Precision DC Source: Low-Noise Remote-Sense Output
← Back to: Test & Measurement / Instrumentation
A precision DC source is “precision” only when its setpoint accuracy, noise, drift, regulation, stability, and protection behavior are controlled at the DUT terminals and can be proven with bandwidth-defined measurements and repeatable wiring (especially 4-wire remote sense). Use it to power or bias sensitive analog circuits and calibration setups without hiding errors behind display resolution or unspecified test conditions.
What a Precision DC Source is (and what it is not)
A precision DC source is not defined by “big power” or “more digits on the display.” It is defined by outputs that remain provably correct and quiet across time, temperature, and wiring—so the DUT sees the intended voltage at the load, not just at the front panel terminals.
Precision means four measurable outcomes
1) Setpoint accuracy
The actual output matches the programmed value within a specified error budget (offset, gain error, and
linearity), not just “fine step size.”
2) Long-term stability
After warm-up, the output holds steady over hours/days (drift/aging/thermal gradients are controlled and
documented), which prevents slow “zero creep” in sensitive measurements.
3) Low noise and ripple
Output noise is specified with bandwidth and conditions (RMS vs peak-to-peak, 10 Hz–10 kHz vs 20 MHz),
because the same source can look “quiet” or “noisy” depending on how it is measured.
4) Tight regulation at the DUT
Load/line regulation and low output impedance keep the voltage stable when the DUT draws dynamic current.
Remote sense makes the value correct at the load, not just at the source terminals.
Do not buy a precision DC source if the job is actually something else
- SMU (Source Measure Unit) need: If the workflow requires four-quadrant operation (source + sink), built-in precision measurement of V/I, or sweep/IV characterization, that is an SMU problem—not a DC source problem.
- Battery / solar simulator need: If the DUT expects an adjustable internal resistance model, fast dynamic impedance shaping, or realistic source droop under pulses, that is a battery/solar simulator problem.
- AC source / linear power amplifier need: If the output must be phase-synchronized sine/arb at power level, or you must emulate mains behavior, that is an AC source/linear amp problem.
Where a precision DC source earns its cost
Analog front-end supply and bias
Noise and drift on the supply can translate into measurement error (offset wander, baseline drift, degraded ADC noise floor).
Reference injection and calibration points
A known, stable, and traceable voltage is required to validate gain/offset, test protection thresholds, or validate converter linearity.
Component evaluation under controlled conditions
When evaluating ADC/DAC/sensors or precision analog chains, stable voltage at the DUT and predictable protection behavior
reduce “measurement system” artifacts.
Quick decision checklist
- If the voltage must be correct at the DUT through long leads, remote sense is non-negotiable.
- If results drift over time/temperature, prioritize drift specs, warm-up behavior, and documented conditions.
- If the DUT is sensitive (AFE/reference), prioritize low-frequency noise (10 Hz–10 kHz) and predictable protection states.
Specs that actually matter
A precision DC source is chosen from its error budget and conditions, not a single headline number. The same instrument can look “excellent” or “terrible” depending on bandwidth, wiring, load dynamics, and warm-up. The goal is to translate each spec into what it breaks and how to test it.
| Spec | What it breaks in real use | Why it happens | How to measure / align conditions |
|---|---|---|---|
| Accuracy (gain/offset/linearity) | Programmed value does not match actual DUT voltage; errors vary across setpoints. | Offset and gain error plus non-linearity (INL), temperature effects, and calibration residuals. | Sweep multiple points (0/10/50/90/100%) with a higher-grade meter/standard; note warm-up time and ambient. |
| Resolution (step size) | False confidence: fine steps but wrong absolute value or poor stability. | A small step is not the same as low error; DAC/trim resolution does not guarantee system accuracy. | Treat resolution as usability; verify accuracy and drift separately with a stable setup and multi-point checks. |
| Noise (RMS / p-p + bandwidth) | ADC noise floor rises; baseline wanders; small-signal measurements become unstable. | Noise integrates over bandwidth; p-p depends on observation time; wiring and probe grounding add artifacts. | Always state BW (e.g., 10 Hz–10 kHz, 100 kHz, 20 MHz), measurement method (FFT or time-domain), and wiring. |
| Drift / tempco (ppm/°C, aging) | Output slowly “creeps” over hours; morning vs afternoon results disagree. | Reference aging, self-heating, thermal gradients in resistors/terminals, and ambient variation. | Log voltage vs time after warm-up under constant load; keep cabling/terminal conditions fixed; record ambient. |
| Load regulation & transient response | Voltage droops or rings on load steps; DUT browns out during current pulses. | Finite closed-loop output impedance; limited control bandwidth; insufficient headroom in post-reg stage. | Use a step load and record droop (ΔV) and recovery time; document step amplitude, edge rate, and sensing mode. |
| Line regulation | Output shifts when mains or pre-reg input changes; repeatability suffers. | Finite PSRR and control loop gain; reference and amplifier sensitivity to supply rails. | Vary input within rated range (or simulate) and measure output delta; note load and thermal steady state. |
| Output impedance (DC to HF) | Oscillation with long leads/large caps; high-frequency droop or ringing persists. | Loop phase margin degrades with cable inductance/capacitance; closed-loop Zout rises at higher frequencies. | Correlate Zout behavior with step response; keep layout consistent; check stability notes for capacitive load limits. |
| Remote sense capability | DUT voltage is wrong due to lead drop; sense wiring causes instability or unsafe behavior if disconnected. | Maximum compensable drop is limited; sense loop picks up noise; open-sense handling varies by design. | Add controlled lead resistance and verify correction range; test open-sense response; document wiring and grounding. |
| Protection behavior (OCP/OVP/OTP) | DUT may be stressed by unpredictable over-current states or recovery surges; start-up may fail. | Different limit modes (constant current, foldback, latch, retry) produce different voltage/current waveforms. | Fault-inject: short, overload, reverse current; observe waveforms and recovery; prefer documented state behavior. |
How to read specs without getting fooled
- Demand conditions: noise is meaningless without bandwidth; accuracy is meaningless without temperature and time since calibration.
- Separate “fine control” from “correct output”: resolution describes step size; accuracy, drift, and regulation describe correctness.
- Predict your failure mode: AFE bias needs low-frequency noise + drift; pulsed loads need transient response + Zout + stable sense wiring.
System architecture map
This map explains how a programmed number becomes the voltage seen by the DUT. A precision DC source is best understood as four coordinated paths: the set path that creates the target, the feedback path that forces the output to match it, the monitor path that proves correctness, and the protection path that overrides behavior during faults.
Set path (what you request)
VREF → DAC/Divider → setpoint node. This defines the target voltage. It impacts resolution and static
linearity, but it does not guarantee the DUT sees the right voltage if the feedback/sense is wrong.
Feedback path (what you actually get)
Sense node → error amp → pass device → output. Loop gain and bandwidth shape output impedance,
transient response, and stability—especially with remote sense, long leads, or capacitive loads.
Monitor path (how precision is proven)
Output/sense → monitor ADC → MCU + calibration table. This is how the instrument verifies setpoint error,
detects drift, and supports self-checks. Without monitoring, “precision” cannot be demonstrated or repeated.
Protection path (fault overrides)
OVP / OCP / OTP / reverse-current → gate control. Protection does not sit “outside” the system; it can
force constant-current, foldback, shutdown, latch, or retry behaviors that change what the DUT experiences.
Typical noise and error injection points (and the symptom they create)
- Reference (VREF): low-frequency noise and drift become slow baseline wander at the output.
- DAC/Divider: code transitions and glitch energy show up as steps, spikes, and long settling tails.
- Error amplifier: offset drift and 1/f noise appear as slow output movement even with a fixed setpoint.
- Pass device / pre-reg supply: finite PSRR lets upstream ripple or load-coupled noise leak into the output.
- Remote sense wiring: pickup and added phase lag can cause ringing/oscillation if wiring and compensation are poor.
A fast way to localize problems using this map
- Start in local sense with short leads. Confirm noise and drift are stable.
- Enable remote sense with the same short leads. If instability appears, the sense loop/compensation is the cause.
- Increase lead length gradually. New ringing points to wiring pickup/phase lag and output impedance limits.
- Add capacitive loading intentionally (within safe limits). If it now oscillates, the feedback path margin is insufficient.
Low-noise reference & drift control
Precision sources cost more because stability is engineered end-to-end: the reference must be inherently stable, the signal chain must not amplify its noise/drift unnecessarily, and the thermal environment must not turn small gradients into large output errors.
Reference families (engineering behavior, not brand names)
Bandgap-based references
Strong integration and cost advantages. System-level filtering and thermal management are often needed to
meet low-frequency noise and drift targets across temperature.
Buried-zener / high-stability arrays
Better long-term stability is achievable, but performance depends heavily on thermal gradients, pre-reg supply
quality, and how the reference is buffered and monitored.
How reference noise and drift become output error
- Noise gain: the output typically scales the reference by a programmed ratio. Any reference noise is multiplied by that ratio, then shaped by the control loop bandwidth.
- Filtering placement: filtering at the reference or setpoint reduces noise but can increase settling time after a step. Filtering in the wrong place can turn DAC glitches into long tails.
- Low-frequency dominance: for precision bias and metrology, low-frequency noise and drift dominate over wideband ripple, so conditions like 10 Hz–10 kHz matter more than a single “20 MHz” number.
Drift is usually a thermal-gradient problem, not a “single-spec” problem
Main heat sources
Pass device dissipation, heatsink gradients, airflow changes, and any upstream pre-reg ripple current.
Sensitive error contributors
Reference package temperature, resistor network temperature coefficient, amplifier offset drift,
and terminal temperature differences (thermal EMF).
Practical mitigation actions
Keep reference and resistor network thermally isolated from power devices, use controlled airflow, and
ensure the terminal region does not experience large gradients during use.
Reference monitoring and self-check hooks
- Reference monitor ADC / window compare: detects out-of-family reference shifts before they appear as output error.
- Temperature sensing near the reference network: links output drift to measured gradients (useful for repeatable calibration conditions).
- Consistency checks: verify calibration table integrity and detect abnormal drift trends across power cycles.
What to record so results are repeatable
- Warm-up time before measurement (and whether airflow/fan speed changes during that time).
- Ambient temperature range and where the instrument sits relative to airflow vents.
- Load level (current) and whether remote sense is enabled; keep lead routing consistent.
Programmable DAC path
A programmable setpoint is only useful when it is predictable. Beyond “bit count,” the DAC path must be evaluated for static correctness (linearity and monotonicity), dynamic usability (glitch and settling), and provability (calibration that holds across time and temperature).
DAC specs that determine whether the setpoint is trustworthy
INL (integral nonlinearity)
INL describes how the programmed curve bends across the range. If INL is large, “correct at endpoints” does not
mean “correct in the middle,” which breaks sweep tests and multi-point comparisons.
DNL & monotonicity
DNL measures step uniformity. Loss of monotonicity creates missing/reversed steps, making small-program updates
unreliable for threshold-based DUT tests.
Glitch impulse (update transient)
During code updates, internal switching can produce a short transient spike. Even if average accuracy is good,
glitch energy can disturb sensitive analog front ends or corrupt time-critical measurements.
Output buffer & reference coupling
The DAC is rarely used bare. Buffer noise, bias currents, and reference input sensitivity can dominate low-frequency
performance, especially when the system filters aggressively for low noise.
Settling time must be tied to a tolerance window (not “looks stable”)
- 0.01% settling means the output stays within a ±0.01% band around the final value. For a 10 V range, that is a ±1 mV window.
- 0.001% settling is a ±0.1 mV window at 10 V—often dominated by low-frequency noise and thermal drift.
- The observed settling depends on measurement bandwidth and filtering. A narrowband meter may hide ringing, while a wideband view exposes it.
Calibration strategies (what each one fixes)
Two-point (offset + gain)
Corrects endpoint errors quickly. It cannot remove mid-range curvature caused by INL, so residual errors vary by voltage region.
Multi-point / segmented linearization
Reduces INL by applying a correction table. Requires stable reference behavior and repeatable conditions to avoid chasing noise.
Temperature binning
Uses different correction sets across temperature ranges. It improves consistency without full closed-loop thermal control, at the cost of complexity.
Low-noise filtering vs update speed (the unavoidable trade)
Filtering the setpoint path reduces noise but increases settling tails after a step. For automated sweeps or threshold tests, the workflow must define a valid read window after each update, rather than sampling immediately.
A practical “valid read window” rule for automation
- Apply the new setpoint (log the target and timestamp).
- Wait at least the specified settling time for the required tolerance (0.01% or 0.001%).
- Only then sample and record. Keep bandwidth and wiring conditions consistent.
Linear post-regulator & pass device design
A linear post-regulator is used because it makes the output quiet and predictable. It can provide high PSRR and well-controlled dynamics, but it also demands headroom and produces heat. A good design balances noise, transient performance, and thermal stability.
Why a linear post-regulator exists in a precision source
- PSRR improvement: attenuates upstream ripple and supply noise before it reaches the DUT.
- Low noise floor: avoids switch ripple injection and enables narrower measurement windows for sensitive DUT rails.
- Controlled dynamics: loop bandwidth and compensation shape output impedance and step response.
The three hard constraints: headroom, heat, and safe operating region
- Headroom: without sufficient (VIN − VOUT), regulation degrades, PSRR collapses, and transient droop increases.
- Heat: dissipation scales as P ≈ (VIN − VOUT) × IOUT. Heat creates thermal gradients that directly increase drift.
- Linear stress: the pass device must tolerate sustained linear-region dissipation. Protection behavior is often designed around this boundary.
Pass device choice (concept-level)
MOSFET pass element
Flexible for wide ranges and fast control. Linear-region dissipation and thermal gradients must be handled carefully to keep drift and stability in check.
BJT pass element
Can offer strong analog behavior and predictable control characteristics. Biasing and thermal management still dominate long-term stability.
Current sensing: shunt placement, Kelvin pickup, and bandwidth/noise trade
- Placement defines meaning: sensing near the output terminals best represents DUT current, while internal placement can include pass-device or protection currents.
- Kelvin pickup: separate sense taps across the shunt reduce lead resistance error and improve repeatability, especially at low voltage drop.
- Bandwidth vs noise: wide bandwidth catches fast transients but adds noise to steady readings; narrow bandwidth stabilizes readings but slows protection response.
Practical decision rules (what to prioritize)
Quiet bias rails / analog validation
Favor low-frequency noise and drift control. Use consistent warm-up and stable airflow to keep ppm behavior repeatable.
Pulsed loads / step testing
Prioritize transient response and output impedance. Ensure headroom remains sufficient during load steps to stay in regulation.
High current operation
Prioritize thermal path and protection behavior. Heat drives drift, so stable gradients matter as much as electrical specs.
Remote sense & cabling
Remote sense is the difference between “the supply output looks correct” and “the DUT actually receives the correct voltage.” It compensates cable and contact drops, but it also extends the control loop through wiring and the DUT input network. A good setup focuses on correct sense points, stable loop behavior, and compensation headroom.
2-wire vs 4-wire (Kelvin): what changes in practice
2-wire (force only)
The supply regulates at its own terminals. Cable and contact resistance cause a current-dependent drop, so the DUT sees a lower voltage:
VDUT ≈ Vset − I × Rlead.
4-wire (force + sense)
The supply regulates using a remote measurement point. If compensation headroom is available, the DUT receives the programmed voltage
even when cable drop changes with load.
Sense stability: why remote sense can oscillate or become noisy
- Pickup and loop area: large wiring loops act like antennas. Sense leads can inject noise directly into the error signal.
- Extra phase lag: cable inductance/capacitance and large DUT input capacitance reduce phase margin, increasing ringing risk.
- Compensation trade: adding a small compensation capacitor or limiting sense bandwidth improves stability but slows response to steps.
Remote compensation limit: “max sense voltage” is a hard boundary
Remote sense can only correct cable drop up to a maximum. When I × Rlead exceeds the available compensation headroom, the force output saturates and the DUT voltage remains low. Symptoms include unexpected droop at higher current and misleading readings if the monitoring point is not truly remote.
Wiring rules that prevent most remote-sense problems
Do
- Twist Sense+/Sense− as a pair and Force+/Force− as a pair.
- Connect sense to the true DUT voltage point (Kelvin point), not a mid-cable node.
- Keep loops small; route sense away from noisy power switching or motor cables.
- Use single-point grounding to avoid ground loops when the DUT has other references.
Avoid
- Shorting sense at the supply terminals (remote sense becomes ineffective).
- Tying sense to force at the wrong location (compensation happens at the wrong point).
- Large loop area wiring (high pickup → noise → ringing/oscillation risk).
Quick debug flow (3 steps)
- Short 2-wire test: verify basic behavior with short leads and stable load.
- Short 4-wire test: confirm correct sense point and stable response before using long cables.
- Scale to real harness: extend cable length and load steps gradually to find the stability boundary.
Protection & fault behavior
Protection is not a checkbox. What matters is how the output behaves after a fault and how recovery works. Different behaviors can be safer for sensitive DUTs or more compatible with automated testing.
Behavior-focused protection checklist (trigger → output shape → recovery)
OVP (over-voltage)
Possible actions include shutdown, clamp, or crowbar-like pull-down. Each changes DUT stress differently:
shutdown removes power, clamp limits voltage while still powering, and aggressive pull-down can create large fault currents.
OCP (over-current / short)
Common modes: constant current, foldback, hiccup retry, or latch-off.
The key question is the terminal voltage waveform during a short and whether the instrument auto-retries.
Reverse current / backfeed
When the DUT is externally powered or has large stored energy, current can flow back into the supply.
Protection may block, shutdown, or transition into a protective state to prevent unexpected internal stress and misleading readings.
Sense/force open detect & OTP
Sense open can be dangerous if the controller tries to “correct” a missing signal by raising force voltage.
Force open detection, temperature sensor placement, and OTP behavior determine whether faults fail-safe or create unstable states.
Practical preferences (DUT safety vs test automation)
- Sensitive DUT: prefer predictable shutdown or foldback and clear user-reset rules, to avoid repeated stress.
- Automated sweeps: avoid behaviors that create hidden retries unless the script explicitly detects and logs the state.
- High-current operation: foldback or well-defined thermal limiting often protects pass devices and reduces drift caused by heating.
A clean recovery policy should be explicit
A precision source should document whether recovery is automatic (retry), requires user reset, or needs a power cycle. This prevents “mystery” behaviors where the DUT repeatedly restarts or measurements silently fail.
Stability, transients & output impedance across frequency
A precision DC source can look perfect on a front panel and still misbehave on certain DUTs: squeal, ringing, oscillation, or large droop during load steps. The practical cause is the interaction between load-step excitation, the source’s output impedance Zout(f), and the control-loop response across frequency.
Load steps: the single relationship that predicts droop and ringing
For a current step ΔI, the initial droop is dominated by the effective output impedance: ΔV ≈ ΔI × Zout(effective band). If Zout(f) has a peak at a certain frequency, that frequency often appears as the ringing tone in the step response. A stable instrument is not “infinitely low impedance” — it is predictably low with controlled peaking.
Complex loads that trigger instability (and the symptoms to look for)
Large Cload at the DUT
- Symptoms: large inrush, slow settling, low-frequency “breathing” or repeated overshoot.
- Why it happens: the loop sees extra phase lag; damping becomes insufficient.
Long cables / remote sense harness
- Symptoms: strong ringing that changes with cable length, or instability that disappears with short leads.
- Why it happens: added L/C and pickup extend the loop and can reduce phase margin.
DC/DC input stages (negative-impedance region)
- Symptoms: squeal/oscillation only at specific operating points; stable on a resistive load.
- Why it happens: the converter can present an effective negative impedance over a band, amplifying Zout peaking.
Compensation strategies (methods + what they trade)
Output capacitor & ESR window (concept)
Improves mid/high-frequency support and reduces droop/ringing. Too little or too much ESR/capacitance can shift the stability boundary.
Damping / isolation resistor (small R)
Adds damping to suppress resonances (especially with cables and input caps). Tradeoff: extra DC drop and power loss at high current.
Bandwidth limiting / slow mode
Reduces sensitivity to high-frequency pickup and phase lag. Tradeoff: slower settling and less agile response to load steps.
How to observe and judge stability (practical tests)
- Scope step response: measure droop, overshoot, ringing frequency, and settling time to a defined window (e.g., ±0.1% or ±0.01%).
- Conceptual loop injection: use it to check trends in stability margin when harness/Cload changes (no deep setup required here).
- Noise bandwidth notes: document measurement bandwidth because “noise” differs dramatically between 10 Hz–10 kHz and wideband views.
Acceptance template (copy/paste for validation logs)
- Load step: ΔI = [ ], from I1=[ ] to I2=[ ], edge time=[ ].
- Droop/overshoot: |ΔV| < [ ] mV, overshoot < [ ] mV.
- Settling: to ±[ ]% in < [ ] ms.
- Harness condition: cable length=[ ], remote sense=[on/off], DUT input C=[ ].
Calibration, verification & traceability
“Precision” is meaningful only when it can be proven. A good workflow separates verification (check against limits) from calibration/adjustment (update correction factors) and preserves traceability with clear records of conditions.
Calibration vs verification (do not mix the terms)
- Verification: measure performance and decide PASS/FAIL against a spec or tolerance band.
- Calibration/Adjustment: change internal coefficients so the output matches a reference more closely, then re-verify.
Key points to calibrate and verify on a precision DC source
- Output voltage accuracy: multi-point sweep (low/mid/high) under defined load and wiring.
- Remote sense correctness: 4-wire configuration checks with documented cable and connection point.
- Protection thresholds: OVP/OCP trigger points plus the observed post-fault output shape and recovery rule.
- Stability validation: step response acceptance under representative harness and DUT input capacitance.
Traceability chain (concept): reference → transfer → instrument under test
Traceability is established when the standards used for comparison have known uncertainty and documented calibration status. A common structure is a higher-grade reference standard, a transfer standard for lab use, and the instrument under test (UUT).
Practical uncertainty contributors (what usually dominates)
- Reference and meter limits: the standard itself and the measurement instrument uncertainty.
- Thermal EMF: terminal temperature gradients and dissimilar-metal junctions create microvolt-level offsets.
- Lead/contact resistance: wiring, clamps, and connectors matter more at higher current and longer harnesses.
- Bandwidth settings: noise and stability conclusions depend on measurement bandwidth and filtering.
Report fields that make results reproducible
- Environment: temperature range, airflow, humidity (if relevant).
- Warm-up: warm-up time for the source and the reference/meter.
- Connection: 2-wire/4-wire, cable length/gauge, fixture/clamp type, Kelvin point.
- Measurement setup: bandwidth/filtering/integration settings, sampling method.
- Load conditions: current points, step conditions for transient checks, input capacitance notes.
- Protection verification: thresholds + observed behavior + recovery method.
Validation & production test checklist
A precision DC source is “done” only when it can be proven across noise, drift, regulation, stability, and protection behavior. Use a three-layer approach: R&D characterization (find boundaries), production limit tests (fast and repeatable), and field self-tests (detect degradation and wiring mistakes).
Release evidence (what “PASS” must include)
- Noise: RMS + spectrum with bandwidth declared (e.g., 10 Hz–10 kHz and a wideband view).
- Drift: warm-up curve, temp tagging, and a repeatable “stability after X minutes” statement.
- Regulation: load and line checks using 4-wire where applicable.
- Stability: load-step acceptance (droop/overshoot/settling) under representative harness and Cload.
- Protection: thresholds + post-fault output shape + recovery rule (CC/foldback/latch/retry).
R&D validation (characterize boundaries, not just typical results)
Noise: RMS + FFT/spectrum (bandwidth declared)
- Measure RMS noise in a defined band (example: 10 Hz–10 kHz) and capture a wideband view for spurs.
- Verify no abnormal tones near control-loop / cable resonance frequencies.
- Pass evidence: RMS < [limit], no unexpected spurs; record BW/filter/integration settings.
Drift: warm-up, thermal gradients, and repeatability
- Run warm-up drift (0–60 min typical) and define a stable window (e.g., “after 30 min”).
- Tag board/terminal temperature; check sensitivity to airflow and heat sources (pass device, heatsink).
- Pass evidence: drift < [ppm] after [T] minutes; tempco < [ppm/°C] under stated conditions.
Regulation: load/line sweeps with 4-wire discipline
- Sweep load and input across corners; compare 2-wire vs 4-wire results.
- Explicitly capture lead resistance/contact effects (Kelvin points, clamp type).
- Pass evidence: load regulation < [µV/A], line regulation < [ppm/V] with wiring declared.
Stability: step response on “difficult” loads
- Test ΔI steps with representative cable lengths and DUT input capacitance.
- Capture droop/overshoot/ringing frequency/settling to ±[X]%. Link ringing to Zout peaking.
- Pass evidence: ΔV droop < [ ], settling < [ ], no sustained oscillation.
Protection: verify thresholds + post-fault waveform + recovery rule
- OVP/OCP: validate threshold accuracy and behavior mode (CC vs foldback vs latch vs retry).
- Reverse feed: verify safe behavior with back-driven outputs (large capacitors / parallel supplies).
- Pass evidence: trip points within limits; waveform matches defined state-machine behavior.
Always record (prevents “precision disagreements” later)
Temperature & airflow, warm-up time, wiring (2-wire/4-wire), cable length/gauge, Kelvin point, measurement bandwidth/filtering,
load condition, and firmware/calibration revision ID.
Production test (fast, repeatable, limit-based)
- 3–5 point voltage check: low/mid/high setpoints at fixed load; PASS/FAIL to limits. Store results with a calibration revision ID.
- Harness & terminal checks: quick lead resistance estimate (Kelvin short fixture) and terminal thermal-EMF sanity check.
- Noise quick screen: RMS window measurement with bandwidth declared; flag abnormal tones or elevated floor.
- Stability spot-check: one representative ΔI step; verify no oscillation and that settling stays within a limit.
- Protection spot-check: verify trip point and behavior (latch/retry) on at least one operating point; log event counters.
Production-friendly design hooks (examples)
- Monitor ADC path: AD7177-2 / AD7175-2 for internal verification and fast limit screens.
- Temperature tagging: TMP117 / ADT7320 near terminals/reference region.
- Nonvolatile log: MB85RC256V (FRAM) or 24AA256 (EEPROM) for cal data + test logs.
Field self-test (power-on confidence + detect wiring mistakes)
- POST (power-on self-test): check reference monitor window, ADC health, and calibration revision ID.
- Sense continuity: detect sense open/short/miswire before enabling aggressive remote compensation.
- Thermal & fan health: verify temperature sensor plausibility and fan tach (if present); safe shutdown path must work.
- Event logging: keep counters for OVP/OCP/OTP trips and repeated retries; trending is often more valuable than a single snapshot.
Concrete PN examples used in self-test chains
- Comparator / threshold: LTC6752 or TLV3201 (speed/voltage dependent) for fast fault detection.
- Reverse-current protection (if required): LTC4359 class ideal-diode control for output back-drive control.
- Supervisor/watchdog: TPS3890 or LTC2937 to keep logs and calibration writes consistent.
Practical BOM anchors (typical part numbers that enable “precision you can prove”)
- Ultra-stable reference: LTZ1000A, ADR1000 (reference class depends on target drift/noise).
- Precision DAC (programmable setpoint): AD5791, AD5781, LTC2758.
- Monitor ADC (verification/log): AD7177-2, AD7175-2.
- Zero-drift amplifier (error/monitor paths): ADA4522-2, OPA189.
- Low-leakage switching (ranges / test routing): ADG1209, ADG1211.
- Temperature sensor: TMP117, ADT7320.
- Calibration/log NVM: MB85RC256V (FRAM), 24AA256 (EEPROM).
- Precision resistors (critical dividers/sense): Vishay VHP/VH102 foil families; current shunts often use Isabellenhütte PBV/BV classes (range dependent).
- Fan control (if needed): MAX31790 class multi-fan controller/monitor (implementation dependent).
Request a Quote
FAQs (Precision DC Source)
These answers focus on what changes real results: accuracy vs resolution, bandwidth-defined noise, drift and warm-up, remote sense wiring, stability with cables/capacitive loads, and protection behavior.
