Radiation (TID/SEE) Monitor for Spacecraft Health Monitoring
← Back to: Avionics & Mission Systems
A Radiation (TID/SEE) Monitor turns the space radiation environment into actionable engineering data—cumulative dose and time-stamped event statistics—so missions can trend lifetime risk, detect bursts, and make mode decisions with confidence. It focuses on a trustworthy sensing chain (sensor → AFE → discrimination/counting → telemetry) with calibration and redundancy, helping distinguish real radiation effects from false triggers.
H2-1 · What a Radiation Monitor really does (TID + SEE, and why both matter)
Extractable answer: A spacecraft Radiation Monitor turns the local radiation environment into two engineering signals: TID dose (a slow, cumulative trend) and SEE events (fast, discrete upsets). Sensors (RADFET/PIN/Si) feed an AFE that integrates dose and shapes pulses for event discrimination, then produces counters, timestamps, and compact telemetry for health and mission decisions.
- Particles / LET / dose interact with a sensing element (RADFET threshold shift, PIN/Si current or pulses).
- AFE conditioning converts tiny signals into measurable quantities: TIA/charge integration for dose, pulse shaping for SEE.
- Quantization & decisions separate slow vs fast: ADC tracks dose-related trends; a discriminator classifies valid events.
- Data products are formed: dose accumulator (krad(Si) trend) plus event counters and timestamps (rate/fluence evidence).
- Redundant power domains keep monitoring and reporting continuous across single-domain disturbances and transient conditions.
Why TID and SEE must be handled differently: TID behaves like a slow variable (cumulative drift and long-term margin), so accuracy depends on leakage/temperature separation and stable integration. SEE behaves like a fast event stream (SEU/SEL/SET/SEFI), so correctness depends on thresholding, debounce/dead-time, and trustworthy timestamps.
- Lifetime margin: track dose accumulation to support derating, scheduling, and end-of-life forecasts.
- Actionable alarms: detect event-rate excursions that justify mode changes or operational constraints (interface-only to protection actions).
- Correlation evidence: time-align upset bursts with subsystem anomalies to separate real radiation from false triggers.
- Mission context: compare measurements across attitude/shielding changes using consistent metrics and reporting fields.
Scope boundary: This page focuses on sensing → AFE → counting/timestamping → reporting and the resulting telemetry products. Power protection, bus architectures, and downlink systems are referenced only as interfaces, not explained in depth.
H2-2 · Radiation metrics engineers actually use (krad(Si), LET, fluence, rate)
Radiation monitoring becomes actionable only when measurements are expressed in metrics that map directly to AFE dynamic range, threshold/debounce policy, and reporting granularity. This section defines the minimum set of terms that keep dose trends and SEE event streams comparable across test and flight.
| krad(Si) (dose) | Cumulative deposited dose referenced to silicon. Design mapping: sets the sensor output span and integration headroom, plus when range switching or controlled discharge is required to avoid saturation and preserve long-term resolution. |
|---|---|
| Dose rate | Dose per unit time. Design mapping: sets integration/update cadence and determines how strongly leakage and temperature drift can masquerade as “dose change” if the AFE is not stabilized and compensated. |
| Reference basis | The metric is meaningful only when the reference is stated (e.g., “Si” in krad(Si)). Design mapping: reporting must include the reference label so trends remain interpretable across missions and test campaigns. |
| LET | Linear Energy Transfer: a proxy for how “strong” a particle strike is in the sensitive volume. Design mapping: motivates multi-threshold discrimination (event “bins”) because event amplitude/shape can vary across LET, affecting how thresholds and hysteresis should be chosen. |
|---|---|
| Fluence | Total particle exposure (integrated flux). Design mapping: requires reporting an exposure window alongside counts; otherwise event totals cannot be compared between orbits, attitudes, or test runs. |
| Cross-section (σ) | Event probability per particle exposure for a given mechanism (e.g., SEU). Design mapping: reinforces that counts alone are insufficient; telemetry should pair counts + time + exposure tag to preserve engineering meaning. |
| Event rate | Events per unit time. Design mapping: sets alarm thresholds and drives debounce/dead-time choices to keep false triggers low without hiding real bursts. Timestamp health must be maintained to support correlation. |
Metric → design constraints (fast reference):
Dose range → sensor span + integrator headroom + drift separation
Dose rate → update cadence + leakage/temperature compensation strategy
LET/fluence → discriminator thresholds + hysteresis + debounce/dead-time
Event rate → alarm policy + reporting granularity (summary vs histogram vs event queue)
Common traps that distort interpretation:
• Focusing on TID only can miss bursty SEE conditions where dose remains low but SEU/SET activity is high.
• Focusing on SEE only can miss slow drift that changes thresholds and biases, causing the monitor itself to lose fidelity over time.
H2-3 · Sensing elements: RADFET vs PIN diode/Si detector vs “victim-based” monitors
A Radiation Monitor is only as good as its sensing element. The correct choice depends on the primary intent: TID lifetime trending, dose-rate/fluence tracking, or SEE proxy evidence. The key engineering rule is simple: the sensor output type dictates the AFE architecture (voltage drift, continuous current, charge pulses, or event/error counts).
| Need: long-term TID margin | Prefer RADFET-style sensing (threshold/voltage drift). It directly represents cumulative ionizing dose, but demands low-drift readout and careful temperature interpretation. |
|---|---|
| Need: dose-rate / fluence changes | Prefer PIN diode / Si detector sensing (current or pulse rate). It reacts quickly to environment changes, but must handle leakage, noise, and saturation across wide dynamic range. |
| Need: “did the platform get upset?” | Use victim-based proxies such as SRAM/FPGA scrub statistics or error counters as SEE evidence. This is closest to mission impact, but it is not a pure environment measurement and must be tagged with exposure time and operating mode. |
How temperature and aging show up in data: RADFET drift can change slope with temperature and annealing, so dose trending requires a temperature context tag. PIN/Si leakage typically rises with temperature, which can look like “dose-rate” unless leakage is bounded or compensated. Victim-based counts can change with workload and scrub policy, so they must be interpreted as platform-upset rate under a defined mode.
- Voltage drift (RADFET): precision readout, low offset drift, stable sampling cadence (slow variable).
- Continuous current (PIN): TIA/integration, range headroom, leakage control and guarding.
- Charge pulses (Si detector): pulse shaping + discriminator + timestamped counting (fast events).
- Error counts (proxy): requires exposure window, timebase health, and operating mode tag to be comparable.
H2-4 · AFE architecture for dose: charge integration, TIA, and drift control
Dose monitoring is a slow-variable measurement: it succeeds when the AFE makes pA–µA-level sensor signals measurable without confusing leakage, bias drift, and temperature effects as “dose change”. For most implementations, the practical architecture is a guarded front-end plus TIA or charge integration, a stable sampling/ADC stage, and a digital accumulator that records both the dose estimate and confidence flags.
| Input domain (pA–µA / charge) | The signal can be comparable to board leakage and amplifier bias. Front-end guarding, clean routing, and bounded leakage paths keep the measurement observable rather than drift-dominated. |
|---|---|
| TIA vs Integrator | TIA converts current to voltage for continuous dose-rate proxy; integration converts tiny current/charge into a measurable ramp for better low-level resolution. The choice is driven by whether the data product emphasizes dose-rate tracking or dose accumulation. |
| Dynamic range controls | Windowing, controlled discharge, and range switching prevent saturation across quiet conditions and storm bursts, while preserving resolution for long-term trending. Each control action should be recorded as a health flag to protect interpretation. |
| Drift control | Auto-zero/chopper techniques reduce offset drift so that multi-hour or multi-day dose trends remain meaningful. The intent is not high speed, but stable long-duration fidelity. |
| ADC + digital accumulation | ADC selection prioritizes low noise and stable gain over bandwidth. Digital accumulation produces dose, dose-rate (optional), and quality indicators (saturation, excessive leakage, temperature out-of-range). |
- Input bias & offset drift: appears as a slow ramp even when environment is quiet.
- Leakage (sensor/package/PCB): temperature-dependent current that can dominate pA signals.
- Temperature coefficient: changes the apparent gain/offset and alters long-term slope.
- Rf noise (TIA): sets the practical low-level resolution; too large a resistor can increase noise and drift sensitivity.
- Integration saturation: produces clipped ramps; without flags, clipped data can be misread as a plateau.
How “dynamic range” becomes real design: A monitor that must survive both quiet periods and burst conditions typically needs at least one of: (1) integration window control, (2) controlled discharge/reset, or (3) range switching. The telemetry should expose which mechanism is active so trend analysis remains trustworthy.
H2-5 · AFE for SEE events: pulse shaping, discrimination, and timestamping
The SEE event path turns fast sensor spikes into trustworthy event evidence. A robust chain can be expressed as a four-step flow: (1) sensor pulses/spikes, (2) pulse shaping, (3) discrimination with threshold policy and dead-time, and (4) counting plus timestamping. The engineering goal is to reject noise-triggered hits without hiding real events, and to preserve enough timing information for correlation.
- Sensor pulse / spike: a short transient current or charge packet appears at the detector output.
- Pulse shaping: bandwidth and noise are traded to produce a stable pulse width and amplitude suitable for a comparator.
- Discrimination: thresholds + hysteresis + debounce/dead-time produce a single clean trigger per physical event.
- Counting + timestamping: counters (or bins) provide statistics; timestamps provide correlation and burst analysis.
| Fixed threshold | Simple and comparable across runs. Best when baseline noise is stable. Needs hysteresis to prevent chatter. |
|---|---|
| Adaptive threshold | Tracks noise floor to stabilize false-trigger rate. Must be bounded and flagged, or it can silently raise the bar and miss real events. |
| Multi-threshold bins | Produces event “levels” (low/med/high) that are ideal for telemetry histograms. Requires clear dead-time and bin definition to avoid double counting. |
False triggers in event chains: Noise or coupled spikes can exceed a comparator threshold and look like an event. The countermeasures are implemented inside the discrimination path: hysteresis (prevents edge chatter), debounce (requires persistence), and dead-time (suppresses re-triggering from pulse tails and ringing).
Counting formats: A timestamp queue supports correlation but costs bandwidth and can overflow during bursts. A histogram (counts per bin per window) is bandwidth-friendly and preserves distribution information. Many systems combine both: summary counters always, and a short timestamp queue only when rates spike.
H2-6 · Data products: what to report (dose, rate, histograms, SEE taxonomy)
Telemetry is the product of a Radiation Monitor. The right reporting set must support engineering decisions (lifetime margin, alarms, correlation) without exhausting bandwidth or storage. A practical approach is to publish two levels: a high-frequency summary for operations and a low-frequency detail set for analysis.
| Level A — Real-time summary |
TID: dose_total, dose_rate (windowed), dose_quality_flags. SEE: counts_by_bin (or type), see_rate (windowed), max_burst, timestamp_health. Purpose: alarms, trending, and quick correlation with subsystem anomalies. |
|---|---|
| Level B — Low-rate detail |
Histograms: counts per threshold/bin over longer windows (bandwidth-friendly). Timestamp queue (limited): short event list during spikes or when triggered (correlation evidence). Purpose: root-cause support without continuous high-rate data. |
- Dose fields: dose_total, dose_rate (windowed), dose_flags (saturation, reset/discharge, range switch).
- SEE fields: counts_by_bin/type, see_rate, max_burst (short-window peak), queue_depth/overflow flag.
- Correlation tags: temperature, operating mode, exposure/shielding tag (metadata only).
- Integrity & redundancy: domain_id (A/B), schema_version, seq_counter, timebase_state, CRC.
Granularity trade-off: Reporting too much detail can break bandwidth and storage budgets, especially during bursts. Reporting only coarse totals can hide whether the monitor saturated, switched range, or lost timestamp fidelity. The two-level approach keeps operations stable while preserving enough evidence for post-analysis.
H2-7 · Calibration & compensation: temperature, annealing, sensor aging
Calibration is what keeps long-duration radiation data meaningful. A monitor that trends dose for days to years must separate true environment change from temperature effects, AFE drift, and sensor aging. The most reliable strategy is a two-part loop: ground calibration produces traceable coefficients, and on-orbit health checks prevent compensation from hiding real events or fabricating “fake dose”.
| Temperature effects | Sensor leakage and sensitivity can change with temperature; AFE offset and bias drift can mimic slow dose slope. Temperature context tags and bounded compensation prevent false trending. |
|---|---|
| RADFET annealing | Apparent “dose rollback” or slope change can occur as trapped charge partially relaxes. Treat it as an expected behavior that must be modeled/flagged, not as an environment improvement. |
| Sensor aging | Long-term sensitivity changes can shift gain and baseline. Trending requires a calibration version and uncertainty grade to keep lifetime estimates conservative. |
- Cross-channel consistency: if multiple channels shift in the same direction and scale, the AFE is a likely contributor.
- Reference/anchor behavior: a stable internal reference path (or a known “quiet” window) helps identify offset/bias drift.
- Temperature correlation: leakage-driven shifts often track temperature; true dose accumulation should not instantly follow temperature swings.
| Core coefficients | gain, offset, tempco (bounded model), plus valid_range or range_id. |
|---|---|
| Traceability | cal_version, cal_date, and an uncertainty grade (e.g., low/med/high or numeric bound). |
| Health linkage | Flags that protect interpretation: saturation, range switch, discharge/reset activity, temperature out-of-range, and “compensation bounded” indicators. |
| Calibratable | Repeatable gain/offset errors, temperature coefficients in defined ranges, predictable baseline drift that can be verified by health checks. |
|---|---|
| Not calibratable (flag instead) | Burst transients, saturation/recovery behavior outside valid range, unexpected coupling that creates rare spikes, and timebase loss during an upset. These require quality flags, not aggressive correction. |
Compensation safety rule: If a correction cannot be validated on-orbit by a self-consistency check, it should be bounded, versioned, and accompanied by a data-quality flag. This prevents “fixing the data” at the cost of hiding real environment change.
H2-8 · Redundant power domains & fault tolerance (monitor must survive the event)
Redundancy exists to keep the monitoring chain alive and trustworthy during SEL-like upsets and transients. The purpose is not to describe a full aircraft/spacecraft power system, but to ensure the radiation monitor can continue measuring, preserve traceability, and avoid fabricating events during recovery.
| Goal | Monitoring continuity through upsets; no silent data loss; consistent timebase and counter meaning across failover. |
|---|---|
| Mechanisms (monitor-side) | Dual domains A/B (each with sensor+AFE+counter), independent health flags, watchdog/state machine for safe reset, and a single output path with controlled failover. |
| Acceptance checks | Failover preserves domain_id, seq_counter continuity, and timebase_state visibility; A/B summaries remain comparable within defined bounds; overflow/saturation states are always flagged. |
- Compare: cross-check A/B summary counters in the same window; if divergence exceeds a bound, raise a quality flag.
- Vote/failover: select the domain with valid timebase and health flags; switch only with explicit state and logging.
- Align timing: preserve timebase_state and domain_id in every packet so ground analysis can stitch records reliably.
Boundary reminder: Protection actions and power topology belong to the system power page. Here, redundancy is treated only as it impacts the monitoring chain: measurement continuity, event integrity, and traceable telemetry.
H2-9 · Placement & shielding for monitors (measurement integrity, not full shielding theory)
Placement determines what the radiation data actually represents. A monitor can be deployed to measure a representative cabin/box environment, to capture a worst-case exposure near a sensitive item, or to provide a correlation point that helps explain anomalies. Local structure and partial shielding can introduce bias (a systematic offset versus the true environment) and lag (delayed response to environment changes), so deployment should be treated as part of the measurement system.
| Representative | Measures typical exposure for trending and long-window comparisons. Best when the goal is stable dose/rate history rather than hotspot detection. |
|---|---|
| Worst-case | Placed near a sensitive component or module to track local peaks and stress. Best for explaining failures and setting conservative margins. |
| Correlation | Positioned near a structural feature, opening, or known gradient path to help interpret changes driven by configuration or shielding variation (metadata-driven interpretation). |
- Bias: partial shielding or nearby structure can make a point read consistently lower/higher than the intended reference environment.
- Lag: if the point is behind structure, step changes in exposure can appear delayed in the recorded dose-rate response.
- Comparability: any mechanical or configuration change should be recorded as a metadata tag, or historical trends become non-comparable.
| Single-point (representative) | Lowest cost. Good for long-term trending. Limited ability to explain localized faults or shielding changes. |
|---|---|
| Two-point gradient | One point near structure/opening, one near the core zone. The difference/ratio tracks shielding variation and improves interpretability without needing full materials modeling. |
| Hybrid (representative + sensitive) | A baseline point plus a hotspot point near a critical module. Best for correlation: “environment changed” vs “local hotspot changed”. |
Scope boundary: This section focuses on measurement integrity (what the data means and how it can be compared over time). It does not attempt to provide a complete shielding design guide.
H2-10 · Verification & qualification: heavy-ion/proton tests and acceptance criteria
Verification proves the monitor is “done” by producing deliverable evidence for both slow TID behavior and fast SEE event integrity. The goal is not to restate full standards, but to define engineering acceptance criteria: linearity/monotonicity for dose trending, predictable threshold response for event chains, traceable timebase behavior, and uninterrupted monitoring across redundancy events.
| What to exercise | Stepped dose exposure, temperature-tagged intervals, and post-step observation windows for recovery behavior. |
|---|---|
| What to deliver | Drift curves (raw vs compensated), residuals vs temperature tags, saturation/range-switch flags, and calibration version used. |
| Acceptance focus | Trend is monotonic within valid range (or within defined piecewise model bounds), compensation is bounded and flagged when outside validity. |
| What to exercise | Heavy-ion / proton exposure with threshold scans (single or multi-bin), burst conditions, and queue/overflow stress. |
|---|---|
| What to deliver | Threshold scan summaries, counts_by_bin/type and rate windows, burst metrics (max_burst), queue depth/overflow statistics, and discriminator settings. |
| Acceptance focus | False-trigger rate controlled by policy; missed-event risk is quantified by comparison to reference injection/conditions; binning remains consistent across runs. |
| Timestamp integrity | timebase_state is always visible; timebase loss/recovery is logged; timestamp accuracy is within a declared bound for correlation use. |
|---|---|
| Counter meaning | Counts are linear/monotonic versus intended stimulus in valid ranges; saturation/overflow never occurs silently (flags required). |
| Redundancy continuity | Failover does not create fake events; domain_id + seq_counter continuity supports stitching; monitoring remains observable during recovery. |
- Setup: sensor type, range_id, threshold policy, window settings, cal_version.
- TID outputs: dose step table, drift curves (raw/compensated), residual summary vs temperature tags, flags summary.
- SEE outputs: threshold scan summary, counts_by_bin/type, rate/burst metrics, queue overflow stats, dead-time settings.
- Timing: timebase_state transitions, timestamp consistency bound.
- Redundancy: failover logs, A/B compare metrics, continuity evidence (seq continuity, data gaps detectable).
H2-11 · Failure modes & field diagnostics: distinguishing real radiation from false triggers
On-orbit anomalies often show up as a sudden rise in event rate, noisy low-threshold bins, or inconsistent telemetry between redundant domains. This section provides a monitor-side diagnostic workflow to separate real radiation-driven changes from false triggers caused by noise/EMI coupling, threshold drift, temperature steps, or single-event upsets affecting configuration and counters.
- Noise/EMI triggering: low-threshold bins jump while higher bins remain flat; event width / dead-time occupancy becomes abnormal.
- Threshold drift: event rate creeps upward over minutes to hours; strongly correlated with temperature or long-term aging.
- Temperature step: abrupt baseline shift (bias/leakage) changes trigger probability; histogram shifts toward lower bins.
- Register upset (concept level): counters/configuration flip (threshold_state, dead_time, bin map) causing discontinuities or impossible values.
- Telemetry inconsistency: packet sequencing/CRC flags, timestamp disorder, or A/B domain mismatch without a plausible physical gradient.
| Spatial consistency | Multi-point monitors rise together (or change with a stable, explainable gradient). Single-point-only spikes suggest local coupling or drift. |
|---|---|
| Shape consistency | Histogram/bin profile and burst statistics look physical (not only the lowest bin jumping). Sudden “all-bins-flat except low” often indicates noise. |
| State consistency | Threshold_state, dead_time, domain_status, and timestamp_health support the interpretation; no silent configuration changes or timebase loss. |
- Freeze a “configuration snapshot”: threshold_state, hysteresis, dead_time, shaping mode/bandwidth, bin map, cal_version.
- Inspect the time shape: is the rise a single spike, a step that persists, or a slow creep?
- Check multi-point correlation: synchronous rise across points, or localized to one point?
- Check histogram shape: do higher bins move, or only the lowest bin? Is there an abnormal “burst” pattern?
- Check discriminator health: dead_time_occupancy (or equivalent), threshold_state stability, overflow/saturation flags.
- Check telemetry integrity: packet_seq continuity, CRC_ok, timestamp ordering, and timebase/timestamp_health flags.
- Do a minimal self-consistency test (small, safe change): apply a minor threshold/dead-time adjustment and observe sensitivity.
Heuristic: noise-triggered rates are often extremely sensitive to small threshold shifts; physical spectra usually change more smoothly and predictably.
| Event statistics | event_rate (per bin/type), max_burst, optional queue_depth/overflow |
|---|---|
| Shape | histogram_bins (low-rate report), or compact bin-count summary |
| Decision state | threshold_state, dead_time (or occupancy), discriminator_mode |
| Time integrity | timestamp_health, timebase_state, monotonicity/ordering flag |
| Redundancy integrity | domain_status (A/B active), domain_id, cross-check/vote status |
| Telemetry integrity | packet_seq, CRC_ok, config snapshot reference (cfg_hash or cfg_version) |
| Observed symptom | Most likely cause | Next check |
|---|---|---|
| Only lowest bin spikes; higher bins flat | Noise/EMI coupling into discriminator | threshold sensitivity test; dead_time_occupancy; multi-point sync |
| All bins increase; stable gradient across points | Real environment change (credible radiation shift) | shape consistency + timestamp_health; confirm metadata tags (mode/attitude) |
| Slow creep over hours; tracks temperature | Threshold drift / bias/leakage drift | temperature tag correlation; threshold_state history; compensation validity |
| Sudden step with domain switch; A/B mismatch | Failover side effects or configuration upset | domain_status timeline; cfg_hash change; packet_seq continuity; cross-check status |
| Impossible counter jump or wrap without flags | Register upset / missing overflow signaling | require explicit overflow flags; verify counter width; enforce plausibility checks |
- MCU (event management / telemetry): Microchip SAMRH71 (rad-hard MCU)
- FPGA (histograms / higher-rate processing): Microchip RTG4 / RT4G150 (rad-tolerant FPGA family)
- Comparator (multi-threshold discriminator): TI TLV1704-SEP (rad-tolerant quad comparator in SEP)
- ADC (monitor acquisition): Renesas ISL73141SEH (rad-hard 14-bit SAR ADC, 1 MSPS class)
- Op amp (monitor chain building blocks): Renesas ISL70444SEH (rad-hard quad op amp)
- Voltage reference (ADC/AFE stability): Renesas ISL71091SEH10 (rad-hard precision reference)
- MUX (range/channel/self-check selection): Renesas ISL73841SEH (rad-tolerant 32:1 analog MUX)
- Electrometer-grade prototype AFE: ADI ADA4530-1 (femtoamp input bias; typically used for ground/prototype validation of pA–nA chains)
Request a Quote
H2-12 · FAQs ×12
These FAQs focus on monitor-side decision rules and deliverable fields (dose/rate, event bins, timestamp health, and redundant-domain status), avoiding system-wide power/bus/EMC deep dives.
