123 Main Street, New York, NY 10001

UVLO Thresholds for Gate Driver ICs (ON/OFF Margins)

← Back to: Gate Driver ICs

UVLO thresholds are not just “undervoltage protection”—they define whether the gate is driven hard enough to stay out of the under-drive danger zone. A correct UVLO_ON/UVLO_OFF window, plus ripple/droop margin and deterministic recovery, prevents half-conduction loss, chatter, and reset storms.

Intent: turn “UVLO” into an engineering threshold for gate effectiveness

What UVLO Thresholds Really Mean (and Why “Half-Conduction” Happens)

UVLO thresholds are not a generic “low supply” warning. In a gate driver, UVLO defines whether the switch is driven inside a fully-enhanced, low-loss region or forced into an under-drive region where conduction loss rises sharply.

DefinitionUVLOON vs UVLOOFF

  • UVLOON: the supply level where the driver is allowed to begin driving the gate.
  • UVLOOFF: the supply level where the driver must stop driving (to avoid under-drive).
  • The gap between them is hysteresis, the minimum immunity against ripple and noise.

Practical rule: always evaluate valley voltage (worst-case droop + ripple), not only the average.

Why it failsHalf-conduction is an under-drive condition

“Half-conduction” does not mean the gate is half-open. It means the gate drive is insufficient to keep the device in the low-loss region:

VDRV ↓ → VG ↓ → RDS(on)/VCE(sat) ↑ → Pcond ↑ → Tj ↑ → reliability risk ↑

A common trap: UVLO may never assert if UVLOOFF is set too low, yet under-drive still occurs. That produces overheating without a clear “UVLO fault”.

System impactChatter amplifies failures

  • PWM interruption: switching bursts become fragmented when the driver toggles enable/disable.
  • Control wind-up: integrators accumulate error during off periods and overshoot at re-enable.
  • Reset/retry storms: repeated threshold crossings cause repeated restarts and secondary faults.

Engineering takeaway: UVLO thresholds must be set to keep the gate drive out of the under-drive zone during worst-case ripple/droop.

Pass criteria template (fill X/Y/N):

Worst-case VG(valley) stays above the device’s recommended minimum by X V across Y load steps and N minutes of operation (no UVLO chatter).

VDRV waveform showing ripple and brownout crossing UVLO thresholds, plus driver state machine.
UVLO thresholds: valley crossings and driver states A VDRV time waveform with UVLO_ON and UVLO_OFF lines, ripple, a brownout dip, and a 3-state driver mode diagram. VDRV vs time (valley matters) time VDRV UVLO_ON UVLO_OFF brownout event ΔVpp hysteresis Driver mode (simplified) OFF ACTIVE FAULT-LATCH cross UVLO_ON active to off path below UVLO_OFF policy-dependent Key idea: design thresholds for worst-case valley (droop + ripple), not average.
Intent: map UVLO numbers to internal blocks that shape behavior

UVLO Inside a Gate Driver (Comparator, Hysteresis, Latch, Output Stage)

UVLO behavior is determined by an internal chain: reference → comparator → hysteresis → logic/latch → output stage. The same datasheet threshold value can behave very differently depending on how these blocks implement noise immunity, timing, and safe output action.

1Reference + Comparator: where the threshold comes from

  • The reference and comparator define the nominal trip point, plus tolerance and temperature drift.
  • High dv/dt and supply spikes can appear at the pins as fast disturbances; if they reach the comparator input path, false trips become possible.
  • Threshold evaluation should use min/typ/max (not a single “typical” number) to build margin.

2Hysteresis: the minimum anti-chatter window

  • UVLOON and UVLOOFF are intentionally separated.
  • If ΔVpp(ripple) ≥ hysteresis, repeated enable/disable transitions become likely during ripple valleys.
  • Hysteresis is a first-order “guardrail”; it does not replace proper supply integrity and decoupling.

3Latch/Logic: what happens after a trip

  • Some drivers simply disable outputs until VDRV recovers above UVLOON.
  • Others combine UVLO with a fault policy (for example: hold-off, auto-retry, or latch until enable reset).
  • Recovery behavior must be defined in an acceptance test to avoid “reset/retry storms”.

4Output stage action: the practical “safe state”

  • UVLO can force different gate actions: strong pull-down, high-impedance, or a controlled/soft transition.
  • Selection should match the external gate network and the required shutdown predictability.
  • For multi-channel drivers, confirm whether each channel has independent UVLO and how close the thresholds match.

What this chapter enables: a consistent interpretation of UVLO numbers, so later chapters can budget ripple/droop and verify behavior with a repeatable bench procedure.

Implementation details for short-circuit protection, Miller clamp, or bootstrap sizing are intentionally excluded to avoid cross-page overlap.

Internal UVLO signal chain: reference, comparator, hysteresis, logic, and output stage.
UVLO internal chain inside a gate driver Block diagram showing reference and comparator with hysteresis feeding latch/logic and output stage, plus optional filter/delay and multi-channel note. UVLO signal chain (what shapes real behavior) VDD / VDRV filter / delay (optional) Reference tolerance Comparator noise sensitivity Hysteresis UVLO_ON / OFF Latch / Logic retry / hold / latch Output stage action pull-down / Hi-Z / soft Gate pin(s) HS / LS channels (may be independent) sense path Measured outcomes trip point chatter / recovery
Intent: define a consistent UVLO margin budget language

Defining UVLOON / UVLOOFF Budget (Ripple, Droop, Tolerance, Temp)

UVLO thresholds must be evaluated against the worst-case valley at the driver supply pins. The goal is not “average VDRV is OK” but: VDRV(valley) stays safely above UVLOOFF (no under-drive) and VDRV(restore) crosses UVLOON (predictable restart).

Budget objective: ensure Margin_OFF = VDRV(valley) − UVLO_OFF remains positive with headroom under worst ripple, droop, tolerance, and temperature.

All terms must be referenced to measurements at the driver VDD/GND pins, not at a remote regulator node.

Supply floorVDRV,min (droop + distribution)

  • Load-step droop: finite regulator bandwidth, current limit foldback, or transformer secondary sag.
  • Distribution loss: trace resistance, connector drop, and return-path inductance (local ground lift).
  • Corner definition: evaluate at worst load, worst duty/burst mode, and worst temperature.

This term sets the “DC floor” that ripple rides on.

AC disturbanceΔVripple (peak-to-peak)

  • Use peak-to-peak ripple measured at the driver pins with proper probing.
  • Translate ripple into valley impact: the worst valley typically tracks about ΔVpp/2 below the average (unless waveform is highly asymmetric).
  • Ripple must be evaluated during the same switching mode that causes field failures (burst, high duty, hard-switch).

When ripple approaches hysteresis size, chatter risk rises sharply.

Threshold uncertaintyMin/Max + temperature drift

  • Use datasheet UVLOON(min/max) and UVLOOFF(min/max), not “typ”.
  • Temperature shifts both the supply behavior (droop/ripple) and the comparator/reference behavior (effective trip point).
  • Budget must hold across the required operating temperature range: Y °C corner(s).

Decision latencyFilter / delay vs short dips

  • Too little filtering: short spikes can cause false UVLO trips.
  • Too much filtering: real brownouts may be detected late, leaving time in the under-drive zone.
  • Define a practical criterion: “dips shorter than X μs must not trip; dips longer than Y μs must trip.”

This is evaluated by time-aligned VDRV waveforms and output-state transitions.

Reusable UVLO Budget Table (fill X/Y/N)

The table below standardizes how UVLO margin is communicated in design reviews and test reports. All entries refer to the driver supply pins and the worst-case operating mode.

Budget item Definition at driver pins How to obtain Pass criteria (X/Y/N)
VDRV,nom Nominal rail at VDD/GND during steady operation Measure at typical load, typical mode Reference only (not a pass limit)
Droop DC/low-frequency sag under worst load step Worst-case load step at mode N Droop ≤ X V
ΔVripple Peak-to-peak ripple at VDD/GND pins Scope with correct probe and bandwidth ΔVpp ≤ X V
UVLO_OFF (worst) Use UVLO_OFF(max) for margin-to-off evaluation Datasheet max + temperature corner check Use max in margin calculation
UVLO_ON (worst) Use UVLO_ON(max) for restart guarantee Datasheet max + temperature corner check Restart crosses UVLO_ON within Y ms
VDRV(valley) Worst observed supply valley in real switching mode Worst duty/burst + worst load + worst temp VDRV(valley) − UVLO_OFF ≥ X V @ Y °C
No chatter No repeated enable/disable toggles around thresholds Log UVLO/outputs for N minutes 0 events / N minutes

Corner pass criteria (fill X/Y): Margin ≥ X V at Y °C, with no UVLO chatter across the defined operating modes.

Budget stack showing nominal rail reduced by ripple, droop, tolerance, and temperature to compare against UVLO_OFF.
UVLO budget stack to protect the valley A step-down budget stack from nominal VDRV subtracting ripple, droop, tolerance, and temperature effects, compared to UVLO_OFF with a margin placeholder. Budget Stack: protect VDRV(valley), not average Nominal VDRV,nom − Ripple ΔVpp/2 − Droop load step − Tolerance UVLO max Temp shift VDRV(valley) worst case UVLO_OFF (max) Margin ≥ X Pass criteria ≥ X V @ Y°C no chatter
Intent: map device gate-drive rails to UVLO thresholds (avoid under-drive)

Matching Device-Recommended Gate Drive Rails to UVLO (Avoiding Under-Drive)

UVLO thresholds must be aligned to the device’s effective drive region. A driver can stay “enabled” while still operating in an under-drive danger zone if UVLOOFF is set too low. The objective is to use UVLO to exclude the region where conduction loss rises rapidly and reliability degrades.

Threshold-setting logic (fill X/Y):

  • UVLOOFF ≥ (upper edge of under-drive danger zone) + Margin(X)
  • UVLOON ≥ (entry point of full-drive region) + Margin(Y)

“Danger zone edge” and “full-drive entry” must be sourced from the switch vendor’s recommended gate-drive guidance and validated at the intended switching conditions.

IGBTUnder-drive quickly increases loss

If gate drive falls below the recommended region, conduction loss rises steeply. UVLOOFF should be positioned to prevent operation in the high-loss region rather than only reacting to deep brownouts.

Detailed short-circuit behavior and DESAT policies belong to the DESAT page.

SiC MOSFETDrive window and turn-off strength matter

SiC platforms often rely on strong turn-off behavior and tight control of the effective gate voltage. UVLO policies must account for supply valleys that reduce the effective drive margin and can increase susceptibility to dv/dt events.

Negative rail design and clamp details are handled in the SiC driver and Miller-clamp pages.

GaN HEMTNarrow gate window demands strict rail control

GaN gate windows are typically narrower; a “safe full-drive region” should be defined explicitly to avoid both under-drive loss and overvoltage stress. UVLO thresholds should work with rail limits to keep operation inside that region.

Technology-specific gate limits are handled in the GaN driver page.

LV MOSFETChatter is common in burst/skip modes

In synchronous bucks and BLDC stages, burst/skip modes can create large ripple valleys. UVLO hysteresis and restart behavior should be chosen to avoid repeated toggles while still blocking true under-drive.

EMI vs loss tradeoffs for slew control belong to the gate-resistor / slew-control page.

Practical alignment steps (review-ready)

  • Step 1: define the device’s “safe full-drive region” and the “under-drive danger region” for the target switching conditions.
  • Step 2: apply the H2-3 valley budget to compute worst-case VDRV(valley) at the driver pins.
  • Step 3: place UVLOOFF above the danger region edge with margin; verify no operation occurs in the danger region.
  • Step 4: place UVLOON so restart only occurs when the rail returns to the full-drive entry with margin.
  • Step 5: validate with time-aligned waveforms: VDRV, output state, and thermal/power symptoms under worst mode.
2D map of under-drive danger, safe full-drive, and overvoltage risk regions with UVLO_ON and UVLO_OFF placements.
Mapping UVLO thresholds to device drive regions A 2D conceptual map with VDRV on x-axis and effectiveness/loss risk on y-axis, showing under-drive danger, safe full-drive, overvoltage risk, and UVLO_OFF/UVLO_ON boundaries. Drive-region map: keep operation inside “safe full drive” VDRV (gate drive rail) effectiveness / loss risk Under-drive danger zone Safe full drive Overvoltage risk zone UVLO_OFF UVLO_ON +X +Y IGBT SiC LV MOSFET GaN Place UVLO_OFF to block the under-drive zone; place UVLO_ON to permit restart only inside the safe full-drive region.
Intent: define UVLO policy for fast events (duration matters)

UVLO Behavior During Fast Events (Brownout, Load Step, Burst Mode)

Field UVLO incidents often occur even when the average rail looks adequate. The root cause is typically a valley crossing driven by fast events: short spikes, load-step droop, or burst/skip modes that increase ripple depth. UVLO policy must therefore be specified in two dimensions: dip depth and dip duration.

Policy statement: define (1) a minimum dip duration below which UVLO must not trigger (T_ignore) and (2) a duration above which UVLO must trigger (T_must_trip).

This prevents false trips on very short disturbances while still blocking sustained under-drive conditions.

Class AShort spike (very brief dip)

  • Signature: sharp, narrow dip; may repeat with switching edges.
  • Desired behavior: should not trigger UVLO when duration is below T_ignore.
  • Risk if mis-handled: repeated enable/disable toggles that look like “random resets”.

Class BLoad-step droop (medium duration)

  • Signature: dip with recovery; duration depends on regulator bandwidth and local decoupling.
  • Key supply-path contributors: distribution resistance, insufficient local capacitance, return-path inductance.
  • Policy note: behavior may be application-dependent; define an allowed dip window explicitly.

Class CBrownout (long duration)

  • Signature: sustained low rail; often tied to power-source collapse or limit/foldback behavior.
  • Desired behavior: must trigger UVLO and force a defined safe output state when duration exceeds T_must_trip.
  • Risk if mis-handled: prolonged under-drive with elevated loss and thermal stress.

Event Classification Table (fill X/Y)

The table below standardizes acceptance criteria for “dip depth vs duration” across brownouts, load steps, and burst modes. Values are defined at the driver supply pins.

Event class Duration window Allowed dip definition Expected UVLO action
Class A Spike < T_ignore (X μs) May cross UVLO briefly, but must not cause a state toggle Ignore (no disable)
Class B Droop X–Y μs/ms Valley must remain ≥ UVLO_OFF + X V (or follow a defined hold-off rule) Policy-defined (documented)
Class C Brownout > T_must_trip (Y ms) Any sustained dip below UVLO_OFF beyond Y ms Must trip (enter safe state)

Pass criteria template (fill X/Y/N):

  • No chatter: 0 unexpected toggles during N minutes of worst-case burst/load-step operation.
  • Required response: Class C events must force safe output state within X μs.
  • Recovery: restart occurs only after crossing UVLO_ON and meeting the defined stability window Y ms.
Three time-domain dip waveforms with UVLO_OFF line and policy boxes: ignore, policy-defined, must trip.
Fast-event UVLO policy by dip duration Three stacked waveforms: short spike, medium droop, and long brownout, each compared to UVLO_OFF with policy decision boxes. Dip duration + depth define UVLO behavior UVLO_OFF UVLO_OFF UVLO_OFF Class A: spike T_ignore Ignore no toggle Class B: droop X–Y Policy defined Class C: brownout T_must_trip Must trip safe state
Intent: define UVLO meaning for split rails (VDD/VEE) and output state

Split Supplies and Negative Rails UVLO (VDD / VEE / Output-State Interaction)

Many gate driver platforms use split rails: a positive drive rail (VDD) and an optional negative rail (VEE). UVLO must be interpreted per-rail: a “rail undervoltage” is not a single condition. The required output behavior must be stated as a rail-to-behavior mapping to avoid ambiguity in reviews and acceptance tests.

Single-railOne UVLO path

  • Only VDD is monitored for UVLO behavior.
  • Trip/recovery semantics are typically simpler: disable below UVLO_OFF, re-enable above UVLO_ON.

Split-railTwo UVLO paths, multiple meanings

  • VDD undervoltage affects drive strength and the ability to hold the device in the full-drive region.
  • VEE undervoltage can reduce turn-off robustness and increase susceptibility to dv/dt-induced events (risk statement only).
  • Systems must define: which rail triggers which output state (OFF / Clamp / Hi-Z) and how recovery is handled.

Clamp and Miller mechanisms are intentionally not expanded here to avoid cross-page overlap.

Rail-to-Behavior Table (fill X/Y/N)

Use this template to document UVLO meaning when multiple rails exist. Output states are left as placeholders to match the chosen driver policy.

Condition at rails UVLO detection Expected output state Pass criteria (X/Y/N)
VDD OK, VEE OK No UVLO asserted ACTIVE Stable operation for N minutes
VDD LOW, VEE OK UVLO_VDD asserted (per device min/max) OFF / Hi-Z (policy) Disable within X μs; no under-drive operation
VDD OK, VEE LOW UVLO_VEE asserted (if implemented) OFF / Clamp (policy) No false turn-on; output state matches policy across N events
VDD LOW, VEE LOW Both UVLO paths asserted SAFE STATE (policy) Safe state maintained until recovery criteria met
Recovery (rails restored) Cross UVLO_ON (per rail) + stability window RE-ENABLE (policy) Restart only after Y ms stable rails; 0 chatter

Pass criteria template (fill X/Y/N):

  • For each rail undervoltage condition, output state matches the documented policy for N injected events.
  • No unintended turn-on behavior during VEE-low conditions.
  • Re-enable occurs only after both rails satisfy UVLO_ON and remain stable for Y ms.
Split-rail UVLO block diagram: VDD and VEE monitored separately, feeding a policy block that selects OFF/Clamp/Hi-Z and optional fault flag.
Split-rail UVLO mapping to output state Block diagram showing VDD and VEE rails entering a driver, with UVLO_VDD and UVLO_VEE monitor blocks feeding a policy block selecting OFF, CLAMP, or HI-Z, plus an optional fault flag. Split rails: define UVLO per rail and output-state policy VDD (+) VEE (−) Gate Driver power domains UVLO_VDD ON / OFF UVLO_VEE ON / OFF Policy select output state Output state options OFF CLAMP HI-Z /FLT optional
Intent: prevent UVLO chatter by controlling crossings near thresholds

Preventing UVLO Chatter (Noise, EMI, Layout, and Filtering)

UVLO chatter is a repeated threshold crossing that causes output-state toggling. It is not a “random reset” problem; it is a measurable interaction among rail ripple, coupled disturbance, and ground reference movement. The objective is to make UVLO behavior deterministic: no false trips and no repeated toggles.

Definition: chatter = more than X UVLO-related enable/disable transitions within Y seconds under a repeatable operating mode.

A compliant design reports 0 unintended toggles during the defined stress window.

Source 1Excess rail ripple at VDD pins

  • Mechanism: the VDD valley repeatedly crosses UVLO_OFF (and hysteresis window).
  • UVLO-view symptom: toggling correlates with ripple envelope and switching mode (burst/skip/high duty).
  • UVLO knobs: local decap layering, short supply loop, rail filtering near VDD pins.

Source 2Common-mode injection into REF/UVLO sense (CMTI-related)

  • Mechanism: dv/dt coupling shifts internal reference/sense nodes, creating an “apparent” rail dip.
  • UVLO-view symptom: chatter appears in high dv/dt tests and disappears when edge rate is reduced.
  • Scope boundary: coupling physics belongs to the CMTI page; this chapter only defines UVLO-facing checks and knobs.

Source 3Ground bounce / return crossing splits

  • Mechanism: PGND and local reference ground move relative to each other during di/dt events.
  • UVLO-view symptom: VDD-to-GND at the pin looks lower even if the remote rail node is stable.
  • UVLO knobs: Kelvin PGND for sensing/control ground, avoid return paths crossing partition boundaries.

UVLO-Specific Mitigation Knobs (actionable, scope-limited)

  • RC filtering (policy-defined): set a defined T_ignore window to reject ultra-short dips while still tripping on sustained brownouts.
  • Decoupling layering: local high-frequency capacitance at VDD pins plus bulk support, minimizing the loop area of the VDD/PGND path.
  • Kelvin PGND reference: keep the UVLO sense reference tied to a quiet local return rather than high-current power return.
  • Hysteresis strategy: ensure UVLO_ON–UVLO_OFF separation is sufficient relative to observed ripple at the pins.

Pass criteria (fill X/Y/N):

  • 0 unintended UVLO toggles during N minutes in worst switching mode.
  • Class-A dips shorter than T_ignore do not toggle the output state.
  • When UVLO legitimately trips, the transition is single-shot (no chatter) and recovery follows the documented policy.
Coupling-path block diagram showing VDD decoupling loop, dv/dt injection to REF/UVLO, and ground-bounce path, with mitigation knobs panel.
UVLO chatter sources and mitigation knobs Diagram of driver VDD decoupling network and supply loop with arrows indicating ripple path, dv/dt common-mode injection path to REF/UVLO, and ground-bounce path. A right-side panel lists mitigation knobs. Chatter = repeated crossings near UVLO thresholds Gate Driver UVLO sense REF / UVLO VDD PGND VDRV source reg / iso bias C_local near pins C_bulk energy ripple dv/dt CMTI path ground bounce Mitigation knobs RC filter local decap Kelvin PGND hysteresis
Intent: define safe output state at UVLO and pin hierarchy (/EN, /FLT)

Safe State Definition (Outputs at UVLO, and How to Wire /EN /FLT)

UVLO must be treated as a predictable system contract. When UVLO is asserted, the output must enter a clearly defined safe state, external control pins must not “fight” the internal lockout, and fault signaling must remain observable. This chapter defines the UVLO-facing behavior only and avoids mixing other fault sources.

ContractOutput state at UVLO

  • Enumerate the safe state: Gate LOW / Hi-Z / Clamp (policy-defined).
  • Timing: enter safe state within X μs after UVLO assertion.
  • Hold rule: remain in safe state until recovery criteria are satisfied (no oscillation).

Hierarchy/EN vs UVLO vs /FLT

  • UVLO: internal lockout layer; rail not valid ⇒ drive not allowed.
  • /EN (/DIS): external strategy layer; rail valid ⇒ system may still keep outputs disabled.
  • /FLT: observability layer; reports lockout state (commonly open-drain, policy-defined).

RecoveryRestart behavior

  • Auto-recover: restart after crossing UVLO_ON and meeting stability window Y ms.
  • Latch policy: restart requires explicit /EN transition or reset action (policy-defined).
  • No-chatter rule: recovery must not re-enter UVLO repeatedly under the same disturbance.

Fault-Path Checklist (fill X/Y/N)

Use this checklist as an acceptance gate to ensure UVLO behavior, external control, and fault reporting remain consistent across boards, labs, and firmware versions.

/FLT pull-up present (X kΩ)
/FLT polarity documented
/EN default state defined
UVLO entry time ≤ X μs
Recovery window = Y ms stable
0 chatter events / N minutes
Safe state = LOW / Hi-Z / Clamp
Logs capture UVLO + /EN + /FLT

Pass criteria (fill X/Y/N):

  • UVLO assertion forces the documented safe output state within X μs.
  • System remains stable for N minutes with 0 unexpected state toggles.
  • Recovery occurs only after rails cross UVLO_ON and remain stable for Y ms.
State machine diagram for Disabled/Enabled/UVLO-Lockout/Recover and pin relationship arrows for /EN input and /FLT open-drain output.
UVLO safe-state contract and pin hierarchy A simple state machine with four states: Disabled, Enabled, UVLO-Lockout, Recover. Arrows labeled with /EN and UVLO conditions, plus a pin mini-map showing /EN input and /FLT open-drain output. Define a predictable UVLO safe state and recovery policy Disabled /EN = 0 Enabled rails OK UVLO-Lockout safe state Recover stability /EN high /EN low UVLO asserted rails OK Y ms stable latch? (policy) Pins /EN /FLT open-drain
Intent: make UVLO measurable and acceptance-ready (bench + corners)

How to Verify UVLO Thresholds (Bench Procedure + Corner Cases)

UVLO verification must use a consistent trigger definition. Thresholds are not “datasheet numbers” until they are measured at the VDD pin and tied to an observable output-state transition. This section defines a reusable bench procedure and a corner matrix that maps directly to acceptance criteria.

Trigger definitions (policy-aligned):

  • UVLO_OFF (falling ramp): VDD(pin) at the moment the output enters the documented safe state (LOW/Hi-Z/Clamp).
  • UVLO_ON (rising ramp): VDD(pin) at the moment the output is allowed to resume drive under the documented recovery policy.

/FLT may be captured as an auxiliary indicator, but output-state transition is the primary trigger to avoid false pass/fail from wiring errors.

Baseline Bench Procedure (slow ramp)

  1. Instrument the correct nodes

    Measure VDD(pin), OUT, and optional /FLT. Avoid remote-supply nodes as the threshold reference.

  2. Set a slow, controlled ramp

    Use dV/dt = X (rising) and dV/dt = X (falling) so the measurement reflects static thresholds rather than fast-event filtering.

  3. Apply the trigger rule

    Record the exact VDD(pin) when OUT enters safe state (UVLO_OFF) and when drive resumes per policy (UVLO_ON).

  4. Repeat for statistics

    Repeat N cycles and report min/typ/max for UVLO_OFF, UVLO_ON, and hysteresis (ON–OFF).

  5. Check for unintended toggling

    Confirm 0 chatter during steady operation around thresholds (no repeated enable/disable flips).

Corner Matrix (temperature, ripple, fast brownout)

CornerTemperature

  • Conditions: T_cold = X°C, T_hot = Y°C.
  • Action: run the baseline ramp at each temperature and record drift in UVLO_OFF/UVLO_ON.
  • Pass: margins remain ≥ X V and recovery policy remains deterministic.

CornerRipple injection

  • Conditions: ripple amplitude X Vpp, frequency Y (sine or switching-like).
  • Action: superimpose ripple at VDD and observe threshold crossings and chatter near UVLO_OFF.
  • Pass: no unintended toggles; ripple headroom remains within the defined budget envelope.

CornerFast brownout pulse

  • Conditions: dip depth X V, duration Y μs/ms.
  • Action: inject dip pulses and verify T_ignore vs T_must_trip behavior.
  • Pass: short dips do not toggle; long dips force safe state within X μs.

Measurement Record Table (min/typ/max + pass criteria)

Use the table below as an acceptance-ready record. Values are measured at VDD(pin) and referenced to the documented safe state.

Condition Ramp dV/dt UVLO_OFF (min/typ/max) UVLO_ON (min/typ/max) Hys (ON–OFF) Pass criteria (X/Y/N)
Room X min/typ/max min/typ/max min/typ/max Margin ≥ X V; chatter=0 / N min
Cold (X°C) X min/typ/max min/typ/max min/typ/max Recovery stable for Y ms; deterministic
Hot (Y°C) X min/typ/max min/typ/max min/typ/max Safe state entry ≤ X μs
Ripple (X Vpp) observe observe 0 unintended toggles
Brownout pulse trip / ignore recover Meets T_ignore/T_must_trip
Bench setup block diagram: programmable PSU to DUT, scope measuring VDD pin and OUT, optional /FLT, with ripple injection and brownout pulse modules and test points.
UVLO bench verification setup Block diagram showing programmable PSU feeding DUT, oscilloscope probing VDD(pin) and OUT, optional /FLT, with ripple injection module and brownout pulse module connected into the VDD path, and labeled test points TP1–TP3. Bench: measure VDD(pin) at output-state transition PSU programmable Ripple inj X Vpp Brownout pulse DUT gate driver UVLO SCOPE VDD / OUT OUT VDD(pin) /FLT (opt) 1 TP 2 TP 3 TP
Intent: convert UVLO requirements into executable gates (Design → Lab → Production)

Engineering Checklist (Design → Bring-up → Production)

This checklist turns UVLO thresholds into a repeatable engineering gate. Each phase produces artifacts that can be reviewed and accepted: budget and policy definitions in design, measured evidence in bring-up, and screening limits in production.

Design Gate

  • Threshold budget completed Margin ≥ X V @ Y°C; includes ripple/droop/tolerance stack.
  • UVLO_OFF / UVLO_ON policy documented Defines safe-state entry and recovery rule (auto vs latch).
  • Fast-event strategy defined T_ignore / T_must_trip thresholds specified for dips.
  • Split-rail behavior mapped (if applicable) Per-rail UVLO conditions → output state (LOW/Hi-Z/Clamp).
  • Chatter risk evaluated Mitigation knobs selected (decap, Kelvin PGND, hysteresis, filtering).

Bring-up Gate

  • Ramp thresholds measured at VDD(pin) UVLO_OFF/ON min/typ/max recorded; hysteresis computed.
  • Ripple injection stays within contract 0 unintended toggles; validated at X Vpp / Y.
  • Brownout pulses meet event policy Short dips ignored; long dips trip and enter safe state ≤ X μs.
  • Safe state is predictable Output state matches definition; recovery window = Y ms stable rails.
  • Observability is consistent /FLT and logs reflect UVLO events (auxiliary indicator, not sole trigger).

Production Gate

  • ATE screening window defined Threshold window limits derived from lab min/max.
  • Pin-state combinations verified /EN and /FLT behavior consistent with UVLO lockout and recovery policy.
  • Recovery policy check included Auto vs latch behavior validated in a shortened production-friendly sequence.
  • Traceability enabled Serial ID → measured thresholds → pass/fail stored for audits.

Phase pass criteria (fill X/Y/N):

  • Design: margin ≥ X V @ Y°C; policies documented and reviewable.
  • Bring-up: measured thresholds within window; 0 chatter / N minutes; safe-state timing ≤ X μs.
  • Production: ATE windows and pin-state checks prevent shipment of units that violate UVLO contract.
Flow diagram with three pipeline stages: Design, Lab, Production, each with checkbox nodes, plus an artifacts column.
Engineering gate flow: Design → Lab → Production Three-stage pipeline diagram. Each stage contains 4–6 checkbox nodes. A right-side column lists artifacts such as budget sheet, waveforms, record table, and ATE limits. Executable gates: Design → Lab → Production Design Lab Production Budget Margin Policy Rail map Chatter risk Ramp test Ripple inj Brownout Safe state Log ATE limits Pin states Recovery Trace
Intent: UVLO-only application playbooks (no system-level detours)

Applications (Where UVLO Thresholds Make or Break You)

Applications differ in how UVLO is triggered and how recovery must behave. Each playbook below stays strictly on a UVLO axis: trigger shapesystem consequencepolicy & guardrails.

Application3-Phase Inverter

  • Trigger shape: auxiliary bias droop or DC-bus disturbances reduce VDD(pin) below UVLO_OFF.
  • Consequence: UVLO chatter breaks PWM determinism, causing repeated enable/disable and restart storms.
  • Policy & guardrails: define recovery window = Y ms (stable rails) and require 0 toggles / N min around threshold; prefer deterministic safe-state entry (LOW/Hi-Z/Clamp).

Example BOM part numbers (gate-driver-focused):

Exact UVLO numbers and behavior depend on variant and must be verified in the datasheet.

TI UCC21750 TI UCC21520 ST STGAP2SICS ST STGAP2HD Infineon 1EDC20I12MH Analog Devices ADuM4121 Silicon Labs Si827x

ApplicationPFC + LLC (Burst / Light-load)

  • Trigger shape: burst/skip operation creates a ripple envelope that crosses UVLO_OFF even when average VDD looks sufficient.
  • Consequence: repeated UVLO entry re-triggers startup and produces non-deterministic restart timing.
  • Policy & guardrails: specify an allowed ripple window such that the VDD valley stays ≥ X V above UVLO_OFF; require hysteresis/filter behavior consistent with T_ignore/T_must_trip.

Example BOM part numbers (gate-driver-focused):

Use these as starting points for UVLO behavior screening; verify UVLO thresholds and recovery mode per datasheet.

TI UCC27714 TI UCC27211 Infineon IRS21867 Infineon IRS2003 onsemi NCP51530 ST L6398 TI UCC24612 (SR)

ApplicationPOL / VR (Multiphase)

  • Trigger shape: one phase experiences larger local droop/ripple and hits UVLO earlier due to decoupling/layout asymmetry.
  • Consequence: phase drop-out forces current redistribution and can create hot-spots and phase imbalance.
  • Policy & guardrails: enforce threshold consistency (ΔUVLO ≤ X mV, placeholder) and a deterministic “phase-safe” response when a single channel locks out.

Example BOM part numbers (driver & multiphase building blocks):

Driver selection here prioritizes channel consistency and predictable lockout behavior; verify UVLO window and recovery policy per datasheet.

TI UCC27284 TI UCC27531 TI UCC27614 onsemi NCP81074 (multiphase ctrl) onsemi NCP302155 (DrMOS) Infineon TDA21490 (DrMOS) Renesas ISL99390 (DrMOS)

Application-to-Risk Mapping (UVLO axis)

Use the mapping below to keep application discussions UVLO-only. Each line indicates which UVLO failure mode dominates: droop, ripple, chatter, restart.

Application mapping diagram. Left side: 3-Phase Inverter, PFC+LLC, POL/VR blocks. Right side: droop, ripple, chatter, restart nodes. Lines connect each application to its dominant UVLO risks.
Applications mapped to UVLO risk modes Left column shows three applications. Right column shows four UVLO risk modes: droop, ripple, chatter, restart. Connection lines indicate which risk modes dominate for each application. UVLO application mapping (risk-only) 3-Phase Inverter bias droop / restart PFC + LLC burst ripple / chatter POL / VR phase UVLO mismatch droop ripple chatter restart Knobs: margin · hysteresis · filter · policy
Intent: UVLO-only decision tree + concrete part numbers

IC Selection Logic (Choosing UVLO Thresholds and Behaviors)

Selection must stay UVLO-specific: threshold shape, hysteresis, filtering/time behavior, safe-state output, recovery policy, and multi-channel threshold consistency. The decision tree below starts from the rail architecture and ends with an acceptance-ready UVLO contract.

Decision tree entry rules (UVLO axis only):

  • Input: rail architecture (single / split ± / isolated bias) and dominant event type (droop / ripple / fast brownout).
  • Output: UVLO type (fixed or programmable), behavior (latch or auto), and required margins (X/Y placeholders).
  • Verification: every selected behavior must be bench-verifiable with the ramp + corner matrix.

UVLO-Related Spec Checklist (only)

Selection item Why it matters (UVLO-only) Acceptance / pass criteria (X/Y/N)
UVLO_ON / UVLO_OFF range Must avoid under-drive region and prevent half-conduction during droop. Margin ≥ X V at valley; meets device drive window.
Independent ON/OFF thresholds Defines an effective window against ripple without sacrificing protection intent. Hysteresis window ≥ X V equivalent headroom.
Hysteresis size Primary anti-chatter mechanism when ripple sits near UVLO_OFF. 0 toggles / N minutes under injected ripple.
UVLO filter / delay / response Defines T_ignore vs T_must_trip for fast events. Short dips ignored; long dips trip within X μs.
UVLO safe-state output Determines what the gate does at lockout: LOW, Hi-Z, or Clamp. Safe state matches system contract; no ambiguous mid-level state.
Recovery policy Auto vs latch defines whether the system chatters or recovers deterministically. Recovery window = Y ms stable; deterministic state transitions.
Multi-channel threshold consistency Prevents “one channel drops first” failures in multiphase and multibridge systems. ΔUVLO_ON/OFF ≤ X mV across channels (placeholder).

Concrete Part-Number Shortlist (UVLO-focused)

The parts below are grouped by UVLO needs. Use them as concrete BOM candidates and filter by required UVLO window and policy in the datasheet.

UVLO need Example part numbers (BOM candidates) Typical fit (UVLO axis)
Isolated driver with strong UVLO contract TI UCC21750, TI UCC21520, ST STGAP2SICS, ST STGAP2HD, Infineon 1EDC20I12MH, ADI ADuM4121, SiLabs Si827x Inverters where droop and restart policy dominate
Bootstrap half-bridge driver with UVLO TI UCC27714, TI UCC27211, Infineon IRS21867, Infineon IRS2003, onsemi NCP51530, ST L6398 PFC/LLC bridges where ripple/chatter must be controlled
Low-side driver with defined UVLO behavior TI UCC27531, TI UCC27614, Microchip TC4420, Microchip TC4427, onsemi NCP81074 (controller-side gating) POL/VR and low-side stages where local droop causes early lockout
SR driver (secondary-side UVLO sensitivity) TI UCC24612, TI UCC24610, onsemi NCP4306 LLC secondary where burst behavior can create chatter windows
DrMOS integrated driver (phase consistency focus) Infineon TDA21490, Renesas ISL99390, onsemi NCP302155 Multiphase VR where per-phase UVLO mismatch is costly

Note: the shortlist is intentionally UVLO-focused. Exact UVLO thresholds, timing filters, and recovery behavior vary by option code and must be confirmed in the datasheet before freezing the BOM.

Decision Tree Diagram (rails → knobs → recommended UVLO shape)

Decision tree diagram. Inputs: single rail, split rails, isolated bias. Middle decisions: ON/OFF independent, hysteresis, filter/delay, safe state, recovery policy, channel consistency. Outputs: fixed/programmable UVLO, latch/auto, margins X/Y, and example parts list.
UVLO selection decision tree Three-column flow. Left inputs are rail architectures. Middle column lists UVLO knobs. Right outputs are recommended UVLO forms and example BOM parts. Arrows connect inputs through decisions to outputs. Rails → UVLO knobs → Recommended UVLO form Inputs (rails) Single rail VDD only Split rails VDD / VEE Isolated bias secondary rail Knobs (UVLO-only) ON/OFF independent? Hysteresis size Filter / delay Safe state Recovery policy ΔUVLO consistency Outputs Recommended UVLO fixed / programmable margin X / Y Behavior latch / auto window Y ms Example parts UCC21750 · STGAP2SICS ADuM4121 · Si827x UCC27714 · IRS2003 NCP51530 · UCC27531

Request a Quote

Accepted Formats

pdf, csv, xls, xlsx, zip

Attachment

Drag & drop files here or use the button below.

FAQs (UVLO Thresholds)

Scope is strictly UVLO: thresholds, ripple, droop/brownout, chatter, and recovery policy. Each answer is fixed to 4 lines: Likely cause / Quick check / Fix / Pass criteria (X/Y/N placeholders).

UVLO never triggers, but the switch clearly overheats—why “half-conduction” still happens?
Likely cause UVLO_OFF is too low versus the device’s under-drive danger zone, so the gate is under-driven without reaching lockout.
Quick check Measure worst-case V_G at the device during maximum droop (probe at driver pins and Kelvin source/return); compare to vendor-recommended minimum drive.
Fix Raise UVLO_OFF or improve rail integrity (reduce droop/ripple). If selecting a driver, screen UVLO behavior on candidates such as UCC21520, STGAP2HD, ADuM4121 families.
Pass criteria Worst-case V_G ≥ (recommended minimum − X V) across Y load steps; no thermal rise beyond limit during N minutes at worst conditions.
UVLO chatters during PWM bursts, average VDRV is fine—what’s wrong?
Likely cause Ripple valley crosses the effective UVLO window; hysteresis is too small for the burst envelope.
Quick check Scope VDRV/VDD(pin) with sufficient bandwidth; capture min valley versus UVLO_OFF during worst burst pattern.
Fix Strengthen local decoupling at driver pins and/or add UVLO-side RC filtering; choose larger hysteresis or shift thresholds so valley margin increases.
Pass criteria No UVLO toggles for Y minutes under worst burst; valley margin ≥ X V to UVLO_OFF for N cycles.
Cold start fails (UVLO stuck), warm start is OK—first suspect what?
Likely cause Cold-condition headroom collapses (converter current limit, diode drop, ESR rise), so VDRV never crosses UVLO_ON.
Quick check Measure VDRV ramp and plateau at Tmin; record whether/when UVLO_ON is crossed at VDD(pin).
Fix Improve cold-start rail (soft-start tuning, higher headroom, lower-drop path, better decoupling near pins); avoid operating near threshold at Tmin.
Pass criteria UVLO_ON crossed within X ms at Tmin with load Y; no stall across N cold-start trials.
UVLO triggers only during hard-switch dv/dt events—supply looks stable on slow scope.
Likely cause Fast common-mode injection creates local spikes at the UVLO comparator/reference node; slow probing or distant probing hides the event.
Quick check Probe directly at driver VDD/GND pins with short ground; correlate UVLO events with dv/dt edges and look for coincident spikes.
Fix Tighten return/decoupling loop at the pins, add local filtering, and enforce partitioning so dv/dt currents do not share the UVLO reference return.
Pass criteria No UVLO events at dv/dt = X kV/µs for Y switching conditions over N cycles; observed pin spikes ≤ X mV (placeholder).
Driver restarts repeatedly after a brownout—system enters “reset storm.”
Likely cause UVLO auto-retry occurs while the rail hovers near threshold; insufficient hysteresis/holdoff causes repeated crossings.
Quick check Log VDRV vs /EN vs output state; confirm repeated UVLO_OFF/ON crossings within short intervals.
Fix Add hysteresis/holdoff and enforce “disable-until-stable” (rail must recover with margin before re-enable); limit retries at system level.
Pass criteria For a brownout event: max retries ≤ X; stable recovery within Y s; no oscillation for N subsequent cycles.
Two channels have different UVLO behavior—one leg turns back on earlier.
Likely cause Threshold mismatch and/or different local supply impedance (layout/decoupling asymmetry) shifts per-channel crossings.
Quick check Measure per-channel VDD(pin) and ground bounce during events; record each channel’s UVLO_ON/OFF crossing points using the same ramp method.
Fix Enforce symmetry in routing/decoupling and Kelvin returns; if needed, choose a driver with tighter channel matching (e.g., UCC21520, STGAP2HD families).
Pass criteria Channel mismatch: ΔUVLO_ON ≤ X mV and ΔUVLO_OFF ≤ Y mV (or ≤ Y%) across temperature; no leg-asymmetry events in N stress cycles.
UVLO_OFF seems “too high,” causing nuisance trips—should we lower it?
Likely cause Rail ripple/droop is the real problem; lowering UVLO_OFF reduces protection margin and increases under-drive loss risk.
Quick check Compare trip rate versus measured valley margin and estimate loss increase at reduced V_G during droop; confirm whether nuisance trips correlate with poor rail integrity.
Fix Fix rail integrity first (decouple/regulate, reduce impedance at driver pins). Only after valley margin is controlled, adjust threshold if still necessary.
Pass criteria Zero nuisance trips across Y minutes and N events while maintaining V_G margin ≥ X V above minimum recommended drive.
Negative gate rail exists, but UVLO only monitors +VDD—what’s the risk?
Likely cause −rail droop weakens turn-off, reducing effective negative bias and increasing dv/dt-induced false turn-on susceptibility.
Quick check Monitor VEE during worst transients and correlate with ringing/false turn-on symptoms (gate rebounds) while confirming +VDD stays above UVLO_OFF.
Fix If available, select/enable −rail UVLO monitoring; otherwise regulate and decouple the −rail so its deviation stays bounded at the driver pins.
Pass criteria VEE stays within X V of target for Y worst events; no false turn-on in N stress cycles.
Datasheet UVLO_ON/OFF is specified, but our measured values differ—why?
Likely cause Definition mismatch (trigger point, load condition, ramp rate, probe location) shifts the observed crossing.
Quick check Standardize ramp dV/dt, define the trigger as “output state changes,” and probe at VDD(pin); repeat across the same temperature points.
Fix Re-run with the agreed method and document it; avoid comparing “board rail” to “pin rail” values; keep probe loop short to avoid artifacts.
Pass criteria Measured UVLO_ON/OFF falls within datasheet min/max when using the defined method; method + ramp rate recorded in the report (no ambiguity).
Bootstrap-based high-side trips UVLO at high duty—what to check first (without doing full bootstrap math)?
Likely cause High-side VDD valley dips below UVLO_OFF due to insufficient refresh time or added droop under high duty.
Quick check Capture high-side VDD(pin) valley versus time at worst duty and switching conditions; confirm whether the valley crosses UVLO_OFF.
Fix Ensure bias headroom/refresh strategy first; then (if needed) follow the dedicated bootstrap sizing workflow. Candidate HB drivers include UCC27714, IRS21867, NCP51530 families.
Pass criteria High-side valley margin ≥ X V at duty Y% across N cycles; no UVLO events in the defined worst-case window.
EMI fix (adding series impedance) increased UVLO trips—why?
Likely cause Added impedance increased local droop/ripple at the driver supply during switching and bursts, shrinking UVLO valley margin.
Quick check Compare VDD(pin) droop and ripple before/after the EMI change under identical load/switching conditions.
Fix Move impedance upstream (away from driver pins) and keep low-Z local decoupling at the driver; re-balance filtering so UVLO window is not crossed.
Pass criteria Pin droop reduced by ≥ X% and valley margin ≥ X V; no UVLO events in N consecutive stress cycles.
How do we write an acceptance criterion for UVLO in a test report?
Likely cause The report lacks a margin-based definition and only quotes typical thresholds, which is not acceptance-ready.
Quick check List worst-case conditions (temperature, ripple, droop/brownout event) and identify the measurement definition (trigger point, ramp rate, pin probe).
Fix Specify UVLO window + margins + no-chatter requirement + recovery time, all tied to a documented bench procedure.
Pass criteria UVLO_ON/OFF within X–Y (datasheet window); valley margin ≥ X V; no chatter (0 toggles) for Y minutes; recovery time ≤ N.
Data structure rule: Each FAQ is exactly four lines and uses placeholders (X/Y/N) so test plans and acceptance reports can reuse the same contract.