123 Main Street, New York, NY 10001

Submit Your BOM

Seismic / Vibration / Low-Frequency Instrumentation Amplifiers

Seismic / Vibration / Low-Frequency Instrumentation Amplifiers

← Back to:Instrumentation Amplifiers (INA)

Low-frequency “fidelity” is not a single noise number: it is the combined result of 1/f noise, drift/warm-up, and thermal/grounding realities across the full chain (sensor → INA → ADC → reference). This page shows how to budget, measure, and pass/fail each contributor so sub-Hz signals stay trustworthy in real wiring and production builds.

What “Low-Frequency Fidelity” Means in Seismic / Micro-Vibration

Low-frequency fidelity is not judged by a single “RMS noise” number. It is the ability to preserve real signal content inside a defined frequency window and time window, while keeping drift, 1/f noise, thermal-gradient artifacts, and mains spurs below measurable pass criteria.

Define the observation window (frequency + time)

Frequency window
  • mHz–10 Hz: long-term stability dominates (thermal gradients, slow drift, leakage).
  • 0.1–100 Hz: vibration + mains lines become prominent; phase and recovery behavior matter.
  • Always state the window explicitly; “low-noise” without bandwidth is not actionable.
Time window
  • Warm-up: define “settled” as a band around the final value (not a fixed time guess).
  • Short-term averaging: choose the averaging time τ that minimizes the combined noise + drift effect.
  • Long-term: specify duration (e.g., 1–24 h) for drift, thermal sensitivity, and repeatability checks.
Requirement sentence template (copy/paste)
Measure ΔX within [flow, fhigh] over T, with pass criteria defined for: 0.1–10 Hz p-p, drift/°C, warm-up, and Allan σ(τ).

Separate noise vs drift vs interference (by observable signatures)

Random noise
Appears as a continuous floor in FFT; decreases with averaging; evaluated by band-limited RMS and p-p in a defined window.
Quick tell: remove trend → remaining amplitude scales with averaging time.
Structured error (drift / thermal EMF / leakage)
Appears as a slow trend in time logs; often correlates with temperature gradients, humidity, and warm-up. Not fixed by “more averaging”.
Quick tell: trend slope changes with airflow, board temperature, or handling.
Interference (mains / ground loops / RFI)
Appears as discrete spurs (50/60 Hz and harmonics) or bursts; strongly depends on wiring geometry and grounding.
Quick tell: rotate/move cables → spur amplitude changes immediately.

Practical success metrics (measurable + pass criteria form)

Metric: 0.1–10 Hz noise (peak-to-peak)
How to measure: band-limit to 0.1–10 Hz; remove DC trend; compute p-p over a defined window length.
What it tells: usable resolution for slow signals (captures 1/f behavior better than wideband RMS).
Pass criteria form: p-p(0.1–10 Hz) < Npp,target
Metric: drift vs temperature (offset or gain)
How to measure: apply controlled temperature steps; log output; fit slope after settling; repeat with reversed airflow direction.
What it tells: sensitivity to thermal gradients and component tempco; key for multi-hour measurements.
Pass criteria form: |dV/dT| < DT,target
Metric: warm-up settling time
How to measure: power-cycle; track output until it stays inside a stability band for a continuous duration.
What it tells: time needed for thermal equilibrium; prevents “first minutes drift” from corrupting data.
Pass criteria form: |ΔV| < Wband after Twarm
Metric: Allan deviation σ(τ)
How to measure: compute σ(τ) from long logs; identify τ where σ(τ) is minimized; confirm stability across repeats.
What it tells: whether white noise, 1/f, or random-walk drift dominates; guides the optimal averaging time τ.
Pass criteria form: min σ(τ) at τopt < Atarget
Diagram: Target windows + success metrics (frequency + time)
Low-frequency target windows and success metrics A frequency axis with two target bands and compact badges for 0.1–10 Hz noise, drift per degree, warm-up, and Allan deviation. mHz 0.1 Hz 10 Hz 50 Hz 60 Hz 100 Hz mHz–10 Hz 0.1–100 Hz mains Success metrics 0.1–10 Hz p-p drift / °C warm-up Allan σ(τ) Frequency (log scale conceptual) + time-window driven validation

End-to-End Error Budget: From Sensor → INA → ADC → Digital

A production-grade low-frequency front end is an end-to-end system. The measured output is the sum of sensor physics, wiring realities, INA behavior, filtering choices, ADC/reference limits, and digital windowing. A usable budget ties every error term to a measurable signature and assigns an owner (sensor, cable, analog, ADC/ref, or DSP).

Error taxonomy that maps to measurements

Random noise
Signature: continuous FFT floor; band-limited RMS and p-p in 0.1–10 Hz. Owner: INA input noise + sensor/source noise + ADC noise.
First check: short input → measure baseline (system noise floor).
Structured error (drift / thermal EMF / leakage)
Signature: slow time trend; temperature slope; humidity sensitivity; warm-up curvature. Owner: thermal gradients, materials, protection leakage, reference drift.
First check: thermal step (airflow or controlled ΔT) → observe correlated output shift.
Interference (mains / ground loops / RFI)
Signature: discrete spurs (50/60 Hz + harmonics), bursts, or step changes when cables move. Owner: wiring geometry, shielding, grounding, nearby fields.
First check: rotate/move cable loop area → spur amplitude should track geometry.

Budget table fields (measurable + owner + risk)

Each row below is designed to be filled from bench data. Numbers are intentionally expressed as targets (placeholders) to avoid false precision before the system-level requirement is fixed.

1) Sensor / source
Term: sensitivity + source impedance
Measure: known stimulus → output scale; sweep source R if configurable.
Risk: wrong scaling hides real resolution limits.
Term: sensor temp sensitivity
Measure: controlled ΔT; log slope after stabilization.
Risk: thermal changes masquerade as seismic drift.
2) Wiring / connection
Term: loop area + shielding
Measure: move/rotate cable; compare 50/60 Hz spur amplitude.
Risk: mains lines dominate low-frequency window.
Term: leakage paths (humidity/flux)
Measure: humidity step or controlled cleaning; observe DC shift and drift slope.
Risk: picoamps become millivolts at high impedance.
3) INA front-end
Term: 0.1–10 Hz noise + 1/f corner
Measure: short input; band-limit; compute p-p and/or Allan σ(τ).
Risk: low-frequency window fails even if wideband looks quiet.
Term: offset/gain drift vs temperature
Measure: ΔT step; fit slope after settling; repeat for airflow reversal.
Risk: slow trends bury micro-vibration content.
4) Analog filtering / notch (if used)
Term: corner frequency + phase impact
Measure: step response + frequency sweep; verify group delay remains acceptable in-band.
Risk: “clean” spectrum with distorted time-domain interpretation.
5) ADC + reference
Term: low-frequency ADC noise + reference drift
Measure: short at ADC input; compare with reference toggles/thermal steps.
Risk: reference behavior sets the floor even with a perfect INA.
6) Digital windowing / logging
Term: detrending + τ selection
Measure: compare metrics with/without detrend; sweep τ; verify Allan minimum aligns with stable operation.
Risk: wrong windowing mislabels drift as “noise” (or the reverse).

Fast diagnosis order (minimum steps, maximum separation)

Step 1 — Short input
Goal: lock the baseline floor; separate sensor/wiring issues from electronics.
Step 2 — Spectrum scan
Goal: identify discrete mains spurs vs continuous floor; prioritize interference fixes first.
Step 3 — Trend check
Goal: detrend logs; quantify drift slope and warm-up behavior using fixed rules.
Step 4 — Thermal step
Goal: correlate output shifts with airflow/ΔT; isolate thermal-gradient artifacts.
Step 5 — Leakage/humidity probe
Goal: validate contamination/leakage by controlled cleaning and humidity changes.
Diagram: End-to-end chain + dominant error ownership
Sensor to digital chain with error ownership A block diagram showing Sensor, Wiring, INA, Filter/Notch, ADC+Reference, and DSP with labels for dominant error types and a legend for noise, drift, and interference. Sensor tempco Wiring loops INA 1/f + drift Filter phase ADC + Ref LF floor DSP / Logging windowing detrend Legend noise drift spurs

Low-Frequency Noise: 1/f Corner, 0.1–10 Hz, and How to Integrate It

Low-frequency noise must be expressed in a windowed way: a frequency window (e.g., 0.1–10 Hz) plus a time window (log length). Wideband density alone is not sufficient, because 1/f behavior and mains spurs can dominate the usable resolution. The goal is to convert datasheet terms (noise density and 1/f corner) into measurable, pass/fail metrics.

Model the spectrum, then integrate to the band result

Key terms (datasheet → model)
  • en,white (nV/√Hz): the flat (white) noise floor at higher frequencies.
  • fc (Hz): the 1/f corner where low-frequency noise begins to rise.
  • [f1, f2]: the validation band (example: 0.1–10 Hz).
  • Twindow: the time window used to compute p-p and Allan σ(τ).
Band integration (conceptual forms)
RMS from PSD: Vrms2 = ∫f1→f2 ein,total(f)2 · df
Peak-to-peak must declare time window: Vpp(0.1–10 Hz, Twindow) < Npp,target
RMS summarizes energy; p-p captures the tail risk of 1/f noise in slow measurement windows. Both are required for low-frequency fidelity.

Combine INA, source, and ADC noise in an input-referred budget

Input-referred total noise (band-wise)
ein,total(f) = √( eINA(f)2 + esource(f)2 + eADC,in(f)2 + … )
Use input-referred terms to compare contributions fairly. Output noise follows: eout(f) = Gain · ein,total(f).
Dominance rules (fast identification)
  • Short-input baseline sets the electronics floor (INA + ADC + reference path).
  • If noise grows strongly with Rsource, source noise and leakage paths are likely dominant.
  • If spectrum shows strong lines at 50/60 Hz, treat them as interference spurs, not “noise density”.
  • Use the same band limits and logging rules for all comparisons (frequency window + Twindow).

Measurement recipes (FFT/PSD + time-domain statistics)

Recipe A — FFT / PSD (floor + spurs)
  1. Run two conditions: short input and representative Rsource.
  2. Compute PSD; mark 0.1–10 Hz band and 50/60 Hz lines.
  3. Record: floor level, spur amplitude, and 1/f rise (corner proximity).
Pass forms: spur@50/60 < Spurtarget; floor(band) < Floortarget
Recipe B — Time domain (RMS + p-p + detrend)
  1. Band-limit to 0.1–10 Hz and apply detrend (remove slow slope).
  2. Compute RMS and p-p over a declared Twindow.
  3. Optionally compute Allan σ(τ) to identify τ that minimizes combined noise + drift.
Pass form: Vpp(0.1–10 Hz, Twindow) < Npp,target
Reproducibility fields (must be logged)
Fs record length window detrend band limits grounding state airflow / ΔT

Common traps (why low-frequency results look “mysterious”)

  • Wideband density-only decisions: a great nV/√Hz number can still fail 0.1–10 Hz due to 1/f rise.
  • RMS-only reporting: p-p over long Twindow can exceed limits even when RMS looks small.
  • Spurs mixed into “noise”: 50/60 Hz and harmonics must be treated as interference and bounded separately.
  • Unlogged processing: filter/decimation/detrend changes the metric; results become non-comparable.
  • Baseline not established: always capture the short-input floor before blaming the sensor or cable.
Diagram: Noise spectrum (1/f + white floor) + integration window
Noise spectrum with 1/f corner and 0.1–10 Hz window A conceptual PSD plot showing 1/f rise at low frequency, a white noise floor, markers for fc, 0.1 Hz and 10 Hz, and a shaded integration window. 0.1 Hz 10 Hz fc 1/f white floor Outputs RMS p-p spurs σ(τ)

Drift & Warm-Up: Offset/Gain Drift, Aging, and What Calibration Can’t Fix

Drift is a time-dependent error that does not necessarily improve with averaging. Warm-up behavior is a drift subtype driven by thermal settling. A production-ready low-frequency design must: (1) quantify drift with repeatable tests, (2) define warm-up with a stability band, and (3) apply calibration only where coefficients remain stable across temperature, time, and assembly variations.

Drift sources (expressed as measurable fields)

  • Input stage mismatch tempco: manifests as dV/dT after thermal settling.
  • Reference drift: correlates with reference temperature or supply state changes.
  • Bias current vs temperature: amplified by high source impedance; appears as a slow offset shift.
  • Materials, stress, humidity: causes hysteresis and long-term aging drift; repeatability degrades across re-assembly.
Drift reporting (recommended)
dV/dT dV/dt warm-up curve hysteresis repeatability

Warm-up: define it by a stability band (not by guesswork)

Stability band definition
Warm-up passes when |ΔV(t)| < Wband continuously for Thold
This definition prevents “early-time drift” from being mistaken asм: warm-up is a thermal settling and material stabilization problem.
Minimal warm-up test (repeatable)
  1. Power-cycle under a fixed environment; log output for a fixed duration.
  2. Compute ΔV(t) relative to the final reference segment.
  3. Report Twarm when the stability band is met (include Wband and Thold).
  4. Repeat with airflow direction reversed to expose thermal-gradient sensitivity.

Calibration policy: what it can fix vs what it cannot

Recommended calibration ladder
  • 2-point (offset/gain): use when coefficients are stable across temperature and time.
  • Multi-point / LUT: use only when the error surface is repeatable and measurement uncertainty is well below targets.
  • Periodic recalibration: required when aging and assembly stress shift coefficients beyond guardbands.
Post-cal pass forms: residual drift < Drifttarget; repeatability < Reptarget
Coefficient stability checks (must be proven)
  • Across ΔT: same calibration repeated at multiple temperatures yields bounded coefficient variation.
  • Across Δt: re-test after soak and time delays; coefficients should not drift beyond a defined limit.
  • Across assembly: reconnect/reattach cables and fixtures; coefficient scatter must stay within guardband.

Calibratable vs must be engineered down (hard boundary)

Often calibratable (when repeatable)
  • Stable offset and gain constants
  • Repeatable temperature-dependent linear terms
  • Fixed-scale errors that track known references
Not reliably calibratable (must be reduced in hardware)
  • Thermal-gradient EMFs that change with airflow direction
  • Ground-loop / mains pickup spurs and wiring-geometry sensitivity
  • Humidity/flux leakage drift at high impedance nodes
  • Assembly stress / connector hysteresis causing non-repeatable offsets
Diagram: Drift timeline (power-up → settle → environment → aging)
Drift and warm-up timeline with calibratable and non-calibratable regions A time-axis diagram showing power-up warm-up, stable region, environment temperature step, and long-term aging, with overlays indicating calibratable versus non-calibratable drivers. Power-up Settled Env ΔT Aging calibratable (repeatable coefficients) engineer down (thermal / wiring / leakage) measured drift air ΔT cable Time (define T_window, T_warm, and long-term soak explicitly)

Thermal-Gradient Control: Thermocouple EMFs, Heat Paths, and Layout Tactics

In low-frequency instrumentation, thermal gradients can generate µV-level false signals through thermocouple EMFs. This is not “noise density” and does not reliably average out. The practical goal is to make the input region isothermal and symmetric, keep heat sources away, and validate sensitivity with repeatable disturbance tests.

Thermal EMF is a “ΔT → µV” converter

Where it shows up
  • Connector / terminal joints: dissimilar metals plus local ΔT create DC/low-frequency offsets.
  • Solder joints: uneven copper and airflow create a persistent gradient.
  • Input network: asymmetric parts or copper makes +/− inputs see different thermal conditions.
How to detect (field signatures)
  • Airflow direction changes the offset sign or slope.
  • Hand proximity causes repeatable steps or slow ramps near the input region.
  • Local heating near one input leg creates differential drift even when common-mode stays stable.

Heat-path priority (fix A before B before C)

Priority A — input asymmetry
  • Make +/− input copper and parts mirror-symmetric.
  • Match vias, lengths, and copper area around both input legs.
  • Keep the connector-to-input path balanced on both sides.
Priority B — heat source coupling
  • Move regulators, ADCs, and digital hotspots out of the input isothermal zone.
  • Route high-current traces away from the input area; avoid local copper bottlenecks.
  • Use keepouts and thermal barriers to prevent hot air paths across the input network.
Priority C — board-level thermal gradients
  • Balance copper density across layers to avoid one-sided thermal mass.
  • Provide controlled heat spreading to a “thermal sink region” away from the input.
  • Ensure mechanical fixtures and shields do not create new one-sided heat paths.

Layout tactics (action checklist)

  • Isothermal zone: keep the entire input network inside a symmetric thermal envelope.
  • Mirror routing: match copper area, vias, and distances on both input legs.
  • Heat spacing: place heat sources outside the input region; avoid hot airflow across inputs.
  • Heat spreading: use copper planes to spread heat away from sensitive joints rather than concentrating it.
  • Shielding / insulation: reduce random convection changes; validate that shielding does not introduce a new gradient.
Required logging fields (for comparable results)
airflow state distance ΔT step time window board orientation

Verification experiments (repeatable disturbances)

Experiment 1 — temperature step
Output: dV/dT and hysteresis under a declared ΔT and time window.
Pass form: |dV/dT| < DT,target
Experiment 2 — airflow (reverse direction)
Output: sign and magnitude change when airflow direction is reversed.
Pass form: airflow sensitivity < Aair,target
Experiment 3 — hand proximity
Output: repeatable step or ramp when a human heat source approaches the input zone.
Pass form: hand-induced step < Shand,target
Diagram: Thermal gradient map (heat paths → input EMF)
Thermal gradient map showing heat paths to the input region A block-style diagram with heat source, connector, solder joints, input network, and INA, plus heat-flow arrows and a ΔT to µV badge near the input. isothermal zone Connector terminals Solder joints Input network R / C / clamps INA Heat source LDO / ADC / MCU ΔT → µV keep away mirror-symmetric copper and parts reduce thermal gradients

Input Bias, Leakage, and Guarding: When Picoamps Become Millivolts

In high-impedance sensor wiring and humid environments, picoamp-level currents can create millivolt-level errors. Bias currents, protection leakage, and surface contamination appear as DC offset and slow drift. The practical workflow is: compute the risk with variables, map leakage paths, apply guarding and cleaning, then verify with humidity stress tests.

Variable math (risk can be computed)

Bias-current error
Verror ≈ Ib · Rsource
The same bias current produces much larger error when Rsource is large or changes with humidity.
Leakage error (surface + protection)
Verror ≈ Ileak · Req
Req includes surface resistance and the leakage behavior of clamps/ESD structures under humidity and temperature.

Leakage sources map (A/B/C layers)

Layer A — protection and components
  • ESD diodes / TVS reverse leakage into high-Z nodes.
  • Series resistors and pads that trap contamination near the input.
  • Connector insulation changes with humidity and residue.
Layer B — PCB surface and process
  • Flux residue, dust, and fingerprints reduce surface resistance.
  • Humidity and condensation create new parallel leakage paths.
  • Conformal coating helps only after cleaning is verified.
Layer C — return paths and geometry
  • Leakage returns that flow through the sensitive input region.
  • Guard discontinuities that let surface currents enter the node.
  • High-Z traces that run near copper edges or solder mask openings.

Guarding playbook (what “good guarding” looks like)

  • Closed guard ring around the high-Z node, with continuous copper (avoid breaks).
  • Clear guard reference: guard is tied to a defined potential near the node, not to noisy returns.
  • Short high-Z routing: minimize exposed surface length and avoid solder mask openings.
  • Keepouts: keep silkscreen, test pads, and unrelated copper away from the guarded area.
  • Return routing: route protection return currents around the high-Z region (not through it).
Manufacturing hooks (must be controlled)
cleaning residue test coating humidity stress IR test

Pass criteria (humidity and insulation verification)

  • Pre/post humidity offset shift: ΔVoffset(pre/post) < Otarget.
  • Insulation resistance: Rinsulation > Rtarget under declared test voltage/time.
  • Slope change: Δ(dV/dt) < Starget after soak and recovery.
  • Reconnect repeatability: re-plug scatter < Spreadtarget.
Required test condition fields
RH% T soak time test V recovery fixture
Diagram: Guard ring top view (high-Z node + leakage paths + return)
Guard ring layout around a high-impedance node A top-view diagram showing a high-Z node surrounded by a guard ring, leakage arrows from contamination, and a controlled return path away from the node. humidity residue High-Z node input Guard ring Return Leakage divert keepout: avoid pads / silkscreen / copper edges near high-Z

Mains & Ground Loops: 50/60 Hz Rejection That Works in Real Wiring

Low-frequency measurement systems often fail on 50/60 Hz spurs and ground-loop currents, not on the INA’s datasheet CMRR. The practical objective is to stop CM-to-DM conversion created by real cable impedance imbalance and uncontrolled shield/return paths. Use a repeatable workflow: break the loop, then decide shield termination, and only then apply filtering.

DC CMRR vs AC CMRR (why “paper CMRR” collapses)

Real-world failure mechanism
  • Common-mode mains + ground potential difference creates large CM voltage on long wiring.
  • Imbalance in cable / connector / input networks turns CM into DM (CM → DM).
  • The result is a narrow spur (50/60 Hz and harmonics), not broadband noise.
Field signatures (quick recognition)
  • Spur amplitude changes when the cable is moved or rotated.
  • Spur changes when ground point is moved (chassis vs system vs acquisition).
  • Shorted input significantly reduces the spur, indicating wiring dominance.

Shield termination decision (use loop-current controllability)

If loop current is NOT controllable
Long runs, multiple grounds, unknown chassis bonds, or large ground potential differences.
Prefer: single-end shield termination (break the loop first)
If loop current IS controllable
Solid bonding, low impedance between grounds, controlled chassis/system reference, and verified low loop current.
Option: two-end shield termination (only after verification)

3-step workflow (break loop → shield → filter)

Step 1 — break the loop
  • Use a single reference point temporarily and move the ground point.
  • Short inputs or use a dummy source to separate electronics vs wiring.
  • Track spur bins: A50/A60 and harmonics.
Step 2 — set shield termination
  • Compare single-end vs two-end with the same cable routing.
  • Rotate and move the cable; log ΔApose.
  • Confirm the shield carries noise return without entering the signal reference.
Step 3 — apply filtering last
  • Filter cannot recover from front-end saturation or clipping.
  • Validate notch and LP/HP choices with step and recovery tests.
  • Pass form: mains spur < Spurtarget in the observation band.

Verification mini-suite (repeatable)

  • Input short test: isolate electronics baseline from wiring pickup.
  • Cable pose test: rotate/move cable and measure ΔApose.
  • Ground point sweep: move ground reference and track Amains.
Pass forms: Amains < Spurtarget, ΔApose < ΔAtarget
Diagram: Ground loop + shield return map
Ground loop and shield return map for 50/60 Hz spur injection A block diagram showing sensor, cable shield, chassis earth, system ground, and ADC/INA ground with a highlighted loop current path and a CM to DM conversion marker. Sensor remote Cable twisted pair Shield INA + ADC acquisition Chassis System GND ADC GND Z+ Z− CM → DM loop current shield termination

Filtering for Low Frequency: HP/LP/Notch Choices Without Killing Phase

Stronger filtering is not automatically better in low-frequency measurement. High-pass choices can remove real ultra-low-frequency content and create long baseline recovery. Notches can reduce mains spurs but may introduce group-delay ripple and long settling. The practical method is to decide by headroom and saturation risk first, then validate phase and recovery with tests.

High-pass corner planning (drift removal vs signal preservation)

  • Define the observation floor: fsignal,min (mHz/0.01 Hz/0.1 Hz).
  • Pick fHP to keep amplitude and phase distortion acceptable in-band.
  • Validate baseline recovery: settle-to-band time < Ttarget.
Acceptance fields
f_signal,min f_HP settle time phase

50/60 Hz notch placement (analog vs digital)

If mains spur threatens headroom
When the front-end may clip/saturate, a purely digital notch cannot restore lost information.
Prioritize: wiring fixes + front-end headroom, then minimal analog suppression
If headroom is sufficient
When the signal chain stays linear, a digital notch can be stable and adjustable with predictable behavior.
Option: digital notch, validated by group delay and settling tests

Filter selection tree (if/else)

  • If mains spur risks saturation → fix wiring/loops first, then consider minimal analog suppression.
  • Else → prefer digital notch; validate notch depth and side-lobes in-band.
  • If ultra-low-frequency content must be preserved → set fHP low and enforce a recovery-time limit.
  • If phase coherence matters → constrain group-delay ripple in the observation band.
Pass forms: settle time < Ttarget, group delay ripple < GDtarget, spur < Spurtarget

Test checklist (do not trust curves alone)

  • Step response: overshoot and settle-to-band time.
  • Group delay estimate: in-band ripple and notch neighborhood.
  • Notch shape: depth, width, and side-lobe impact on nearby frequencies.
Minimum reporting fields
f_HP f_LP notch settle GD ripple side lobes
Diagram: Magnitude + phase risk map (HP + LP + notch)
Combined HP, LP, and mains notch with group-delay risk zone A simplified frequency-response diagram showing HP corner, LP corner, a 50/60 Hz notch, and a highlighted group-delay risk zone near the notch. Magnitude f_HP 50/60 f_LP Phase / GD GD risk near notch headroom settling phase

ADC + Reference Pairing for Low-Frequency Integrity

Low-frequency fidelity is a system property: INA noise/drift, ADC low-frequency behavior, reference noise/drift, and the sampling + statistics window combine into the final observable floor. The selection goal is not a single “best part” number, but a consistent measurement definition and a repeatable validation plan that matches the observation band.

ADC low-frequency noise (use one consistent definition)

  • 0.1–10 Hz p-p: describes visible slow fluctuation in the observation band.
  • Shorted-input noise: isolates chain baseline from sensor wiring and interference.
  • Digital filtering/decimation: changes the effective time constant and window correlation; verify with the intended window length.
Minimum reporting fields
0.1–10 Hz p-p shorted noise filter mode window length

Reference pairing (noise + drift are injected through real paths)

  • Reference noise sets an output floor when it is not correlated or not canceled by the measurement strategy.
  • Reference drift + thermal sensitivity appears as long-term baseline shift; layout thermal symmetry matters.
  • Remote sense (if used) must be validated for wiring-induced errors and unintended return paths.
Reference selection fields (system-facing)
noise drift warm-up thermal sense

Sampling + statistics (window definition controls the result)

Oversampling & decimation
Use OSR and decimation only with a defined observation window; verify that the chosen filter mode does not create unacceptable recovery time.
Window statistics
Define window length, overlap, and detrending policy; separate slow drift from noise and spurs using consistent reporting fields.
Low-frequency anti-alias meaning
Band-out energy that reaches the ADC can fold into the observation band; validate with spur bins and headroom checks instead of assumptions.

Selection fields (ADC + reference + system) & risk mapping

Must-have fields
  • ADC: LF p-p, shorted noise, filter/decimation modes
  • REF: noise, drift, warm-up, thermal sensitivity, sense option
  • System: observation band, window length, headroom, spur targets
Risk mapping (what to suspect first)
  • Shorted LF p-p fails → baseline is insufficient; wiring is not the first suspect.
  • Warm-up drift dominates → define effective measurement time after power-up.
  • Sense option used → validate loop/shield paths and spur behavior with the same window.
Pass forms: LF p-p < Ptarget, drift < Dtarget, spur < Spurtarget

Validation methods (minimum experiment set)

  • Experiment A — input short: record LF p-p and spur bins in the intended observation window.
  • Experiment B — constant input, long run: record warm-up behavior and drift slope over time.
  • Experiment C — reference thermal stimulus: apply a controlled ΔT/airflow and record dV/dT and hysteresis.
Report fields: window length, filter mode, LF p-p, drift slope, Amains, dV/dT
Diagram: Noise + drift composition (INA + ADC + REF + window)
Noise and drift composition at the digital output A block diagram showing INA noise, ADC noise, reference noise, and a statistics window converging to LF p-p and drift outputs. INA noise ADC noise REF noise Output sum in the band window Digital output LF p-p drift use one definition

Application Patterns: Seismic, Micro-Vibration, Structural Monitoring

Application coverage is organized as patterns: each pattern defines a dominant failure mode, the key chain requirements, and the validation tests. The focus is on needs and acceptance fields (band, headroom, drift, spur, phase/recovery), not on circuit topologies.

Pattern entry (quick classification)

  • If the observation floor reaches mHz → use Pattern A (long observation).
  • If phase and waveform integrity in 1–200 Hz matters → use Pattern B (micro-vibration).
  • If low-frequency baseline + event trigger is required → use Pattern C (structural monitoring).

Pattern A — Ultra-low-frequency long observation (mHz–1 Hz)

Target
mHz–1 Hz, long-term baseline stability
Main enemy
drift, thermal gradients, mains spurs as spurs
Design hooks
thermal symmetry, defined windows, verified warm-up
Tests
24 h drift, window metrics, airflow sensitivity
Pass
drift < Dtarget, LF p-p < Ptarget, spur < Spurtarget

Pattern B — Micro-vibration / vibration monitoring (1–200 Hz)

Target
1–200 Hz, waveform amplitude + phase fidelity
Main enemy
group delay ripple, recovery time, fold-in spurs
Design hooks
phase-aware filtering, headroom, verified settling
Tests
step + recovery, sweep amplitude/phase, GD estimate
Pass
GD ripple < GDtarget, settle < Ttarget, spur < Spurtarget

Pattern C — Structural monitoring (LF baseline + event trigger)

Target
low-frequency baseline with event detection
Main enemy
baseline drift causing false triggers, recovery after events
Design hooks
detrend/HP policy matched to trigger logic, window definition
Tests
false-trigger rate, baseline recovery, threshold stability
Pass
false triggers < Ftarget, recovery < Ttarget, threshold drift < TDtarget
Diagram: Three pattern comparison (same template, different variables)
Three low-frequency application patterns comparison A three-column card diagram comparing long observation, micro-vibration, and structural monitoring with band, dominant issue, hooks, and tests. Pattern A mHz–1 Hz dominant: drift hooks: thermal tests: 24 h Pattern B 1–200 Hz dominant: phase hooks: GD tests: step Pattern C LF + event dominant: trigger hooks: baseline tests: FTR same template, different dominant terms

Engineering Checklist: Layout, Cabling, Thermal, and Verification Tests

This checklist is designed for low-frequency integrity reviews and sign-off. Each item includes a verification method and an acceptance field so design changes can be audited, reproduced, and released with guardband.

A) Layout checklist (highest priority)

Symmetry & isothermal placement
Keep differential input paths geometrically symmetric and thermally matched to reduce thermal-gradient artifacts.
Verify
Airflow/hand-near test and thermal step test; observe baseline shift and recovery.
Pass
|ΔV| < Vtherm (system budget), recovery < Trec
Guarding for high-impedance nodes
Add guard rings and keep high-impedance nodes away from flux residues, soldermask breaks, and humid airflow paths.
Verify
Humidity soak (or controlled mist) before/after offset comparison; insulation resistance check if available.
Pass
|ΔVoffset| < Vleak, Riso > Riso,target
Kelvin routing & return continuity
Maintain continuous returns under sensitive routing; avoid split-plane crossings and unintended current sharing in analog returns.
Verify
Ground move test (change single-point location); observe mains spur and baseline change.
Pass
Amains < Spurtarget, |Δbaseline| < Vloop
Protection placement vs leakage budget
Clamp/ESD parts and series resistors can introduce leakage and bias-induced offsets; place and size them with a leakage budget.
Verify
Temperature sweep; compare offset slope and bias-related drift with protection populated vs bypassed (if safe).
Pass
dV/dT < Starget, |VIb·R| < Vbudget

B) Cabling & shielding checklist

Cable type selection (twisted pair / coax / shielded pair)
Use geometry that controls loop area and maintains consistent impedance; prioritize repeatable routing and strain relief.
Pass
Cable movement produces no new spurs above Spurtarget in the observation window
Shield termination rule (single-end vs both-ends)
Use single-end termination when loop current is not controlled; use both-ends only when the return path is defined and verified.
Verify
Change shield termination and one ground point; compare Amains and baseline stability.
Pass
Amains reduction is repeatable; baseline shift stays < Vloop
Connector & contact material consistency
Keep dissimilar-metal junctions near sensitive inputs minimized and thermally matched to reduce thermocouple EMFs.
Pass
Thermal perturbation produces |ΔV| < Vtherm (budgeted)

C) Thermal checklist (control gradients, not just temperature)

Heat source isolation and spreading
Keep DC/DC, CPUs, and power resistors away from the input front-end; use copper spreading to reduce local ΔT.
Verify
Controlled power step on nearby loads; record baseline transient and settle time.
Airflow sensitivity (fan, vents, hand-near)
Treat airflow as a thermal gradient stimulus; validate with repeatable distance, duration, and direction.
Pass
|Δbaseline| < Vair, recovery < Tair

D) Verification tests checklist (sign-off oriented)

Shorted-input baseline
Record LF p-p, Amains, and spur bins with the intended filter mode and window length.
Pass: LF p-p < Ptarget, Amains < Spurtarget
24-hour drift
Separate warm-up interval from steady-state; report drift slope and repeatability.
Pass: drift < Dtarget, warm-up settle < Twarm
Temperature sweep slope
Sweep temperature across the intended range; report dV/dT and hysteresis.
Pass: |dV/dT| < Starget, hysteresis < Htarget
Repeat mounting / fixture repeatability
Re-clamp or re-install the sensor/fixture and re-run short-window statistics; confirm reproducibility.
Pass: repeatability error < Rrep, no new spurs above Spurtarget
Diagram: Verification loop (Plan → Measure → Diagnose → Fix → Re-test)
Verification loop flow for low-frequency measurement projects A five-step loop showing plan, measure, diagnose, fix, and re-test with short labels for each step. Plan window Measure shorted Diagnose spurs Fix one change Re-test same setup Keep window + conditions constant

IC Selection Logic (INA + ADC + Reference) for Low-Frequency Projects

Selection should follow a budget-driven order: define the observation band and dynamic range, then allocate noise and drift budgets, then verify interference behavior and layout constraints. Supplier data must include test conditions (temperature, source impedance, filter mode, and window).

1) Selection fields (request these with test conditions)

INA 0.1–10 Hz p-p 1/f corner offset/drift Ib(T) AC CMRR protection leakage
Required conditions: gain, input source R, input common-mode, temperature points, bandwidth definition, and time window length.
ADC LF noise shorted noise filter modes latency headroom
Required conditions: sample rate, output data rate, digital filter selection, window length, and spur bin definition.
REF noise drift warm-up thermal sense
Required conditions: load, buffering, temperature profile, and physical placement constraints (isothermal region).

2) Decision order (budget → verify → guardband → release)

  1. Observation band → define flow…fhigh, required window length, and dynamic range.
  2. Noise budget → allocate LF p-p and wideband density contributions to an input-referred budget.
  3. Drift budget → allocate warm-up settle and drift slope; define when a measurement is “valid”.
  4. Interference reality → verify mains spurs, ground loop sensitivity, and cable movement behavior.
  5. Guardband → reserve margin for process, temperature, humidity, and installation variability.
Diagram: Selection flow (Budget → Candidate → Verify → Guardband → Release)
Selection flow for low-frequency measurement chain A five-step flow from budget to candidate selection, verification, guardband, and release. Budget Candidate Verify Guardband Release Verify with the intended window + filter mode

3) Inquiry template (ask suppliers for comparable data)

  • Provide 0.1–10 Hz p-p with gain, bandwidth definition, filter mode, and window length.
  • Provide shorted-input noise under the same output data rate and digital filter setting.
  • Provide drift and warm-up definition: time axis, temperature points, and whether warm-up interval is included.
  • Provide AC CMRR vs frequency around 50/60 Hz and relevant harmonics for real wiring conditions.
  • Provide input protection leakage and bias behavior across temperature and humidity (typical and max if available).
  • Provide reference noise/drift with load and buffering conditions; specify any remote-sense constraints.

Reference examples (part numbers; starting points only)

These parts are listed to speed up datasheet lookup and lab verification. Final selection should follow the budget and verification flow above.

INA candidates
  • TI INA188 — zero-drift INA starting point for low-frequency precision.
  • TI INA333 — micropower zero-drift INA starting point for battery nodes.
  • TI INA828 — precision INA starting point when low noise and practical integration are needed.
  • ADI AD8421 — low-noise instrumentation amplifier family starting point.
ADC candidates
  • TI ADS1262 — precision ΣΔ ADC starting point for sensor measurement chains.
  • ADI AD7177-2 — ΣΔ ADC starting point for low-frequency DAQ configurations.
  • ADI AD7124-8 — integrated low-frequency measurement front-end style ADC.
  • TI ADS1282 — high-resolution ΣΔ ADC family often used in seismic-class systems.
Reference candidates
  • TI REF70 — low-noise, low-drift reference family starting point.
  • ADI ADR4550 — precision reference family starting point.
  • ADI LTC6655 — low-noise reference family starting point.
Passive / protection examples
  • VPG Foil resistors — starting point category for ultra-low TCR / thermal EMF sensitive designs.
  • TI TPD4E1B06 — ESD protection device starting point when leakage must be managed (verify in-system).
  • BAV199 — low-leakage diode class starting point for clamp concepts (verify over T/H).

Request a Quote

Accepted Formats

pdf, csv, xls, xlsx, zip

Attachment

Drag & drop files here or use the button below.

FAQs: Seismic / Vibration / Low-Frequency Fidelity

Short, actionable answers that close long-tail field questions without expanding the main content. Each item follows the same 4-line structure: Likely cause → Quick check → Fix → Pass criteria.

Why does the output drift for the first minutes after power-up?
Likely cause: Warm-up thermal settling changes internal offsets and PCB gradients, creating a predictable baseline ramp.
Quick check: Log baseline vs time after power-up under a fixed setup; repeat with airflow/cover change to see time-constant shifts.
Fix: Isolate heat sources, enforce isothermal symmetry near inputs, and gate “valid data” until warm-up criteria is met.
Pass criteria: Within Twarm, |V(t) − Vfinal| < Vsettle for Thold, then drift_rate < Dtarget.
How to tell drift from 0.1–10 Hz noise in logs?
Likely cause: Drift is a slow trend (slope/curvature), while 0.1–10 Hz noise is window-local fluctuation around that trend.
Quick check: Compute (a) slope over Twin and (b) p-p/RMS after detrending in the same window; compare dominance.
Fix: Separate reporting into “trend” and “noise” metrics; suppress drift via thermal control and wiring, not via heavier low-pass.
Pass criteria: |slope| < Dtarget and p-p(0.1–10 Hz) < Ptarget using the defined Twin.
Why does touching/moving the cable change the reading?
Likely cause: Cable motion changes ground-loop current, shield coupling, and connector micro-thermal gradients, injecting sub-Hz baseline error and mains spurs.
Quick check: Compare (1) input shorted vs real sensor, and (2) shield termination A vs B; observe Δbaseline and A50/60.
Fix: Control the return path (single-point ground), choose a stable cable geometry (twisted/shielded), and add strain relief; minimize dissimilar-metal junctions.
Pass criteria: Under a defined “touch/bend” stimulus, Δbaseline < Vcable and added spurs < Spurtarget.
What is a practical way to verify 50/60 Hz rejection on the bench?
Likely cause: Real wiring introduces impedance mismatch and parasitic coupling that bypasses datasheet CMRR.
Quick check: Measure A50/60 with (a) input shorted at the connector and (b) sensor connected; repeat while moving only the ground point.
Fix: Break uncontrolled loops, enforce shield termination rules, and restore symmetry in both signal and return paths before adding any notch.
Pass criteria: A50/60 < Spurtarget and ΔA50/60 under ground-point change < ΔSpurtarget.
Guard ring: when is it mandatory and how to validate it worked?
Likely cause: High source impedance plus humidity/contamination creates leakage currents that appear as DC offset and slow “drift”.
Quick check: Compare offset before/after humidity soak (or controlled mist) and after cleaning; look for large, repeatable ΔVoffset.
Fix: Add guard ring around high-Z nodes, increase creepage, enforce cleaning process, and consider conformal coating where needed.
Pass criteria: After soak, |ΔVoffset| < Vleak and Riso > Riso,target (or equivalent leakage limit).
Why do “great CMRR specs” collapse in real wiring?
Likely cause: Source-impedance mismatch and parasitic capacitance asymmetry convert common-mode into differential error, especially around mains frequencies.
Quick check: Introduce a controlled ΔRsource or change cable routing; observe A50/60 sensitivity and baseline stability.
Fix: Restore symmetry (routing + returns), reduce mismatch exposure (3-op-amp INA is typically more tolerant), and control shield/ground paths.
Pass criteria: With ΔRsource within expected field tolerance, Amains < Spurtarget and Δbaseline < Vloop.
How to pick a high-pass corner without losing sub-Hz content?
Likely cause: An overly aggressive high-pass removes drift but also distorts amplitude/phase near the lowest signal band.
Quick check: Test a known sub-Hz stimulus (or replayed waveform) and measure gain/phase error at flow plus step recovery time.
Fix: Set fHP based on required flow and allowable phase error; prefer drift reduction at the source (thermal/wiring) before raising fHP.
Pass criteria: |Gain_error(flow)| < Gtarget and |Phase_error(flow)| < φtarget, recovery < Trec.
How to budget thermocouple EMFs from connectors/solder joints?
Likely cause: Dissimilar-metal junctions under a temperature gradient generate microvolt-level EMFs that look like real low-frequency signals.
Quick check: Apply a controlled thermal stimulus (airflow/heat step) and correlate baseline shift with measured ΔT near connectors and input joints.
Fix: Minimize dissimilar junctions near inputs, enforce isothermal symmetry, and move thermal sources away; use copper spreading to reduce ΔT.
Pass criteria: Under ΔTstep, |ΔV| < Vtherm,budget and the shift is reduced after symmetry/isothermal fixes.
Does chopper/zero-drift always win for seismic signals?
Likely cause: Zero-drift improves offset/drift, but system performance can still be limited by wiring, thermals, and spur artifacts inside the observation window.
Quick check: Measure spectrum and time logs in the intended mode; look for fixed spurs or window-dependent artifacts versus pure noise reduction.
Fix: Choose parts by the budget (LF p-p + drift + spur limits), then validate with the full wiring/thermal setup rather than relying on “zero-drift” labels.
Pass criteria: In the observation window, p-pLF < Ptarget, drift_rate < Dtarget, and spurs < Spurtarget.
How much warm-up time is “enough” and how to define it quantitatively?
Likely cause: Warm-up “time” depends on thermal paths and load conditions, so fixed minutes are not transferable.
Quick check: Define Vsettle and evaluate the first time the baseline enters and stays within the band for Thold.
Fix: Implement a validity gate based on settle criteria; reduce warm-up by moving heat sources and improving isothermal symmetry.
Pass criteria: Twarm is the earliest time when |V − Vfinal| < Vsettle for Thold and then drift_rate < Dtarget.
Why does humidity or flux residue create a DC offset that looks like drift?
Likely cause: Surface contamination and moisture form leakage paths; the resulting bias currents create slow baseline changes that mimic drift.
Quick check: Compare offset before/after cleaning and a humidity stimulus; confirm correlation with humidity level and high-Z node location.
Fix: Enforce cleaning/handling, add guarding and creepage, and apply conformal coating when the environment is uncontrolled.
Pass criteria: After humidity exposure, |ΔVoffset| < Vleak,budget and the result is repeatable across builds/boards.
What pass criteria proves “production-ready low-frequency fidelity”?
Likely cause: Production failures usually come from uncontrolled wiring/thermal/leakage variability, not from typical datasheet numbers.
Quick check: Run a minimal sign-off set: shorted-input noise, A50/60, 24h drift, temperature slope, and repeat mounting.
Fix: Lock the setup (window + filter mode + wiring), add guardband, and require test conditions in supplier data and internal reports.
Pass criteria: p-pLF < Ptarget, Amains < Spurtarget, drift < Dtarget, |dV/dT| < Starget, repeatability < Rrep.
icnavigator

ICNavigator helps engineers find verified ICs, alternatives, and fast quotes.

Explore
Categories
  • Power Management
  • MCU
  • Automotive ICs
  • Sensors
  • Interface & Transceivers
  • Analog Front-End
Get in Touch
  • Email: hello@icnavigator.com
  • WeChat/WhatsApp: +86-xxxx

© 2025 ICNavigator. All rights reserved.