I'm using an alternator to charge a 300Ah lead-acid battery bank. The cable has a 60A fuse, yet measuring a shunt suggest that the current is about 70-90A for the first 10-20 seconds, and after a few minutes it is down to about 30-35A.

The alternator provides 14.0V at the car battery, and cable losses further reduce the voltage to about 13.8V at the battery bank.

I'm trying to introduce an automatic switch using a P-channel mosfet.

The problem I'm facing is that the voltage drop over the MOSFET should be minimal, since a lower voltage will significantly reduce the charge current. A better alternative would be a boost circuit to 14.4V, but at 30-50A, they get very expensive.

The P-channel MOSFETs tend to have higher on-resistance that N-channel. But I've found this: AP6681GMT

The package is undesirable, and heatsinking is only possible through the PCB.

I built a circuit, but when I tried it, black smoke came out of it, and it failed open.

I'm wondering what I did wrong, and what my alternatives are. I understand a relay has a high on-resistance?

Would it make sense to buy a starter relay for a car? The specification is typically not provided and perhaps the on-resistance will be too much.


simulate this circuit – Schematic created using CircuitLab

For testing the circuit, I connected the drain-side to a blower motor, and switched the gate-signal on/off every 1000ms. The circuit worked correctly, even with peak currents of 20A and sustained current of 7A.

I then connected the drain-side to the battery bank. When I turned on the alternator, it jumped to 14.0V on the source side. Is it possible that the 1k resistor didn't raise the Vgs fast enough, such that it continued to conduct without pulling-up? e.g. a runaway scenario.

UPDATE: Using an N-channel MOSFET instead?

N-channel MOSFET: IRFB4410Z (100V, 97A, 9mOhm/10V)

or AP99T03GP (30V, 200A, 2.5mOhm/10V 4mOhm/4V)


simulate this circuit

This would allow me to use a heat sink, to have lower on-resistance, bigger package. But with the inconvenience of a boost-circuit.

  • \$\begingroup\$ That's a P (or N) channel) MOSFET. A PNP (and an NPN) is a BJT. \$\endgroup\$ Commented Dec 29, 2016 at 15:53
  • \$\begingroup\$ I think you have violated the safe operating area curve; at 2V Vds, the maximum DC current is 40A (assuming that appropriate heat sinking is used). \$\endgroup\$ Commented Dec 29, 2016 at 16:08
  • 1
    \$\begingroup\$ Alternator output current is normally controlled using the alternator field current. Simple automotive regulators switch the field current off and on, and let the battery "average" the voltage. \$\endgroup\$ Commented Dec 29, 2016 at 17:30
  • 1
    \$\begingroup\$ When you disconnect the alternator from the battery, as you are trying to do, the alternator output voltage will rise, and will probably blow the rectifier diodes in the alternator. You must NEVER disconnect a running alternator from its battery! \$\endgroup\$ Commented Dec 29, 2016 at 18:13
  • \$\begingroup\$ Ah, I forgot to mention that this is a circuit in addition to the car battery. I will update the diagram. Thanks for the input, it's good to know. \$\endgroup\$ Commented Dec 29, 2016 at 18:15

3 Answers 3


The basic formula for smoking parts depends on the thermal resistance of temp rise per power lost in ['C/W] for the rating of the heatsink.

Pout = 50A*14.4V= 770 Watts

However when supplying a sine wave into a rectifier with a battery acting as almost a short circuit the peak currents can be be >>10x times as much, thus increasing the requirement now to >>500A pk

  • With 2.2mΩ and short term currents of >>100A for a drained battery I^2R = 22Watts with good heatsink to chassis with insulator and paste.

To minimize the Vds drop, AND have sufficient Vgs now you need a Pch switch with gate to ground control AND have accurate gate control.

Use a Pch and decide what threshold to disable the switch. I roughly chose 13.2V R divider with a poor-man's 1.1V Darlington comparator. ( it works) The threshold may need to be corrected slightly to 13.4V to as a threshold. Since the drain current is low the input threshold to the Darlington is also low at 1.1V @1mA

$2.20(1pc) 2.2mΩ Pch type 170A


simulate this circuit – Schematic created using CircuitLab

The Vbat threshold to Vgs voltage gain is > 1k so the threshold amplifies Vgs to quickly move past turn ON threshold from 1.2 to 2.2V with just a few mV change in V+.. the only drawback is the threshold is not very precise (5%) and has a NTC of -2.1mV/'C which translates to 21mV/'C of battery threshold shift or a drift of 13.2 +/-0.4V for +/-20'C so 13.5V is advised.

  • \$\begingroup\$ I've updated the question with a boost circuit using an N-channel MOSFET. I hope my design is not too ridiculous. \$\endgroup\$ Commented Dec 29, 2016 at 16:35
  • \$\begingroup\$ it's getting close to there. \$\endgroup\$ Commented Dec 29, 2016 at 20:59
  • \$\begingroup\$ This is great. I can use this concept with the Nch approach as well. However, doesn't this design still suffer from the problem, that if Vs suddenly rises from 12.5V to 14.0V, the gate will still be at a lower voltage due to R1, and therefore Vgs will be -1.5, and cause the Pch to turn on slightly, etc. causing thermal runaway? \$\endgroup\$ Commented Dec 29, 2016 at 21:17
  • \$\begingroup\$ NO Q1 switches on at 13.2V which rarely happens on startup Its either <13V or >13.5V . better to have 1V hysteresis... for Nch MUST BE on Low side. V- with PNP Darlington!!! \$\endgroup\$ Commented Dec 29, 2016 at 21:20
  • \$\begingroup\$ Why can't the Nch be on the high side, as long as the Vg is 10V higher than Vs? \$\endgroup\$ Commented Dec 29, 2016 at 21:45

I'm wondering what I did wrong

What you did wrong was not understand the peak voltages that can come from an alternator and how these will easily exceed the gate-source maximum voltage for that device (+/- 20 volts).

You should have used a 15 volt zener protection diode across the gate-source and a small collector resistor to limit current into the zener when the alternator produces spikes. The BC337 is also a risky device to use given that it's peak voltage spec is only 45 volts.

Automotive circuits like this are really difficult to get reliable without going for 100 volt rated devices.

  • \$\begingroup\$ Aha. I thought the car battery would help sustain the voltage at lower levels? I'll update the circuit as you've suggested. How big should the zener-resistor be? E.g. a 10w 10 ohm power resistor? I imagine that if the resistor is too big, it won't absorb the over-voltage completely? Especially given that it competes with an 8mOhm MOSFET? \$\endgroup\$ Commented Dec 29, 2016 at 16:02
  • \$\begingroup\$ You should probably look for a mosfet with 100 volt ratings rather than try and fix this. The resistor neededn't be 10 watts - a 1 watt device would probably do and a 5 watt zener but, without knowing the type of voltage spike it's really hard to be precise. \$\endgroup\$
    – Andy aka
    Commented Dec 29, 2016 at 16:05
  • \$\begingroup\$ Actually I'm a bit worried that the overvoltage from the alternator could harm my devices, e.g. laptop. So I think I'll definitely need this kind of overvoltage protection. I suppose the resistance to ground via the batteries is much higher than via 5 ohm resistor. But would this be sufficient protection? \$\endgroup\$ Commented Dec 29, 2016 at 16:32
  • \$\begingroup\$ The over voltage protection I'm talking about applies exclusively to the MOSFET. BTW your new circuit might work in that it will cut-down on spikes entering the battery but that power will be disspated in the MOSFET. What functionality is your automatic switch providing? \$\endgroup\$
    – Andy aka
    Commented Dec 29, 2016 at 16:39
  • 1
    \$\begingroup\$ You might find that you can connect the SP directly across the battery with very little leakage back into the alternator because of the bridge rectifiers in the alternator but do some research on this. \$\endgroup\$
    – Andy aka
    Commented Dec 29, 2016 at 18:24

Deciding against P-channel: Unintended activation and thermal runaway

In retrospect, I think a P-channel with pull-up, where the source voltage changes, is not a good design approach, because the voltage change can trigger a small voltage across the Vgs, which in turn allows current but at high resistance, thus potentially causing thermal runaway as well unintended activation.

Using high-side NPN instead: The low-side cannot be connected via an N-channel, because a constant chassis connection is desired for better earthing and easier cabling.

A high-side N-channel thus seems the best option, however requiring a boost circuit to provide a gate voltage 10V above source.

Reducing thermal problems: In addition, I've concluded that using several N-channel MOSFETs will allow me to reduce current effects significantly. For example, using two MOSFETs, each experiences half the current, but 1/4th of the power dissipation.

And since power dissipation is the biggest problem, I'll probably get three MOSFETs with heat sinks. At 33A each and 2.5 mOhm, the heat gain will be 2.7w.

Solving overvoltage problems: I'll also try to add a Zener diode at 20V through a 500mA fuse to see if overvoltage occurs despite the main car battery being in parallel. If it turns out to be a problem, I'll add a 16V zener with a resistor and test again, to see if the problem has been resolved.

  • \$\begingroup\$ Vgs is less stable than Vbe which may be sufficient for a simple design. How are you planning to generate 24Vdc? \$\endgroup\$ Commented Dec 29, 2016 at 21:35
  • \$\begingroup\$ Would there be a problem of the Nch approach supporting reverse current as well? Max 30A (charging car battery from solar panel charge controller) \$\endgroup\$ Commented Dec 29, 2016 at 22:04
  • \$\begingroup\$ Suitable blocking diode is essential \$\endgroup\$ Commented Dec 29, 2016 at 22:06
  • \$\begingroup\$ The charge controller has a built-in blocking diode. I read that MOSFETs have much lower reverse current tolerances, e.g. 100A should support ~30A reverse current. \$\endgroup\$ Commented Dec 29, 2016 at 22:10
  • 1
    \$\begingroup\$ normally a design starts with all the known limits for V+/- tolerance, current +/- polarity and range. ESR of each part and thermal resistance. then choose suitable conduction and protection as required. Your diagram and specs are not well defined with no ESR and no PV. \$\endgroup\$ Commented Dec 29, 2016 at 22:14

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.