Reverse polarity protection: diode, MOSFET or ideal diode?

Three circuits protect a board from a supply connected backwards, and the choice between them comes down to two numbers and one question. The numbers are the voltage lost in the forward direction and the power that loss burns: a Schottky diode drops about 0.4 V and, at 10 A, dissipates 4.65 W; a MOSFET in its place drops 35 mV and dissipates 0.35 W. The question is whether the circuit must also stop current flowing backwards out of the board when the input collapses — which a diode does, a bare P-channel MOSFET does not, and an ideal-diode controller does in under a microsecond. The reverse polarity protection calculator gives the drop, the dissipation and the gate limits for all three at your own supply and current.

For a few hundred milliamps, use the Schottky. For amps, use a P-channel MOSFET and read the two ways it fails. For a supply that can be shorted, reversed or hit with automotive transients while holdup capacitors sit behind it, use an N-channel MOSFET with an ideal-diode controller. Every figure below comes from TI’s SLVAE57B, the application note that measures the three side by side, and SNVA683, which opens with the honest version of why the subject exists:

Most engineers, at some point in time, have connected a power supply backwards to a circuit, and a majority of those engineers have discovered that in unprotected systems a huge amount of current will flow where it was not designed to.

What a reversed supply actually breaks

The damage is not abstract. SLVAE57B’s first two figures name the victims: the ESD diode inside an MCU or DC/DC converter, which conducts hard when its supply pin is pulled below ground, and the polarised electrolytic capacitor at a converter’s input, which is now charged the wrong way. Neither survives a battery for long.

The reversal is also not always a human error at a connector. The same note lists what an automotive rail does on its own: ISO 7637-2 pulse 1 puts −150 V on a 12 V line through a 10 Ω source for 2 ms, and −600 V on a 24 V line through 50 Ω for 1 ms, when an inductive load is disconnected. SNVA683 adds the other direction — load dump takes the rail past 40 V. A protection circuit that only handles a battery installed backwards has solved the easy half.

The Schottky diode: one part, one voltage drop

A diode in series with the supply is the whole circuit. Forward, it conducts; reversed, it blocks. SLVAE57B shows the response to a 12 V input snapped to −20 V: the output does not follow it down, because the diode is reverse-biased, and a bulk capacitor on the output side holds the load for a while.

The cost is the forward drop, and it is paid continuously. The note’s own comparison uses an STPS20M60S at 10 A:

$P_{\text{diode}} = I \cdot V_F = 10\ \text{A} \times 0.465\ \text{V} = 4.65\ \text{W}$

That is a heatsink, or a lot of copper, for a part whose only job is to do nothing most of the time. The drop also eats headroom exactly when there is least of it — SLVAE57B points out that during a cold crank a 12 V battery sags to 3 or 4 V, and 0.4 V taken off that forces a wider-input converter downstream. And because forward voltage has a negative temperature coefficient, two diodes ORing two supplies never share current evenly: the hotter one takes more, gets hotter, and takes more.

For small loads none of this matters, and the diode is the right answer. The arithmetic says where “small” ends.

What the drop costs at your current

Three ways of doing the same job, at 12 V, with the numbers SLVAE57B measured on real parts. The MOSFET on-resistance is the 3.5 mΩ implied by the note’s 35 mV at 10 A — measured on a DMT6007LFG in its forward-voltage plot, while the text under its power plot names a DMT6005LPS-13 for the 0.35 W; two similar parts, and 35 mV × 10 A = 0.35 W ties the plots together. The regulated forward voltage is the 20 mV its linear control loop holds.

load     Schottky (0.465 V)   P-FET, 3.5 mΩ, on   ideal diode, regulated
 1 A     0.465 V   0.47 W     3.5 mV    3.5 mW    20 mV     20 mW
 3 A     0.465 V   1.40 W     10.5 mV   32 mW     20 mV     60 mW
 5 A     0.465 V   2.33 W     17.5 mV   88 mW     20 mV    100 mW
10 A     0.465 V   4.65 W     35 mV    350 mW     35 mV    350 mW

Two things in that table are not obvious from the formulas. The ideal-diode column is worse than a fully-enhanced MOSFET below 5.7 A, because the controller deliberately holds 20 mV across the FET rather than letting it fall to I·R_DS(on) — that regulated drop is what lets it detect reverse current, and SLVAE57B marks the crossover at exactly 20 mV / R_DS(on) = 5.7 A. And the diode column treats V_F as constant, which it is not; at 1 A a Schottky drops less than at 10 A, so the top rows overstate it somewhat. The bottom row is measured.

What 350 mW means for a SOT-23 or a DPAK in still air is a thermal question, and the LDO thermal calculator answers it for any package — the arithmetic is the same whether the dissipating part is a regulator or a pass FET.

The P-channel MOSFET: how it turns itself on

Replace the diode with a P-channel MOSFET whose body diode points the same way the Schottky did, and tie the gate to ground. SLVAE57B describes the sequence:

During normal operation of the battery, the body diode from MOSFET is forward biased and conducts for a very short time until the MOSFET is turned ON when gate voltage is pulled below source. When the battery polarity is reversed, gate-source voltage swings positive and the MOSFET is turned off, protecting the downstream circuits from negative voltage.

That is the entire mechanism. The battery itself provides the gate drive: with the source at +12 V and the gate at 0 V, V_GS is −12 V and the channel is fully on. Reverse the battery and V_GS goes positive, the channel is off, and the body diode is reverse-biased. No controller, no charge pump, one resistor to limit gate current, and a forward drop of I·R_DS(on) instead of a diode’s.

SNVA683 is candid about why this is nonetheless not the common choice in serious designs:

A p-type MOSFET can provide reverse protection with minimal effort, however, the increased cost and on-resistance makes a p-type MOSFET less desirable than a comparable n-type. The common choice is an n-type MOSFET, driven with a controller, of which there are many options.

A P-channel part carries more on-resistance than a comparable N-channel one — SNVA683 leaves it at “increased cost and on-resistance” — and SLVAE57B quantifies what that costs in board area for a load above 6 A: 140 mm² for the P-channel solution against 37.1 mm² for an N-channel FET driven by its LM74500-Q1, “three times smaller”. That controller is a reverse-polarity protection controller, not an ideal-diode controller — the note is explicit that it does not block reverse current — but the gate drive, and therefore the FET and the area, are the same story.

Where the P-channel MOSFET fails

It fails at low input voltage, and it fails to block reverse current. Both are consequences of the gate being driven by nothing but the battery.

Cold crank. The gate drive is the battery voltage. SLVAE57B: “During severe cold crank where battery voltage falls below 4 V, P-channel MOSFET series resistance increases drastically” — the FET is barely on, the drop rises, and “with higher gate to source threshold this can sometimes lead to system reset due to turning off of the P-channel MOSFET.” The controller it compares against — the LM74500-Q1 again, driving an N-channel FET — keeps that FET fully enhanced down to 3.2 V, because a charge pump, not the rail, drives the gate.

The gate rating. The same coupling works the other way on a 24 V system: V_GS is now −24 V continuously, and that is beyond the gate-source maximum on most MOSFET datasheets. The fix is a zener from gate to source and a series resistor to feed it, which is two more parts and a standing current — check the V_GS(max) line on the datasheet before assuming a 12 V circuit scales.

No reverse current blocking. This is the one that surprises people. The P-channel FET is on whenever V_GS is negative, and V_GS is negative whenever the input is positive — regardless of what the output is doing. Short the input while the output’s holdup capacitors are charged, and the FET stays on until V_GS crosses threshold near 0 V, by which time the capacitors have discharged backwards into the short. SLVAE57B’s figure 4-5 shows the output collapsing for exactly this reason, where the Schottky in figure 4-4 held it up. The note’s summary:

The holdup capacitors are discharged as this P-Channel MOSFET protection does not block reverse current from flowing back into the input.

The same property means the P-FET does not rectify an AC disturbance on the supply line, so the ripple current lands in the output electrolytics’ ESR instead of being blocked — SLVAE57B’s AC-superimposed test, 2–4 V peak-to-peak from 20 Hz to 30 kHz, is where that shows up.

Reverse polarity protection is not reverse current blocking

SLVAE57B separates the two jobs by name, and the distinction is the most useful thing in the document:

Reverse polarity protection, also referred to as reverse hookup protection (RHP), prevents the load from getting damaged due to negative voltage at the input during a reversely connected battery or dynamic reverse polarity conditions during a inductive load disconnect from battery. Reverse polarity protection does not necessarily block reverse current flowing into the battery from the load or downstream DC/DC converters.

A Schottky does both, expensively. A P-channel MOSFET does the first only. An ideal-diode controller does both, cheaply. Which of the two jobs a given rail needs decides the circuit, and SLVAE57B gives the rule by example: a logic path — regulators, converters, an MCU — needs both, because its holdup capacitance is the whole point of it. A load-driving path — relays, a horn, wiper motors — needs only the first, because nothing behind it stores energy worth protecting.

The ideal-diode controller: an N-channel FET and a comparator

An ideal-diode controller drives an N-channel MOSFET, wired so its body diode blocks reverse current when the channel is off, from an internal charge pump that lifts the gate above the source. In the forward direction it holds the FET on; the moment the voltage across it reverses, a comparator pulls the gate down and the body diode takes over. SLVAE57B describes the two control schemes:

• Linear regulation holds the forward drop at a fixed 20 mV by adjusting the gate, so R_DS(on) rises at light load and the FET is only fully enhanced above 20 mV / R_DS(on). The regulated drop is what makes a DC reverse current of exactly zero detectable.
• Hysteretic on/off fully enhances the FET above a forward threshold and cuts it at a reverse one, so the drop is set by the FET alone — but a reverse current smaller than V_REV / R_DS(on) cannot be seen.

Either way the forward loss is the MOSFET’s, not a diode’s. The measured comparison at 10 A, MOSFET against Schottky, is 35 mV against 465 mV and 0.35 W against 4.65 W — SLVAE57B calls it “more than 10x power saving”.

How fast the reverse path closes

The number that matters is the time from the input collapsing to the FET being off, because that is how long the holdup capacitors have to drain backwards. It is a comparator delay plus a gate discharge:

$t_{\text{off}} = t_{\text{comparator}} + \frac{C_{\text{gate}} \cdot V_{GS}}{I_{\text{pulldown}}}$

SLVAE57B’s figures for the LM74700-Q1 are a 0.75 µs maximum comparator delay and a 2.37 A gate pull-down, and it works the example: a 5 nF gate is discharged in 21 ns, for 0.77 µs total. (Solving the discharge term back gives a gate that was sitting about 10 V above the source.) The comparator, not the gate, is the whole delay — which is why a strong pull-down alone does not make a slow controller fast.

Against an ISO 7637-2 pulse 1, the note’s oscilloscope capture shows the gate off within a microsecond and the output never going negative, with the input TVS clamping at −42 V. That TVS is not optional; SLVAE57B states it is “required to clamp the voltage from exceeding the absolute maximum ratings” of controller and MOSFET. Sizing it is the TVS clamping tool’s job, and the placement rules in the ESD layout article apply unchanged — the pulse is longer, the loop-inductance argument is the same.

The N-channel low-side variant

There is a fourth circuit: an N-channel MOSFET in the ground return, gate tied to the positive rail through a resistor. It has the N-channel’s lower on-resistance and needs no controller — the rail turns it on the way the P-FET’s rail turned it off. SLVAE57B gives it a paragraph and one caveat that decides whether it is usable at all:

However, a jump in the system ground voltage during turn ON/OFF or load current transients may not be tolerated by all systems and needs to be considered during system designs.

Anything that shares a ground with the outside world — a connector shield, a sensor return, a communications bus — sees that jump as a signal. It is a circuit for a board that is electrically an island.

What the automotive tests actually demand

The reason SLVAE57B measures against named tests rather than a reversed bench supply is that the tests are what a design is certified to. In its pages:

• ISO 7637-2 pulse 1: −150 V through 10 Ω for 2 ms on 12 V systems; −600 V through 50 Ω for 1 ms on 24 V. The dynamic reversal.
• Cold crank: battery to 3–4 V, from the drawbacks of the Schottky and the P-FET alike.
• AC superimposed: 2–4 V peak-to-peak, 20 Hz to 30 kHz, on the supply line, with no loss of function. Up to 200 kHz in mild-hybrid power trees.
• Load dump: 35 V suppressed peak on 12 V systems; unsuppressed, 101 V on 12 V and 202 V on 24 V per ISO 16750-2. SNVA683’s “exceed 40V” is the same event.

A circuit chosen for a reversed battery alone will meet the first of these and may meet none of the rest. The list is the specification; the reversed battery is one line of it.

Choosing between them

                          Schottky     P-channel FET    ideal diode ctrl
forward drop, 10 A        465 mV       35 mV            35 mV (20 mV < 5.7 A)
dissipation, 10 A         4.65 W       0.35 W           0.35 W
blocks negative input     yes          yes              yes
blocks reverse current    yes          NO               yes, < 1 µs
works at 3-4 V input      drop hurts   FET starves      yes, to 3.2 V †
24 V without extra parts  yes          gate clamp       yes
parts                     1            1-3              controller + FET
board area, > 6 A         heatsink     140 mm²          37 mm² †

† Measured by SLVAE57B on the LM74500-Q1, a reverse-polarity protection controller driving an N-channel FET without reverse current blocking. The two figures depend on the charge-pumped gate drive, which the ideal-diode controllers share; the note does not repeat them for the LM74700-Q1.

• Under about an amp, or wherever 0.4 V and its heat are affordable: the Schottky. It is the only one of the three with no failure mode beyond its own dissipation.
• Amps, a 12 V rail, nothing behind it that stores energy: the P-channel MOSFET, with the gate rating checked and the cold-crank behaviour accepted.
• Amps, holdup capacitance downstream, an input that can be shorted or hit with a transient: the N-channel MOSFET and an ideal-diode controller, with the TVS SLVAE57B says it needs.

The decision is rarely about the forward drop, even though that is what the datasheets lead with. It is about whether a shorted input is allowed to empty the capacitors behind the protection — and that is a question about the load, not the protection circuit.

Sources

• TI, SLVAE57B — Basics of Ideal Diodes (revised October 2021). The Schottky, P-channel and ideal-diode comparison, the 10 A dissipation figures, the cold-crank and reverse- current-blocking behaviour, the LM74700-Q1 timing figures, and the ISO 7637-2 and ISO 16750-2 test levels.
• TI, SNVA683 — Surge Stopping and Reverse Voltage Protection with the LM5069 (April 2013). The −24 V reversal and 40 V load-dump figures, and the note on P-channel versus N-channel cost and on-resistance.