MOSFET heat calculation

[sobrinho]

Hi there,

I'm trying to understand how much a MOSFET will heat up given a load of 5V/3A DC.

My understanding so far is that formula:

P = RDS_on * I ^ 2
T = P * R_thJA + TC

RDS_on

My first question is how to interpret RDS_on.

Looking at IPP80P03P4L04AKSA1 datasheet it says that it will be around 4mΩ.

P = RDS_on * I ^ 2
P = 4 / 1000 * 3 ^ 2
P = 0.004 * 9
P = 0.036

T = P * R_thJA + TC
T = 0.036 * 62 + 25
T = 27.232ºC

This means a heat up of ~2ºC over TC assuming -5V of V_GS and TC of 25ºC.

Are that formula and calculations correctly?

Alternatives

What I'm curious about those calculations is that if they are correct, I'm not finding any P-Channel MOSFET for sale in Brazil that is below 0.200Ω.

One that is widely available here is the IRF9540 and by those formulas, it would heat up to 137,5º even that the datasheet says it supports up to 19A.

It seems odd to the MOSFET heat almost to its maximum operating temperature with only 15% of its maximum load.

Thanks!

[jremington]

It seems odd to the MOSFET heat almost to its maximum operating temperature with only 15% of its maximum load.

Depends on the the heat sink. You would need a very efficient one, with an excellent thermal connection, to support the maximum current.

[sobrinho]

Well, I was looking for a MOSFET to be used without a heatsink due to space restrictions.

If the formula and calculation are correct, IPP80P03P4L04AKSA1 is a very good option.

[MarkT]

sobrinho:
Hi there,

I'm trying to understand how much a MOSFET will heat up given a load of 5V/3A DC.

My understanding so far is that formula:

P = RDS_on * I ^ 2

Yes, although if using PWM you have to consider switching losses too

T = P * R_thJA + TC

No, its power divided by thermal resistance that gives temperature difference. Perhaps you meant
thermal conductivity, not resistance?

RDS_on

My first question is how to interpret RDS_on.

Looking at IPP80P03P4L04AKSA1 datasheet it says that it will be around 4mΩ.

At logic levels (5V) the worst-case on-resistance is 7 milliohms. Always look down the table for the worst-case figure
at the gate drive voltage of interest. Any graph is for "typical device", don't use those for design, always
use the worst-case.

P = RDS_on * I ^ 2
P = 4 / 1000 * 3 ^ 2
P = 0.004 * 9
P = 0.036

T = P * R_thJA + TC
T = 0.036 * 62 + 25
T = 27.232ºC

This means a heat up of ~2ºC over TC assuming -5V of V_GS and TC of 25ºC.

Are that formula and calculations correctly?

3A is nothing for a low on-resistance like this - but yes, I-squared-R is the way to figure this out always.

Alternatives

What I'm curious about those calculations is that if they are correct, I'm not finding any P-Channel MOSFET for sale in Brazil that is below 0.200Ω.

p-channel FETs are both rarer and intrinsically 3 times worse in performance all else being equal - this is
a physical property of silicon.

One that is widely available here is the IRF9540 and by those formulas, it would heat up to 137,5º even that the datasheet says it supports up to 19A.

Never go anywhere near the maximum current rating of a MOSFET (except for short pulses). Its completely
infeasible as you'd need water-cooling to get there. Max current specification is defined by max power dissipation on
infinite heatsink... You only ever need to check the max current for pulse operation.

Some manufacturers lie about the figure too, quoting a max current rating higher than the package's
maximum current handling!!

It seems odd to the MOSFET heat almost to its maximum operating temperature with only 15% of its maximum load.

Yes, it catches people out - power MOSFETs are very different from BJTs

[DVDdoug]

Its been awhile since I've done thermal calculations and the real-world specs for heatsinks, insulators, and thermal compound, etc., always seemed a little sketchy. And then there's the ambient temperature which can vary, especially if the heatsink is inside a box, and it can depend on airflow, etc. If the air around the heatsink is stagnant, the immediate-ambient temperature increases.

But... A small fraction of a Watt isn't going to overheat a T0-220 device. Above 1W it's probably going to be too hot to touch (assuming no heatsink) and it might overheat and burn-up.

It seems odd to the MOSFET heat almost to its maximum operating temperature with only 15% of its maximum load.

If you can keep the case cool with a BIG heatsink, or water cooling, or whatever it takes, it should be safe running at it's maximum-continuous current rating when switched fully-on.

If it's used "linearly" (as an amplifier or linear voltage regulator, etc.) you have to be aware of it's maximum power (Wattage) rating, and if you are "pushing it", you WILL need heatsinking.

[MarkT]

Never even contemplate trying to run a MOSFET at its max current rating, I assure you it
will be a disaster. For one thing it probably assumes 10V gate drive.

[sobrinho]

Yes, although if using PWM you have to consider switching losses too

So, I'm calculating P correctly but I should use RDS_on(max) instead.

I don't need to use PWM if I want to turn the MOSFET just on and off, right?

My plan is to use a digital pin from an Arduino to turn on/off a resistive load.

No, its power divided by thermal resistance that gives temperature difference. Perhaps you meant
thermal conductivity, not resistance?

What would be the thermal resistance?

Essentially I'm trying to figure out how much degrees the MOSFET is going to be above TA.

If it's used "linearly" (as an amplifier or linear voltage regulator, etc.) you have to be aware of it's maximum power (Wattage) rating, and if you are "pushing it", you WILL need heatsinking.

Never even contemplate trying to run a MOSFET at its max current rating, I assure you it
will be a disaster. For one thing it probably assumes 10V gate drive.

I'm not trying to use it on its maximum, not even barely near.

I'm trying to use it just to turn on and off a resistive load of 5V/3A DC, a digital switch.

[MarkT]

I'm not trying to use it on its maximum, not even barely near.

I was replying to DVDdoug...

What would be the thermal resistance?

Thermal resistance is measured in degC/watt, thermal conductivity in watt/degC.
I think I might have been confused above, since for heat conduction power is treated as current,
temperature as voltage. The easy way to get calculation right is to follow the units. If a heatsink
is given as 3 degC/watt, multiply the power by that figure to get the temperature rise.

[sobrinho]

The easy way to get calculation right is to follow the units. If a heatsink
is given as 3 degC/watt, multiply the power by that figure to get the temperature rise.

This means never use a MOSFET without a heatsink even with small loads?

[JohnRob]

This means never use a MOSFET without a heatsink even with small loads?

I wouldn't go that far but in your situation, you are running the MosFet (IRF9540) with a gate voltage that does not fully turn the MosFet ON (i.e. conducting). Also note that Rds goes up with increase in device temperature.

To make matters worse, heat is the result of expended energy and must go somewhere else it builds up.

As a though experiment; consider you put your MosFet in a well insulated foam housing. As the Mosfet dissipates energy it will heat the local air, however due to the foam not letting any heat out, the MosFet will continue to get hotter until it fails.

For your project you need to get the heat out of your device. You did not describe your project but if I assume your device is in a Plastic housing of some sort then you need to get rid of the heat:

If mounted alone on a board with no heatsink, the heat path is:

Mosfet --> Air around Mosfet --> plastic housing --> ambient air. This is a pretty poor path to get rid of heat.

I suggest you make one side of the housing out of aluminum and attach the MosFet to the aluminum. The path then becomes:

Mosfet --> aluminum --> ambient air. Decidedly better.

I suggest you obtain one of your best MosFet options and perform a test. Often MosFets will perform better than the spec.

[sobrinho]

I'm not using the IRF9540, I'm trying to understand how to calculate the heat to be able to find a good option.

Understanding how to calculate that I can try to find one that does not require a heatsink or a very small one that wouldn't' require too much ventilation since this is going to be enclosure a box.

If my formula is right, the IPP80P03P4L04AKSA1 is going to heat only 2ºC above ambient air, that wouldn't require a heatsink at all.

[MorganS]

It's most highly dependent on the gate drive. The headline specification on the first page of the datasheet is R(DS)on for 10V drive (or -10V for a P-type). You always have to look at the charts to find the actual operating condition you are using.

But yes, after that point then it's simple multiplication and division. The heatsink calculations are very simple.

[sobrinho]

Around 0.004Ω for -5V on Gate for IPP80P03P4L04AKSA1.

How to calculate the heating?

[charliesixpack]

The spice model probably contains temperature information. You can download it here...
Infineon P2 device spice model

LTspice is a free simulator. If you are unfamiliar...
Sparkfun LTSpice tutorial

[MarkT]

Spice normally

MorganS:
It's most highly dependent on the gate drive. The headline specification on the first page of the datasheet is R(DS)on for 10V drive (or -10V for a P-type). You always have to look at the charts to find the actual operating condition you are using.

But yes, after that point then it's simple multiplication and division. The heatsink calculations are very simple.

The charts are not the place, they are "typical device", and devices vary a lot for anything to do with
gate voltages in MOSFETs - checkout the maximum Rds(on) spec.

Many MOSFET gate voltage parameters are quoted as +/-1V, ie a 2V variation between devices/temperatures/aging.

[MorganS]

"maximum Rds(on)"????

There is no maximum.

Yes, there is variation part-to-part and over temperature. That information (if it is in the datasheet) is in the charts, not in the simple table of maximums.

[MarkT]

There is always a maximum Rds(on) in the datasheets, its about the most important spec in the whole
datasheet.

Look at the V(thr) spec for a clue to the variation in gate voltages - MOS thresholds are extremely sensitive to ion concentrations and position in the ultra-thin gate oxide layer, and this drifts over time
due to applied voltage and varies between devices. Modern devices are much better than older
in this respect but its still important to know that the typical graphs can be a volt or so different
from your device at some point in its life...

The maximum Rds(on) figure is the one you can rely on as guaranteed by the manufacturer.