← Back to: eFuse / Hot-Swap / OR-ing Protection
Inrush & EMI Control — What This Page Solves
This page focuses on three practical levers before you head to the chamber. First, shape soft-start dV/dt to keep the inrush current under a fixed budget, so plug-in events do not brown out upstream rails. Second, damp ringing through a combination of spread-spectrum and a right-sized RC snubber so pre-scan peaks fall below typical limits. Third, hold the load-end voltage over long harnesses by enabling remote-sense cable-drop compensation with a “do-not-over-compensate” limiter to avoid light-load oscillation. Typical pilot-build symptoms include plug-in resets, upstream droop, pre-scan failures, and visible drop at the load after a harness change. Start with an inrush budget and derive a safe dV/dt; identify the ringing frequency to size an RC snubber; estimate ΔV across the cable and enforce a capped compensation window. Boundary: this page intentionally excludes SOA/derating and hiccup details (Hot-Swap), as well as dual-source OR-ing and priority windows (Ideal-Diode/OR-ing).
- Plug-in resets and upstream rail droop in early builds;
- EMI pre-scan peaks fail without clear root cause;
- Load-end voltage sags with longer cables or new harness SKUs;
- Light-load overshoot when compensation is aggressive.
Decision path: set inrush budget → derive dV/dt; measure ringing frequency fr → size RC snubber; estimate ΔVcable → enable remote-sense with a limiter.
Soft-Start dV/dt and Inrush Shaping
Start from a peak-current budget and derive the soft-start slope. Use the gate network (R_g/C_ss) and the driver’s source/sink capability to translate that slope into real hardware values, while checking upstream UV/OV monitors and the distribution of C_load (output, harness capacitance, and load input caps).
Inrush: I_inrush ≈ C_load * (dV/dt)
Gate charge time: t_g ≈ Q_g / I_chg, where I_chg = (V_drv − V_th) / R_g
Cap-limited slope: dV/dt ≈ I_src / C_load (controller current limited into load cap)
- Budget: define
I_inrush,maxand the allowed time window. - Estimate C_load: output cap + harness capacitance + load input cap (10–470 µF typical).
- Compute dV/dt: ensure
dV/dt ≤ I_inrush,max / C_load. - Map to R_g / C_ss: pick initial values from driver source/sink specs; iterate from scope traces.
- Check UV/OV: verify slow ramp doesn’t trigger upstream monitors.
- Thermal check: estimate MOSFET transient loss & Tj peak during inrush.
- Too-slow dV/dt can trip upstream monitors into false faults.
- Large
C_load+ cold start can still overshoot a cap-only strategy. - Gate loop coupling “modulates” slope — stabilize locally with RC and tight return.
Bench checklist:
- Confirm
dV/dt ≤ I_inrush,max / C_loadfrom scope traces (10–90% rise, Ipk). - Verify upstream UV/OV stays inactive over the full ramp.
- Estimate transient loss and Tj peak for the inrush window.
Spread-Spectrum and RC Snubber
Standardize the flow from ringing identification to RC snubber sizing and thermal check, then apply spread-spectrum within a safe window. Keep layout and return paths tight to make the math hold on hardware.
Step A — Identify ringing frequency
- Use a near-field probe + oscilloscope FFT on the switch node; typical fr is 20–80 MHz.
- If hot-plug adds higher-frequency edges, lock the dominant mode first.
Step B — Snubber initial values
- Corner frequency:
f_c = 1/(2π·R_s·C_s) ≈ 0.5–0.7 · f_r(start here, then trim on bench). - Energy:
E = 0.5 · C_s · V^2; Power:P ≈ E · f_event(steady-state vs. plug-in). C_slimited by EMI benefit and thermal budget;R_ssets damping and dissipation.
Step C — Spread-Spectrum (controller option)
- Safe window (starting point): depth ±3–6%, modulation 10–40 kHz.
- Re-measure loop phase margin (Bode) after enabling; verify sampling/sync behavior.
Layout & return strategy
- Place the snubber at the switch node with shortest return; bundle vias; close the return locally.
- Segregate gate drive vs. power loop; keep measurement pads near pins.
- At connectors, keep beads on the source side; TVS/RC on the load/interface side to avoid self-oscillation.
Cable-Drop Compensation & Remote Regulation
Keep load-end voltage within target across harness SKUs, temperature, and current. Prefer hardware remote-sense and add a limiter in firmware/PMBus so compensation cannot overshoot. Reduce gain at light load to avoid oscillation and protect against sense open/reversal.
Model — ΔV = I · (R_out + R_return); copper tempco α ≈ 0.0039/°C. Reserve ±15% initial tolerance for harness SKUs and drift.
Remote-Sense first
Kelvin or PGOOD-sense with small RC; clamp and limit the sense path to survive open/reversal.
Limiter
Enforce a max compensation window in MCU/PMBus; drop gain under light load or disconnect.
Cross-brand migration
Keep the compensation window invariant; widen only with bench evidence.
Seven-Brand Mapping (Field-Driven)
Fields: Soft-Start (internal/external), Config dV/dt, Gate Driver (src/sink mA), Spread-Spectrum (depth/rate), Snubber App Note, Remote/Load-Side Sense, Cable-Drop Compensation, AEC-Q100, I²C/PMBus Telemetry. Spread-spectrum and remote-sense are typically provided by downstream regulators/PMICs rather than the eFuse/Hot-Swap itself; this is noted as “via regulator” where applicable.
TI
- TPS25982 / TPS2595x — eFuse with programmable dV/dt, accurate ILIM, clear PG/FAULT; AEC-Q options. SS: via regulator.
- LM5069 / LM5060 — Hot-Swap controller; slope/ILIM adjustable, robust with large Cload. Snubber AN available.
- LM74700-Q1 — Ideal diode controller for reverse blocking and low drop; pairs well with eFuse in hot-plug paths.
- TPS546D24 / TPS544C25 — PMBus bucks for SS/telemetry/remote sense; unify compensation window.
ST
- STEF01 / STEF12 — eFuse with programmable soft-start, UV/OV, ILIM; good for consistent inrush profiles.
- STPMIC1 / L5965 family — PMIC/controllers; remote sense and SS depend on rails.
- Ideal-diode solution — ST controllers/diodes for priority path and reverse blocking.
NXP
- PF8100 / PF5020 — Automotive PMIC; multi-rail bucks with SS/telemetry; remote sense by rail.
- PCA9420 — Low-power PMIC; controlled startup/ILIM; SS depends on converter block.
Renesas
- ISL6146 / ISL6144 — Hot-Swap controllers; adjustable slope/ILIM; strong app notes incl. snubber tips.
- ISL68224 / RAA229xxx — Digital multiphase controllers with PMBus; SS and remote sense supported.
- ISL28022/28030 — Power monitors (I²C) to attach validation data to cloud logs.
onsemi
- NIS5021 / NIS5135 — eFuse with programmable startup/ILIM; automotive variants available.
- NCP/NCV buck families — SS and remote-sense capabilities by device; pair with eFuse for EMI + compensation.
Microchip
- MIC2545A / MIC20xx — Power distribution/load-switch with controlled startup/ILIM; good for interface rails.
- MCP19124/5 — PMBus digital power controllers; SS/remote-sense/telemetry to unify compensation window.
- MIC28514/6 — Synchronous bucks with optional jitter (by device); pair with eFuse.
Melexis
- MLX91216 / MLX91220 — Hall current sensors (high bandwidth/accuracy) to capture Iinrush and light-load oscillation signatures.
- System role: sensing/telemetry augmentation; primary power-path devices come from the other six brands.
Request a Quote
Frequently Asked Questions
How do I choose dV/dt to keep inrush under a fixed limit without triggering upstream UV/OV monitors?
Derive the maximum slope from Iinrush,max = Cload·dV/dt, then add 10–20% margin for cold starts. Map the slope to gate R/C or controller current limit and verify PG timing. Finally, sweep VIN and temperature to ensure UV/OV filters do not chatter during the soft-start ramp under worst-case cable and load conditions.
Should I shape inrush by gate R/C or by current-limit mode—and when do they conflict?
Gate R/C sets a predictable dV/dt and low EMI but can lengthen startup and stress upstream UV filters. Current-limit mode clamps Iinrush regardless of Cload, yet raises MOSFET dissipation and thermal risk. Use gate shaping for moderate capacitance; prefer current-limit for large or variable Cload. Avoid combining aggressive limits that trap the device in foldback.
What’s a practical method to identify the ringing frequency before sizing an RC snubber?
Probe the switch node with a short ground spring and run FFT to find the dominant mode during transition. Confirm with a near-field H-probe over the loop. Capture both steady-state and hot-plug events. Use fr to target R·C ≈ 1/(2π·fc) with fc ≈ 0.5–0.7·fr, then validate energy and thermal limits at the intended event rate.
How much spread-spectrum depth/range is effective without breaking control-loop stability?
Begin with ±3–6% depth at 10–40 kHz modulation. This usually lowers peak emissions without significant phase-margin erosion. Always re-run Bode plots and load-step tests because modulation smears energy and can degrade gain/phase near crossover. If margin narrows below target, reduce depth or shift the modulation rate away from compensation poles or sampling notches.
Where should I place the snubber physically to tame switch-node ringing on compact boards?
Place the RC snubber directly between the switch node and its quiet return, minimizing loop area and via count. Keep it on the same layer as the MOSFET and catch diode, away from signal traces. Avoid sharing the return via with high di/dt currents. Short, wide traces beat long narrow ones; verify improvement with near-field scans.
Can spread-spectrum replace a snubber, or are both required in high-Cload hot-plug cases?
Spread-spectrum reduces spectral peaks but does not dissipate localized ring energy. High-Cload hot-plug often excites a hard resonance that needs real damping. Use SS to lower emissions and an RC snubber to absorb energy at the switch node. Validate by comparing peak amplitude, ring decay time, and temperature rise across representative plug/unplug cycles.
How do I prevent remote-sense cable-drop compensation from overshoot or oscillation at light load?
Cap the compensation (±% window), reduce loop gain at light load, and filter sense leads with a small RC close to the controller. Add a fail-safe clamp so open or reversed sense does not drive overshoot. Test with load steps, open-sense injection, and long harnesses at temperature extremes to confirm stability and overshoot limits.
What’s the safe margin for cable resistance variation across harness SKUs and temperature?
Use copper’s temperature coefficient (≈0.0039/°C) to bound resistance drift and add SKU tolerance from supplier data. A practical starting window is ±15% when SKUs or ambient vary widely. Validate by sweeping current across −20/25/85 °C with the longest harness, logging load-end voltage error. Tighten limits once production measurements show narrower dispersion.
How do I pre-qualify inrush and EMI on the bench before sending units to the chamber?
Build a mini-matrix: Cload (10–470 µF) × cables (0.5–3 m). Capture Ipk, 10–90% rise, PG order, and switch-node FFT with SS on/off and snubber before/after. Accept only configurations showing Ipk ≤ budget and ≥6 dB peak reduction. Archive CSV and screenshots to seed chamber runs and speed failure triage.
Which PG/FAULT semantics should tie to EMI pre-checks to block risky firmware states?
Gate firmware states that disable snubbers, expand compensation windows, or turn off spread-spectrum behind PG conditions. Require a passing near-field pre-scan signature before releasing PG to the system. Log FAULT if SS depth or limiter settings are out of policy. This prevents shipping images that raise emissions or destabilize the power path.
How do I compare cross-brand devices when only some list “soft-start dV/dt” explicitly?
Use proxies: availability of an external Css pin, programmable gate-current sources/sinks, and hot-swap controllers with defined fault timers. App notes describing inrush waveforms are strong evidence. If dV/dt is not specified, measure with a standard Cload and cable set. Favor devices whose gate-drive current and fault handling stay within thermal limits.
What’s the minimum dataset for reporting inrush/EMI events into a cloud mapper for audits?
Log timestamp, VIN, Vout, Ipk, 10–90% rise, ringing frequency, SS depth/rate, snubber state, PG/FAULT, pass/fail, operator, and temperature. Attach scope screenshots and raw waveforms for edge cases. Use consistent units and naming so replacements compare directly across brands. This enables regression dashboards and traceable sign-offs during audits.
