Source Measure Unit (SMU): Four-Quadrant Source/Measure Basics
← Back to: Test & Measurement / Instrumentation
A Source Measure Unit (SMU) combines a programmable source with synchronized precision measurement and compliance protection, enabling trustworthy I–V characterization, ultra-low current tests, and repeatable automated sweeps in one instrument.
This page explains the internal source/measure architecture, four-quadrant behavior, autorange and settling rules, Kelvin/guard techniques, and the logging/triggering practices that keep results accurate and auditable.
What an SMU is (and when it beats PSU + DMM)
A Source Measure Unit (SMU) is a single instrument that forces a programmed voltage or current while measuring the resulting current or voltage in tight synchronization, with a built-in safety envelope called compliance. It is designed for I-V characterization and automated test loops where “source + measure + protect + log” must behave like one coherent system.
The 3 capabilities that make an SMU “different”
- Programmable forcing with defined limits: voltage-source or current-source output, with a measurable transition into compliance (current limit or voltage limit) rather than an undefined “something happened” event.
- Coherent timing: the output step, the settling delay, and the measurement window are coordinated inside one timing model (trigger/holdoff/integration), which reduces “false I-V shapes” caused by unsynchronized instruments.
- Wide dynamic range by design: range switching and autorange logic are part of the measurement chain (including metadata like range state and compliance flags), enabling reliable sweeps from pA/µA to A without manual reconfiguration.
Typical tasks where SMU is the right tool
- I-V sweeps (diodes, MOSFETs, sensors, materials): step forcing + controlled measurement window + compliance guarding against over-stress during curve tracing.
- Leakage and pre-breakdown monitoring: low-current measurement with guarding awareness, plus predictable behavior when the DUT approaches limit conditions.
- Low-resistance or contact-sensitive measurements: Kelvin 4-wire Force/Sense reduces lead/contact errors that dominate 2-wire setups.
- Automated test scripts: list sweeps, triggerable measurements, and logged flags (range/compliance/trip) make results reproducible and debuggable.
Why “PSU + DMM” often fails for characterization
- Timing mismatch: the DMM integrates during a different moment than the PSU step, producing apparent “hysteresis” or “kinks” that are actually measurement timing artifacts.
- No unified compliance behavior: current limit, foldback, or transient overshoot is not consistently measured and logged across two devices, risking DUT damage or confusing data.
- Range switching is opaque: manual range changes and settling delays are easy to forget, so sweeps show jumps that are not real DUT behavior.
- 2-wire errors dominate: lead/contact resistance and thermals distort low-ohm and low-voltage results unless Force/Sense is handled correctly.
Figure F1 focuses on the practical difference: an SMU ties forcing, measurement, compliance, and range behavior to one timing and logging model.
System architecture: source path + measure path + control plane
An SMU becomes easy to reason about when it is split into three planes: the source path that creates the output, the measurement path that converts terminal conditions into numbers, and the control plane that coordinates modes, timing, range switching, protection, and calibration. Good SMU performance is not one “magic component” — it is the consistency of these three planes under real DUT and cabling conditions.
1) Source path (forcing chain)
- Setpoint generation: a precision DAC creates a voltage/current command, often supported by fine/coarse ranges to cover wide output spans.
- Loop control: an error amplifier enforces CV or CC behavior by closing the feedback loop around the output stage and sense feedback.
- Power stage: implements the required compliance, output drive, and (when needed) source/sink behavior. Stability is shaped by output impedance, cable inductance, and capacitive DUTs.
2) Measurement path (reading chain)
- Voltage measurement: terminal voltage is routed through protection/filtering and scaling (divider/buffer) into an ADC. Accuracy is shaped by offset, gain error, noise, and temperature drift.
- Current measurement: current is measured through a shunt (higher currents) or TIA-style front ends (low currents), often with range switching. Thermal effects and leakage paths matter most at low currents.
- Integration/sampling window: integrating ADC behavior (NPLC-like) trades noise for speed; digitize modes trade speed for higher noise and stricter settling requirements.
3) Control plane (mode, timing, calibration, metadata)
- Mode management: V-source vs I-source selection, and defined compliance behavior when limits are hit (clamp/hold/log).
- Timing model: triggers, delays, holdoff, and measurement windows that align “step → settle → measure” for repeatable sweeps.
- Range switching & protection: relays/switches, comparators, and SOA logic that act quickly and report “what happened” in flags.
- Calibration injection & tables: internal references and injection points allow offset/gain corrections and temperature-aware compensation.
A practical “architecture checklist” for selection and debug
- Does the instrument report range state and compliance state for every reading?
- Is there a defined settling strategy (delay/holdoff/integration window) for sweeps and autorange transitions?
- Are there clear mechanisms for remote sense fault detection (sense open/reversed) and safe behavior?
- Is calibration traceable at the points that matter: source setpoint, terminal voltage, and measured current?
Figure F2 acts as a map for the entire SMU page: later chapters will deepen the four-quadrant stage, autorange switching, compliance/protection, and calibration injection points.
Four-quadrant operation: what it really means electrically
“Four-quadrant” is not a marketing phrase. It is a statement about the SMU output port: both terminal voltage and terminal current can be positive or negative, so the instrument can both deliver power to a DUT and absorb power from it. The sign of power is set by P = V × I: positive power means the SMU is sourcing energy; negative power means the SMU is sinking (absorbing) energy.
The 2×2 quadrant map (what you should expect at the port)
- Q1 (V>0, I>0): source power into the DUT (common I–V sweep region).
- Q4 (V>0, I<0): sink current while holding positive voltage (absorbing energy from the DUT).
- Q3 (V<0, I<0): source power with reversed polarity (negative voltage bias cases).
- Q2 (V<0, I>0): sink power with negative voltage (the port absorbs energy under negative bias).
Why four-quadrant matters in real tests
- Devices can drive the instrument back: capacitive discharge, inductive kick, or DUT-generated voltages can push energy into the SMU. A four-quadrant stage can absorb that energy in a controlled way rather than letting the port overshoot.
- Negative resistance / hysteresis regions become measurable: some DUT behaviors appear only when the port can hold voltage while controlling current direction. Four-quadrant capability expands the “safe, measurable” region without redesigning the setup.
- Compliance boundaries become the real borders: the practical quadrant is defined by limits (V limit, I limit, power/SOA). Understanding this boundary prevents “fake curves” created by hidden limiting.
Two implementation styles (what you can verify, without guessing topology)
Linear four-quadrant stage
- Typical strengths: very low noise/low ripple at the port, clean settling for sensitive I–V work.
- Typical trade-off: high internal dissipation; usable region is strongly shaped by power limits and thermal SOA.
Switching + linear hybrid stage
- Typical strengths: higher power/efficiency; wider usable quadrant at higher currents.
- Typical trade-off: output ripple/residual switching artifacts require careful bandwidth/filters and measurement timing.
Where does the “sunk” energy go?
- Internal dissipation: the absorbed power is converted to heat inside the instrument. This is common and limited by thermal design and SOA.
- Regenerative return (if supported): the instrument routes absorbed energy back to a supply bus. Verification is done by checking the datasheet for “sink power” and “regenerative” behavior and confirming it with controlled tests and logged power direction.
Figure F3 ties the quadrant concept to real limits: compliance lines define where the SMU can safely source or sink while holding the commanded condition.
Source modes & control loops: CV, CC, and compliance behavior
An SMU can behave like a voltage source or a current source, but the key difference from a generic power supply is that compliance is part of the measurement story. A reading is only meaningful when you know whether the instrument is in normal regulation or in compliance. For reliable sweeps, always treat V/I values + compliance flag + range state as one dataset.
CV source (voltage set, current limited)
- What is controlled: terminal voltage is regulated to the programmed setpoint while the required load current stays below the I-compliance.
- What changes in compliance: once the DUT demands more than the current limit, current is clamped and the terminal voltage may droop or be reduced (behavior depends on instrument policy).
- Data interpretation rule: in I-compliance, the “V setpoint” is no longer the guaranteed terminal condition—use measured V and the compliance flag.
CC source (current set, voltage limited)
- What is controlled: output current is regulated to the programmed setpoint while the required terminal voltage stays below the V-compliance.
- What changes in compliance: once the DUT requires more voltage than the limit, voltage is clamped and the current may fall below the setpoint.
- Data interpretation rule: in V-compliance, the “I setpoint” is not delivered—use measured I and the compliance flag to label the region.
What can happen after compliance triggers (common behaviors)
- Clamp: the limited variable “sticks” close to the compliance boundary while the other variable departs from its setpoint.
- Mode transition: the effective behavior shifts (CV→CC or CC→CV) to keep the port within safe bounds.
- Foldback / protection policy: output backs off further to reduce stress; readings show an abrupt drop with a protection indication.
- Hold/freeze window: measurement timing or output hold behavior changes; confirm by timestamps and sweep step metadata.
A universal verification method (works across vendors)
- Use a known resistor (and, if needed, a switchable resistor bank) as a controlled DUT so the expected V–I line is predictable.
- In CV mode, step the voltage up while holding a fixed I-compliance, until compliance is reached. Record Vmeas, Imeas, compliance flag, range state.
- In CC mode, step the current up while holding a fixed V-compliance. Repeat the same logging.
- Confirm that the boundary region is labeled consistently: compliance flags should align with “flat” behavior at the limit, and the measured values should follow the expected clamp policy rather than drifting silently.
- If results are inconsistent, repeat with different cable lengths and with remote sense enabled/disabled to reveal stability and settling dependencies.
Stability: cables and capacitive loads are part of the loop
- Why oscillation happens: lead inductance and DUT capacitance add phase lag, reducing loop margin and causing ringing or sustained oscillation.
- What it looks like in data: “never-settling” readings, compliance flags flickering near thresholds, or sweep points that depend on cable routing.
- Practical mitigations: reduce output slew rate (if supported), shorten leads, avoid large capacitive steps, and validate settling with step tests before trusting sweep curves.
Figure F4 highlights the practical rule: compliance acts as a limiter in the drive path, and the resulting compliance flag must travel with the measurement data.
Measurement chain: accuracy, noise floor, and digitize mode
In an SMU, “accuracy” is not owned by the ADC alone. It is the combined behavior of front-end scaling, offset and gain drift, integration window, and range state. A trustworthy reading is a bundle: V/I value + measurement mode + integration/average settings + range state + compliance/overload flags.
Where “good numbers” come from (a practical view)
- Noise floor: set by front-end noise density, bandwidth, and the integration / averaging window.
- DC accuracy: set by offset + gain error + reference stability + temperature drift, not by resolution alone.
- Repeatability: depends on stable range state, predictable settling, and consistent measurement timing.
- Auditability: requires logging the flags (range/compliance/overload) together with the reading.
Voltage measurement (V-meas): buffer, scaling, protection, and ADC roles
- Input buffer + protection: protects the ADC path and prevents overload recovery from contaminating slow, low-noise readings. At low levels, leakage and input bias can become measurable errors.
- Scaling network (divider / gain): sets the voltage range and the temperature drift profile. Ratio drift and thermals matter for long sweeps.
- ADC roles inside an SMU: integrating / ΣΔ paths support low noise and strong mains rejection in “reading mode”; SAR paths support higher update rates and digitize capture, trading lower noise for speed.
Current measurement (I-meas): shunt vs TIA and what limits low-current work
- Shunt-based sensing (higher currents): robust and linear, but accuracy is shaped by shunt self-heating, contact resistance, and thermals along the current path.
- TIA-based sensing (µA to nA ranges): enables tiny currents, but the true enemies are leakage paths, amplifier bias/1/f noise, and feedback resistor noise and temperature drift.
- Key implication: at low currents, the measurement setup behaves like a system of “unwanted resistors” (leakage) plus drift, so guarding, stable temperature, and consistent range state often matter more than ADC resolution.
Integration and sampling time: NPLC, averaging, and filters
- NPLC / integration window: longer windows reduce noise by narrowing effective bandwidth and can reject mains interference, but reduce throughput and can smear fast transients.
- Averaging: improves stability for slow-changing points, but adds latency and can hide range-switch or compliance events if not logged.
- Filters: useful for display and slow sweeps, but they create time correlation—sweep step timing must include settling plus filter behavior.
Digitize mode (fast capture): why it exists and what it costs
- Why it is needed: to see step response, switching transients, or range-change glitches that are invisible in slow reading modes.
- What it costs: higher noise per sample, more dependence on trigger timing, and larger datasets that require post-processing.
- Practical boundary: digitize is for transient visibility and validation (settling, overshoot, glitches), not a replacement for full waveform instruments.
Figure F5 shows why “accuracy” is a chain: the dominant error source can move between front-end drift, leakage, range state, and timing window.
Range switching & autorange: making µA and A share one box
Range switching is the SMU’s hardest “hidden” engineering problem. The instrument must change physical signal paths (relays, resistor banks, amplifier gains) without injecting enough disturbance to create false readings or to trigger compliance. Autorange is not simply “pick the best range” — it is a policy that trades stability, settling, and throughput.
Why range switching is difficult (real error mechanisms)
- Thermal EMF (relay thermocouples): µV-level offsets can dominate low-level voltage and low-ohm work.
- Contact resistance + contact noise: impacts both accuracy and repeatability, especially when current paths change.
- Charge injection and parasitics: switching capacitances kick nodes, creating transient spikes in current or voltage.
- Recovery + settling: amplifiers and ADCs need time to recover from overload and to re-stabilize after path changes.
Typical range hardware blocks (what shares the same port)
- Shunt resistor bank: multiple current paths for mA/A ranges (trade heat and burden vs resolution).
- TIA feedback resistor bank (Rf bank): ultra-high R values for µA/nA ranges (trade noise and drift vs sensitivity).
- Voltage divider / gain network: sets voltage measurement span and drift profile, tied to relay/switch selection.
Autorange strategy (the rules that prevent misreads)
- Thresholds: switch before overload, leaving headroom for peaks and settling.
- Hysteresis: prevents “ping-pong” range chatter near boundaries.
- Minimum dwell time: keeps a range stable long enough for thermal EMF and filters to settle.
- Settling criterion: defines when a reading becomes valid after switching (not just “wait a bit”).
“How long until it is trustworthy?” (a practical validation recipe)
- Use a stable, predictable DUT (precision resistor or a simple fixed load), and configure steps that will force a range transition.
- Log reading plus range state, overload/compliance flags, and timestamp for every sample.
- After each range switch, sample fast (digitize or high-rate reads) to observe transient spikes, then switch to integrating reads to confirm final stability.
- Define a “settled” rule: for example, N consecutive readings inside a tolerance band (±X%) with no overload/compliance events and no further range changes.
- Bake the proven wait + settle rule into scripts so sweeps do not create false kinks at range boundaries.
Figure F6 makes the key point visible: µA and A ranges are different physical paths. Autorange policy exists to avoid switching artifacts becoming “data”.
Kelvin 4-wire, remote sense, guarding: fighting lead errors & leakage
In real SMU work, the biggest measurement errors often come from what sits between the instrument and the DUT: leads, contacts, fixtures, and leakage paths. The purpose of Kelvin sensing, remote sense, and guarding is to separate the “force current path” from the “sense voltage path,” then prevent unwanted surface currents from becoming part of the reading.
When 4-wire Kelvin is mandatory (not “nice to have”)
- Low resistance / high current: 2-wire includes lead drop and contact resistance in the “DUT” voltage.
- Contact-sensitive setups: clamping force or oxidation changes contact resistance and creates drift.
- Repeatable characterization: Kelvin sensing reduces fixture variability when comparing lots or materials.
Kelvin wiring in one sentence: Force carries current, Sense measures voltage
- Force HI/LO carries the load current, so lead resistance and contact resistance create voltage drops on the force path.
- Sense HI/LO draws negligible current, so the SMU measures the DUT terminal voltage with minimal error from lead drops.
Remote sense stability: why long leads and capacitive loads can misbehave
- What changes electrically: remote sense moves the feedback point to the DUT, so cable inductance and load capacitance become part of the loop.
- Typical symptoms: overshoot, slow settling, reading jitter, or protection trips even when setpoints look reasonable.
- Wiring risks: a sense lead that is open, swapped, or connected to the wrong point can create silent errors or unstable behavior. Always confirm sense integrity before trusting automated sweeps.
Guard / driven shield: how leakage becomes “current”
In high-resistance and low-current measurements, surface contamination and humidity create an unintended leakage resistance (Rleak). That leakage path sits in parallel with the DUT and adds extra current. Guarding works by driving nearby surfaces close to the sensitive node potential, reducing the electric field across insulators and dramatically lowering leakage current.
- Most common leakage sources: dirty insulators, flux residue, fingerprints, moist connectors, long unshielded pigtails.
- Practical check: measure “open fixture” drift (no DUT). If the reading moves with humidity or with cable movement, leakage is dominating.
Fixture cleanliness: a fast, executable troubleshooting routine
- Baseline: run a short measurement with no DUT (open fixture) and record drift vs time.
- Swap test: replace only the cable or only the fixture. If the behavior follows the hardware, the DUT is not the cause.
- Environmental clue: log humidity/temperature. If the reading tracks humidity, leakage dominates.
- Clean + dry: clean insulators and connectors, dry the setup, then re-run the open-fixture baseline to confirm improvement.
Figure F7 highlights the field reality: if leakage paths exist, they become part of the measured current unless guarding and cleanliness are controlled.
Protections & safe operating area: saving DUT and saving SMU
Protection features are only useful when they are treated as an engineering checklist: define limits, verify the instrument’s response, and log the trip reason. “Setpoint within limits” does not guarantee safety, because the internal safe operating area (SOA) depends on voltage, current, range path, and thermal conditions.
Protection map (what each item prevents)
- OVP / OCP: clamps or shuts down when voltage/current exceed configured bounds.
- OTP: protects internal power devices and resistors when thermal limits are reached.
- Reverse / back-drive: prevents damage when the DUT forces current back into the SMU output.
- Inductive kickback: addresses transient overvoltage when current is interrupted into an inductive load.
Why “it should be safe” still trips (SOA intuition)
- Power is two-dimensional: V and I can look “under limit,” but V×I can exceed internal dissipation.
- Range path matters: different ranges use different resistors/switches with different thermal margins.
- Transient stress: charging capacitors or interrupting inductive current can exceed limits briefly.
- Thermal derating: higher ambient temperature reduces usable SOA even with the same setpoints.
Relay and switch lifetime: rules for safe switching
- Prefer open-circuit switching: reduce output, then switch ranges, then ramp back up.
- Avoid rapid toggling: frequent range switching turns measurement policy into a lifetime limiter.
- Respect surge conditions: large capacitors and inductors create stress even when steady-state looks safe.
DUT-friendly output behavior (default-safe playbook)
- Set compliance conservatively first, then ramp setpoints toward the target.
- Limit slew rate to reduce overshoot and avoid triggering protective clamps.
- Precharge / controlled charge for capacitive DUTs, and add a discharge path if residual voltage can mislead later steps.
- Log trip reason (and range/compliance flags) so “mystery trips” become diagnosable events.
Figure F8 makes verification concrete: confirm where the SOA boundary is, then confirm the protection chain produces a clear, logged trip reason.
Settling, transients, and measurement integrity
“Unstable readings” usually mean the measurement is being taken before the system has finished responding to a change. In an SMU, settling is shaped by control-loop behavior, range switching events, cable inductance, load capacitance, thermal drift, and the time delay introduced by averaging and filters. Measurement integrity comes from choosing the right sampling mode and defining an explicit “valid sample window.”
Where instability comes from (and how it shows up)
- Loop behavior: overshoot, ringing, slow tails, sensitivity to long leads and capacitive loads.
- Range switching: step-like jumps, spikes, or “kinks” near autorange boundaries.
- Cable L and load C: transient currents and voltage excursions that can hit compliance briefly.
- Thermal drift: slow movement over seconds to minutes, especially on low-level ranges.
- Filter / averaging delay: smoother numbers, but time-shifted response and longer effective settling.
Decide the sampling strategy: trigger capture vs integrating reads
The correct question is not “which mode is better,” but “what is being proved.”
- Use integrating reads (NPLC + averaging) to report steady-state points (IV curves, leakage, resistance, drift trends). This maximizes repeatability and mains rejection.
- Use triggered capture (fast read / digitize) to reveal transients (overshoot, ringing, range-switch spikes, compliance events). This maximizes visibility of worst-case behavior.
- Best practice workflow: first capture the transient to learn the worst-case settling time, then switch to integrating reads for the final reported value.
Three practical tests that define a reliable wait rule
- Step response: apply a small voltage or current step, capture the response, and measure overshoot and time-to-band (e.g., within ±X%).
- Open/short sanity: verify that open and short conditions produce expected behavior and clear flags (overload/compliance), not “mystery numbers.”
- NPLC sweep: repeat the same stable measurement at multiple NPLC settings and log noise (standard deviation) vs throughput (points per second).
Avoid false interpretation: compliance changes the meaning
When compliance is active, one variable is being clamped (current in CV mode or voltage in CC mode). The other variable may still be measured, but it no longer represents a “free” DUT response. For trustworthy automation, log the compliance flag, range state, and timestamps with every reading, and treat compliance-active points as a distinct condition, not a normal steady-state point.
Figure F9 turns settling into a scriptable rule: step → holdoff → enter band → sample in a defined window.
Calibration, self-cal, and traceable performance (practical)
Practical calibration for an SMU means two separate things: the measurement chain must report correct V/I, and the source chain must produce the correct output. Self-calibration can correct offset and gain drift, but it cannot “remove noise,” and it cannot undo dynamic errors caused by relay thermal EMF, range switching artifacts, or unstable wiring and fixtures.
Self-cal can fix vs cannot fix
- Can fix: offset, gain, and some repeatable nonlinearity using internal references and correction tables.
- Cannot fix: noise floor, leakage-dominated drift, relay thermal EMF during switching, unstable cabling effects.
- Key takeaway: self-cal improves DC correctness; it does not guarantee transient integrity.
Calibration paths: source chain and measure chain must be treated separately
- Measure chain calibration: reference → injection point → front-end/ADC → correction table.
- Source chain calibration: DAC/output stage → terminal output → correction table, validated through the measurement chain or an external reference.
Field verification: a minimal closed-loop routine for confidence
- Short: check zero behavior and drift trend on relevant ranges (watch thermal settling).
- Open: check leakage dominance (fixture and humidity sensitivity) before trusting pA–nA readings.
- Standard resistor(s): validate a few points across ranges and confirm consistent, repeatable results.
- Record traceability fields: cal table ID/version, temperature, range state, NPLC, and flags with the results.
Figure F10 clarifies the real workflow: inject known references, correct source and measure chains separately, and record traceability fields with verification.
Automation, triggering, and test scripting (SMU as a test engine)
In automated characterization and production test, an SMU is most valuable when it behaves like a deterministic test engine: every source step has a defined timing budget, every measurement has a valid sample window, and every exceptional condition (compliance or trip) is captured as structured data. The goal is repeatable results that can be replayed and audited—not just a sweep that “looks fine.”
List sweep: define a single-point timing budget that never lies
Treat a sweep as a loop of identical “points.” Each point must explicitly define when the instrument is allowed to measure, otherwise transients, autorange events, and filter delay will leak into the data.
- Source step: apply next V or I setpoint (V-sweep or I-sweep).
- Dwell: minimum dwell time for loop response and load behavior to settle.
- Measure delay: push the measurement window beyond overshoot / switching artifacts.
- Measure: integrating read (NPLC/average) for steady-state, or fast capture for transients.
- Store: write a complete row including flags (range state, compliance, trip reason).
Compliance behavior: handle it as a state, not a comment
During a sweep, compliance is not “noise.” It is a real operating condition where one variable is clamped (current clamp in voltage-source mode, or voltage clamp in current-source mode). A robust script makes compliance visible and consistent across runs.
- Continue-with-flag: keep sweeping, but mark points as compliance-active.
- Re-measure: apply extra holdoff, then re-measure (useful near thresholds).
- Stop-on-trip: if protection trips (OVP/OCP/OTP/back-drive), stop and log the trip reason.
Trigger model: Arm → Trigger → Delay/Holdoff → Measure, plus marker events
The smallest trigger model that scales from a single SMU to multi-instrument setups includes four concepts: Arm (ready), Trigger (start an action), Delay/Holdoff (avoid transients), and Marker (signal other equipment). Marker events are how external switches, chambers, cameras, or other instruments stay aligned with the SMU’s internal timeline.
Practical multi-instrument synchronization (baseline approach):
- One master trigger: one device (or controller) defines the step cadence.
- Aligned step index: every device records the same step/event ID per point.
- Per-device delay trim: compensate fixed latency differences using delay/holdoff values.
Interfaces and scripting: keep communication separate from test logic
USB, LAN, and GPIB typically converge to the same command layer (often SCPI). Reliability comes from a clean test structure: keep instrument transport and error handling in one layer, and keep sweep/trigger policy in another. This prevents “changing interface” from changing the measurement.
- Session layer: connect, identify, reset, configure, query errors.
- Plan layer: sweep list, dwell/delay, measurement mode, compliance policy.
- Run layer: state machine execution and row-based data writing.
Data output: minimum columns that make results explainable
A sweep file that only contains “measured V and I” cannot be debugged. The following columns should be treated as mandatory:
- timestamp (or time-from-trigger), step_index, set_V, set_I, meas_V, meas_I
- range_state (V-range, I-range), autorange_event (if any), NPLC/avg settings
- compliance_flag (active/inactive), mode (V-source or I-source)
- trip_reason (OVP/OCP/OTP/back-drive/other), and whether the script continued, re-measured, or aborted
Example parts (for automation I/O and synchronization blocks)
The exact platform differs by vendor, but these part numbers are commonly used when building internal automation I/O, isolation, and timing fanout around an SMU-class instrument (choose per bandwidth, isolation rating, and system constraints):
- USB isolation: ADI ADuM4160
- USB hub (internal expansion): TI TUSB8041
- USB-to-multi-UART/MPSSE tooling: FTDI FT4232H
- GbE PHY: TI DP83867E / Microchip KSZ9031RNX
- Low-jitter fanout buffer (trigger/clock distribution concept): TI LMK1C1104
Figure F11 is designed to be “scriptable”: the main path defines timing per point, while compliance and trip become explicit branches with logged outcomes.
Request a Quote
FAQs (SMU practical questions)
These answers focus on repeatable SMU setup: what “four-quadrant” means in real tests, how to set compliance safely, how to treat autorange settling, and which metadata must be logged for trustworthy automation.
