In June 1998, a gyro calibration on SOHO left one gyroscope disabled, another stuck in high-gain mode, and a $1B+ spacecraft tumbling uncontrolled. The investigation board traced the failure to three compounding operator errors on a compressed timeline with no review board sign-off. That incident appears in satellite engineer interviews to this day because how you walk through the failure chain tells the panel how you think about systems, procedures, and human factors.
This guide organizes by the subsystem you’ll own — ADCS/GNC, EPS, TT&C, CDH, and Thermal — and separates the NewSpace constellation-ops loop from the legacy GEO/MEO mission-assurance loop. Subsystem specialists: jump via the TOC. Generalists: read top-to-bottom, then work through the prep matrix.
In this article, we’ll cover the following 28 questions:
- Walk me through how you’d derive a pointing accuracy budget for a 500-kg LEO Earth-observation satellite.
- When would you pick a star tracker over a sun sensor + magnetometer pair for attitude determination?
- Explain how a Kalman filter combines gyro and star tracker measurements, and what happens during a star tracker outage.
- What lessons does the SOHO 1998 attitude loss carry for modern ADCS design and operations?
- Which ADCS simulation tools have you used, and what are the tradeoffs between STK, Cosmos, and OREKIT for control loop validation?
- How do you size a solar array for a LEO sun-synchronous mission with 35% eclipse fraction?
- Why did the industry move from NiCd to NiH₂ and then to Li-ion for satellite batteries? When is NiH₂ still preferred?
- Walk through a peak-power-tracker (PPT) vs direct-energy-transfer (DET) decision. When is each appropriate?
- If you were specifying a 6U CubeSat EPS, what tradeoffs would you make on bus voltage, regulation topology, and load shedding?
- What are the most common EPS failure modes on LEO satellites and how do you design against them?
- Walk me through a complete link budget for a S-band TT&C downlink from a 500-km LEO orbit.
- Compare S-band, X-band, and Ka-band for satellite communications. When would you pick each?
- How would you design a ground station network for a 100-satellite LEO constellation with global coverage requirements?
- Why is BPSK still common on TT&C links when higher-order modulations exist? When would you move to QPSK or 8PSK?
- What’s the difference between command authentication and command encryption, and which would you require for a commercial Earth-observation satellite?
- Explain the difference between rad-hard and rad-tolerant components. When can you accept rad-tolerant for a LEO mission?
- Walk through the tradeoffs between SRAM, SDRAM, MRAM, and Flash for satellite flight software memory.
- Centralized vs distributed avionics architecture: when would you pick each on a modern satellite design?
- Describe how you’d design a fault detection, isolation, and recovery (FDIR) tree for a single-string CubeSat vs a redundant prime-built satellite.
- What’s your experience with NASA cFE / cFS or other open-source flight software frameworks, and when would you build custom?
- When do you transition from passive thermal control (MLI + radiator sizing) to active control (loop heat pipes, fluid loops)?
- How do you size a radiator for a payload dissipating 100 W in a GEO orbit with seasonal beta angle variation?
- What’s the difference between a 10-layer and a 30-layer MLI blanket, and when does adding layers stop helping?
- Walk me through the thermal challenge of a satellite entering and exiting eclipse, and how you’d design margin for the worst-case cold case.
- Which thermal analysis tools have you used (Thermal Desktop, SINDA, SystemaTHERMICA), and what are their tradeoffs?
- Walk me through a design review you led or contributed to (PDR / CDR / TRR). What was the major issue surfaced and how was it resolved?
- Describe a satellite anomaly investigation you participated in. What was the root cause and how was it confirmed?
- Walk me through a trade study you led where the obvious answer turned out to be wrong.
What satellite engineer interviews actually test
Satellite engineering interviews test whether you understand the orbital environment as a design constraint — radiation, thermal cycling, atomic oxygen — and whether you can close budgets at subsystem boundaries and reason about failure propagation. Generic systems engineering is necessary but not sufficient; panels expect domain depth in the subsystem you’re interviewing into.
Two distinct hiring tracks probe different competencies. The table below maps them — built from public job postings, mission cadence data, and process documentation. Treat it as directionally useful, not confirmed fact.
| Dimension | NewSpace constellation-ops (Starlink, Kuiper, OneWeb, Planet) | Legacy GEO/MEO mission-assurance (Lockheed Martin, Northrop Grumman, Boeing, Maxar) |
|---|---|---|
| Hiring filter | Manufacturing cadence, rapid iteration, software-first, on-orbit ops at scale | Mission assurance rigor, formal review gate experience, heritage components, decade-long lifecycle ownership |
| Lifecycle | Months per satellite; design evolves batch-to-batch; factory-floor integration | 5–15 years per program; PDR/CDR/TRR enforced; single-spacecraft build |
| Review process | Lighter gates; anomalies resolved via SW patch or next-batch hardware change | NPR 7123 gates: SRR → PDR → CDR → MRR → TRR → FRR; exit criteria formally tracked |
| What they probe | Anomaly resolution speed, FDIR tradeoffs, constellation link budget, manufacturing design choices | Design review participation, trade study methodology, single-string fault tolerance, heritage qualification |
| Where to find real signal | Job postings (“manufacturing cadence”, “ops at scale”); LinkedIn from constellation operators | Glassdoor for the specific prime; NASA LLIS; INCOSE CSEP holder accounts |
For a systems-level aerospace role spanning aircraft and spacecraft, see the aerospace engineer interview guide. For legacy prime prep, see the Lockheed Martin, Northrop Grumman, and Boeing interview guides — all three run formal gate-based technical screens that differ from NewSpace loops.
Attitude determination & control (ADCS / GNC) interview questions
ADCS is the most commonly probed subsystem in generalist satellite-engineer interviews — close a pointing budget from first principles, make sensor selection tradeoffs, and reason about attitude knowledge chain failures. For the control theory foundation, see the controls engineer interview guide.
Walk me through how you’d derive a pointing accuracy budget for a 500-kg LEO Earth-observation satellite.
Concept: pointing error budget decomposition | Difficulty: mid/senior | Stage: technical
What they’re probing: Whether you structure an error budget from mission need down to component specs, not recite sensor numbers in isolation.
Start with the mission requirement — say, 0.05° — and allocate backward: attitude determination error (star tracker noise, gyro drift, filter lag), control error (reaction wheel deadband, structural flexibility), and disturbance torques (gravity gradient, aerodynamic drag, SRP) — worst-case and RSS. Blue Canyon achieves 0.002° pointing knowledge on ESPA-class buses. (Source: NASA SmallSat SoA 2026)
Strong answer hits: Allocate explicitly — “40% determination, 40% control, 20% unmodeled” — then justify against the hardware catalog. Listing specs without rolling them up to the mission requirement is the red flag.
When would you pick a star tracker over a sun sensor + magnetometer pair for attitude determination?
Concept: sensor selection tradeoff | Difficulty: mid | Stage: technical
What they’re probing: Whether you think in terms of the full sensor chain — nominal ops, degraded ops, and safe mode — not just peak-performance specs.
Star trackers deliver 3-axis absolute attitude knowledge to arc-second precision but go blind in eclipse or stray light. A sun sensor + magnetometer pair is lightweight and low-power — coarse two-axis attitude only. Pick the star tracker for sub-0.1° pointing: Earth observation, optical comms, laser ranging. Pick the coarser pair for CubeSats or safe-mode fallback. (Source: NASA STEMGATEWAY)
Strong answer hits: Explain the sensor-mode handoff logic and the attitude uncertainty during the transition window.
Explain how a Kalman filter combines gyro and star tracker measurements, and what happens during a star tracker outage.
Concept: state estimation under sensor failure | Difficulty: senior | Stage: technical / system design
What they’re probing: Whether you understand the propagate-update cycle and what it means for attitude knowledge when one measurement source disappears.
Gyros propagate the attitude quaternion forward but accumulate drift. Star tracker measurements correct the state and reset drift. During a star tracker outage the filter enters gyro-only propagation: covariance grows; pointing error degrades linearly with drift rate and outage duration.
Strong answer hits: SOHO operators disabled Gyro B — removing the last measurement source — because they skipped the hypothesis-discrimination step. FDIR must require a working backup before any primary sensor goes offline. (Source: NASA/ESA SOHO Report)
What lessons does the SOHO 1998 attitude loss carry for modern ADCS design and operations?
Concept: postmortem analysis / procedural safety | Difficulty: senior | Stage: behavioral / system design
What they’re probing: Whether you can extract transferable design and process lessons from a historical anomaly — not just recite what happened.
Three compounding errors: gyro calibration omitted the re-enable command for Gyro A; Gyro B left in high-gain mode triggered Emergency Sun Reacquisition (ESR); ground operators commanded Gyro B off — removing the only working gyroscope. The board (Prof. Massimo Trella, ESA; Dr. Michael Greenfield, NASA) cited compressed-timeline procedures with no review board sign-off. (Source: NASA/ESA SOHO Final Report)
Strong answer hits: ESR must stay enabled during maintenance. FDIR must require a working backup before commanding any primary sensor off. First-time procedures on compressed timelines need independent verification.
Which ADCS simulation tools have you used, and what are the tradeoffs between STK, Cosmos, and OREKIT for control loop validation?
Concept: simulation toolchain judgment | Difficulty: mid | Stage: technical
What they’re probing: Whether you understand the boundary between orbital mechanics tools, control design tools, and FSW test frameworks.
STK: orbit propagation and access analysis — control-loop depth requires MATLAB/Simulink integration. OREKIT (open-source, Java): high-fidelity propagators for Monte Carlo and CI/CD. Cosmos (Ball Aerospace): ground software and FSW interface tool — not a dynamics simulation environment. MATLAB/Simulink: workhorse for control law prototyping and HIL test. (Source: Tustin Recruiting SE Guide, 2025)
Strong answer hits: Calling Cosmos a control-loop sim tool is a red flag. Correct decomposition: STK for orbit geometry, OREKIT or MATLAB for the control loop, Cosmos or YAMCS for FSW command interfaces.
For a side-by-side view of all five subsystems, see the prep matrix below.
Electrical power system (EPS) interview questions
EPS questions test closing a power budget from solar input through storage to load, making defensible battery chemistry choices, and selecting a charge controller architecture for the mission profile. The ScienceDirect 2025 CubeSat EPS framework frames this as “sizing, design, integration, and qualification” end-to-end. (Source: El Hachimi et al., ScienceDirect 2025)
How do you size a solar array for a LEO sun-synchronous mission with 35% eclipse fraction?
Concept: solar array power budget | Difficulty: mid | Stage: technical
What they’re probing: Whether you can walk from orbit geometry to a design number, not just state the formula.
Key equation: array power = (load × eclipse fraction / charge efficiency + load) / (sun fraction × degradation factor). BOL vs. EOL output differs 15–30% with radiation dose. At low beta angles, maximum eclipse coincides with depressed cell efficiency — model that interaction explicitly. Size to EOL, verify positive margin at BOL. (Source: ScienceDirect EPS framework 2025)
Strong answer hits: Walk the power budget by mode — nominal, safe, peak. EOL sizing prevents depth-of-discharge violations in year one.
Why did the industry move from NiCd to NiH₂ and then to Li-ion for satellite batteries? When is NiH₂ still preferred?
Concept: battery chemistry tradeoffs | Difficulty: mid | Stage: technical
What they’re probing: Whether you understand that battery chemistry selection is driven by mission duration, cycling profile, and risk tolerance — not just energy density.
NiCd: ~50 Wh/kg, memory effect. NiH₂: ~60 Wh/kg, 40,000+ LEO cycles at 30% DoD — GEO standard through the 2010s. Li-ion: 150–200 Wh/kg, no memory effect, excellent LEO cycle life — but sensitive to overcharge/overdischarge. (Source: Tustin Recruiting SE Guide, 2025)
Strong answer hits: NiH₂ remains preferred for 15+ year GEO missions. Recommending Li-ion without 15-year GEO cycle-life data is the red flag — GEO eclipse seasons run ~90 cycles per year.
Walk through a peak-power-tracker (PPT) vs direct-energy-transfer (DET) decision. When is each appropriate?
Concept: EPS architecture selection | Difficulty: mid/senior | Stage: technical / system design
What they’re probing: Whether you understand the efficiency vs. complexity tradeoff and can map it to a real mission scenario.
A PPT operates the array at its maximum power point via DC-DC converter — more energy extracted across temperature and degradation variation. A DET connects the array directly through a shunt regulator — simpler, fewer components, array voltage fixed to the bus. PPT wins when beta-angle swings drive large temperature variation; DET wins when the thermal environment is stable and reliability dominates. (Source: ScienceDirect 2025)
Strong answer hits: GEO satellites with well-characterized eclipse seasons typically choose DET for heritage and simplicity. Early Simulink EMC analysis catches conducted emissions before manufacturing.
If you were specifying a 6U CubeSat EPS, what tradeoffs would you make on bus voltage, regulation topology, and load shedding?
Concept: CubeSat EPS design tradeoffs | Difficulty: mid | Stage: system design
What they’re probing: Whether you can navigate the CubeSat power-mass tradeoff with real design choices rather than generic principles.
Bus voltage: 3.3V/5V unregulated for small payloads; 12V regulated for higher-power payloads but adds conversion stages. Topology: most 6U designs use an unregulated battery bus with regulated rails per subsystem. Load shedding: nominal → safe → emergency; flight computer and radio shed last. PC/104 form factor constrains regulator count and bulk capacitor volume. (Source: NASA SmallSat SoA 2026)
Strong answer hits: Name commercial EPS modules (EnduroSat, GomSpace, Clyde Space) and justify the topology choice for this orbit and payload. Module-level familiarity is the senior answer.
What are the most common EPS failure modes on LEO satellites and how do you design against them?
Concept: EPS fault tolerance design | Difficulty: mid/senior | Stage: technical
What they’re probing: Whether you think about failure modes as design inputs from day one, not as afterthoughts during test.
The ESA ESOC anomaly dataset flags solar array power regulator switch-offs as a recurring category. (Source: ESA ESOC Anomaly Dataset) Key failure modes: regulator latch-up → overcurrent protection per string, SEL-immune regulators; battery overdischarge → low-voltage cutoff with autonomous load shedding; bus failure → OR-diode isolation between battery strings; ground fault/ESD → bonding per ECSS-E-ST-20.
Strong answer hits: Each failure mode is a design input with a specific mitigation — not a test-phase discovery.
For a side-by-side view of all five subsystems, see the prep matrix below.
Telemetry, tracking & command (TT&C / comms) interview questions
TT&C interviews probe closing a link budget from first principles, defending frequency band and modulation choices quantitatively, and reasoning about ground segment architecture. Expected debugging methodology: link budget margins, antenna pointing, noise floor, BER trends, and frequency offsets. (Source: Tustin Recruiting SE Guide, 2025)
Walk me through a complete link budget for a S-band TT&C downlink from a 500-km LEO orbit.
Concept: link budget derivation | Difficulty: mid/senior | Stage: technical
What they’re probing: Whether you can walk the full signal chain — transmitter to bit error rate — without skipping a term or hiding behind a tool.
EIRP = transmitter power − cable losses + antenna gain. At S-band (~2.2 GHz), worst-case slant range (~2,000 km at 5° elevation) gives ~165 dB free-space path loss. Received power = EIRP − path loss − atmospheric attenuation + receive antenna gain. Divide by system noise temperature → C/N₀; compare to required Eb/N₀. Target: +3 dB minimum link margin. (Source: Dr. Frances Zhu, University of Hawaii OER)
Strong answer hits: Name all noise sources: thermal noise (dominant), cosmic background ~2.7 K, atmospheric attenuation. Justify the margin target — programs with access gaps may target +6 dB.
Compare S-band, X-band, and Ka-band for satellite communications. When would you pick each?
Concept: frequency band selection | Difficulty: mid | Stage: technical
What they’re probing: Whether you understand frequency selection as a systems tradeoff — not just a higher-is-better bandwidth question — with regulatory, rain-fade, and aperture-size implications.
S-band (~2–3 GHz): established allocations, low rain fade — the TT&C workhorse. X-band (~7–8 GHz): higher bandwidth for Earth observation and deep-space science data downlinks. Ka-band (~23–28 GHz): very high throughput for commercial broadband constellations; heavy rain attenuation (30+ dB) requires site diversity or ACM. (Source: Dr. Frances Zhu, University of Hawaii OER)
Strong answer hits: S-band for TT&C robustness. X-band for science data (KSAT, SSC, Leaf Space all have X-band infrastructure). Ka-band for high-throughput where site diversity or ACM is affordable.
How would you design a ground station network for a 100-satellite LEO constellation with global coverage requirements?
Concept: ground segment architecture | Difficulty: senior | Stage: system design
What they’re probing: Whether you understand orbit geometry driving contact frequency, the owned-vs-commercial tradeoff, and how latency requirements constrain the architecture.
At 500 km LEO, contact windows are 5–10 minutes per 90-minute orbit. Global coverage needs polar anchor stations (Svalbard, McMurdo, Punta Arenas) plus mid-latitude stations on multiple continents — minimum 3–4 for one contact per orbit. Commercial alternative: KSAT and AWS Ground Station offer per-pass pricing at 30+ sites globally.
Strong answer hits: 100 satellites at 10 kbps = 1 Mbps aggregate — telemetry routing and ops automation matter as much as antenna count. NewSpace automates at scale; legacy primes staff a mission ops center per spacecraft.
Why is BPSK still common on TT&C links when higher-order modulations exist? When would you move to QPSK or 8PSK?
Concept: modulation selection tradeoff | Difficulty: mid | Stage: technical
What they’re probing: Whether you understand the fundamental power-bandwidth tradeoff in link design rather than defaulting to “higher order is better.”
BPSK has the lowest Eb/N₀ threshold for a given BER — the most power-efficient choice when link margin, not throughput, is the binding constraint. Move to QPSK when payload data rate requires it and margin allows the 3 dB Eb/N₀ penalty. Move to 8PSK or higher for Ka-band payload downlinks where spectrum licensing limits bandwidth.
Strong answer hits: Safe-mode command links must use the lowest-order modulation with FEC — losing 3 dB can break command contact at low elevation angles.
What’s the difference between command authentication and command encryption, and which would you require for a commercial Earth-observation satellite?
Concept: command security architecture | Difficulty: mid/senior | Stage: technical / system design
What they’re probing: Whether you understand these as separate security properties with different threat models, and can articulate the minimum-viable security posture for a commercial operator.
Command authentication verifies the command source via MAC or digital signature. Command encryption protects command content from interception and replay. For a commercial EO satellite, authentication is mandatory: an unauthenticated link lets any transmitter with sufficient power issue commands — including deorbit or safe-mode disable. Encryption is highly recommended; CCSDS provides the standard framework.
Strong answer hits: Authentication without encryption = unforgeable but visible. Encryption without authentication = hidden but source unverified. Treating these as equivalent is the common candidate error.
For a side-by-side view of all five subsystems, see the prep matrix below.
Command & data handling (CDH / avionics) interview questions
CDH interviews test the radiation environment as a component-selection constraint, memory tradeoffs, and fault management architecture differences between a constrained CubeSat and a redundant prime-built platform. Satellite CDH is one of the most demanding embedded systems environments — flight software must be deterministic, radiation-tolerant, and capable of autonomous recovery for days.
Explain the difference between rad-hard and rad-tolerant components. When can you accept rad-tolerant for a LEO mission?
Concept: radiation hardening selection | Difficulty: mid/senior | Stage: technical
What they’re probing: Whether you understand the full mitigation stack — hardware qualification + software mitigation + FDIR — rather than treating rad-hard as a binary pass/fail.
Rad-hard: RHBD techniques (guard rings, TMR, specialized process nodes) — qualified to 100 krad TID or higher. Rad-tolerant: commercial/industrial parts tested at 10–30 krad TID, combined with ECC, memory scrubbing, and watchdog timers. For low-inclination LEO at 500–600 km, TID over 3 years may be only 5–10 krad — within the rad-tolerant COTS envelope with appropriate derating. (Source: NASA SmallSat SoA 2026)
Strong answer hits: High-inclination LEO and GEO accumulate TID an order of magnitude faster — rad-hard is not optional there. Document risk acceptance at PDR.
Walk through the tradeoffs between SRAM, SDRAM, MRAM, and Flash for satellite flight software memory.
Concept: space-grade memory selection | Difficulty: senior | Stage: technical
What they’re probing: Whether you understand endurance and retention as distinct axes, not just “which memory is best for space.”
Four axes: write endurance, data retention, radiation susceptibility, power. SRAM: unlimited endurance, volatile — standard for working RAM. Flash: ~1×10⁵ write cycles, non-volatile — flight software image and telemetry log storage. MRAM: ~1×10¹³ write cycles, non-volatile — superior for write-intensive telemetry buffering; confirm SEU and TID specs with the manufacturer. (Source: NASA SmallSat SoA 2026)
Strong answer hits: FSW execution → SRAM; SW image → Flash; telemetry ring buffer → MRAM (8 orders of magnitude better endurance than Flash). SRAM always needs non-volatile backup loaded at boot.
Centralized vs distributed avionics architecture: when would you pick each on a modern satellite design?
Concept: avionics architecture tradeoffs | Difficulty: mid/senior | Stage: system design
What they’re probing: Whether you can map an architecture decision to the mission’s actual fault-tolerance and development-team requirements.
Centralized CDH: single OBC for all processing — simpler integration, fewer interfaces, but single point of failure. Distributed: dedicated processors per subsystem (ADCS, EPS, payload) over SpaceWire, CAN, or Ethernet — better fault isolation and parallel development at the cost of interface count. Modern SmallSat trend: distributed architectures with standardized backplane buses (SpaceVPX, CompactPCI) and hot-swappable modules. (Source: NASA SmallSat SoA 2026)
Strong answer hits: Legacy GEO uses centralized with redundant OBC strings — heritage qualification beats redesign. SpaceWire for high-data-rate science payloads; CAN bus for slow telemetry on CubeSats.
Describe how you’d design a fault detection, isolation, and recovery (FDIR) tree for a single-string CubeSat vs a redundant prime-built satellite.
Concept: FDIR architecture | Difficulty: senior | Stage: system design
What they’re probing: Whether you understand that FDIR design is driven by the hardware availability model — a single-string design cannot recover from hardware failures, only software transients — and that this must be explicitly accepted at PDR.
FDIR is hierarchical: detect, isolate, recover. Single-string CubeSat: watchdog timers and conservative power-on defaults — shallow tree: detect → reset → safe mode → wait for ground contact. Single-string hardware failures cannot be recovered; only software transients can. Redundant prime-built: switch to redundant OBC, reconfigure power buses, reroute comms — deeper tree with cross-strapping logic and telemetry confirmation.
Strong answer hits: Document the FDIR tree at PDR even for CubeSats — undocumented recovery assumptions are a common on-orbit surprise. (Source: Shiotani et al., AIAA SPACE 2014)
What’s your experience with NASA cFE / cFS or other open-source flight software frameworks, and when would you build custom?
Concept: flight software framework selection | Difficulty: mid/senior | Stage: technical
What they’re probing: Whether you have hands-on FSW experience and a principled view on build vs. reuse — not just awareness that cFS exists.
NASA cFE / cFS: message routing (software bus), time management, telemetry services, plug-in application architecture — flight heritage on VxWorks, RTEMS, and Linux. Build-vs-reuse turns on: team familiarity, target processor compatibility (cFS is well-supported on heritage rad-hard CPUs; novel SoCs may need custom frameworks), and qualification value (NASA programs often require cFS because its review history is part of the qualification package).
Strong answer hits: cFS handles scheduling and telemetry plumbing; AI/ML inference models publish outputs as telemetry messages outside cFS’s service layer — not a replacement for it.
For a side-by-side view of all five subsystems, see the prep matrix below.
Thermal control system (TCS) interview questions
Thermal interviews test closing a heat balance from orbit geometry to component temperature, choosing passive vs. active control based on dissipation levels, and sizing a radiator and MLI system with defensible margin. The beta angle ties thermal, power, and ADCS design together — expect to discuss it regardless of which question comes first.
When do you transition from passive thermal control (MLI + radiator sizing) to active control (loop heat pipes, fluid loops)?
Concept: passive vs. active thermal architecture | Difficulty: mid/senior | Stage: system design
What they’re probing: Whether you treat passive control as the first choice and can articulate the threshold at which active control becomes necessary.
Passive control — MLI, radiators, surface coatings — handles most satellites below ~500 W total bus power. The transition is driven by dissipation density exceeding conduction path capacity, or payload stability requirements (e.g., ±2°C) passive control can’t maintain. LHPs and MPFLs move heat across structural joints to remote radiators. MDPI 2024 showed heat pipes improving equipment deck temperatures by +34.2°C on a 1,175 km LEO satellite with 448 W SSPA dissipation. (Source: Huang and Bu, MDPI Aerospace 2024)
Strong answer hits: Payloads below ~100 W in LEO typically close the thermal budget with passive elements alone.
How do you size a radiator for a payload dissipating 100 W in a GEO orbit with seasonal beta angle variation?
Concept: radiator sizing with beta angle variation | Difficulty: mid/senior | Stage: technical
What they’re probing: Whether you can identify both bounding cases — hot and cold — and explain which drives radiator area vs. heater power budget.
Q = ε × σ × A × (T⁴ − T_sink⁴). For 100 W in GEO: worst-case hot — high beta (|β| > 58°) = no eclipse, constant solar illumination; worst-case cold — low beta = maximum eclipse (~72 min/orbit at equinox), highest heater duty cycle. Beta varies ±27° seasonally in GEO. Radiator area is set by the hot case; heater power is set by the cold case. (Source: Huang and Bu, MDPI Aerospace 2024)
Strong answer hits: Model both extremes in Thermal Desktop or SINDA. At maximum beta angle, GEO has no eclipses — heater power requirement is zero.
What’s the difference between a 10-layer and a 30-layer MLI blanket, and when does adding layers stop helping?
Concept: MLI blanket design and diminishing returns | Difficulty: mid | Stage: technical
What they’re probing: Whether you understand the physical mechanism behind MLI and can identify where adding layers is counterproductive.
MLI reduces radiative exchange through aluminized Mylar or Kapton layers with low-conductance spacers. Effective emissivity decreases with additional layers but with diminishing returns: at low layer counts, radiation dominates; at high layer counts, conduction through spacers and edge effects becomes the limit. Practical ceiling: ~20–30 layers — beyond that, mass added exceeds thermal benefit. (Source: Huang and Bu, MDPI Aerospace 2024)
Strong answer hits: As-tested MLI performance runs 20–40% worse than models predict — allocate margin against effective emissivity, not the theoretical minimum. Key degradation drivers: ground handling, spacer compression, seam conductances.
Walk me through the thermal challenge of a satellite entering and exiting eclipse, and how you’d design margin for the worst-case cold case.
Concept: eclipse thermal transient design | Difficulty: mid/senior | Stage: technical
What they’re probing: Whether you understand that thermal design is governed by extremes — hot case (radiator sizing) and cold case (heater sizing) — and that margin must be explicitly allocated against the worst-case input combination.
Eclipse cuts solar input to zero in seconds; LEO eclipse can exceed 35 minutes per orbit. Worst-case cold: maximum eclipse duration (low beta, below ~30°) + minimum solar input (EOL surface degradation) + minimum dissipation (safe mode). Set heater power to maintain survival temperature under all three — add at least 10°C margin above the minimum operating limit. (Source: Huang and Bu, MDPI Aerospace 2024)
Strong answer hits: Thermal-balance testing validates the thermal model — model-to-test discrepancies trigger a model update cycle before flight.
Which thermal analysis tools have you used (Thermal Desktop, SINDA, SystemaTHERMICA), and what are their tradeoffs?
Concept: thermal simulation toolchain | Difficulty: mid | Stage: technical
What they’re probing: Whether tool choice is driven by program pedigree and customer requirements — not just technical merit — and whether you can explain the boundary between the geometric modeler, orbital heating calculator, and thermal network solver.
Thermal Desktop (C&R Technologies): industry-standard US aerospace — geometric modeling, orbital heating, and finite-difference thermal network solving. SINDA/FLUINT: solver for fluid loop and heat pipe modeling with two-phase flow capability. SystemaTHERMICA: ECSS-compliant, used in European programs and by ESA contractors.
Strong answer hits: US prime/NASA contractor → Thermal Desktop. European prime/ESA → SystemaTHERMICA or ESATAN-TMS. COMSOL is not a substitute for mission-level orbital thermal analysis. (Source: ScienceDirect 2025)
For a side-by-side view of all five subsystems, see the prep matrix below.
Cross-subsystem and behavioral interview questions every satellite engineer gets
Behavioral questions test process judgment and systems thinking. Concrete, attributable examples are required — vague STAR answers that could apply to any engineering role signal the candidate hasn’t worked in a real spacecraft program.
Walk me through a design review you led or contributed to (PDR / CDR / TRR). What was the major issue surfaced and how was it resolved?
Concept: design review process fluency | Difficulty: mid/senior | Stage: behavioral
What they’re probing: Whether you understand design review gates as specific decision points with formal exit criteria — not generic “team meetings” — and whether you’ve contributed substantively to surfacing and resolving a real technical issue.
Name the gate and what was at stake: PDR freezes the design baseline and authorizes long-lead procurement; CDR freezes drawings, BOM, and ICDs and releases manufacturing; TRR confirms procedures and facility readiness. The issue should be technically specific — ICD conflict, a margin that closed only under optimistic assumptions, a test gap. Resolution should name who was in the room and what the corrective action was. (Source: Anywaves Space Project Milestones)
Strong answer hits: Describing a PDR without knowing its formal exit criteria — or describing a “design review” outside the SRR-PDR-CDR-MRR-TRR-FRR sequence — signals peripheral participation. (Source: Shiotani et al., AIAA SPACE 2014)
Describe a satellite anomaly investigation you participated in. What was the root cause and how was it confirmed?
Concept: anomaly investigation methodology | Difficulty: senior | Stage: behavioral
What they’re probing: Whether you apply structured diagnostic thinking under uncertainty — not whether you’ve experienced a dramatic anomaly.
Strong answers follow a postmortem structure: timeline reconstruction, hypothesis list, discriminating data (what rules each hypothesis in or out), and root cause confirmation. ESA ESOC anomaly dataset categories: solar array regulator switch-offs, DPU resets, attitude disturbances — all requiring multi-subsystem telemetry correlation. (Source: ESA ESOC Anomaly Dataset)
Strong answer hits: SOHO 1998 is the canonical failure of this structure — operators skipped hypothesis discrimination and disabled the last functioning gyro. (Source: NASA/ESA SOHO Final Report) “Closed as unexplained” is legitimate if you explain why further discrimination wasn’t possible.
Walk me through a trade study you led where the obvious answer turned out to be wrong.
Concept: trade study rigor and intellectual honesty | Difficulty: senior | Stage: behavioral
What they’re probing: Whether you challenge first-order intuitions when data points the other direction, and whether your methodology is formal enough to surface that kind of finding.
Strong answers name competing options, criteria (mass, cost, schedule, reliability, heritage), how each was weighted, and what the non-obvious finding was. Example: “The obvious choice was Li-ion on energy density, but the GEO cycle-life analysis at actual depth-of-discharge favored NiH₂ on 15-year calendar life, and the program chose NiH₂ to avoid heritage qualification risk.”
Strong answer hits: Documented figures of merit, traceable weighting rationale, formal recommendation memo. The signal is whether you treat trade studies as a decision tool or a post-hoc rationalization.
For a side-by-side view of all five subsystems, see the prep matrix below.
Red flags satellite interviewers watch for (and how to avoid them)
Panels have short lists of answers that immediately signal a candidate hasn’t worked in a real spacecraft program.
- Citing TLE propagation as an ADCS simulation tool. TLEs are orbit state inputs for position propagation, not attitude dynamics tools. STK handles orbit geometry; MATLAB/Simulink or OREKIT handle the control loop.
- Recommending Li-ion for a 15-year GEO mission without cycle-life data. Li-ion’s GEO eclipse-season data is thinner than NiH₂’s. Recommending it without citing depth-of-discharge, cycle count, and calendar-life qualification data means you haven’t run the EPS trade.
- Calling CCSDS optional for any government mission. CCSDS is mandatory on most NASA and ESA programs and required for DSN or ESTRACK ground station support.
- Describing a design review without knowing its exit criteria. “We did a CDR” without knowing CDR exits with frozen drawings, BOM, and ICDs signals peripheral participation.
- Treating thermal margin as a fixed percentage. “We added 10%” is not a methodology. Margin must trace to specific worst-case hot and cold inputs.
- Conflating command authentication with command encryption. Authentication prevents unauthorized sources; encryption prevents content disclosure. They address different threat models.
The subsystem-to-prep matrix: what to study cold by the role you’re interviewing for
Identify your primary subsystem row, study the top questions cold (no notes), and be ready to discuss the named postmortem at depth. For adjacent subsystems, read the “what to know” column to speak credibly at the system boundary.
| Subsystem | What the role owns in 2026 | Top 2 questions to nail | Postmortem or trade study to know | Tools to know cold |
|---|---|---|---|---|
| ADCS / GNC | Pointing budget, sensor chain, control law design, attitude-loss FDIR | Derive a pointing budget from mission requirement; explain Kalman filter propagate-update and gyro-only degradation | SOHO 1998 — three compounding operator errors, ESR disable, gyro chain failure (NASA/ESA) | STK (orbit geometry), MATLAB/Simulink (control loop), OREKIT (Monte Carlo) |
| EPS | Solar array sizing, battery selection, charge controller architecture, power mode management | Size a solar array for LEO SSO with 35% eclipse fraction; walk the PPT vs DET decision | NiCd → NiH₂ → Li-ion transition and the 15-year GEO cycle-life question (ScienceDirect 2025) | MATLAB/Simulink (EPS sizing + EMC), COMSOL (component thermal evaluation) |
| TT&C / Comms | Link budget, frequency band selection, ground station architecture, command security | Walk a complete S-band LEO link budget with margin; explain authentication vs. encryption | Ground station architecture trade for 100-satellite LEO — owned vs. commercial-as-a-service | STK (link geometry), STK RF module, MATLAB (link budget spreadsheet) |
| CDH / Avionics | Processor/memory selection, rad-hardening strategy, FDIR tree, flight software framework | Rad-hard vs. rad-tolerant for 3-year LEO; SRAM vs. MRAM vs. Flash by use case | CubeSat review omission — interfaces discovered at integration after skipped PDR/CDR (Shiotani et al. 2014) | NASA cFE/cFS, SpaceWire/CAN bus specs, MATLAB (FDIR modeling) |
| Thermal | Radiator + MLI sizing, passive vs. active control, eclipse transient management | Size a radiator for 100 W GEO with seasonal beta variation; when MLI layers stop helping | Heat pipe improvement — +34.2°C deck temperature on LEO constellation satellite (Huang & Bu, MDPI 2024) | Thermal Desktop (US), SINDA/FLUINT (fluid loops), SystemaTHERMICA/ESATAN-TMS (European) |
For propulsion, payload, and structures, the ESA ECSS standards library and AIAA Space Systems Engineering resources are the right starting point.
Questions to ask the interviewer (satellite engineer specific)
Reverse questions are how an experienced engineer reads the program’s actual culture. Generic questions signal a candidate who hasn’t thought about what makes satellite programs specifically challenging.
- What’s the cadence of design reviews on your current program, and who chairs them? (Substantive vs. ceremonial.)
- When an on-orbit anomaly happens, who leads the investigation — standing tiger team or per-incident? (Institutional process vs. ad-hoc.)
- Which subsystems are growing on the team and which are shrinking? (Program phase — hiring thermal engineers suggests Phase C/D.)
- How do you balance mission assurance rigor with launch cadence? (The central NewSpace tension.)
- What’s the most significant technical risk still open, and what’s the mitigation plan? (Whether the mitigation is credible or aspirational.)
- Is there a test-like-you-fly process for the ground-to-on-orbit transition? (Configuration control maturity.)
- What’s the team’s relationship with mission assurance — independent oversight or embedded support? (Legacy primes run formal independent verification; NewSpace often doesn’t.)
Satellite engineer interview prep: a 14-day roadmap
The plan assumes a subsystem specialty and adjacent knowledge gaps to shore up.
- Days 1–3: Master your primary subsystem cold. Work through your subsystem questions — ADCS, EPS, TT&C, CDH, or Thermal — until you can answer each from first principles without notes. Derive the key budget equation on paper.
- Days 4–7: Study one adjacent subsystem at depth-1. Pick the subsystem sharing the most interfaces with yours — enough to speak credibly at the boundary.
- Days 8–10: Master one real postmortem. SOHO 1998 (NASA/ESA Investigation Report) or Aeolus 2023 (ESA Mission Report). Read the actual report. Know the timeline, root cause chain, three concrete lessons, and the investigation board chairs by name.
- Days 11–13: Rehearse behavioral and reverse questions. Use specific examples from your project history. Rehearse five reverse questions until they feel conversational.
- Day 14: Company-specific calibration. NewSpace: scan job postings for “manufacturing cadence,” “rapid iteration,” “on-orbit ops at scale.” Legacy prime: confirm the program’s current design review gate. Read the NASA Lessons Learned System (LLIS) for legacy-prime interviews.
The matrix, postmortems, and red-flag list are worth revisiting the morning of the interview. Close the budget, explain a failure, ask a question that signals you’ve thought about the program.