Pico/Femto Ammeter: Ultra-Low Leakage Current Measurement
← Back to: Test & Measurement / Instrumentation
Pico/femto-amp measurement is mainly about controlling “fake currents” from leakage, cable/fixture charging, and drift—not chasing more ADC bits. When guarding, triax wiring, settling rules, and humidity/cleanliness are done correctly, the remaining current reading becomes stable, repeatable, and trustworthy.
H2-1 · What a Pico/Femto Ammeter Really Measures
Core takeaway:
A pico/femto ammeter measures ultra-small current, but the real job is controlling everything that can impersonate current—surface leakage, cable charge, bias currents, and drift—at the input node. In this regime, “better resolution” matters less than “better leakage management” and a stable measurement method.
What “pA…fA” means in practice
Picoampere (10−12 A) to femtoampere (10−15 A) measurement is the domain where the measured signal is often comparable to parasitic currents created by humidity films, contamination, dielectric absorption, and mechanical motion of cables. That is why these instruments behave more like a controlled front-end system than a simple series ammeter.
Typical targets (what this page focuses on)
- Device leakage currents: off-state leakage, gate leakage, junction leakage.
- Sensor/AFE input bias currents: tiny DC currents that shift high-impedance nodes.
- Photodiode dark current: low-light detection and spectroscopy front-ends.
- Surface leakage on fixtures: contamination + humidity creating a resistive film across insulators.
Why a “normal current range” often fails here
At fA/pA levels, the measurement can be dominated by the instrument’s input environment (cables, connectors, insulators, and guard implementation), not the ADC digit count. A standard current range is designed for broader ranges and protection trade-offs; a pico/femto ammeter is designed to minimize the driving voltage across leakage paths and to keep the input node stable enough to separate true current from parasitics.
Two specs that should be “gatekeepers”
- Noise floor: the smallest current that can be distinguished after the chosen bandwidth/integration time.
- Error sources: leakage and offset mechanisms that add a DC (or slowly varying) current component.
Next section breaks the readout into these components and shows how “fake current” is created.
H2-2 · Current at fA Level: Where “Errors” Come From
At pico/femto levels, the display value is rarely “just the DUT current.” A useful way to keep the work engineering-driven is to treat the readout as a sum of true current plus several parasitic components that behave like current.
A practical readout model
Iread = Itrue + Ileak(surface) + Ibias + Itribo + Iabsorption + Ioffset/drift
The dominant terms depend more on fixture, humidity, cabling, and time history than on digit resolution.
Why “leakage management” outranks ADC resolution
The input node is extremely high impedance. A thin contamination or humidity film can create a parallel leakage path that is comparable to the DUT signal. Mechanical motion of cables can inject charge that looks like spikes or slow tails. Dielectric absorption can create memory effects where the current “relaxes” long after a range change or connection step.
Error tree cards (symptom → mechanism → quick check → fix)
1) Surface leakage (contamination + humidity)
- Symptom: reading is “always higher than expected,” often drifts with humidity.
- Mechanism: resistive film across insulators creates a current driven by small voltage differences.
- Quick check: compare open-input behavior after cleaning and drying; log humidity/temperature.
- Fix: clean fixtures/connectors, keep insulators dry, use guard rings to reduce the voltage across leakage paths.
2) Bias current + input offset drift
- Symptom: slow drift over minutes to hours; sign may stay consistent.
- Mechanism: front-end bias/offset changes with temperature and time; extremely visible at high source impedance.
- Quick check: measure shorted-input and open-input baselines to separate instrument drift from DUT behavior.
- Fix: allow thermal stabilization, use appropriate zeroing method, reduce thermal gradients at terminals and fixtures.
3) Triboelectric cable charging (motion-induced)
- Symptom: spikes or bursts when the cable is moved; sometimes followed by a decay tail.
- Mechanism: friction in dielectric generates charge; the front-end interprets charge flow as current.
- Quick check: hold cable still vs gently tapping or flexing and observe correlation to events.
- Fix: use low-noise triax, secure cables (strain relief), minimize motion and vibration near the DUT/fixture.
4) Dielectric absorption / “memory” effects
- Symptom: after a range change or connection step, current slowly relaxes toward a new value.
- Mechanism: stored charge in insulators and cable dielectric releases over time, mimicking current.
- Quick check: repeat the same step sequence; a reproducible tail suggests absorption rather than random noise.
- Fix: longer settling time, consistent pre-conditioning steps, and fixtures/materials designed for low absorption.
5) Switch/connector leakage and charge injection (preview)
- Symptom: jump at connection or range transition; the jump may depend on history.
- Mechanism: tiny leakage paths and charge injection at interfaces; visible because the node impedance is extreme.
- Quick check: compare behavior across different connectors/fixtures; watch if events align with switching steps.
- Fix: keep interfaces clean/dry; use low-leak switching methods and controlled transition/blanking windows (detailed later in auto-range section).
A minimal 3-step separation workflow (fast and repeatable)
- Step 1 — Shorted input: establishes instrument noise floor and drift baseline.
- Step 2 — Open input: exposes leakage, absorption, and environmental sensitivity in cabling/fixtures.
- Step 3 — Connect DUT: the remaining change is more likely dominated by true DUT current (still verify with settling and repeatability).
H2-3 · Front-End Architectures: Resistor TIA vs Charge Integrator
Decision goal:
Select the front-end that matches the signal shape and the tolerated “cost” (bandwidth, settling, reset artifacts, and sensitivity to cable/fixture capacitance). At fA/pA levels, architecture choice often dominates measurable performance more than display digit count.
Scope note: this section focuses on measurement front-ends (TIA vs integration). Sourcing/top-level SMU behavior is intentionally out of scope.
Resistor transimpedance amplifier (Rf-TIA)
Converts current to voltage continuously (conceptually: Vout follows I × Rf). This is the most “intuitive” path for tracking trends and steps.
- Strength: continuous readout; easier correlation with time events.
- Cost: Johnson noise of Rf and strict leakage requirements around the feedback network.
- Watch-outs: larger input capacitance and longer cables can raise noise gain and complicate stability/settling.
Charge integrator (Cf) / “charge method”
Converts current to accumulated charge over a defined window (conceptually: voltage ramps with I / Cf over time). The integration window provides built-in averaging.
- Strength: can reach a lower effective noise floor for very small currents by integrating longer.
- Cost: requires reset cycles; reset injection and dielectric absorption can create tails and memory.
- Watch-outs: update rate is limited by the integration window plus post-reset settling.
Selection boundaries (engineering language)
- If the goal is to observe continuous change or step response, then Rf-TIA is usually the first pick (accept higher sensitivity to Rf noise and leakage control).
- If the goal is the lowest practical noise floor for near-DC ultra-low current, then a charge integrator is often better (accept reset cycles, injection, and memory effects).
- If the input capacitance is large (long cable / large fixture), then TIA stability and noise gain become harder; integrator behavior is often more predictable but slower.
- If a “trustworthy number” is needed rather than a fast waveform, then prioritize integration time and stable baseline management (auto-zero/zero sequences).
- If range transitions are frequent, then compare how each path handles settling: TIA may settle after Rf switching; integrator must recover from reset injection + absorption tails.
Engineering comparison (not a price table)
| Dimension | Rf-TIA (resistor transimpedance) | Cf-Integrator (charge method) |
|---|---|---|
| Noise floor | Often limited by Rf Johnson noise + amplifier noise; leakage around Rf matters immediately. | Often improves by extending the integration window; random noise averages down, but drift/absorption can remain. |
| Bandwidth / response | Supports continuous tracking; bandwidth set by stability and filtering choices. | Update rate is set by window length + reset/settling; inherently “slow” when chasing the lowest noise floor. |
| Stability vs Cin | Large input capacitance can increase noise gain and complicate settling; cable/fixture management is critical. | Integration is less about loop bandwidth; still sensitive to dielectric behavior, but often more predictable than TIA at large Cin. |
| Reset / settling | No mandatory reset, but range switching and step changes can require settling time to trust the baseline. | Reset is mandatory; reset injection and dielectric absorption can create tails that need defined “trust windows.” |
| Best-fit targets | Leakage current trends, slowly varying bias/leak changes, cases where continuous monitoring is needed. | Ultra-low near-DC currents, dark current characterization, and applications prioritizing lowest stable average. |
Where auto-zero / chopping fits (brief)
These techniques exist to reduce baseline drift and low-frequency error terms that averaging cannot remove. They improve long-term trust in the zero and slow current readings without turning the section into a nanovoltmeter discussion.
H2-4 · Noise Floor Engineering: What Sets Your Resolution
Key point
Display digits are not the same as true resolution. The practical resolution is set by the noise floor and by low-frequency error terms (drift, absorption, environment). Averaging reduces random noise, but it cannot remove drift-driven limits.
What contributes to the noise floor (in engineering terms)
Rf thermal (Johnson) noise
The feedback resistor itself produces random noise. Larger Rf increases signal gain but also raises the resistor noise contribution.
Amplifier current noise (in)
Appears directly as an equivalent input current noise and can dominate at extremely low currents.
Amplifier voltage noise (en) × input capacitance
Input capacitance converts voltage noise into a frequency-dependent current term and can raise noise gain and stability cost.
1/f noise and baseline drift
Sets the “averaging stop point.” When drift dominates, longer averaging no longer improves the trusted resolution.
Bandwidth vs noise: the core trade
- Narrower bandwidth / longer integration reduces random noise and makes the readout calmer.
- Longer integration also slows response and can expose slow error terms (drift, absorption tails).
- Averaging helps when random noise dominates; averaging stops helping when baseline drift dominates.
A formula-lite way to think about it
- Noise density → RMS noise: integrate the effective current noise density over the measurement bandwidth to estimate RMS current noise.
- Integration time effect: increasing integration time reduces random noise, but does not remove drift or memory effects.
Practical interpretation: first choose the response time the application can accept, then optimize noise within that allowed bandwidth/time window.
Tuning knobs that actually move the outcome
- BW limit / digital filtering: reduces random noise but increases response delay.
- Averaging / integration window: improves stability until drift becomes dominant.
- Auto-zero / baseline management: targets low-frequency drift and improves long-term trust.
- Cin and cabling: cable length/fixture capacitance can raise noise gain and degrade settling stability.
- Range strategy: avoid unnecessary switching events that create injection and settling tails.
H2-5 · Guarding & Driven Shields: The Non-Negotiables
Why guarding exists
At pA/fA levels, many “mystery currents” are not from the DUT, but from leakage paths around the input node. Guarding works by driving nearby surfaces to nearly the same potential as the high-impedance node, making the leakage-driving voltage ΔV ≈ 0 so surface leakage collapses.
Guarding is not the same thing as shielding. Shielding blocks external fields; guarding removes the voltage that drives leakage.
Driven shield / triax roles (must be correct)
Center conductor = signal (HI)
Carries the measured current from DUT to the front-end input node.
HI signalInner shield = driven guard
Actively driven near the HI potential to reduce ΔV across surface paths.
Driven GUARDΔV ↓Outer shield = ground / noise shield
Provides EMI shielding and a stable reference boundary; not the same as guard.
Shield / GNDLO return ≠ guard
LO is the measurement return path; guard is a driven equipotential surface.
Do not shortGuard driver stability (symptom → fix)
| Symptom | Likely mechanism | Practical countermeasure |
|---|---|---|
| Reading “wobbles” periodically | Guard driver loop becomes marginal with large cable capacitance; phase margin drops. | Reduce guard bandwidth, add series resistance at guard output, use a defined compensation option. |
| Spikes after touching/moving cable | Tribo-electric charge and capacitance modulation; guard not behaving as a stable equipotential. | Use low-noise triax, immobilize cable, reduce motion near the input node. |
| Range changes create large tails | Charge redistribution and dielectric absorption interact with guarded surfaces and Cin. | Use blanking windows + settle checks after switching; avoid frequent toggling near thresholds. |
Fixture & mechanics: make leakage paths short, clean, and predictable
- Insulating supports: prefer stable, low-absorption materials and keep the high-impedance node physically separated from contamination-prone surfaces.
- Air gaps: avoid tight creepage across dirty surfaces; do not route the input node near flux residue or fingerprints.
- Guard ring around the node: surround the sensitive pad with a driven ring to collapse ΔV across surface films.
- Clean & dry: surface contamination + humidity creates a conductive film; cleaning and drying often gives the largest improvement-per-effort.
Wiring checklist
- Center conductor → HI
- Inner shield → GUARD (driven)
- Outer shield → Shield/GND (defined strategy; avoid accidental shorts to guard)
- Confirm GUARD is not tied to GND
- Keep the guarded path continuous from instrument to fixture
Setup checklist
- Fix cables and fixtures to minimize motion
- Let the setup thermally stabilize before trusting the baseline
- Keep hands, airflow, and moisture away from the input node
- Define a consistent “zero” routine (baseline management)
- Inspect connectors for contamination (especially triax interfaces)
Top 5 mistakes (and how they show up)
- Guard tied to ground → leakage stays high; readings track humidity and surface condition.
- Inner and outer shields shorted → driven shield becomes a passive shield; ΔV across surface films returns.
- Wrong cable type → motion-induced spikes, inconsistent baseline, “touch sensitivity”.
- Dirty/damp fixture → slow “creep” in the reading; large offsets after handling.
- Over-aggressive guard drive bandwidth → oscillation-like wobble or extra noise on the measurement.
H2-6 · Auto-Ranging Without Glitches: Range Switching Done Right
Why auto-range can “jump”
Range changes reconfigure high-impedance networks. Switching can inject charge and redistribute parasitic capacitances, creating a transient that looks like current. A robust auto-range system treats switching as a controlled event: switch → blank → verify settle → then report.
Goal: minimize switching events, prevent ping-pong near thresholds, and define a trustworthy post-switch window.
How ranges are implemented (in practice)
Rf decade ladder + low-leak relays
A stepped feedback network (Rf1/Rf2/Rf3…) is selected by low-leakage relays designed for high resistance paths.
Low leakageRepeatableSolid-state switching (use with care)
Leakage, injection, and temperature-dependent paths can dominate at pA/fA levels. If used, verification must focus on injection and tails.
Injection riskTemp pathsWhere glitches come from (physics, not software)
- Charge injection: the switch event deposits charge at the sensitive node, creating a spike.
- Capacitance redistribution: parasitic capacitances re-partition voltage, creating steps and overshoot.
- Dielectric absorption: the transient can leave a slow tail, which looks like a drifting current.
Stop ping-pong: overlap + hysteresis
Auto-ranging should avoid switching at the decision boundary. A hysteresis/overlap rule prevents repeated toggling: switch only when the reading has exceeded a threshold for a defined condition, and do not switch back until a separate return threshold is met.
Where settling time comes from (and what controls it)
| Contributor | What it looks like | Control knob |
|---|---|---|
| Input capacitance (cable / fixture) | Long tails and slow return after switching or motion; sensitivity to handling. | Shorten/immobilize cables, reduce fixture capacitance, keep guarding continuous. |
| Rf/Cf network reconfiguration | Step + overshoot; spike polarity may flip across ranges. | Minimize switching frequency; use overlap rules; verify injection hot spots. |
| Guard driver response | Extra noise or wobble if marginal; can extend “trust window”. | Set guard bandwidth and stability; avoid over-aggressive drive. |
| Digital filtering / averaging | Calmer output but delayed response; can mask incomplete settling. | Apply filters after blanking; define a settle check before reporting. |
Recommended auto-range sequence (event-managed)
- Decision: determine if a new range is truly needed (with hysteresis).
- Switch: execute the range change.
- Blanking window: ignore samples during the injection transient.
- Settle check: confirm stability (slope/RMS in a window) before trusting the value.
- Report: publish the reading and record the range-change event for traceability.
Range-switch verification checklist (what “good” looks like)
- Peak glitch amplitude: switching spike stays bounded and consistent for the same transition.
- Recovery time: time-to-trust is predictable and meets the application response needs.
- Repeatability: repeating the same switch yields similar spike + tail shape.
- No ping-pong: threshold region does not trigger rapid back-and-forth switching.
- Baseline integrity: after settling, the baseline returns without permanent offset.
H2-7 · Burden Voltage & Input Protection: Measuring Without Disturbing
Key idea
In real circuits, an ammeter can change the operating point. Two things decide whether the measurement stays honest: (1) how much burden voltage the instrument imposes at the input, and (2) whether the protection network survives ESD/transients without adding leakage, capacitance, or long recovery tails.
Objective: keep the DUT node voltage nearly unchanged while keeping the input alive and clean.
Burden voltage (what it means in practice)
Definition: the effective voltage developed at the ammeter input while measuring current.
At pA/fA levels, the DUT node is often high impedance. Even a small input voltage shift can reroute leakage paths and change the true current.
Measurement disturbanceWhat it looks like: the “leakage” changes when the meter is connected, or changes with range.
If the node potential is altered, the measured current may represent a new condition rather than the original DUT behavior.
Node voltage shiftRange-dependentHow to keep burden low (engineering levers)
Virtual-ground / feedback hold
Use a feedback-controlled input so the sensitive node stays near a defined potential, reducing the voltage disturbance to the DUT.
Node held stableLower ΔVLow bias + low drift front-end
Minimize effective input bias/offset that can appear as a parasitic current. Auto-zero methods aim to reduce slow drift that pollutes long integrations.
Bias controlDrift controlBaseline management
Establish a repeatable zero/baseline state and validate stability before trusting results, especially after range changes or reconnects.
Repeatable zeroMeasure without forcing new paths
Keep the guarded geometry continuous and avoid adding extra leakage paths (contamination, unguarded surfaces, wrong cabling) that create new currents.
Guarded geometryFewer parasiticsInput protection at fA/pA: the three trade-offs
| Protection side-effect | What you see | Mitigation strategy |
|---|---|---|
| Leakage | Baseline creeps with humidity/temperature; “zero” changes after handling. | Use staged protection; keep the sensitive inner stage low-leak and place it inside the guarded zone. |
| Capacitance | Spikes and steps on connect/switch; long tails that look like leakage. | Keep high-C devices outside; isolate the sensitive node; use blanking + settle checks before reporting. |
| Recovery time | After ESD/transient, readings stay “dirty” for a while; the trust window shifts. | Define a post-event recovery procedure; validate baseline stability; avoid measuring immediately after disturbances. |
Protection architecture that survives and stays clean
- Staged protection: outer stage handles the energy (ESD/transients); inner stage prioritizes low leakage and low capacitance.
- Switchable protection: a rugged mode for field handling and a low-leak mode for final measurement can reduce compromises.
- Guard-zone placement: components tied to the sensitive node should be inside the driven-guard equipotential region to collapse ΔV across surface films.
- ESD philosophy (input-level): survive the event, then allow a clear path back to a clean baseline without long tails.
Quick verification checklist (trust before numbers)
- Burden check: confirm the DUT node voltage does not shift noticeably when connected or when ranges change.
- Protection “pollution” check: compare baseline stability with protection modes (if switchable) under the same environment.
- Recovery check: after a reconnect or disturbance, measure time-to-trust (blanking + settle criteria).
- Repeatability: repeat the same action multiple times; spikes/tails should be consistent in shape and timing.
H2-8 · Cabling, Fixtures, and Environment: The Hidden Dominant Factor
Reality check
At femto/picoamp levels, the dominant error is often not the ADC or the display digits. Cables, fixtures, humidity, and motion can create currents larger than the DUT leakage. The fastest path to stable readings is usually: cabling + fixture + humidity control, then filtering and averaging.
Objective: make the measurement setup behave like a controlled component, not an uncontrolled leakage generator.
Cable effects that masquerade as current
Tribo-electric effect (motion → spikes)
Cable movement and friction can generate charge that appears as transient current. “Touch sensitivity” is often a cable problem first.
Motion spikesFix cablesStress / piezo-like behavior (bend → drift)
Bending or mechanical stress changes electric fields and charge distribution, shifting the baseline and creating slow tails.
Bend sensitivityStrain reliefMoisture absorption (humidity → leakage)
Moisture forms a conductive surface film on insulators and connectors. Baseline creep that tracks humidity is a classic signature.
Humidity filmDry controlWhy ordinary coax is often insufficient
Without a driven guard, surface leakage sees a larger ΔV. Triax with a driven inner shield turns the cable into a controlled guarded structure.
Triax + guardΔV ↓Fixtures: isolate, guard, and keep surfaces clean
- Insulating supports + air gaps: reduce long surface creepage paths that turn contamination into leakage.
- Guard rings and guarded geometry: collapse ΔV across surface films around the sensitive node.
- Clean & dry workflow: residue (flux, oil, fingerprints) + humidity becomes a conductive film.
Environment: humidity, gradients, airflow, and static
Humidity
Drives surface film conductivity. If baseline creep follows humidity, fix enclosure/drying before tuning filters.
Record RHDry enclosureTemperature gradients & airflow
Slow drift can be dominated by thermal changes in materials and surfaces. Airflow can introduce both thermal variation and static disturbance.
StabilizeShield from draftsStatic and handling
ESD events can force long recovery time. Define a handling routine and avoid measuring immediately after reconnects or disturbances.
Handling routineRecovery windowVibration / motion
Motion-induced charge often dominates in the field. Immobilize cables and fixtures before changing instrument settings.
ImmobilizeStrain reliefExecutable SOP (field-ready)
- Warm-up & stabilize: allow the setup to reach a steady thermal state.
- Clean & dry: connectors and fixture surfaces; minimize contamination films.
- Fix geometry: immobilize cables and fixture; apply strain relief.
- Connect triax + guard: ensure driven inner shield is active and continuous to the fixture.
- Baseline/zero: establish a repeatable baseline before measurement.
- Measure in a trust window: avoid reading during spikes/tails; apply settle checks.
- Record environment: log humidity and temperature; note handling events.
Priority rule: fix cabling/fixture/humidity first, then tune filtering/averaging.
H2-9 · Zeroing, Settling, and Measurement Method: Getting Trustworthy Numbers
Goal
Trustworthy femto/picoamp readings come from a repeatable method: the right zero state, a defined settling test, and reporting only inside a validated time window. Digits alone do not define measurement truth.
Focus: turn “watching numbers” into pass/fail criteria and a repeatable workflow.
Zeroing is not one thing: short-zero vs open-zero
Short-zero (shorted input)
Used to establish the instrument’s internal offset/baseline in a controlled state. It is a health check for the measurement chain itself.
Instrument baselineOffset controlOpen-zero (open input)
Not equal to “zero current.” It includes cable/fixture leakage, surface film conduction, and charge release effects that can look like current.
System baselineEnvironment sensitiveWhy open input can read non-zero: charge redistribution, dielectric absorption release, humidity-driven surface films, and guard recovery can all create a slowly decaying “tail” that appears as leakage.
Practical implication: open-zero must be treated as a baseline with a defined settling test and recorded environment.
Settling: define “stable” with criteria (not intuition)
| Criterion | What it tests | When to use |
|---|---|---|
| Slope threshold | Rate of change stays below a limit for a sustained duration (tail has “flattened”). | After range switching, reconnects, or any condition that produces a decay tail. |
| Window variance | Short-term variability (RMS/variance) stays below a limit within a defined window. | When random noise dominates and “noise floor stable” matters. |
| Repeatability | Repeating the same action yields similar stable values and similar time-to-trust. | To separate setup problems (cable/fixture/humidity) from instrument behavior. |
Measurement levers: integration, filtering, reporting
Integration time
Longer integration reduces random noise, but can amplify the visibility of slow drift and charge-release tails. Choose based on the trust window, not just noise.
Noise ↓Response ↓Digital filtering
Filtering can clean noise but can also smear real short events into long tails. Filter strength should match the event dynamics being measured.
Clean displayTail riskSampling vs report rate
Internal sampling and external reporting are different. Reporting should occur only after blanking and settling checks have passed.
Valid windowLog disciplineOutlier handling (use with care)
Blind outlier rejection can delete real leakage events. Prefer labeling events (switch, touch, humidity change) before deciding to exclude data.
Avoid false cleanupSetup wizard (repeatable workflow)
- Define target range: expected current scale and whether events are slow drift or fast transients.
- Select range & method: set the range with enough headroom to avoid chattering near thresholds.
- Set integration/reporting: pick integration time and a report rate consistent with the expected dynamics.
- Establish zero state: short-zero for instrument baseline; open-zero for system baseline (cable/fixture involved).
- Apply settling tests: blank the immediate transient; then use slope/variance criteria to decide stability.
- Record & report: log only inside the valid window with full configuration and environment fields.
Minimum record fields (for reproducibility)
- Range, integration time, filter mode, report rate
- Guard status (on/off), cable type (triax/coax) and length, fixture ID
- Environment (humidity/temperature) and event notes (switching, reconnects, handling)
H2-10 · Calibration & Self-Test: Proving fA Performance
Purpose
“Seeing femtoamps” is not proof. Proof requires a known stimulus, a predictable response, and repeatable pass/fail checks across ranges. Calibration and self-test should confirm both accuracy directionally and the full behavior: noise, settling, and recovery.
Scope: instrument-level verification and sanity checks, without expanding into a full metrology system.
Known-current generation (three practical approaches)
High-value R + known V
Simple and intuitive for pA verification and range consistency checks. Requires clean surfaces and stable conditions to avoid leakage dominating the stimulus.
Static checkSetup-sensitiveCharge injection (C-step)
Validates dynamic response and the settling method by injecting a known charge step. Watch for parasitic capacitance and switching tails.
Dynamic checkTail-awareInternal reference / injection path
Fast and repeatable for field checks. Demonstrates internal path health and range network consistency, but cannot replace external transfer checks.
Fast BISTRepeatableTraceability (concept only)
A defensible result links stimulus and measurement back to controlled references. At this page level, focus on controlled inputs and repeatable validation gates.
ConceptNot a full systemSelf-test / BIST gates (what to verify per range)
| Gate | What it proves | What to record |
|---|---|---|
| Short-zero | Instrument baseline and internal offset health in a controlled state. | Range, integration, filter, baseline value. |
| Open drift | System baseline stability including cable/fixture influence and environment sensitivity. | Humidity/temperature, guard status, time-to-stable. |
| Known injection response | Stimulus-to-reading transfer and expected settling behavior (including tails). | Stimulus method (R+V / C-step / internal), response magnitude, settling time. |
| Range consistency | Range network behaves consistently; no hidden leakage paths appear in certain ranges. | Adjacent-range comparison, hysteresis behavior, repeatability. |
| Recovery time | After a disturbance (reconnect, transient), the system returns to a clean baseline predictably. | Blanking window, settle criteria pass time. |
Field sanity check (fast diagnosis when readings drift)
- Run internal injection (if available): confirms internal chain and range network health quickly.
- Verify open baseline + environment: if baseline tracks humidity or handling, fix setup before tuning settings.
- Repeat a controlled reconnect action: the spike/tail signature should be repeatable; non-repeatability suggests uncontrolled setup.
- Re-validate per range: pass the same gates after changing range or cable/fixture components.
Rule: prove the instrument path first, then prove the setup, then trust the DUT result.
H2-11 · Troubleshooting: When the Reading Looks Wrong
Fast triage rule
Separate “instrument chain” from “cable/fixture/environment” before touching settings. A stable short-input but unstable open-input almost always points to setup leakage, charging, or cabling effects.
This section is written as on-bench actions: symptom → likely causes → do this now → confirm.
90-second three-step isolation (do this first)
Step 1 — Disconnect DUT
If the symptom persists with the DUT removed, the DUT is not the main cause.
Step 2 — Short the input
If noise/spikes/drift remain, suspect range/front-end/guard stability/settings first.
Step 3 — Open the input
Open is not zero. Watch for tails, humidity tracking, and touch sensitivity.
Then swap one variable
Swap cable/fixture or toggle guard state. If the symptom follows the swap, it is setup-driven.
Symptom index (tap to jump)
Large drift / baseline wander (slowly drifting over minutes)
Quick signature
- Reading slowly walks in one direction, often after reconnects or environmental changes.
- Drift correlates with humidity, temperature gradients, or recent range switching.
Most likely causes
- Surface contamination + humidity film leakage on fixtures, insulators, or connectors.
- Thermal gradients (airflow, warm instruments nearby) causing slow baseline changes.
- Dielectric absorption / charge release tail after switching or handling.
- Guard coverage incomplete (guard not surrounding the sensitive node end-to-end).
Do this now
- Run the isolation flow: disconnect DUT → short input → open input (watch whether the drift remains).
- Log humidity and temperature; watch if baseline tracks RH.
- Freeze geometry: fix the cable and fixture to remove micro-movement.
- Increase blanking + enforce a “slope threshold” settling rule before reporting.
- Toggle guard ON/OFF once (a strong change indicates guard/wiring/coverage involvement).
Example parts to swap / verify (material numbers)
- Low-noise triax cable: Keithley 7078-TRX-12 (or shorter 7078-TRX variants).
- Low-noise triax alternative: Keysight PX0102A cable options.
- Triax bulkhead jack: Trompeter/Cinch BJ77 (panel triax connector).
- Electrometer-grade front-end amp examples: ADI ADA4530-1; TI OPA129; TI LMC6001 (for “known-good behavior” A/B checks in a prototype chain).
Use part numbers as known references. Actual leakage behavior depends strongly on temperature, cleanliness, layout, and guarding.
Related diagrams: Leakage path map (Figure F2), Guard wiring (Figure F5), Blanking/settle/valid windows (Figure F9).
Spikes / bursts (touch- or movement-triggered)
Quick signature
- Spikes appear when the cable is touched, flexed, or moved; sometimes followed by a decay tail.
- Spikes appear at the same moment as relay/range switching.
Most likely causes
- Triboelectric charging in the cable (motion → charge → apparent current).
- Electrostatic discharge events or poor grounding of the outer shield.
- Charge injection from switching elements during range changes.
- Guard transient/instability when driving a large cable capacitance.
Do this now
- Stop movement: strain-relief the cable, avoid hanging loops, and immobilize the fixture.
- Perform a “touch test”: lightly tap the cable/connector and check repeatability of spike timing.
- If spikes align with auto-range: extend blanking, add range hysteresis, and report only in the valid window.
- Re-check shielding: outer shield must be solidly referenced (ground strategy must be consistent).
- Toggle guard: if spikes worsen with guard ON, check guard wiring and guard driver stability first.
Example parts to swap / verify (material numbers)
- Low-noise triax cable: Keithley 7078-TRX-12 (low-noise cable assembly family).
- Low-noise triax alternative: Keysight PX0102A.
- Low-leakage switching reference: Coto Technology reed relay 9007-05-00 / 9007-05-01 (for range network injection comparisons).
Spikes are often dominated by cabling and switching transients; swapping to a known low-noise cable is the fastest discriminator.
Related diagrams: Guard + triax wiring (Figure F5), Blanking/settle/valid windows (Figure F9).
High noise / unstable (fast jitter within a band)
Quick signature
- Reading rapidly jitters; changes immediately when integration time or filter is adjusted.
- Guard ON makes it worse (a red flag for guard wiring or stability).
Most likely causes
- Too much measurement bandwidth (short integration, weak filtering, high report rate).
- Guard driver oscillation or marginal phase stability with cable capacitance.
- Miswiring (triax inner shield not driven, guard tied to ground, HI/LO confusion).
- Environmental agitation (cable micro-motion, airflow temperature shifts).
Do this now
- Make a conservative baseline: increase integration time, reduce report rate, enable stronger averaging.
- Short-input check: if short-input is noisy, fix instrument/guard/settings before blaming setup.
- Guard sanity check: verify center conductor = signal, inner shield = driven guard, outer shield = shield/ground.
- Reduce cable length and immobilize it; re-test noise level and repeatability.
Example parts to swap / verify (material numbers)
- Electrometer-grade reference amp: ADI ADA4530-1 (includes an integrated guard buffer conceptually useful in guarded front ends).
- Low-bias reference amps: TI OPA129, TI LMC6001 (for behavior comparison in a guarded prototype chain).
- Triax connector: Trompeter/Cinch BJ77 (verify consistent triax termination and isolation of shields).
Related diagrams: Guard wiring (Figure F5), Blanking/settle/valid windows (Figure F9).
Negative readings (unexpected sign)
Quick signature
- Open-input baseline is negative, or the sign flips after changing mode/range.
- Negative value appears immediately after switching, then slowly returns toward baseline.
Most likely causes
- Zero reference mismatch: using open-input baseline as if it were true zero.
- Polarity/connection interpretation: HI/LO meaning differs from the assumed current direction.
- Leakage path establishes a “reverse” effective drive when shielding/guard reference is wrong.
- Not settled yet: a recovery tail is being read as a valid measurement.
Do this now
- Write down which zero is active: short-zero vs open-zero (do not mix them).
- Run the isolation flow and locate the step where the sign changes (disconnect → short → open).
- Re-check triax wiring and shield references; then enforce the valid window after settling.
Example parts to swap / verify (material numbers)
- Known low-noise cable: Keithley 7078-TRX-12 (to remove cable-driven polarity artifacts).
- Triax connector: Trompeter/Cinch BJ77 (to avoid shield shorting/mis-termination issues).
Related diagrams: Leakage path map (Figure F2), Guard wiring (Figure F5), Blanking/settle/valid windows (Figure F9).
Jumps after auto-range (steps, long recovery, chatter)
Quick signature
- Each range change produces a step or spike; reading takes a long time to become believable.
- Near thresholds, the instrument “chatters” between ranges and never stabilizes.
Most likely causes
- Charge injection and redistribution inside the range network.
- Insufficient blanking/settling policy (reporting too early).
- Range overlap/hysteresis too tight (chatter near the decision boundary).
- Guard transient when switching a high-impedance node.
Do this now
- Increase blanking time and require a settle criteria pass before logging.
- Add more hysteresis or force a fixed range during debug to prevent chatter.
- Confirm repeatability: identical switching should create a similar transient signature.
- If the transient changes with cable/fixture swap, treat it as setup-dominated.
Example parts to swap / verify (material numbers)
- Low-leakage reed relay reference: Coto Technology 9007-05-00 / 9007-05-01 (for stable range switching comparisons).
- Known low-noise triax cable: Keithley 7078-TRX-12 (reduces motion-related artifacts during range switching tests).
Related diagrams: Range switching injection hotspot (Figure F6), Blanking/settle/valid windows (Figure F9).
Bench checklist to keep with the test record
- Range, integration time, filter mode, report rate
- Guard state, cable part number, fixture ID
- Humidity/temperature, and events (switching, handling, reconnect)
If the record fields are complete, “mystery drift” becomes diagnosable instead of repeatable pain.
Request a Quote
H2-12 · FAQs (Pico / Femto Ammeter)
These FAQs target common “reading looks wrong” and “how to measure fA/pA correctly” questions. Each answer gives a practical rule, the main physical reason, and a next action that helps verify the result on the bench.
Tip: for faster on-bench use, open the closest matching symptom/question and apply the “do this now” step first.
