OrbitDSP
Signal Processing + Fault Injection demo on Raspberry Pi, built with NASA JPL F´
Make signal processing “mission-real” inside Flight Software.
OrbitDSP is a compact F´ component that turns noisy / imperfect measurements into robust, testable behavior. It supports a controllable noise model, multiple filter families, and fault injection — all driven through commands, visible via telemetry and events, and mirrored on hardware status (LED).
Why this matters
In real spacecraft and avionics, sensor data is rarely clean: random noise, vibration, EMI spikes, saturation, dropouts, and stuck-at faults. If we feed raw signals directly into decision logic, we risk false alarms, missed detections, or unstable control actions.
- Parameterized noise model (vibration + random + spikes) for repeatable testing
- Filter choices that map to common real-time DSP patterns (EMA, Median, 1st-order LPF)
- Fault injection + expiry for demoing FDIR-style thinking
- Mission-ops style observability: telemetry + events + deterministic scheduler loop
Video (2 min)
Tip: embed the video inside your PowerPoint so the 5-minute talk stays frictionless.
Command / Telemetry loop (demo-ready)
Example sequence that highlights start marker, robust filtering under noise, fault injection, and burn/fuel monitoring.
Filters (quick intuition)
- EMA: lightweight smoothing for white noise (fast, but sensitive to spikes)
- Median: robust outlier rejection (best for spikes/glitches)
- 1st-order LPF: physics-friendly low-pass (removes high-frequency noise with predictable lag)
Noise model
- vib_amp / vib_hz: sinusoidal vibration component
- rand_sigma: Gaussian-like random sensor noise
- spike_rate: stochastic spike process (approx. spikes/sec)