MOSFET questions

I'm looking to take a signal (from a car ECU) to a MOSFET (it is PWM), then if the MOSFET fails (i.e. due to inadequate heatsinking, or else), that it fails in a closed state so that the output (Radiator Fans) stay on.

Can I sink the heat off the MOSFET to the car frame, given that it is also ground, or will that affect operation?

A component can fail short or open or anywhere inbetween there is no way of telling.

You can heat sink to another potential if you use an insulated heat sink mounting kit consisting of a few Insaluted washers and some heat sink compound.

I'll probably stick relays in with it - MOSFET, and when the load is greatest, I'll use the relays.

When do MOSFETs produce the greatest heat ? I recall reading somewhere that when switching (i.e. not fully on), they get their hottest - is that right?

Heat is produced when they dissipate power.that is when the product of voltage dropped across them times current through them is at a maximum. Not turning them on fully is only one situation, not turning them on fast enough is another, especially in thisng like PWM.

tocpcs:
I'll probably stick relays in with it - MOSFET, and when the load is greatest, I'll use the relays.

You could put a manual override switch on the dashboard as well.

tocpcs:
When do MOSFETs produce the greatest heat ? I recall reading somewhere that when switching (i.e. not fully on), they get their hottest - is that right?

Yes, which is why PWM signals need special care with MOSFETS. The MOSFET driver has to be able to sink/source as many amps as possible to keep the transition times short.

If it fails you simply need to be lucky lol

How much power is being disapaited ? I was thinking maybe a sense resistor to determine if the fet is still behaving, by enabling, disabling you can measure the voltage drop via a sense resistor and if nothing changes signal a warning.

cjdelphi:
If it fails you simply need to be lucky lol

How much power is being disapaited ? I was thinking maybe a sense resistor to determine if the fet is still behaving, by enabling, disabling you can measure the voltage drop via a sense resistor and if nothing changes signal a warning.

If it's for a car radiator then maybe the existing temperature sensor is enough...

Yeah a simple ntc thermistor would do the job :slight_smile:

A thermistor can't be connected to the MOSFET, now can it.. :slight_smile:

tocpcs:
A thermistor can't be connected to the MOSFET, now can it.. :slight_smile:

On top of it... sure.
Using the thermistor with a voltage divider to the gate of the fet... it could be

I was thinking of it to measure the radiator. ..

Insulting posts deleted. I trust I won't have to visit this particular thread again in a hurry.

Let me check this is right..
8A x 14V is 112W
112W x 0.45 deg C/W Junction to Case is 50.4 deg C.
50.4 deg C + 30 deg expected ambient is 80.4 deg.

80.4 deg C is less than the datasheet Operating Junction Max Temp of 170 deg C.
It's Junction to Ambient is 62 deg C/W.

No heatsink needed - or have I got it completely backwards?

8A x 14V is 112W

It is but you will never have 8A flowing through the FET with 14V across it.
When there is 14V across it there is no current down it.
When there is 8A flowing through it the voltage across it is determined by the "on resistance" of the FET. This will be in the data sheet. Suppose it is 0.05R then 8A will give 0.4V across the FET. Therefore the FET will be dissipating 0.4 * 8 = 3.2W.

tocpcs:
Let me check this is right..
8A x 14V is 112W
112W x 0.45 deg C/W Junction to Case is 50.4 deg C.
50.4 deg C + 30 deg expected ambient is 80.4 deg.

80.4 deg C is less than the datasheet Operating Junction Max Temp of 170 deg C.
It's Junction to Ambient is 62 deg C/W.

No heatsink needed - or have I got it completely backwards?

If it was dissipating 112W without a heatsink then the temperature to ambient would
be 62 * 112 = 6.7kW !! But of course its not dissipating 112W.

You have to add the thermal resistances together, and if there is no heatsink you have
to add the 62 C/W (otherwise you add the heatsink's thermal resistance to ambient).
The 0.45 C/W is really only relevant if you bolt the device onto a water-cooled mass of
copper.

During switching the MOSFET might briefly be carrying about 4A at 7V (assuming
a resistive load) when the channel resistance matches the load resistance, and this is
the maximum dissipation during switching (28W or 25% of the full load). However this
might only be for 1us, so its only an issue when the switching is at a high frequency
so that the proportion of time spent switching is non-negligible.

http://www.jaycar.com.au/products_uploaded/zt-2468.pdf

That's the MOSFET i'm looking at - 0.005 ohms.

The math I've applied is 10A x 0.005 = 0.05, this times by the amps is 0.5 W, so the MOSFET will dissipate 0.5W when switching 10A?

Yes that is the static dissipation. As mentioned before at the point of switching over it is more than this which is only a problem if you are switching it on and off rapidly like with PWM.

Note that is not a logic level FET and you need a gate voltage of at least 10V to turn it on that hard.

It will be with PWM, so I've gone and got some heatsinks.
To solve the 10V issue, I plan to use some 2n3904 transistors to drive them on / off, so it'll be 5V on to the base, which will have 12V at it's emitter and the gate of the MOSFET at it's collector..

That will turn the MOSFETs off quickly, but not On quickly.

Define quick?
I'd still get a slow down in speed if I was switching them say fastest 400ms timebase (so 50% duty cycle = 200ms On period)?