Automotive protection switches: load limits, reverse blocking and energy budgets

1. Allocate functions by fault location, not by a protection label

Battery entry faces reverse connection, backfeed, supply transients and ECU startup; branch outputs face harness shorts, load inrush, inductive turn-off and open-load diagnosis. An architecture can need both an entry ideal-diode/disconnect stage and branch smart switches. In series, review startup order, voltage drop, fault selectivity and which stage disconnects first, not just each component in isolation.

TPS1HB16 and TPS2HB16 have a normal/nominal 6–18 V supply range. Their datasheets also specify extended ranges with conditions, so 18 V is not an absolute maximum. Neither an extended range nor 40 V load dump establishes an unconditional continuous rating. During their specified suppressed load-dump test up to 40 V, the switch retains its enabled/disabled state; device survival does not imply that the load is disconnected or clamped to 18 V.

Scroll wide tables horizontally; column headers remain visible within the table.

2. A/B/F current-limit thresholds are not continuous load ratings

Select the current-limit version by full orderable. The table shows programmable threshold ranges or an internally fixed threshold from TI's comparison tables, not recommended load current or all tolerances. The listed overcurrent action is to disable the switch immediately, so these devices are not sustained constant-current supplies for arbitrarily charging capacitors. Configure LATCH, enable and retry behavior from the exact device's fault-handling section.

Establish two tolerance-aware boundaries: the minimum trip threshold must accommodate legitimate startup current and timing, while the maximum threshold and turn-off time must protect the harness, connector and load. If no overlap exists, revise startup, load, wiring or protection architecture rather than hiding the conflict with a higher threshold. For dual-channel devices, also evaluate shared thermal/diagnostic effects when one branch shorts and the other operates normally.

Scroll wide tables horizontally; column headers remain visible within the table.

3. Separate steady heating, startup inrush and repeated faults

The 16 mΩ on-resistance is typical at TJ=25°C, not an all-temperature maximum. As initial arithmetic only, one 5 A channel dissipates approximately I²RON=0.4 W, or about 0.8 W for two channels at 5 A each, before hot-resistance increase, switching and diagnostic losses. This does not establish continuous current capability on a board. Use maximum resistance at the relevant conditions, shared-package heating and actual copper area to estimate and measure junction-temperature margin.

For capacitive startup, begin with I=C×dV/dt. Charging an assumed 1000 µF from 0 to 12 V in 2 ms requires 6 A of ideal capacitor current alone, plus operating load and tolerances. A lower trip threshold can misclassify startup as a fault. Stored capacitor energy is ½CV²=72 mJ, not the exact energy necessarily dissipated by the switch. Determine switch energy from ∫VDS(t)×I(t)dt, including source impedance, ESR and the actual ramp.

An external dV/dt capacitor on LM7480 HGATE can control startup slew, but is not a closed-loop short-circuit limiter. During a slow ramp the external MOSFET can operate linearly, requiring an SOA check at actual VDS, ID, pulse width, repetition and starting temperature, not merely low RDS(on) or a pulse-current headline. If retries repeatedly release fault energy, also budget average heating and cooling intervals.

4. Automatic turn-on during reversal is not reverse blocking

TPS1HB16/TPS2HB16 can automatically turn on the internal FET during reverse battery to avoid excessive body-diode dissipation, not to disconnect the load path. A low EN signal therefore does not guarantee zero load current. Automatic turn-on also requires the specified SEL pin's path to module ground: check the single-channel SEL condition and dual-channel SEL2 condition, including protection resistors and MCU-pin paths.

Draw three distinct cases: reversed battery, a normal battery with an externally driven output, and loss of input while output capacitors remain charged. Mark body-diode direction, conducting FETs, MCU clamps and ground paths in each. TI notes that a diode in the smart switch's GND path can protect its controller circuitry without stopping load current. A load that cannot tolerate reverse current needs a blocking stage in the relevant power path.

The TPS2HB16 reverse-battery section explicitly states that protections such as thermal shutdown are unavailable in that condition. Test current, duration, heating and recovery, not just whether one event causes damage. An LM7480 ideal-diode stage can provide a different blocking function, but the result still depends on MOSFET orientation, common-source/common-drain topology and external bypass paths.

5. LM74800 versus LM74801: control law and overvoltage policy

LM7480-Q1 is the family entry. LM74800-Q1 uses A–C differential-voltage linear regulation plus a comparator for reverse-current control; LM74801-Q1 uses comparator-based control. The 10.5 mV typical regulated forward drop belongs to LM74800, not LM74801, and does not mean that the entire back-to-back path always drops only 10.5 mV at any current. External on-resistance and gate-drive conditions still determine high-current voltage drop.

The family drives external back-to-back N-channel MOSFETs over a 3–65 V operating input range. A selector's 300 A entry is not an integrated-switch rating. Conduction loss is approximately I²×(RDS1+RDS2), with temperature effects, connectors, copper and fuses also considered. Include the external power stage in the BOM and validation; without specified MOSFETs and thermal paths there is no complete '300 A design'.

Separate overvoltage behavior by topology and feedback. In SNOSD95C section 10.3's LM74800 common-source application, R2 sensing VBATT gives cutoff; moving it to VOUT permits output clamping. Section 10.3.2.5 requires checking external-MOSFET linear SOA and dissipation for clamping. This application also includes an external resistor/Zener VS-protection network; it does not establish a 200V IC rating or clamp operation for every 00/01 configuration. The ideal diode's 10.5mV forward-drop regulation is a different function. Manage cutoff delay, overshoot, hold-up and remaining energy separately. A common-drain midpoint load retains a body-diode path, so EN low does not isolate every node.

6. Design turn-off energy and diagnostic observability together

An inductive load maintains current at turn-off, and the smart high-side output clamp limits voltage across the FET. Start with EL=½LI²: an assumed 10 mH at 3 A stores 45 mJ. The supply can contribute additional energy during demagnetization, so integrate actual switch dissipation. Include inductance tolerance, mechanical state, temperature and repetition; an integrated clamp does not remove the energy review. External freewheel/TVS networks change demagnetization voltage and release time and must still meet actuator-response requirements.

SNS is a mode-dependent analog diagnostic output, not an independent, instantaneous, full-range current meter. Firmware must track whether it selected a channel current, temperature or fault indication, wait the specified selection/enable settling interval, then sample the ADC and distinguish off-state, open-load, short-to-battery and normal low current. Dual-channel polling introduces observation gaps; fast protection cannot rely solely on software polling.

At minimum, fault records should associate supply voltage, channel commands, SNS selection, fault latch, retry count and timestamp. A generic overcurrent flag cannot distinguish a harness short, cold-filament startup, capacitor charging or an unsettled sample. Acceptance should produce reproducible fault classification and recovery state, not merely a switch that turns on again.

7. Practical checklist from BOM to fault injection

Use applicable project requirements, controlled energy and appropriate laboratory protection for these tests. The numerical examples prepare budgets; they do not replace bench measurements, device SOA or ECU compliance evaluation.

• Specify TPS1/TPS2, A/B/F or LM74800/01 full OPNs; record threshold resistors, tolerances and fault actions, with no substitutions based on family maximum amperes.
• Freeze the 6–18 V normal range, extended-range restrictions and 40 V load-dump waveform separately; check load survival while the switch remains on.
• Estimate heating with worst-case hot RON, simultaneous channels, actual PCB and ambient, then measure; 16 mΩ is only a typical 25°C starting point.
• Budget normal load, cold-start peaks, capacitor charging, shorts and repeated-retry energy separately, checking minimum/maximum trip thresholds against harness protection.
• Draw reverse-battery, output-backfeed and input-loss paths; check SEL/SEL2, MCU clamps and ground diodes; automatic FET turn-on is not reverse blocking.
• Check LM7480 external MOSFET voltage, gate drive, linear SOA, thermals and post-disconnect energy paths; OV cutoff does not replace TVS clamp analysis.
• Verify the healthy channel during a single-channel short, SNS selection/settling, off-state open-load diagnosis and fault-latch/retry states.
• Retain voltage, current, temperature and recovery waveforms; keep device qualification, functional-safety documentation and system validation distinct.

Frequently asked questions

Related parts and complete OPNs

Continue reading