |
The LEON family of SPARC-V8 processors was developed by ESA/ESTEC and is now maintained by Frontgrade Gaisler (formerly Cobham Gaisler). The GRLIB IP library provides a complete SoC platform including the LEON processor core, memory controllers, UART, SpaceWire controller, MIL-STD-1553 interface, and PCI. The LEON3-FT implements hardware fault-tolerance through:
-
Register file with parity and retry
-
Cache memory protected with EDAC
-
On-chip scrubbing controller
-
Processor lock-step with comparison
The RAD750 from BAE Systems is derived from the IBM PowerPC 750 and is manufactured using a radiation-hardened process, with a total ionizing dose tolerance of the order of \(1\,\mathrm{Mrad(Si)}\). It has accumulated an exceptional heritage record: Mars Reconnaissance Orbiter, Curiosity rover, Mars 2020 Perseverance, Lunar Reconnaissance Orbiter, and many others.
10.2.3 FPGAs in Space¶
Field-Programmable Gate Arrays (FPGAs) are used extensively in space for:
-
High-speed data routing and signal processing (where CPUs are too slow)
-
Custom peripheral controllers (SpaceWire routers, codec implementations)
-
Reconfigurable payload processing pipelines
The Xilinx Virtex-4QV and Virtex-5QV are radiation-hardened versions using configuration SRAM cells designed to be SEU-immune. The ProASIC3 from Microsemi uses a flash-based configuration memory that is inherently immune to SEU (since flash cells require erase voltage to change state), making it “zero configuration scrubbing” required.
For commercial/NewSpace missions with lower radiation requirements (LEO, \(<2\) years), Commercial Off-The-Shelf (COTS) FPGAs (Xilinx Zynq, Intel/Altera Cyclone/Arria) are used with Triple Modular Redundancy (TMR) implemented in the FPGA fabric. In TMR, the logic is triplicated and a majority voter selects the output, tolerating a single SEU in any one copy:
where \(F_{\mathrm{single}}\) is the probability of failure for a single module over the mission period.
10.3 Memory Technologies and Error Correction¶
10.3.1 Error Detection and Correction (EDAC)¶
SRAM memories in space are susceptible to SEU-induced bit flips. The standard mitigation is Error Detection and Correction (EDAC) based on a Hamming code with Single Error Correction, Double Error Detection (SECDED). For a data word of \(k\) bits, the minimum number of parity check bits \(p\) satisfying the SECDED condition is:
For \(k = 32\): \(p = 6\) (total codeword: 38 bits). For \(k = 64\): \(p = 7\) (total codeword: 71 bits). The parity check matrix \(\mathbf{H}\) defines the syndrome \(\mathbf{s}\):
where \(\mathbf{r}\) is the received codeword. A non-zero syndrome indicates an error; the syndrome value identifies the position of a single-bit error, which is then corrected by flipping that bit.
10.3.2 Memory Scrubbing¶
Even with EDAC, over time multiple SEUs may accumulate in the same word, eventually overwhelming the single-error correction capability. Memory scrubbing is the periodic re-reading and re-writing of memory to correct accumulated errors before they compound. The scrubbing rate must be set such that the probability of a second SEU in the same word between scrub passes remains negligible:
where \(\lambda_{\mathrm{SEU}}\) is the per-word SEU rate and \(\tau_{\mathrm{scrub}}\) is the scrub interval. Typical scrub rates are \(0.1\)–\(1\,\mathrm{Hz}\) for program memory, \(0.01\)–\(0.1\,\mathrm{Hz}\) for mass memory.
10.3.3 Non-Volatile Memory¶
Non-volatile storage is required for:
-
Boot software (typically PROM or radiation-hardened EEPROM)
-
Non-volatile parameters (orbit ephemeris, attitude reference, mission timeline)
-
Payload science data (mass memory)
EEPROM (electrically erasable PROM) stores \(1\)–\(8\,\mathrm{Mbit}\) of configuration data with \(10^5\) erase/write cycles and \(>10\) year data retention. NAND flash provides \(>1\,\mathrm{Tbit/chip}\) density but requires wear-levelling algorithms and ECC to manage finite program/erase cycles (\(10^4\)–\(10^5\)) and ECC to handle multi-bit errors characteristic of NAND flash.
10.4 Spacecraft Data Buses¶
10.4.1 MIL-STD-1553B¶
MIL-STD-1553B is the dominant spacecraft data bus for command and housekeeping data. The bus architecture is:
-
Bus Controller (BC): one BC manages all transfers on the bus, polling Remote Terminals in turn.
-
Remote Terminals (RT): up to 31 RTs (each addressed 0–30), each representing a subsystem (ADCS, power, propulsion, etc.).
-
Bus Monitor (BM): passive listener for bus traffic logging.
-
Dual redundant bus: two physically independent twisted-shielded pairs (Bus A and Bus B); the BC switches to Bus B on failure detection.
Data encoding uses Manchester biphase-level encoding, providing inherent DC balance and clock recovery without a separate clock signal. Each 1553 word is 20 bits: 3 sync bits + 16 data bits + 1 parity bit. The bus operates at a fixed \(1\,\mathrm{Mbps}\) data rate over twisted shielded cable with a characteristic impedance of about \(70\)–\(85\,\Omega\).
A typical BC-to-RT command transfer consists of:
-
BC transmits Command Word: RT address (5 bits) + T/R bit + subaddress (5 bits) + word count (5 bits)
-
RT transmits Status Word after a response time of \(4\)–\(12\,\mu\mathrm{s}\)
-
RT transmits \(1\)–\(32\) data words (if command was “Transmit”)
The total bus cycle time for 16 RTs each exchanging 32 words is \(\sim 5\,\mathrm{ms}\), giving an effective bus utilisation of \(200\,\mathrm{Hz}\) polling rate per RT. This is more than adequate for housekeeping data but insufficient for high-speed payload data.
10.4.2 SpaceWire¶
SpaceWire (ECSS-E-ST-50-12C, based on IEEE Std 1355) is a high-speed serial data bus designed for spacecraft payload data. Key characteristics:
-
Serial, full-duplex, point-to-point links (or via routers)
-
Low-voltage differential signalling (LVDS): \(\pm 200\,\mathrm{mV}\) differential, robust to EMC
-
Data rate: \(2\)–\(400\,\mathrm{Mbps}\) (auto-negotiated per link)
-
Packet-switched network: SpaceWire routers switch packets between ports
-
Character-level encoding: Data-Strobe (DS) encoding provides clock recovery
-
End-to-end error control via Error Control Codes (ECC) at the packet level
The SpaceWire character encoding transmits data and strobe signals on two LVDS pairs. The strobe signal \(S\) is defined as:
where \(D_n\) is the current data bit. The receiver recovers the clock by XOR-ing data and strobe. This eliminates the need for a separate clock signal and provides inherent tolerance to skew between the signal pairs. SpaceWire networks are used on many ESA and NASA missions (for example Gaia, BepiColombo and the James Webb Space Telescope).
10.4.3 CAN Bus and Low-Speed Interfaces¶
The Controller Area Network (CAN) bus, operating at up to \(1\,\mathrm{Mbps}\) with inherent multi-master arbitration and error detection, is increasingly used in small satellites (\(<100\,\mathrm{kg}\)) where the lower mass and cost compared to 1553 are attractive. I2C (up to \(3.4\,\mathrm{Mbps}\)) and SPI (up to \(25\,\mathrm{Mbps}\)) are used for board-level sensor interfaces (temperature sensors, ADC/DAC converters, gyroscopes on the same PCB as the OBC).
10.5 CCSDS Standards for Space Data Systems¶
The Consultative Committee for Space Data Systems (CCSDS) develops internationally agreed standards for space data communications, adopted by ESA, NASA, JAXA, ISRO, CNSA, and all major space agencies. The key standards form a layered protocol architecture:
| CCSDS Standard | Short Name | Function |
|---|---|---|
| CCSDS 132.0-B-3 | TM Space Data Link | Telemetry framing: Virtual Channels, FECF, Frame Header |
| CCSDS 232.0-B-4 | TC Space Data Link | Telecommand framing: CLTU, Frame Error Control |
| CCSDS 131.0-B-3 | TM Synchronization | Frame sync, pseudo-randomisation |
| CCSDS 401.0-B-31 | Radio Frequency/Modulation | TM/TC modulation standards |
| CCSDS 133.0-B-2 | Space Packet | Packet structure: Primary Header + User Data Field |
| CCSDS 727.0-B-5 | CFDP | File delivery: reliable transfer, segmentation, error recovery |
| CCSDS 122.0-B-2 | Image Data Compression | Lossless image compression for EO data |
| CCSDS 123.0-B-2 | Hyperspectral Compression | Near-lossless hyperspectral data compression |
| CCSDS 350.0-G-3 | Security Overview | CCSDS Space Data Link Security (SDLS) framework |
10.5.1 Space Packet Structure¶
The CCSDS Space Packet (CCSDS 133.0-B-2) is the fundamental data unit used in spacecraft telemetry. The packet structure is:
| Field | Content | Bits | Value/Range |
|---|---|---|---|
| Version Number | Always 000 | 3 | 0 |
| Packet Type | 0 = TM, 1 = TC | 1 | 0 or 1 |
| Secondary Header Flag | 1 if secondary header present | 1 | 0 or 1 |
| APID (Application ID) | Source/destination identifier | 11 | 0–2047 |
| Sequence Flags | 11 = standalone packet | 2 | 01, 10, 11 |
| Packet Sequence Count | Rolling counter per APID | 14 | 0–16383 |
| Packet Data Length | Total packet length \(-7\) (octets) | 16 | 0–65535 |
| Secondary Header (optional) | variable | APID-specific | |
| User Data Field | variable | Science / HK data | |
Packets are multiplexed into Transfer Frames by the TM Space Data Link layer. Each Transfer Frame carries a fixed-length payload of \(\leq 65536\) octets, with a mandatory 6-octet frame primary header (spacecraft ID, virtual channel ID, frame counter) and optional Frame Error Control Field (FECF, CRC-16-CCITT). Multiple Virtual Channels (VCs) within one physical link allow multiplexing of science, housekeeping, and playback data streams.
10.5.2 CCSDS File Delivery Protocol (CFDP)¶
CFDP (CCSDS 727.0-B-5) provides reliable file transfer over lossy space links. Two classes of service exist:
-
Class 1 (Unreliable): no acknowledgement, used for low-priority downlink
-
Class 2 (Reliable): positive/negative acknowledgement, automatic retransmission of lost segments, used for critical command files and software uploads
CFDP handles segmentation of large files into Protocol Data Units (PDUs) that fit within the available Transfer Frame data field, managing reassembly autonomously at the receiving end. It also provides file checksum verification and directory listing services.
10.6 Flight Software Design¶
10.6.1 Real-Time Operating System Selection¶
The flight software executes on a Real-Time Operating System (RTOS) that provides:
-
Deterministic task scheduling with bounded worst-case execution time (WCET)
-
Priority-based preemption (higher-priority tasks immediately preempt lower)
-
Semaphores, mutexes, and message queues for inter-task communication
-
Memory protection (optionally via MMU)
-
POSIX-compatible API for portability
| RTOS | Licence | Heritage | Notes |
|---|---|---|---|
| VxWorks (Wind River) | Commercial | ISS, Hubble, MRO, Curiosity, many GEO | DO-178C certified, POSIX |
| RTEMS | Open source (BSD) | ESA: Rosetta, Mars Express, BepiColombo | Free, POSIX, active community |
| LynxOS-178 | Commercial | Military satellites | DO-178C Level A |
| FreeRTOS | MIT open source | CubeSats, university missions | Lightweight, large ecosystem |
| INTEGRITY (GHS) | Commercial | Various defence missions | High-assurance, POSIX |
| Nucleus RTOS | Commercial | Small satellites | Low footprint |
10.6.2 Scheduling Theory¶
Flight software tasks have hard real-time deadlines. Rate Monotonic Scheduling (RMS) assigns fixed priorities based on task period: shorter period = higher priority. The schedulability condition for \(n\) tasks is:
where \(C_i\) is the worst-case execution time (WCET) and \(T_i\) is the period of task \(i\). As \(n \to \infty\), the bound converges to \(\ln 2 \approx 0.693\). Earliest Deadline First (EDF) is an optimal dynamic scheduler that achieves full processor utilisation (\(U \leq 1\)) but has more complex implementation and less predictable behaviour under overload.
10.6.3 Software Development Standards¶
Space flight software is developed to rigorous standards:
-
ECSS-E-ST-40C : ESA software engineering standard, covering software life cycle, planning, requirements, design, testing, and documentation.
-
NASA NPR 7150.2: NASA software engineering requirements, defining software classes (Class A through H) by criticality.
-
MISRA-C:2012: a set of coding guidelines (rules and directives) for ISO C that restrict undefined behaviour and common error patterns. Applied to safety-critical software.
-
DO-178C: aviation software certification standard increasingly applied by analogy to spacecraft software for highest assurance levels.
Software complexity is managed through the Architecture Framework and the concept of software partitioning: functionally independent modules (FDIR monitor, attitude control law, command decoder) execute in separate protected address spaces, preventing a bug in one from corrupting another.
10.7 Autonomy and Failure Detection, Isolation, and Recovery¶
10.7.1 FDIR Hierarchy¶
FDIR is structured in levels of increasing severity:
-
Level 0 – Local FDIR: hardware-level protection (current limiters, watchdog timers, redundancy switching). Response time: \(\mu\)s–ms.
-
Level 1 – Unit FDIR: unit-level self-tests and resets. Response time: ms–s.
-
Level 2 – Mode FDIR: OBC-executed function checks; transitions between nominal and degraded operational modes. Response time: 1–60 s.
-
Level 3 – Safe Mode FDIR: autonomous entry into Safe Mode triggered by critical anomalies. Response time: \(<30\,\mathrm{s}\).
Safe Mode entry is triggered when any of the following conditions are detected:
-
Attitude error exceeding threshold (e.g., \(>10^{\circ}\) from Sun-pointing): risk of solar array depowering or thermal damage
-
Battery state of charge (SoC) below critical threshold (e.g., \(<30\%\)): risk of deep discharge and permanent damage
-
Critical bus voltage out of range
-
Communication loss for more than a specified number of orbits
-
OBC hardware watchdog timer expiry
In Safe Mode, the spacecraft achieves a minimum viable configuration: solar arrays oriented toward the Sun via magnetorquers or Sun-search manoeuvre, non-essential loads powered off, beacon telemetry transmitted on omni-directional antenna at a slow, robust downlink rate (\(\leq 2\,\mathrm{kbps}\)), and the ground notified to re-establish contact.
10.7.2 Watchdog Timer¶
The watchdog timer is a hardware counter that counts down from a preset value (\(\sim 1\)–\(30\,\mathrm{s}\)). If the flight software fails to “kick” the watchdog within the timeout period (indicating software has hung or entered an infinite loop), the timer reaches zero and asserts a hardware reset signal to the OBC. The reset vector then reloads the boot software from non-volatile memory and restarts the system.
Proper watchdog design requires that the “kick” be performed only after completing a health check of critical software tasks (not simply in an interrupt service routine), to ensure the kick reflects genuine software liveness.
10.8 Payload Data Handling and Mass Memory¶
10.8.1 Mass Memory Sizing¶
The on-board mass memory must store payload data accumulated during the orbit arc where no ground contact is available, until the next downlink pass. The memory requirement is:
where \(R_{\mathrm{payload}}\) is the payload data rate (bits/s), \(t_{\mathrm{store}}\) is the maximum storage interval between downlinks, and \(C_{\mathrm{margin}} \geq 1.25\) is a design margin factor. For an illustrative multi-spectral imager generating \(450\,\mathrm{Gbits/orbit}\) with 2 orbits between downlinks:
Typical modern EO satellite mass memories range from 256 GB to 2 TB, implemented with radiation-tolerant NAND flash arrays in a Reed-Solomon or LDPC ECC-protected configuration.
10.8.2 On-Board Data Compression¶
On-board compression reduces downlink volume and thus ground station contact time requirements:
-
CCSDS 122.0-B-2 Lossless: based on a 2D discrete wavelet transform with adaptive entropy coding, achieving compression ratios of \(1.5\)–\(3\times\) for typical Earth imagery with no information loss. Suitable for scientific data where radiometric accuracy is essential.
-
JPEG-2000 Lossy: wavelet-based with quantisation; compression ratios of \(5\)–\(20\times\) at acceptable image quality degradation. Used for browse products and non-scientific commercial imagery.
-
CCSDS 123.0-B-2: near-lossless hyperspectral data compression, exploiting spatial and spectral correlation, achieving \(3\)–\(10\times\) compression.
10.9 On-Board AI Processing¶
The increasing computational demands of autonomous on-board processing—object detection, change detection, cloud screening, anomaly detection—have driven the development of dedicated AI inference accelerators for space:
| Processor | Architecture | Performance | Power | Status |
| (TOPS, approx.) | (W, approx.) | |||
| Intel Movidius Myriad 2/X | SHAVE VPU | 1 | 1–2 | COTS, flown (e.g. PhiSat-1) |
| NVIDIA Jetson Nano | 128-core Maxwell GPU | 0.5 (TFLOPS) | 5–10 | COTS (limited radiation tolerance) |
| AMD/Xilinx Versal AI | ARM + AI Engine | tens | 10–30+ | COTS, with mitigation |
The Intel Movidius Myriad family has flown on small Earth observation satellites: the ESA PhiSat-1 CubeSat (launched 2020) used a Myriad 2 to run a convolutional neural network (CNN) for on-board cloud screening, so that cloud-contaminated scenes need not be downlinked (Giuffrida et al. 2020).
10.10 OBDH System Budgets and Typical Parameters¶
| Spacecraft Class | OBC Mass | OBC Power | Memory | Example |
| (kg) | (W) | (GB) | ||
| 3U CubeSat | 0.05–0.1 | 0.5–2 | 4–16 | ISIS IOBC, GomSpace NanoCom |
| 6U–12U CubeSat | 0.1–0.3 | 1–5 | 16–128 | SpaceCube Mini, Gomspace |
| Small satellite (50–200 kg) | 0.5–2 | 3–10 | 64–512 | Astrobus S, SSTL 100 |
| Medium satellite (200 kg–1 t) | 2–5 | 10–25 | 256–2000 | Copernicus-class EO |
| Large GEO (>3 t) | 5–15 | 20–50 | 32–256 | Eurostar E3000, Boeing 702 |
| Deep space (planetary) | 3–8 | 10–30 | 8–64 | Curiosity (RAD750) |
10.11 Design Example: Illustrative SAR Satellite OBDH Architecture¶
The following architecture is illustrative (it is not the design of a specific mission) and shows how the elements of this chapter combine in a medium-sized radar satellite. It uses a dual-string design for cold redundancy:
-
OBC: Dual radiation-tolerant processors (for example LEON3-FT class), with EDAC-protected SRAM and non-volatile memory for boot software.
-
Data handling backbone: A SpaceWire network connecting the OBC, the instrument control unit, the mass memory unit, the TT&C transponder interface and the ADCS computer.
-
Mass Memory Unit: A solid-state NAND flash recorder sized for several orbits of radar data between downlink passes.
-
Housekeeping bus: MIL-STD-1553B connecting the remote terminals for platform housekeeping.
-
Flight software: A real-time operating system (for example RTEMS) with software developed to ECSS-E-ST-40C, including FDIR, ADCS control laws and CCSDS protocol stacks.
The SAR instrument generates raw data at rates of the order of gigabits per second during imaging windows. This data is written to the mass memory in real time by a dedicated data-processing unit, independent of the main OBC, which manages the SpaceWire data routing.
10.12 Summary¶
The OBDH subsystem integrates the computational intelligence of the spacecraft, bridging sensor data acquisition, command management, autonomous operations, and payload data handling. The radiation environment imposes fundamental constraints on processor selection, memory protection, and software design that distinguish space computing from terrestrial embedded systems.
The convergence of high-performance COTS processors, TMR-protected FPGAs, and AI inference accelerators is transforming on-board data handling capabilities, enabling spacecraft to perform increasingly sophisticated on-board science analysis and autonomous decision-making that reduces dependence on continuous ground contact. This trend toward edge computing in orbit is a defining technological shift of the 2020s in spacecraft engineering.
Further Reading¶
-
Larson, W. J., Wertz, J. R. (Eds.): Space Mission Analysis and Design, 3rd edn., Microcosm Press, Torrance (1999).
-
ESA: ECSS-E-ST-40C: Software Engineering Standard, ECSS Secretariat, Noordwijk (2009).
-
CCSDS: Space Packet Protocol, CCSDS 133.0-B-2, Washington DC (2020).
-
Giuffrida, G., et al.: “CloudScout: A Deep Neural Network for On-Board Cloud Detection on Hyperspectral Images”, Remote Sensing, 12(14), 2205 (2020).
-
Nicolaidis, M. (Ed.): Soft Errors in Modern Electronic Systems, Springer (2011).