Thermocouple/RTD AFE Op Amp: Low-Noise, Low-Drift
← Back to:Operational Amplifiers (Op Amps)
Thermocouple and RTD measurements become accurate and repeatable when the AFE is designed around the real error sources—thermal gradients, low-frequency noise, lead resistance, leakage, and recovery after faults—not just “low offset” on a datasheet. This page shows how to choose the right op-amp traits, CJC/excitation strategy, calibration hooks, and verification tests so the whole temperature chain stays stable from the bench to the field.
What this page solves for thermocouple & RTD AFEs
Temperature front-ends are rarely limited by “headline noise” alone. Thermocouples live in the µV world and are dominated by offset, drift, thermoelectric EMFs, and EMI-to-DC errors. RTDs are resistance sensors, so accuracy depends on excitation stability, lead resistance, self-heating, and ratiometric sampling. This page helps you turn those realities into a stable, calibratable AFE that survives cables, connectors, airflow, and production spread.
Thermocouple vs RTD: what changes the design
- Thermocouple: microvolts at very low frequency → offset/drift and thermo-EMFs often dominate; CJC and thermal gradients become part of the measurement.
- RTD: resistance measurement → the AFE is defined by excitation, 2/3/4-wire sensing, lead resistance, and self-heating control.
- Both: long cables and harsh environments make EMI behavior, input leakage, and overload recovery critical selection dimensions.
Why “lowest noise” is not the whole answer
In temperature chains, these traits often set the real accuracy floor. A design that looks great on a bench can drift or “jump” on a real panel because cables, protection networks, thermal gradients, and recovery from faults create error mechanisms that wideband noise specs do not capture.
What you get from this page
- A practical architecture map for thermocouple and RTD AFEs.
- An error-budget field list (what to track, what dominates, and why).
- Calibration hooks for offset/gain/CJC/excitation and how to validate them.
- A verification & production checklist that catches drift, EMI, and recovery failures early.
System architectures: thermocouple vs RTD front-end topologies
This section maps the practical topology choices and the minimum validation hooks that prevent “bench-good, field-bad” outcomes. Each branch ties a sensor scenario to a recommended front-end structure and the op-amp traits that actually protect accuracy.
Thermocouple topology families
- Microvolt-level signal with slow dynamics (temperature changes).
- Long cable runs, connector blocks, or grounded junctions that invite common-mode stress and EMI.
- Production needs stable offset/drift and predictable recovery after faults (open/short/ESD).
- Input protection + RC sized for EMI and fault survival, with leakage accounted for.
- Bias return path for input bias currents (avoid floating inputs that “wander”).
- Low-drift gain + low-pass shaping matched to update rate (do not amplify noise bandwidth unnecessarily).
- CJC path: sensor placement and thermal coupling at the terminal region.
- Low offset/drift and low 0.1–10 Hz noise → verify with slow logging (minutes to hours), not only spot measurements.
- EMI-robust input behavior → inject RF/fast transients and check for DC shifts or “stuck” offsets.
- Fast overload recovery → force open/short events and measure settle-to-within-spec time.
Not covered here: internal INA topology derivations and ADC driver design details; link those topics to the dedicated INA and ADC-driver pages to avoid content overlap.
RTD topology families (2/3/4-wire + ratiometric)
- High absolute accuracy or long leads require controlling lead resistance and excitation errors.
- Update rate and power budget matter (self-heating becomes a system error term).
- Field conditions demand stable readings across supply and ambient variations.
- Excitation source (current or voltage) with measured/trackable stability.
- 2/3/4-wire sensing selected by lead resistance sensitivity and installation realities.
- Ratiometric sampling: measure RTD and reference resistor under the same timing and reference domain.
- Amplification + LPF sized for the required temperature update bandwidth.
- 2-wire: simplest; lead resistance directly adds error → best for short leads and modest accuracy.
- 3-wire: cancels first-order lead effects if leads match → requires disciplined wiring and symmetry.
- 4-wire: true Kelvin sensing removes lead resistance impact → preferred for long runs and high accuracy.
- Low bias/leakage (especially with large resistances and protection parts) → verify across temperature and humidity.
- Low drift and strong low-frequency stability → log drift vs ambient and supply steps.
- Recovery behavior → step excitation on/off and confirm no long settle tails or bias memory.
When to pre-amplify vs go direct to a high-resolution ADC
- Pre-amplify when the sensor signal in the intended bandwidth is close to the ADC’s practical noise floor, or when the system needs analog bandwidth shaping before digitization.
- Go more direct when update rate is low and the system can rely on narrow effective bandwidth (digital filtering/averaging), but still keep robust input protection and bias return paths.
- Prioritize input integrity (EMI, leakage, recovery) before chasing gain: many “accuracy” failures originate at the connector and protection network, not in the amplifier core.
Error budget that actually matters: offset, drift, thermoelectric EMFs
Temperature front-ends are often limited by structured error terms that look like “drift” or “jumps” in real logs. For thermocouples, microvolt-level signals make offset, drift, thermoelectric EMFs, and EMI-to-DC the usual dominant contributors. For RTDs, the chain is defined by lead resistance, excitation stability, self-heating, and reference-domain drift. A good budget does not start with one headline number; it starts with a list of terms that can be assigned ownership, measured, and tracked.
Thermocouple: why “µV-world” errors dominate
- Offset & drift map directly into temperature error because the signal is small and low-frequency.
- Thermoelectric EMFs appear at dissimilar-metal junctions (connector, solder, plating) when there is a thermal gradient; the resulting µV terms can look like airflow sensitivity or “touch-to-shift.”
- EMI pickup and rectification can create apparent DC offsets; validate with the intended cable and protection network.
- Ownership points: terminal region (thermal gradients), input network (leakage/EMI), amplifier stage (offset/drift), and CJC path (placement and thermal coupling).
Chopper/zero-drift parts can introduce ripple or EMI-sensitive behavior in some conditions; treat this as a validation item rather than assuming it is always safe. Deep mechanism details belong on the dedicated zero-drift page.
RTD: error structure is set by wiring + excitation
- Lead resistance becomes measurement error unless the wiring strategy cancels it (3-wire) or removes its impact (4-wire Kelvin).
- Excitation stability (current/voltage source + reference resistor) maps into apparent temperature drift unless measured in the same domain as the RTD.
- Self-heating makes accuracy depend on power, mounting, airflow, and duty cycle; treat it as a system error term, not a sensor property.
- Reference drift is minimized by ratiometric sampling (RTD and reference measured with shared timing and reference domain).
Reusable error-budget field list (no numbers, only ownership and relationships)
- Sensor & wiring: TC type + junction style; RTD type + 2/3/4-wire; cable length and connector type.
- Input domain: source impedance, lead/contact resistance, bias return path, protection network leakage budget.
- Amplifier terms: offset, drift, 0.1–10 Hz noise, bias/leakage sensitivity, overload recovery behavior, EMI behavior.
- Excitation/Reference: excitation method, reference resistor strategy, ratiometric domain (shared timing/reference), supply sensitivity.
- Thermal domain: terminal temperature gradients, CJC placement, airflow sensitivity, self-heating risk factors.
- Validation hooks: open/short injection, EMI stimulus, supply step, long-log drift vs ambient, coefficient/version tracking.
- Temperature error is the sensor-domain error divided by sensitivity (avoid mixing electrical and temperature units in logs).
- Thermo-EMF correlates with thermal gradients at junctions; changes in airflow or mounting can look like electrical drift.
- Lead resistance varies with wire temperature and contacts; wiring strategy determines whether that variation becomes error.
- Excitation/reference drift becomes apparent temperature drift unless measured in the same domain (ratiometric timing).
- EMI can cause DC shifts by rectification or recovery effects; digital filtering cannot repair front-end corruption.
Noise in temperature measurements: 0.1–10 Hz, mains, and bandwidth shaping
Temperature changes are slow, so the effective measurement bandwidth is usually narrow. That shifts attention from wideband noise to low-frequency noise (0.1–10 Hz) and mains interference (50/60 Hz). Good results come from defining update rate and effective bandwidth first, then shaping noise with a combination of input filtering, bandwidth limiting, and averaging—without relying on digital filtering to repair front-end corruption.
Why 0.1–10 Hz noise often matters more than wideband noise
- In narrow-band temperature logging, the displayed “stability” is driven by low-frequency noise integrated over the effective bandwidth.
- Wideband noise can be reduced with bandwidth limiting and averaging; low-frequency noise shows up as slow wandering or “breathing” in the readout.
- Validate with long time logs under controlled airflow and thermal conditions; spot measurements rarely reveal the real floor.
50/60 Hz and harmonics: analog RC vs digital notch (the boundary)
- Analog RC first when mains pickup or fast transients can push the front-end into overload or rectification; preventing corruption beats filtering symptoms.
- Digital notch/averaging works best when the chain stays linear and mains is a clean sinusoid in the sampled domain.
- Digital filtering cannot repair errors created by front-end overload, EMI rectification, leakage shifts, or recovery tails; those must be prevented or designed out.
Practical bandwidth-shaping workflow (no deep ADC filter details)
- Define update rate and required readout stability (what “settled” means in seconds/minutes).
- Set effective bandwidth from the update rate and averaging window (avoid capturing unnecessary noise bandwidth).
- Limit energy at the input with RC and layout so EMI and mains do not create DC shifts.
- Apply averaging / notch only after confirming the analog chain remains linear and recovers quickly from faults.
- Verify by logging: compare stability with and without airflow, cable movement, and supply perturbations.
Input bias, source resistance, and protection without corrupting accuracy
Many temperature front-ends fail on real cables because a protected input is not only a signal path—it is also a DC leakage network. Accuracy is corrupted when input bias/leakage currents create an equivalent input error across sensor/source resistance and when clamp devices unintentionally become the only bias return path. A robust design defines the return path, limits fault energy, and verifies recovery after open/short/over-voltage events.
Bias/leakage × source resistance = equivalent input error
- Signal is microvolt-level, so small DC errors look like temperature drift.
- Equivalent error is created when bias/leakage flows through the effective input resistance of the sensor + wiring + protection network.
- Uncontrolled bias return (floating input) often shows up as slow wandering, touch sensitivity, and recovery tails after faults.
- Lead/contact resistance and protection leakage can appear as resistance (and temperature) error.
- Wiring strategy (2/3/4-wire) determines whether lead and contact variation becomes a measurement term.
- Input networks must keep leakage and bias currents out of the sensing nodes, especially when long cables or humid environments exist.
Do not rely on TVS/ESD clamps to provide bias return. Their leakage is temperature- and stress-dependent and can turn a stable input into a drifting one.
Protection blocks: what helps, what can silently break accuracy
- Limits fault current and isolates the input from capacitive loading and RF energy.
- Creates bias-current error across the resistor and increases settling time with input capacitance.
- Sets a place to control the “noise pickup loop” by keeping high-energy cable currents out of the analog core.
- Shapes bandwidth and reduces EMI energy before it reaches sensitive nodes.
- Can add leakage paths and long recovery tails if dielectric absorption or contamination exist.
- Must be placed to minimize loop area from connector to filter and to keep return currents predictable.
- Protects against over-voltage, ESD, and cable faults, but leakage is often temperature- and humidity-sensitive.
- Stress events can change leakage; always validate post-event offset and settle behavior.
- Clamp current paths must not lift analog ground or inject into reference nodes; route returns with intent.
Controlled failure and minimal “open-wire” hooks
A high-accuracy AFE must fail in a detectable way. Open/short/reverse connections should not create a long, silent drift state. Use lightweight hooks so faults become visible to firmware and production tests.
- Inject open/short conditions and measure time-to-within-spec after the fault is removed.
- Verify that clamps do not become the bias return path; check offset drift across temperature and humidity.
- Validate that cable movement and touch do not create step shifts (a common sign of uncontrolled leakage paths).
Not covered here: system-level surge/EMC compliance and cable shielding strategies (handled in EMI-hardened and system pages).
Cold-junction compensation and reference strategy (thermocouple)
Cold-junction compensation (CJC) is a thermal + electrical system, not a single sensor choice. Accuracy depends on whether the CJC sensor measures the temperature of the actual cold junction region under real airflow and board heating, and whether the measurement chain uses a stable reference domain so reference drift does not masquerade as temperature drift.
Placement: measure the cold-junction region, not an arbitrary board temperature
- The cold junction is created at the terminal region where dissimilar metals meet; this is where gradients matter most.
- Airflow, connector heating, and nearby power dissipations create temperature gradients that can change over time.
- Thermal coupling is a design choice: copper area, sensor distance, and mechanical construction determine whether CJC tracks reality.
CJC sensor options (system impact only)
- IC temperature sensor: integration is simple; accuracy depends heavily on placement, copper coupling, and response time to terminal changes.
- NTC: very sensitive and low cost; system work shifts to calibration curve handling, leakage/self-heating control, and stable biasing.
- RTD: stable and linear; system cost is added excitation/sensing and routing discipline to keep gradients and lead effects controlled.
Reference strategy: keep drift from becoming apparent temperature error
- Thermocouple voltage is small, so reference drift can look like temperature drift if the measurement domain is not stable.
- Use a consistent reference domain and timing so supply/reference variations are not interpreted as sensor changes.
- Ground and return paths must keep high currents away from the terminal and reference nodes; measure stability with long logs.
Compensation math (tables/polynomials) is typically handled in firmware; this section focuses on placement, thermal coupling, and reference-domain behavior.
Minimal validation hooks (quick to run, high diagnostic value)
- Airflow test: blow across the terminal area and confirm the compensated reading does not shift abnormally.
- Local heating step: apply a controlled heat step near the terminal block and compare CJC response vs reading response.
- Cable touch test: touch the terminal/cable and check for step offsets (often a gradient/thermo-EMF signature).
- Long log: record for hours and correlate with board temperature and supply events to expose reference-domain drift.
RTD excitation: current vs voltage, self-heating, and ratiometric sampling
RTD accuracy is set by a closed loop: excitation defines the measurement domain, self-heating couples measurement energy into the sensor, and ratiometric + synchronous sampling determines whether drift terms cancel or become apparent temperature drift. A good design starts from update rate and allowed self-heating, then chooses excitation amplitude, duty cycle, and sampling timing.
Current excitation vs voltage excitation: how errors map into temperature
- Measured voltage is proportional to I × RRTD, so excitation drift can appear as temperature drift unless handled by ratiometric sampling.
- 2-wire wiring includes lead/contact resistance as a direct resistance term; 3-wire cancels only when lead mismatch is controlled; 4-wire Kelvin largely removes lead effects.
- Good fit when the system can keep excitation stable and can sample a reference resistor in the same domain and timing.
- Current changes with total resistance, so sensor power can vary with temperature and wiring; this can turn into self-heating dependent drift.
- Lead/contact resistance changes the total resistance domain more directly, so wiring and contact stability become stronger terms.
- Good fit when the voltage domain is stable and the measurement structure explicitly tracks reference and wiring terms.
Avoid “better by default” thinking. Dominant terms are set by wiring, environment, update rate, and allowed measurement energy.
Self-heating: control by amplitude, duty cycle, and windowing
RTD self-heating is not a static spec. It depends on excitation power, mounting, airflow, and the measurement schedule. Treat self-heating as a system error term that can be shaped by measurement timing.
- Lower excitation amplitude reduces power, but may require more gain and filtering.
- Duty-cycled excitation limits average power while preserving short measurement windows for low noise.
- Averaging reduces noise bandwidth, but the excitation schedule must not create a slow temperature ramp in the element.
- Use consistent timing so any remaining self-heating becomes repeatable and calibratable, not random drift.
- Change update rate and duty cycle and check for systematic readout shifts with slow recovery (self-heating signature).
- Step excitation amplitude and verify that the reading change is bounded and repeatable.
Ratiometric + synchronous sampling: conditions for cancellation (and common failure modes)
Ratiometric sampling cancels drift only when RTD and reference measurements share the same reference domain and occur within a timing window where drift terms are effectively constant.
- Same reference domain: RTD and Rref are measured against the same ADC reference or tracked baseline.
- Same timing: samples are synchronous or close enough that reference/supply drift does not change between them.
- Controlled environment for Rref: reference resistor and key nodes are kept away from strong thermal gradients.
- Rref is thermally different from the RTD loop, so ratiometric math injects a thermal drift term.
- RTD and Rref samples are separated by a large mux timing gap, so transient reference/supply movement becomes error.
Not covered here: Howland current-source derivations and transconductance circuit theory (handled in the Howland/Transconductance page).
Layout & thermal design: the hidden dominant error in real boards
In temperature measurement boards, layout and thermal design are not “optimizations”—they are part of the sensor. Dissimilar-metal junctions plus temperature gradients create parasitic thermocouples, and connector/terminal construction can dominate drift even when the schematic is unchanged. The goal is an isothermal terminal region, symmetric sensitive paths, and predictable return currents.
Thermo-EMF and gradients: layout creates (or removes) the error source
- Dissimilar metals at terminals, solder joints, plating, and screws form unintended thermocouples when a gradient exists.
- Airflow and nearby heat sources move gradients over time, creating drift that looks like sensor change.
- Symmetry matters: unmatched copper environments or via counts can create unequal gradients between the two input paths.
Practical isothermal and keep-away rules (board-level)
- Create an isothermal island around the terminal/CJC region with continuous copper and controlled exposure to airflow.
- Keep sensitive input paths symmetric in length, via count, and nearby copper environment.
- Separate the temperature AFE from hot zones (DC/DCs, power resistors, high-current traces) with distance and copper partitioning.
- Route return currents so high-current loops do not pass through the terminal and reference areas.
Why “same circuit, different terminal” drifts: a real-world cause chain
Changing terminal blocks, connector plating, solder alloy, or mechanical construction can change the set of junctions and the thermal coupling. That alters gradients and creates new parasitic thermo-EMF sources, often dominating long-term drift and touch/airflow sensitivity.
- Verify with airflow and local heating steps near the terminal area and observe drift direction and recovery.
- Use long logs to correlate drift with board temperature and supply/heat-source activity.
Minimal validation hooks (layout/thermal)
- Airflow: apply airflow to the terminal and check for step shifts.
- Touch: touch terminal and nearby copper and check for consistent polarity shifts.
- Heat-source step: toggle DC/DC load and confirm the temperature channel is not modulated by board heating.
- Long log: correlate drift with board temperature gradients rather than only ambient temperature.
Offset calibration & drift tracking: practical two-point and in-field strategies
“Calibratable” only matters when calibration is a closed loop: measure → extract structured error → create coefficients → apply correction → verify stability. Two-point calibration is powerful for offset/gain-type errors, but temperature systems often fail because the dominant term is a gradient-driven drift or a reference-domain movement that cannot be fixed by a single constant coefficient.
Two-point calibration: what it fixes and how to choose reference conditions
- Offset: input-referred constant terms that appear at “zero stimulus” in the chosen fixture/path.
- Gain: end-to-end scale error in the selected measurement domain (excitation/reference/gain chain).
- Best results require that the calibration path matches the real measurement path (same protection, wiring, and reference domain).
- Zero: short the input at the same physical boundary used in the product (terminal vs front-end) to avoid fixture-only leakage paths.
- Known stimulus: use a precision resistor / known voltage equivalent in the same reference domain for gain.
- Known temperature point: use a stable, equilibrated setup when system-level calibration is required (avoid gradients and rushing soak time).
- Verify immediately after calibration and again after a delay; constant errors stay corrected, while drift-dominated systems reappear over time.
- Inject a small controlled disturbance (airflow / load change) and confirm recovery does not create a new offset baseline.
When two-point is not enough: trigger conditions and safe extensions
- If the residual error tracks board/terminal temperature, the dominant term is not a constant offset.
- If the residual error changes with power mode, update rate, or excitation schedule, add logging and condition-aware correction rather than more points.
- If the calibration fixture is not repeatable (leakage, contact, gradients), more points can turn into coefficient noise.
- Use a small set of temperature points only when the system is thermally stable and gradients are controlled.
- Use “versioned coefficients” and a rollback rule when field data shows instability.
- Prefer correlation-based tracking (logs + stability checks) before expanding to dense LUTs.
Drift tracking: make the problem measurable with a minimal log schema
Drift becomes diagnosable when measurements carry enough context to separate temperature-correlation, time-correlation, and state-correlation. Record the minimal set below so field behavior can be mapped back to a dominant error source.
- timestamp, serial/lot, coeff_version
- board temp (or terminal/CJC temp when available)
- supply mode, power state
- excitation mode (I/V, amplitude tier, duty cycle)
- update rate, averaging window
- raw code, filtered, compensated outputs
- event flags (open-wire, overload, ESD/OVP, airflow/fan state if relevant)
Use correlations to decide whether the dominant term is thermal gradient, reference-domain movement, leakage changes, or measurement schedule effects.
Calibration boundaries: what cannot be “calibrated away”
- Calibration corrects structured error terms; it does not remove noise floors, interference pickup, or thermal gradients.
- If coefficients require a better stimulus chain than the system can reproduce, coefficients become another noise source.
- Always include a “stability gate”: reject coefficient updates when repeatability is not proven in the same operating state.
Part selection logic: how to pick the op-amp for TC/RTD AFE
A good selection process maps system conditions into risk terms, then selects op-amp traits and must-run validation tests. Avoid choosing by a single headline number. For temperature AFEs, the decisive terms often include overload recovery, leakage/bias interaction with protection networks, and EMI behavior under long cables.
Three-bucket framing (traits + tradeoffs + must-run tests)
- Tradeoffs: can be more sensitive to practical EMI and contamination leakage if the input network is not controlled.
- Must-run: temperature sweep stability, low-frequency noise, source-R / protection leakage sensitivity.
- Tradeoffs: DC noise/drift may not be the absolute best; power can be higher.
- Must-run: cable-touch/line pickup tests, overload and open-wire recovery, high-EMI injection checks.
- Tradeoffs: slower settling and higher low-frequency noise may force longer windows or lower update rates.
- Must-run: mode switching stability, duty-cycled excitation interaction, long-term drift log under real power states.
Must-check traits (mapped to system risks)
- Offset + drift: drives slow temperature error; requires realistic thermal and soak conditions.
- 0.1–10 Hz noise: controls stability of slow-changing temperature readouts.
- Input bias/leakage behavior: interacts with source resistance and protection networks; verify across temperature and humidity.
- EMI behavior: check RF rectification tendency and stability with long cables and RC filters.
- Input common-mode range: ensure real operating CM plus interference stays inside linear region.
- Overload recovery: determines how fast the channel returns after open/short/OVP/clamp conduction.
- IQ: sets the power budget; verify accuracy/noise at the actual operating current.
What to ask vendors (conditions, not just numbers)
- Offset/drift test conditions: temperature points, soak time, input source resistance and bias return setup.
- 0.1–10 Hz noise measurement conditions: bandwidth, filtering, and “zero input” connection method.
- Bias current and leakage vs temperature and input voltage: include worst-case curves when available.
- EMI behavior evidence: RF rectification tests or application conditions relevant to long cables.
- Overload recovery: overload amplitude/duration and time-to-within-accuracy after clamp conduction.
- Common-mode limits on single-supply: identify regions where protection or internal structures conduct.
No part-number recommendations are provided; selection is driven by traits and validation under realistic conditions.
Verification & production checklist: what to test and how to log it
Temperature AFEs fail in the real world when verification only checks a single “noise number”. A complete gate must cover temperature, supply, EMI, and fault injection, and it must produce logs that make drift/noise/steps diagnosable across production and field returns.
A. Verification tests (minimum set)
- Open-wire detection: thermocouple open, RTD sense open (2/3/4-wire), Kelvin sense open.
- Short/miswire: terminal short, pair short, swapped sense leads.
- Log: fault flag, recovery time, residual offset after recovery.
- 50/60 Hz + harmonics: validate both electric-field coupling (cable pickup) and magnetic coupling (loop area).
- Cable touch/move test: verify reading stability when cable position changes.
- Log: interference level, window/averaging settings, residual ripple at output.
- Bidirectional sweep (up and down) to expose hysteresis and gradient-driven offsets.
- Soak stability check: confirm readings return after gradients settle (not just after reaching setpoint).
- Log: board/terminal/CJC temperature, soak time, drift slope, step changes.
- Supply ripple and droop: verify no offset steps when rails shift within expected limits.
- Power-mode transitions and excitation duty changes: verify no baseline steps and predictable settling.
- Log: supply mode, transition event, settling time, step magnitude (if any).
- Input overvoltage/reverse events: force clamp conduction and verify time-to-within-accuracy after release.
- RTD excitation faults: over-current/over-voltage and recovery to normal resistance reading.
- Log: overload amplitude/duration, recovery time, residual offset shift.
- Measure immediately after stress and after a delay to catch slow drift shifts.
- Verify whether coefficients remain valid or a re-calibration gate is required.
- Log: event count, drift trajectory post-event, pass/fail with the same rules used in production.
B. Production log schema (minimal fields)
- Traceability: serial, lot/date, PCB revision, fixture ID.
- Conditions: temperature point, soak time, supply mode, excitation mode (I/V, duty tier), update rate/window.
- Calibration: coeff version, calibration executed (Y/N), before/after delta.
- Key metrics: input-referred offset, drift slope, low-frequency noise (window RMS), recovery time.
- Flags: open/short, clamp conduction, out-of-range, ESD/EFT event, retest count.
A pass/fail rule is only useful when it can be reproduced with the same logging context in both production and field investigations.
C. Failure triage: drift vs noise vs steps (where to look first)
- Check thermal paths first: terminal block, copper isothermal area, proximity to heat sources, airflow changes.
- Check reference/CJC domain: CJC placement and coupling, reference drift mapping, ground/return stability.
- Check leakage sensitivity: protection devices, contamination, humidity, bias return path stability.
- Check mains pickup: cable shielding/grounding, loop area, RC bandwidth vs windowing strategy.
- Check digital coupling: update timing, communication bursts, ground bounce paths.
- Check source-R/bias interaction: bias return and protection leakage behavior under temperature.
- Check clamp conduction and recovery: TVS/diode conduction events and time-to-stable baseline.
- Check state transitions: excitation duty changes, power-mode switches, mux/range actions.
- Check contacts: terminal resistance changes, connector thermoelectric effects, mechanical movement.
D. Project gate checklist (P0/P1/P2)
- P0 must: open/short behavior, overload recovery, mains pickup under cable motion, temperature sweep (up/down), supply mode switching.
- P1 recommended: ESD/EFT post-event accuracy, humidity/leakage sensitivity screening, long soak stability at extremes.
- P2 optional: stress aging/thermal cycling, extended EMC campaigns beyond the product environment.
E. Field log subset (for fast diagnosis)
- Version: firmware, coeff version, hardware revision.
- Temperature: at least one meaningful sensor (board or terminal/CJC).
- State: supply mode, excitation mode and duty tier, update rate/window.
- Outputs: raw, filtered, compensated readings.
- Events: open/short flags, overload/clamp flags, ESD/EFT counters, retest triggers.
Reference parts (MPNs) for fixtures & verification setups (examples)
These MPNs are examples to describe fixture-grade categories (connectors, TVS/ESD, precision references). Selection for production must consider availability, temperature range, package, and qualification.
- Thermocouple connector (Type-K, mini): OMEGA SMP-K-M (miniature plug).
- TVS diode (example): Littelfuse SMF5.0A (SMF series, unidirectional option).
- ESD protection diode (example): Nexperia PESD1CAN (note: some listings mark NRND; use a current alternative for new designs).
- Precision reference resistor (fixture / ratiometric reference): VPG Foil Resistors Y407810K0000V9L (10 kΩ foil, example grade).
Request a Quote
FAQs: thermocouple & RTD AFE op-amp design (CJC, excitation, drift, noise)
These FAQs close common long-tail questions without expanding the main scope. Answers focus on practical checks for thermocouple/RTD front-ends: op-amp traits, CJC, excitation/self-heating, low-frequency noise, cables/grounding, protection leakage, and drift triage.
Tip: If a symptom changes with airflow, start with thermal gradients. If it changes with cable movement, start with mains pickup and grounding. If it appears as steps, start with recovery and state transitions.
