Polyfuse protection of mosfet shortcircuit

[Psi]

I'm working on adding polyfuse protection to a circuit which supplies 6A 12V pwm to a solenoid using a P-channel mosfet. I've never used a polyfuse before and i want to check that i understand all the things that need to be taken into account.

The mosfet is   NVD5117PL
DPak 60V 16mR 61A

What i know so far is..
- A current limit resistor is needed so that the short doesn't exceed the mosfet Imax (61A).
- The current limit resistor needs to dissipate a very large amount of energy while the polyfuse trips. (5-10 seconds)
- The mosfet temp needs to remain under max-junction-temp during this 5-10 seconds
- There's going to be a Vdrop introduced in normal operation due to the current limit resistor.
- If I have a higher limit resistor it reduces the strain on the resistor and mosfet but also causes more vdrop, however a lower resistor will cause the polyfuse to react faster.

I can test most of that practically but the thing i'm not sure about is the temperature change over time of the mosfet.
How can i work out the rate of mosfet junction temperature change for short term high current?

If i put 100W into a mosfet and the Junction-to-Case is 1.3C/W does the mosfet junction get to 130degC instantly?
Basically I'm not sure how I can get the mosfet to handle the large energy needed without violating some datasheet spec

[metalphreak]

Polyfuses have two useful ratings. The long term maximum trip current and the instantaneous trip current. A "1.1A" polyfuse will happily pass 1A all day long. If you put 1.2A through it, it'll still take a while before tripping as it slowly increases in resistance. Instantaneous is typically twice the rated current, so a 1.1A would trip pretty quick at 2.2A. Trip times are all explained in the datasheets.

That mosfet can do 419A pulsed current. You don't really need a current limiting resistor to limit current as the polyfuse will have well and truly tripped before then.

[lewis]

You can calculate the 'instantaneous' power dissipation in the die by using I_MAX² * RDSon. You can size the current limiting resistor accordingly to limit I_MAX if it causes excessive power dissipation in the die.

There will be some thermal inertia, so dissipating 100W in the die will not heat it up instantaneously. But, as near as damn it - it will only take a few microseconds to respond to changes in power dissipation. There are usually thermal models in the datasheet for some devices that treat the thermal properties of the die as a network of resistors and caps. The network always forms a low pass filter.

For very short term overload (< 1ms), where the heat generated by the die does not even have time to get through the package, you need to absolutely ensure that the peak/instantaneous/maximum/worst-case power dissipation (whatever you want to call it) is within the safe operating area curve of the fet. Heatsink size doesn't come into it. For longer duration overloads (> a few ms), you still need to ensure the power dissipation is within the SOA curve, but you also need to ensure the die does not exceed its maximum temperature. Heatsink size and thermal resistances and all that are now very important.

My experience with semiconductors is that they invariably, wilfully and honourably sacrifice themselves, laying down their lives to protect the fuses - polyswitch or otherwise. I do design using polyswitches, but I only treat them as a secondary layer of protection to prevent catastrophe/fire/death if the mosfet fails short. For primary overcurrent protection, I use either a high-side current sense circuit to 'instantaneously' cut the fet off under overcurrent conditions, or, better still, use a high side smart switch like the VN920 which has a diagnostic output proportional to the load current and is fully protected, but is only good for 1-2 KHz PWM. (VNB35N07 works on the low side and is much faster).

[metalphreak]

lewis: would the power source not also be a limit for the maximum short circuit current? Surely such a massively overrated mosfet could handle the short circuit surge current from what I assume is a standard 12V supply? More info about the rest of the circuit is required to make any helpful suggestions

[Short Circuit]

...
- A current limit resistor is needed so that the short doesn't exceed the mosfet Imax (61A).
...

Don't forget to look at the Imax of the PTC. Usually these have rather low values (compared to normal fuses)!

[lewis]

would the power source not also be a limit for the maximum short circuit current?

Absolutely, if it was fast enough, but even the smoothing capacitors in a decent sized SMPS can pack a punch when short-circuited.

In an automotive application with essentially zero supply impedance, the let-through current of even a 3A blade type fuse is enough to blow the shit out of most semiconductors.

[Psi]

Thanks guys, some really good info.

Don't forget to look at the Imax of the PTC. Usually these have rather low values (compared to normal fuses)!

The 6A hold polyfuses i was looking at were
http://www.digikey.co.nz/product-detail/en/MF-R600/MF-R600-ND/259978
40A max

or

http://www.digikey.co.nz/product-detail/en/LR4-600XF/LR4-600XF-ND/1113478
100A max

would the power source not also be a limit for the maximum short circuit current?

Absolutely, if it was fast enough, but even the smoothing capacitors in a decent sized SMPS can pack a punch when short-circuited.

In an automotive application with essentially zero supply impedance, the let-through current of even a 3A blade type fuse is enough to blow the shit out of most semiconductors.

Yep, its automotive, so massive short circuit current is available.
I've yet to measure the total resistance of the loom wiring + pcb tracks to see just how much current would flow under short circuit conditions.
If its under 50A i may not need a resistor.

[Psi]

Surely such a massively overrated mosfet could handle the short circuit surge current from what I assume is a standard 12V supply? More info about the rest of the circuit is required to make any helpful suggestions

It's not that overrated.
The 419A Imax mentioned earlier is for a 10us pulse. (The polyfuse will take effect in seconds, so much longer than that)
For continuous operation mounted to FR4 the Imax is 11A.
 

This is the existing circuit, no polyfuse yet.
There's a 10A fuse in the battery line, (not shown)

[sorin]

Try to do some test and see how many time need the polifuse to blow.

[Psi]

Yeah, i have a run of tests i want to do and that's one of them.

The goal of this thread was to check my understanding of polyfuse usage and their circuit design requirements before proceeding.

[lewis]

Another way to implement cheap-arse current sensing is to measure the voltage across the drain-source junction of the fet with a suitably configured comparator (not the ADC in the MCU, it's too slow). This way, you can use the fet's own RDSon as a current sense resistor. When the voltage drop is too much, the current must be too high, so quickly turn off the mosfet and latch it off. Or check out the VN920 I mentioned.

Automotive design can be extremely tricky. It's good to overspecify (within budget) and build in redundancy. So overspecify the fet so it can handle the s/c current, back that up with active overcurrent protection to cut the fet off as soon as there's a short, back that up again with a polyswitch in the enclosure, and back that up yet again with a mechanical fuse in the fusebox. I never rely on one thing alone.

While you're at it, consider transient voltage immunity. In an automotive environment it's a very good idea to design your product to comply with ISO7637-2 pulse 5. That's the load dump pulse where the battery is disconnected from the alternator when the engine is running. It is an enormously difficult pulse to handle - 87V peak, 400ms decay time, supply impedance 0.5 - 4R (the peak voltage may affect the rating of your pfet depending on how you handle the pulse). If you google "ISO7637-2" there's a lot of stuff to plough through. The standard itself is available for free here: http://www.infovusam.sk/pdf/normy/iso/nove2009/3-14.pdf

edit - that version of the standard doesn't have pulse 5 because it's since been merged into other standards, but this version does: http://xenona.com/xenonaupl/AHDPO/EMC/ISO/ISO_7637-2_2004.pdf

[Psi]

With so many devices in the car containing TVS diodes wouldn't the 87v pulse get chopped to ~24V ?

[SeanB]

Not likely, that pulse will blow car lamps quite nicely, even he headlights if they are on.

I did see a pair of shadetree mechanics starting a car by using a battery, but they did not have jump leads, just placed the battery in place of the other, then pulled the terminals off and placed the flat battery back about a minute later. That the CVH engine kept runnning is a testament to how tough the OEM ignition modules are, along with almost all alternators now having TVS reducing diodes in the bridge, basically just using 100A 30V zener diodes for 3 of the diodes. try that on a more modern car and it stops with a wisp of smoke coming from every electronic box.

[lewis]

With so many devices in the car containing TVS diodes wouldn't the 87v pulse get chopped to ~24V ?

Unfortunately not, the inductance of the wiring loom can cause significant differences in the supply voltage between devices under transient conditions. So the transient can appear at your product even if it is clamped by other devices in the car. There is no 'global transient suppressor' in vehicles, to meet the standard each electronic device has to be designed to handle the pulse in isolation.

[SeanB]

Some cars have a 5A fuse in the fuse box for the ECU, along with a plug in surge suppressor diode ( expensive and agents only) to protect the expensive ECU. This often fails short and blows the fuse, and then is removed and not replaced, which ultimately results in the destruction of the ECU.

[Psi]

I'm quite liking these high side driver chips VN920-E and similar. (Thanks lewis)
They look to be perfect for what i'm doing. Also having the ability to monitor current would be very useful in the product.

With so many devices in the car containing TVS diodes wouldn't the 87v pulse get chopped to ~24V ?

Unfortunately not, the inductance of the wiring loom can cause significant differences in the supply voltage between devices under transient conditions. So the transient can appear at your product even if it is clamped by other devices in the car. There is no 'global transient suppressor' in vehicles, to meet the standard each electronic device has to be designed to handle the pulse in isolation.

Good point.

[Corporate666]

I'm quite liking these high side driver chips VN920-E and similar. (Thanks lewis)
They look to be perfect for what i'm doing. Also having the ability to monitor current would be very useful in the product.

With so many devices in the car containing TVS diodes wouldn't the 87v pulse get chopped to ~24V ?

Unfortunately not, the inductance of the wiring loom can cause significant differences in the supply voltage between devices under transient conditions. So the transient can appear at your product even if it is clamped by other devices in the car. There is no 'global transient suppressor' in vehicles, to meet the standard each electronic device has to be designed to handle the pulse in isolation.

Good point.

I'm using the VN920-E, very nice little device.  I've beat the shit out of some of them in testing and they do exactly what they say - take a beating and keep on going   Haven't found the current sensing to be too accurate but haven't need it to be either.

One thing about polyfuses... if you are using big FET's then you might be switching big loads and might have big thermal considerations that lead you to making a very thermally efficient board.  In my experience, when you have very thermally efficient boards, PTC fuses are very unreliable.  They need to get hot to trip and they need to heat up quick.  That's especially true if you are dealing with mild overcurrents (above the PTC trip threshold but not a dead short either). 

In one case, we had an LED board on a metal core PCB, mounted to a copper slug which then was mounted to the aluminum housing.  The PTC in that case was right near the LED's and was useless when it came to blowing from mild overcurrents (even though they were well in excess of the PTC rating). 

[poorchava]

You can use thermal relief near the polyfuse. What matters is not to keep the trace width at certain minimum level, but rather provide enough thermal mass to the trace. Which means that you can connect polyfuse with rather thin thermals and everything should be okay as long as the track gets much wider next to the thermal relief.

As for TVS diodes: each electronic unit should have a high voltage (100-200V) ceramic capacitor, preferably in 10nF range in parallel with hefty TVS diode to ground. This serves two purposes: mnimizes transients and ESD and also protects your circuit from accidental reverse battery polarity conditions (TVS will conduct full battery current and blow the fuse very fast).

[lewis]

Just a word of warning - TVS diodes are generally ineffective against the ISO7637 pulse 5 unless you get an absolutely enormous TVS. Paralleling TVSes is also very bad, they can never be perfectly matched and one of them will hog the transient. There's a good paper here: http://www.radiocad.co.uk/_downloads/LoadDumpPaper-final.pdf