basic schematic of the circuit

I'm trying to build the circuit in picture above. unfortunately the load on high side of the step up transformer is unpredictable and the input impedance of transformer is very low so I'm guessing the current through the MOSFET is going to be very high, basically the power supply will be short circuit to the ground while it's on. Should I add a resistor to the source of the MOSFET? If so, how big? How do people usually protect these kind of circuit? I can't add a high power resistor due to space constraints, so what's the smart way to do this?

  • 1
    \$\begingroup\$ "basically it will be short circuited to ground" no, it won't. There's a large inductivity in there – that's the whole idea of a switch mode power supply like yours. I'd say: read a better introductory text on SMPSes, and you'll find out essential details like why we use flyback diodes and so on. \$\endgroup\$ – Marcus Müller Nov 3 '18 at 14:57
  • \$\begingroup\$ So what would happen if the MOSFET stays on due to an error for like 2 seconds or more? \$\endgroup\$ – sadegh satari Nov 3 '18 at 15:00
  • \$\begingroup\$ I read very low ohm value on primary side, lower than 1 ohm. The secondry side is around 1.2 kohm. \$\endgroup\$ – sadegh satari Nov 3 '18 at 15:01
  • 1
    \$\begingroup\$ "How do people usually protect these kind of circuit?" They don't. During development you could use a current limited bench supply but that might not be enough. I have blown a few transistors developing 120W class-B amplifiers. \$\endgroup\$ – Oldfart Nov 3 '18 at 15:50
  • \$\begingroup\$ You can use a fuse or PTC. This could protect the mosfet if you turn it on too long. BUT there may still be other faults that could cause it to blow up. Like if you switch it too fast, or if you let the gate float so that the MOSFET operates in a linear region. Note: you also need a diode across the primary winding so that the drain voltage does not spike when you turn off the mosfet. \$\endgroup\$ – mkeith Nov 3 '18 at 16:30

This is a problem when you use something like an MCU to control a device which exists on the edge of blowing up, depending on timing, since the MCU can lock up from EMI or whatever (even when the firmware is working perfectly) and destroy the MOSFET and/or transformer. Hardware-based circuits are less of an issue.

An ordinary watchdog timer may be insufficient to protect the MOSFET. During development you could rig something (like a one-shot and gating) that would limit the on-time of the MOSFET but that's probably not practical for a production circuit.

I have used the nonlinear characteristics of a tungsten light bulb as a series resistor during development. Depending on the actual currents involved, you may be able to use an incandescent automobile lamp (interior lamp, brake light or even a headlamp) as a series limiting PTC (positive temperature coefficient) resistor during development. If the light illuminates, you have a problem.

  • \$\begingroup\$ I always find it fascinating seeing a lightbulb used in this manner. \$\endgroup\$ – Hearth Nov 3 '18 at 16:16
  • \$\begingroup\$ "Hardware-based circuits are less of an issue." I thought of something like 555 but I need more control on switching frequency. I used IR2103 to limit intereference between mosfet/transformer and MCU, and I'll isolate their ground and power via routing and inductors. I can use lightbulb during development, that's an awesome idea, but is a ptc enough for the final product? \$\endgroup\$ – sadegh satari Nov 3 '18 at 19:58
  • \$\begingroup\$ That's a design decision. There are much better switchmode controller chips than 555s, UC3842 is a classic, and is very cheap, but there are many others. \$\endgroup\$ – Spehro Pefhany Nov 3 '18 at 20:19

[summary: plan on responding to over-current in 1.16 microsecond]

The thermal time constant of a cubic meter of silicon is 11,600 seconds (the inverse of the thermal_diffusion), computed by multiplying the specific-heat of a cubic meter with the thermal-resistance of a cubic meter.

The thermal time constant of 0.1 cubic meter (4" on a side) is 100X faster, at 11600/100 or 116 seconds.

The thermal time constant of 0.01m (1centimeter or 0.4 inches) is another factor of 100X faster, at 11600/10000 = 1.6 seconds.

This size --- 1cm --- is much larger than the junction depth of a MOSFET. Which encapsulated the channel where the heat is generated, and will vary depending on the MOSFET physical design.

The active part of the FET will be about 100 microns, so the heat can be easily dumped into a COPPER TAB and pulled out of the package.

We now have a size we can compute with ------ 100 microns, or 0.1 milliMeters of depth.

What is the thermal time constant of 100 microns?

That ---- 100 microns ---- is 10*10 thinner than our previous numbers at 1cm, thus the thermal time constant is

11600 / (100 * 100 * 100 * 100) = 11600 / 100,000,000

== 11600 sec no (this is for 1 cubic meter)

== 116 sec no ( this is for 0.1 cubic meter)

== 1.16 sec no (this is for 1cm cube)

== 0.0116 sec no (this is for 1mm cube)

== 0.000116 sec YES (this, 116 microsecond, is for 100 micron cube)

Thus your protection circuit needs to be AT LEAST THIS FAST.

However, some FETs use just the top 10 microns or so for the FET action, and you need to protect that 10micron region from overheating, by protecting in 1.16 microseconds.

Thus, in summary, you need to aim for 1.16 microsecond response time.


PTCs and fuses won't protect MOSFETs reliably, as analogsystemsrf's answer outlines, semiconductors can be destroyed very quickly indeed.

The best thing to do is to monitor the current through the MOSFET and turn it off if the current starts to leave the area you've identified as safe. There are a couple ways of implementing this current limiting function.

The simplest and probably most reliable is to use an IC that has current limiting built-in to control your switching converter. If you need to have that switching converter controlled by an MCU (ie, to change the voltage setpoint), this is often possible by biasing a feedback node or steering an error amplifier with an external voltage.

If you really need to have the PWM signal under the MCU's control, then you might be able to find a suitable protected switch IC, which is essentially a MOSFET with a built-in driver, current sense, and other protection features built-in. Some of these are even available in standard MOSFET packages, making them drop-in replacements for standard transistors. However, you may have a hard time finding a device that is suitable for use in a switching converter.

A slightly more complex solution is to use an external comparator to disable the PWM signal when the current threshold is exceeded. This can be implemented such that the comparator limits the on-time to give you a current-limited operating mode, or so that the comparator latches the PWM signal off and effectively shuts the converter down entirely. Many modern MCUs have sophisticated PWM units with external fault inputs that can be used to implement either scheme within the peripheral itself (which means your firmware is off the hook for protecting the MOSFET, which is good!). Many MCUs also have comparators built-in, that may allow you to implement this strategy by simply adding a current sense resistor (especially if the MCU has a DAC you can use for the other side of the comparator).

Except for some switching controllers and protected switches that use the resistance of the MOSFET itself to measure current, any current limiting scheme is going to require an external current sense resistor, which you mention may be difficult given your space constraints. Resistor selection really depends on the current range, and lower currents are of course easier to deal with as they mean less power dissipation for a given sense output voltage range. You would have to carefully examine the specs of your comparator and take a close look at your layout to really see how low you can go in terms of sense resistance.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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