Vibration / IEPE Conditioning (Signal Chain & Sync Sampling)
← Back to: Test & Measurement / Instrumentation
IEPE vibration conditioning is about delivering a stable constant-current bias, removing the DC offset cleanly, and preserving low-noise, flat, time-aligned signals into synchronized sampling. Done right, it prevents hidden clipping, long shock recovery, and cable-induced hum/spikes so FFT and coherence results remain trustworthy.
H2-1 · What IEPE conditioning solves (scope & outcomes)
IEPE conditioning turns a “sensor + cable” into a measurement-ready channel that can be trusted for FFT, time-domain events, and multi-channel coherence. The goal is not to describe the sensor type, but to define what modules are required and how to prove the channel is correct in bandwidth, noise, timing, and diagnosability.
The “3 jobs” of an IEPE input chain
- Constant-current excitation: bias the IEPE sensor correctly while keeping enough compliance/headroom for the expected vibration amplitude.
- Bias removal / signal coupling: separate the DC bias from the AC vibration signal without losing required low-frequency content.
- Measurement-grade capture: low-noise differential amplification, anti-alias filtering, and synchronized sampling for accurate amplitude and phase.
Acceptance outcomes (how “done” is verified)
- Usable dynamic range: no clipping at max expected vibration; adequate margin to ADC full-scale.
- Low noise floor: channel noise stays well below the smallest spectral lines of interest.
- Bandwidth & flatness: amplitude response meets the target band with controlled ripple/roll-off.
- Phase integrity: group delay is predictable; phase distortion is compatible with the analysis method.
- Channel-to-channel timing: sampling skew is small enough for coherence/phase comparisons.
- Diagnosability: open/short/saturation/cable faults can be detected and separated from real vibration.
Page navigation (the 6 questions this page answers)
Figure F1 shows the minimum set of blocks needed to make an IEPE vibration channel measurable, comparable across channels, and diagnosable in the field.
H2-2 · IEPE sensor electrical model (what the front-end must respect)
An IEPE sensor presents a powered, low-impedance output that carries vibration as an AC waveform riding on a DC bias. Conditioning must preserve the AC content while keeping the DC operating point in a healthy range and maintaining enough compliance voltage so the current excitation remains constant under real vibration amplitudes and cable loading.
Minimal working model (the only equations needed for design decisions)
- Output composition: VOUT(t) = VBIAS + vAC(t) (bias is a DC operating point; vibration is the AC component).
- Compliance requirement (conceptual): the excitation source must keep enough voltage headroom so the current remains constant across bias + AC swing + cable drop. If compliance is insufficient, the current “collapses” during peaks and the waveform clips or distorts.
- Cable capacitance effect: cable capacitance draws AC current at higher frequencies (and can interact with the driver loop), altering high-frequency amplitude/phase and creating stability risk if the excitation/conditioning front end is not designed for it.
Three “windows” that define correct operation
- Bias window: VBIAS should land in a normal range; drift high/low is an early fault clue (sensor electronics, wiring, or excitation problems).
- Headroom window: the remaining swing for vAC(t) before clipping at either end of the chain (sensor → coupling → amplifier → ADC).
- Compliance window: the maximum voltage the constant-current source can provide while maintaining the specified current into the sensor + cable.
Symptom → likely cause mapping (fast field diagnosis)
- Clipped peaks / sudden harmonic rise at high vibration → compliance or headroom is insufficient; current is no longer constant during peaks.
- High-frequency roll-off or phase mismatch that worsens with longer cables → cable capacitance loading dominates; driver stability or filter placement needs review.
- Intermittent spikes when the cable is touched or moved → cable microphonics / poor connector shielding / mechanical stress coupling into the signal path.
- Bias out of expected range or “stuck” → open/short wiring, sensor electronics failure, or excitation fault; treat bias as a health indicator.
Parameter card (8 terms worth knowing before designing hardware)
- Bias voltage (VBIAS)
- DC operating point used for sensor health checks; abnormal bias often correlates with wiring or sensor electronics faults.
- Constant current (mA)
- Sets sensor excitation and affects headroom, dissipation, and noise coupling; higher current is not automatically better.
- Compliance voltage
- Maximum voltage the current source can develop while maintaining the target current; if exceeded, distortion and clipping appear under load.
- Headroom
- Remaining usable signal swing before any stage saturates (sensor, coupling network, amplifier, or ADC input range).
- Cable capacitance (Ccable)
- Dominant cable loading that influences high-frequency amplitude/phase and can destabilize the driver if not accounted for.
- AC coupling / HPF corner
- Bias removal is typically AC coupling; the high-pass corner sets the lowest recoverable vibration frequency and impacts overload recovery time.
- Overload recovery
- How quickly the channel returns to valid readings after a shock/overrange event; coupling networks often dominate this behavior.
- Fault modes (open/short/sat)
- Health diagnosis relies on bias and excitation behavior to separate wiring faults, sensor failure, and real vibration events.
Figure F2 emphasizes that IEPE design decisions come from three constraints: bias window, headroom window, and the constant-current compliance window, plus the frequency-dependent cable loading that can reshape amplitude and phase.
H2-3 · Constant-current excitation design (noise, compliance, stability)
In an IEPE channel, the constant-current source is not “just power.” It sets the sensor operating point (bias), defines available headroom for the vibration waveform, and can dominate the noise floor or cause distortion if compliance collapses under peak signal swing or cable loading. A robust design therefore starts from three constraints: setpoint current, compliance budget, and stability/noise.
Key trade-offs that must be decided first
- Current setpoint (typically 2–20 mA): higher current can improve robustness against cable loading, but increases dissipation and can reduce usable headroom depending on the sensor’s bias behavior. The best setpoint is the smallest current that keeps the sensor biased correctly across temperature and cable length.
- Compliance voltage: the supply must cover the sensor bias plus the required AC swing (peak vibration) and any protection/series drops. If compliance runs out during peaks, the excitation current stops being constant and the waveform distorts.
- Noise & stability: current-source noise appears as output voltage noise through the sensor/cable impedance. Cable capacitance becomes a heavy load at high frequency and can reduce phase margin, so stability must be checked for the worst-case cable.
What “insufficient compliance” looks like (field-observable symptoms)
- Peak clipping or “flattened” waveform at high vibration levels, even when the ADC range is not yet full.
- Sudden rise of harmonics in FFT (distortion increases rapidly once compliance collapses).
- Bias behaving non-stationary (bias shifts or droops during peaks), indicating the current source is no longer regulating.
Noise coupling: why “current noise” becomes “voltage noise”
Any noise in the excitation current is converted into a voltage disturbance across the sensor/cable impedance, which then rides on the measured signal path. This is why a current source can set the noise floor even if the amplifier and ADC are excellent. At higher frequency, cable capacitance changes the effective impedance and can reshape both amplitude and stability margins—so the noise and loop behavior must be evaluated for the expected cable range, not only a short bench cable.
Design criteria checklist (copy-ready)
- Setpoint range: supports sensor requirements (2–20 mA typical) with margin over temperature and supply tolerance.
- Compliance budget: supply covers bias + peak AC swing + any series/protection drops, with headroom margin at maximum amplitude.
- Noise target: excitation noise does not dominate channel noise floor in the analysis band (especially where coherence/FFT is sensitive).
- Worst-case cable stability: stable regulation for maximum cable capacitance and expected EMI environment; no oscillation or ringing.
- Startup & hot-plug: controlled ramp/limit prevents overshoot and false fault flags when sensors are connected/disconnected.
- Multi-channel consistency: channel-to-channel current accuracy and temperature drift are tight enough to keep bias/headroom comparable.
Implementation paths (selection by criteria, not by “best topology”)
- Op-amp + transistor closed-loop CCS: flexible compliance and current setting; focus on loop stability with cable capacitance and protection networks.
- Dedicated current regulator IC: simpler implementation; verify noise floor and compliance range match the sensor/cable use case.
- Programmable current (DAC-controlled): enables per-channel trimming or mode switching; ensure the programming path does not inject noise into the analog chain.
See also: Calibrator / Reference Source (link label only).
Figure F3 compares common IEPE constant-current approaches. The correct choice is the one that meets compliance, noise-floor, and worst-case cable stability requirements with predictable startup behavior.
H2-4 · Bias removal & coupling network (keep LF, protect headroom)
IEPE signals arrive as an AC waveform riding on a DC bias (often around 8–14 V). The coupling network must remove the DC component so the amplifier/ADC stays in range, while preserving the lowest vibration frequencies that matter (speed-related content and structural modes) and maintaining predictable overload recovery after shocks or accidental overrange.
Three-step design method (practical and testable)
- Define the lowest frequency to preserve (based on the measurement goal: speed/ordering vs modal content).
- Set the HPF corner so low-frequency attenuation and phase shift are acceptable for the analysis method.
- Verify overload recovery: after a large shock or input overrange, the channel returns to valid readings within a defined settling time.
Bias-sense path (health diagnosis without corrupting the signal)
- Purpose: monitor the DC bias as a health indicator (open/short/sensor electronics issues often show up as abnormal bias).
- Non-interference rule: the bias-sense input impedance must be high enough that it does not change the coupling RC or load the IEPE output.
- Noise containment: bias monitoring should be slow and well-filtered so it does not inject digital noise into the vibration band.
Overload recovery (why “settling time” becomes a measurement limit)
Shock events or accidental overrange can push the coupling capacitor and amplifier stages into saturation. Recovery is governed by the discharge path and by how quickly internal nodes return to their linear region. A good design defines an explicit settling requirement (e.g., “valid within X seconds after a Y-level shock”) and validates it with repeatable injection tests.
Figure F4 separates the IEPE DC bias from the AC vibration waveform, adds a high-impedance bias-sense branch for diagnostics, and highlights overload recovery as a first-class acceptance requirement.
H2-5 · Differential / instrumentation amplifier stage (CMRR, noise, headroom)
After bias removal, the amplifier stage defines how much of the sensor’s usable information reaches the ADC without being buried by noise, corrupted by common-mode coupling, or clipped during shocks. A correct design balances three things: input-referred noise, real-world CMRR, and headroom + overload recovery.
Noise budgeting: sensor noise vs amplifier input noise (why gain placement matters)
- Input-referred noise is the most useful metric for amplifier comparison because it tells how much noise is added before any gain. If amplifier noise dominates the total noise floor in the band of interest, moving gain earlier helps ADC utilization but raises overload risk.
- Gain placement should be decided by a measurable goal: make typical vibration signals occupy a healthy portion of ADC full-scale without saturating during peak events, while keeping the combined noise floor below the smallest spectral lines that matter.
- Practical rule: increase early gain only until the amplifier noise is no longer the dominant floor (then stop and keep headroom).
CMRR in the real channel: why datasheet CMRR can collapse
- Impedance imbalance: any mismatch between the +/− input paths (protection parts, coupling networks, connector/contact resistance) converts common-mode interference into differential error, directly reducing effective CMRR.
- Shield/return coupling: shield termination that creates unintended return currents can inject common-mode noise into the input network, making the amplifier “see” interference as differential content.
- Symptom mapping: channel-to-channel differences in a fixed hum tone (e.g., 50/60 Hz) usually indicates input path mismatch or return coupling, not an amplifier spec problem.
Headroom, common-mode range, and overload recovery (do not ignore shocks)
- Input common-mode range must cover the post-coupling operating point over temperature and tolerances; otherwise small disturbances can cause saturation.
- Output swing + ADC range: reserve margin so peak vibration or cable microphonics do not clip the amplifier before the ADC reaches full-scale.
- Overload recovery: specify and validate settling time after overrange. Slow recovery often dominates “time to valid data” in real measurements.
Decision tree (gain & noise placement) — copy-ready
- Set ADC utilization target: typical vibration should occupy a strong portion of full-scale while preserving peak-event headroom.
- Compare floors: if amplifier input-referred noise would dominate, place more gain earlier; if sensor noise dominates, keep early gain moderate.
- Check CMRR risks: verify symmetry of input parts and routing; treat mismatch as “CMRR killer.”
- Validate recovery: inject an overrange event and confirm the channel returns to valid readings within a defined settling time.
Limiter/buffer can be added only if it does not create leakage/capacitance imbalance that harms noise, bandwidth, or CMRR.
Figure F5 shows how sensor noise and amplifier input-referred noise set the achievable floor, and how gain placement trades ADC utilization against overload margin.
H2-6 · Input protection & survivability (ESD/OVP, hot-plug, miswire)
Field vibration setups see hot-plugging, cable swaps, accidental miswires, and ESD events. Protection must prevent damage and false readings, but protection parts can easily degrade noise and bandwidth. The correct approach is to apply a clear priority order: survivability first, then performance, then diagnosability.
Protection part side effects (what they secretly break)
- Leakage: creates bias drift and low-frequency errors, especially visible as slow baseline movement or “false vibration” at very low frequency.
- Capacitance: reduces high-frequency bandwidth, adds phase shift, and can upset stability when combined with cable capacitance.
- Clamp dynamics: introduces distortion near large-signal peaks, raising harmonics and corrupting FFT measurements.
- Asymmetry: unequal protection on +/− paths converts common-mode into differential error (effective CMRR loss).
Hot-plug and open-cable handling (compliance rise must be controlled)
- Soft-start / ramp control: reduces transients during connection and prevents sudden bias jumps.
- Limit voltage excursions: keep the IEPE node from rising uncontrollably when the sensor is open (compliance rise).
- Fault latch + reporting: explicitly classify open/short/overrange so protection events are not mistaken for real vibration.
Priority card (copy-ready)
- Survivability: survive ESD/OVP/miswire without damage and without permanent parameter drift.
- Performance: minimize added capacitance/leakage; keep +/− path symmetry to protect CMRR and phase integrity.
- Diagnosability: bias and fault flags must clearly distinguish wiring faults from real vibration events.
Channel-level isolation (scope-limited)
Protection should not create shared return paths that couple one channel into another. Keep protection and sensing local to each channel, and avoid asymmetrical loading that turns common-mode interference into differential errors.
Figure F6 shows a channel-level protection approach that survives hot-plug and ESD/OVP while explicitly calling out the capacitance/leakage trade-offs that can degrade noise, bandwidth, and effective CMRR.
H2-7 · Anti-alias filtering strategy (cutoff, order, phase)
Anti-alias filtering in a vibration/IEPE channel is not only about “blocking out-of-band.” It must protect FFT amplitude accuracy near band edge and preserve time-domain alignment for impacts and modal work. A practical design therefore starts from sampling rate, required analysis bandwidth, then allocates transition band, order, and phase/group-delay behavior.
Cutoff vs sampling rate (fs): leave a transition band on purpose
- Never place cutoff too close to Nyquist. Real filters need a transition band; pushing cutoff up forces high order and increases phase distortion.
- Budget amplitude & phase error at the top of the analysis band. If amplitude accuracy near band edge matters, keep ripple/attenuation small in-band and ensure sufficient attenuation before Nyquist to suppress foldback.
- Order is not a trophy: higher order helps stopband attenuation but increases group-delay variation and can worsen overload recovery if stages saturate.
Filter type logic for vibration analysis (Bessel vs Butterworth)
- Bessel: smoother group delay and more time-domain friendly. It is typically preferred when impacts, time alignment, and waveform fidelity (modal/impulse work) are important.
- Butterworth: flatter magnitude response in-band, often preferred when steady-state spectral amplitude accuracy is the priority and group-delay variation is acceptable.
- Practical selection: if downstream methods rely on precise time alignment, prioritize group delay; if the goal is stable magnitude spectra, prioritize amplitude flatness.
Distributed filtering across stages (avoid overload + avoid phase surprises)
- Early gentle filtering: tame out-of-band energy so later stages and ADC inputs do not see large high-frequency content that can cause clipping.
- Final anti-alias stage near ADC: define the channel’s official bandwidth and attenuation into Nyquist, where aliasing is created.
- Consistency across channels: use matched filter structures/values per channel to keep phase and group delay aligned for multi-channel analysis.
Text-based selection guide (no formula dump)
- Impulse / time-domain alignment → prioritize smooth group delay → typically Bessel-like behavior.
- Steady-state spectral magnitude → prioritize amplitude flatness → typically Butterworth-like behavior.
- Near-Nyquist content matters → increase transition-band margin and verify attenuation before Nyquist to suppress foldback.
- Large out-of-band energy exists → distribute filtering so no single stage overloads or rings excessively.
Figure F7 compares the practical trade-off: Butterworth-like responses favor magnitude flatness, while Bessel-like responses favor smoother group delay.
H2-8 · Sync sampling & time alignment (skew, aperture, triggering)
Multi-channel vibration work (phase, coherence, modal shapes, order tracking) assumes channels represent the same physical timebase. Two timing errors dominate: channel-to-channel skew (sample timing mismatch) and aperture jitter (uncertainty of the sampling instant). Skew produces deterministic phase error; jitter raises the high-frequency noise floor.
Two timing errors and what they do
- Skew (Δt between channels) → phase mismatch for the same tone. The effect grows with frequency, so high-frequency phase/coherence is the first to degrade.
- Aperture jitter → noise floor rise at higher frequency. Even with perfect skew, jitter turns fast edges/high-frequency content into added noise.
Acquisition boundary: simultaneous-sampling vs MUX scan (scope-limited)
- Simultaneous-sampling ADC: best for phase/coherence work because all channels capture the same instant by design.
- MUX scanning: acceptable for slower trends, but phase comparisons become ambiguous as the channel sampling instants are inherently different.
Deeper trigger/backplane routing details belong to the Trigger/Marker & Event Routing page (link label only).
Error → symptom → verification (practical acceptance checks)
Figure F8 shows how a small channel-to-channel timing offset creates phase error on the same waveform; the effect becomes more severe as frequency increases.
H2-9 · Cable, grounding & shielding (real-world failure modes)
IEPE vibration channels often look clean on the bench but degrade in the field when longer cables, moving harnesses, and nearby power equipment are introduced. Most real-world failures fall into three buckets: cable capacitance loading, shield termination / ground-loop coupling, and microphonics from cable motion. The goal is not “EMC theory,” but repeatable rules and fast fault localization.
Cable capacitance loading (why long cables change bandwidth and stability)
- Capacitance is a frequency-dependent load. As cable length increases, the effective capacitance rises, which can bend high-frequency amplitude/phase and make the anti-alias budget harder to meet at the top of the band.
- Capacitance can amplify “edge problems”. Combined with protection and input network capacitance, it can create ringing or occasional spikes when events occur (plug/unplug, motion, nearby switching noise).
- Field clue: performance loss appears first near the band edge (phase/coherence) and scales strongly with cable length.
Shield termination and ground loops (what “wrong shield” looks like)
- Rule of thumb: avoid letting the shield become a signal return path. A shield that carries unintended return currents can inject low-frequency interference and create hum/drift that varies with equipment placement and power outlet choice.
- Effective CMRR can collapse when the +/− paths are not symmetric (connector resistance, protection parts, coupling networks). Common-mode noise then converts to differential error, showing up directly in the measured spectrum.
- Field clue: touching the cable/shield or the chassis changes the noise floor; moving the cable near a motor drive/VFD changes the hum pattern.
Microphonics (motion-induced spikes that look like real impacts)
- Cable motion can generate false events. Friction, bending, and loose connectors can produce impulse-like spikes that contaminate impact tests and inflate broadband noise in FFT results.
- Repeatable isolation method: fix the cable routing, then move one segment at a time. If spikes correlate with a specific bend or connector, the issue is mechanical coupling, not electronics.
- Field clue: spikes appear when the cable is touched, flexed, or rubs against structures; fixing the cable reduces events immediately.
Field troubleshooting card (symptom → check order)
Keep this page channel-scoped: only IEPE cable/port behaviors are addressed here, not facility-wide grounding or EMC theory.
Figure F9 highlights where noise couples in practice: cable capacitance, shield return loops, and motion-induced microphonics.
H2-10 · Calibration & self-test hooks (gain/phase, reference injection, fault detect)
A maintainable IEPE channel must be able to self-prove its gain, phase/time alignment, bias health, and noise floor without relying on external guesswork. The most effective design pattern is to add a reference injection path, plus minimal bias/fault sensing, and then define pass/fail criteria that can be checked on demand or at service time.
Calibration targets (engineering-acceptance view)
- Gain: channel-to-channel gain consistency and drift control (verify across a small set of frequencies and amplitudes).
- Phase / time alignment: estimate per-channel delay/skew from injected in-phase reference, then correct or flag out-of-range.
- Bias window health: confirm IEPE bias is within a valid range; use it to detect open/short and intermittent wiring faults.
- Noise floor: measure in-band noise metric to catch leakage/capacitance damage, aging, or shield/return issues that raise floor.
Self-test hooks (reference injection + fault detection)
- Reference injection: inject a known small signal at one or multiple frequencies so the channel response can be verified end-to-end (or partial-chain, depending on injection point).
- Open/short detect: use bias window and simple fault flags to classify wiring faults and prevent “false vibration” interpretations.
- Noise floor self-check: measure a quiet-input condition (or defined injected condition) and compare to a stored baseline.
Injection point choice is a coverage trade-off: earlier points include more of the analog chain; later points isolate cable/sensor effects.
Calibration flow (5–7 steps)
- Enter service mode: hold measurement output stable and enable injection path.
- Select stimulus: choose one or more reference tones/levels that cover the analysis band.
- Capture response: acquire synchronized samples across channels under identical conditions.
- Compute metrics: gain error, phase/Δt, bias window, and noise-floor metric.
- Decide pass/fail: compare against thresholds and either store trims or flag the channel.
- Exit mode: disable injection and confirm normal measurement path is restored.
- Record summary: store minimal channel results for maintenance traceability.
Pass / fail criteria (channel-minimum)
- Gain: within allowed % error and stable across the chosen reference points.
- Phase / skew: Δt estimate within allowed limit for the highest analysis frequency.
- Bias window: bias value inside valid range; open/short classified cleanly.
- Noise floor: in-band metric below threshold compared to baseline/reference condition.
Minimal record fields (IEPE channel only)
- Cal version / timestamp (service traceability).
- Gain trim (or gain error metric at the reference point).
- Phase / Δt estimate (per channel, relative to reference channel).
- Bias value + status (OK / open / short / intermittent).
- Noise floor metric + status (OK / high / unstable).
- Last fault code (channel-scoped only).
Figure F10 shows a practical reference injection path with selectable injection point and isolation considerations for gain/phase and health checks.
H2-11 · Debug checklist & acceptance tests (prove it’s done)
This acceptance checklist turns an IEPE channel into measurable outcomes. The test order is intentional: (1) excitation & bias health must pass before (2) noise & headroom, before (3) frequency response & phase, before (4) synchronization. If a later test fails, return to the earliest prerequisite step instead of “tuning blindly.”
A) Electrical health — excitation, compliance, bias window, open/short
- IEPE current (I_IEPE): verify per-channel current value and channel-to-channel matching.
- Compliance margin: confirm headroom at worst-case bias + max AC swing (no clamp/rail hit).
- Bias window: confirm bias stays within the sensor-valid range and remains stable.
- Open/short classification: force OPEN and SHORT conditions and confirm correct detection and safe behavior.
- Measure bias voltage at the port (or bias-sense node) with a high-impedance meter/ADC channel.
- Use a known-good IEPE sensor and a known-good short cable as a baseline, then swap to the field cable.
- For OPEN/SHORT: use a controlled plug/fixture to avoid accidental damage and to reproduce consistently.
- I_IEPE: within ±2% (or tighter if multi-channel matching is required).
- Compliance margin: ≥ 2 V margin at worst-case bias + peak AC (avoid intermittent clipping).
- Bias window: within sensor-valid range (common field window is 8–14 V, but follow sensor datasheet).
- Open/short: detection occurs within 1 s and the channel enters a defined safe state (no latch-up, no uncontrolled rail hit).
B) Noise & dynamic range — noise floor, hum, spikes, headroom to full-scale
- In-band RMS noise: measure within the declared analysis band (example: 10 Hz–20 kHz).
- Noise spectral density: verify 50/60 Hz + harmonics, broadband floor, and discrete spurs.
- Spurious spikes: check for motion-induced events (microphonics) and clamp-trigger artifacts.
- Headroom: confirm max expected vibration + shock events do not clip the analog chain or ADC.
- Quiet condition: use a short cable and a stable sensor mount; avoid cable motion during baseline measurement.
- Spectrum view: use the same windowing and bandwidth settings each run; record settings with the result.
- Headroom test: apply a controlled injected sine (or shaker level) and step amplitude until near-clip; record the clip threshold.
- Noise (10 Hz–20 kHz): ≤ 10 µVrms (example baseline for a low-noise channel; adjust by sensor sensitivity and target g-resolution).
- Mains hum: ≤ −80 dBFS equivalent at 50/60 Hz (or “not visible above broadband floor” for a defined FFT span).
- Headroom: ≥ 6 dB margin to clip at the maximum declared operating vibration level.
C) Frequency response & phase — flatness, group delay, channel phase error
- Amplitude flatness: verify within the declared band (relative to a mid-band reference frequency).
- Group delay consistency: verify trend is stable and channel-to-channel matched.
- Phase error vs frequency: inject the same in-phase tone into all channels and compute relative phase.
- Multi-tone sweep: use the reference injection path (H2-10) to step through a small set of frequencies (low/mid/high band).
- Same cable condition: keep cable length and routing fixed during FR/phase tests to avoid capacitance and motion changing results.
- Match filters: ensure all channels use identical AA filter components/configuration before comparing.
- Flatness: within ±0.5 dB across the analysis band (or a tighter band if calibration is applied).
- Group delay match: channel-to-channel variation ≤ 1% of a sample period (starting target for time alignment work).
- Phase error: ≤ 2° at the top of the analysis band (starting target; depends on modal/coherence requirements).
D) Synchronization — skew measurement, calibration, residual error
- Skew (Δt) estimation: compute from phase vs frequency using a shared in-phase stimulus.
- Post-calibration residual: apply correction and verify residual Δt stays within limit.
- Repeatability: repeat across power cycles (and temperature points if possible) and confirm stability.
- Use simultaneous-sampling ADC when phase/coherence is required; avoid MUX scan for strict phase acceptance.
- Inject one tone near the upper band and one tone in mid-band to make Δt visible and stable.
- Compute Δt per channel relative to a selected reference channel and store results with calibration metadata.
- Skew (pre-cal): record the measured value (do not hide it).
- Skew residual (post-cal): ≤ 50 ns (starting target for many vibration coherence use cases; tighten if needed).
- Repeatability: Δt change ≤ 20 ns across repeated runs (starting target; depends on clocking/jitter regime).
Common failure reasons → fastest localization order
- Noise/hum appears only with long cable → check shield/return and cable motion → then check cable capacitance loading.
- Spikes that look like impacts → reproduce by bending/handling cable → then check clamp/protection dynamics.
- FR/phase mismatch between channels → verify AA filter component matching → then verify gain staging and overload recovery.
- Coherence/phase breaks at high frequency → verify skew/jitter regime → then apply skew calibration and re-check residual.
Concrete part numbers (examples for implementing acceptance hooks)
These examples support the acceptance workflow (injection, switching, bias sensing, simultaneous sampling). Final selection depends on bandwidth, noise target, and fault requirements.
- ADI AD7768-1 (multi-channel simultaneous sampling ΣΔ ADC family option).
- ADI AD7606B (8-ch simultaneous-sampling DAS class device option).
- TI ADS131M08 (8-ch simultaneous-sampling 24-bit ΔΣ ADC option).
- ADI AD9833 (DDS tone generator option for stepped-frequency injection).
- TI DAC80501 (precision DAC option for controlled-level injection).
- ADI AD5686R (multi-channel precision DAC option for calibration stimuli).
- ADI ADG5412F (fault-protected analog switch for safer injection routing).
- ADI ADG5408F (fault-protected mux option for channel-level switching).
- Omron G6K (small-signal relay family option for hard short / routing fixtures, when mechanical switching is preferred).
- Panasonic TQ2 (signal relay family option for controlled routing/shorting fixtures).
- ADI LT3092 (programmable current source building block option).
- TI REF200 (precision current source option for bias/health circuits where applicable).
- TI OPA189 (zero-drift op amp option for bias sensing/health measurement paths).
- ADI AD8421 (instrumentation amplifier option for differential measurement front-ends).
Note: part numbers above are examples to make the acceptance hooks concrete; they are not mandatory.
Record fields (field name + unit + pass range) — ready for a DataModel later
| Field | Unit | Pass range (starting target) | Notes |
|---|---|---|---|
| IEPE_I_mA | mA | Target ±2% | Measure per channel; matching matters for multi-channel comparability |
| Bias_V | V | Within sensor-valid window | Store window bounds used for classification |
| ComplianceMargin_V | V | ≥ 2 | Evaluate at worst-case bias + peak AC swing |
| OpenShort_State | enum | OK / OPEN / SHORT / INTERMIT | Use bias + behavior to classify |
| Noise_uVrms_10_20k | µVrms | ≤ 10 (example baseline) | Tie to a declared bandwidth and measurement condition |
| Flatness_dB | dB | ±0.5 (starting target) | Relative to mid-band reference frequency |
| PhaseErr_deg_at_Fmax | deg | ≤ 2 (starting target) | Use shared in-phase stimulus; store the Fmax used |
| Skew_ns_pre | ns | Record only | Do not hide pre-cal spread; it explains coherence limits |
| SkewResid_ns_post | ns | ≤ 50 (starting target) | Residual after applying correction |
| CalVersion / CalTimestamp | — | Present | Traceability for service and regression comparisons |
ALT: Acceptance test flow for IEPE conditioning channels covering excitation, noise, frequency response, and synchronization.
Request a Quote
H2-12 · FAQs (IEPE Conditioning)
Each answer gives a decision rule, practical criteria, typical symptoms, and a quick verification step for real IEPE vibration channels.
