Hi, we are designing a 48V high side MOSFET switch and we realised that the our mosfet gets too hot. Our solution is to connect additional mosfets in parallel to distribute the load/heat. But I'd like to ask your how to properly connect PMOS in parallel. I read somewhere that the gates should have separate resistors (i dont really understand why)
but im not sure how to do this with our current circuit:


im thinking to have separate zener diodes, 10k and 22K for each additional pmos, then connect them all at the drain of the NMOS.
is this the right way to do this or is there a better way? or can i get away with connecting the gates together and have them share the same zener and resistors?

 

You can use one set of resistors and one Zener for all the MOSFETs in parallel.
No reason for separate components.
Is the switch operating frequency high, as those resistor values will give a slow on and off of the MOSFETs which will increase dissipation during switching?

 

You can use one set of resistors and one Zener for all the MOSFETs in parallel.
No reason for separate components.
Is the switch operating frequency high, as those resistor values will give a slow on and off of the MOSFETs which will increase dissipation during switching?

No, the switch is constantly ON. It is OFF only when we want to power OFF the whole device.
so does this mean that the separate gate resistors only applies when switching the mosfet at high frequency?

 

so does this mean that the separate gate resistors only applies when switching the mosfet at high frequency?

Yes, but even then it may not be needed.

 

You may not need multiple FETs.
The problem may be the FET specifications.

What is the part-number of your current FET ?
What is the Voltage on the Gate when "On" ?
What is the Character of the Load ?, Maximum-Amperage ?, Motor ?
.
.
.

 

You may not need multiple FETs.
The problem may be the FET specifications.

What is the part-number of your current FET ?
What is the Voltage on the Gate when "On" ?
What is the Character of the Load ?, Maximum-Amperage ?, Motor ?
.
.
.

We are using SUP90P06-09L-E3 https://www.vishay.com/doc?73010
the gate zener diode is 11V.
We tested our circuit with an electronic load, drawing a constant 25A as this is the max expected current our load will draw.
we originally had 1 fet but we notice that it gets very hot and affects the surrounding components. we tried adding a heat sink but it has no significant change in temp

 

Excellent FET choice,
but it does require a proper Heat-Sink.

What is the actual MEASURED Voltage at the Gate ?
What is the actual MEASURED Heat-Sink-Tab-Temperature ?
.
.
.

 

What is the actual MEASURED Voltage at the Gate ?
What is the actual MEASURED Heat-Sink-Tab-Temperature ?

We havent measured the Gate voltage, sorry. Should i measure from Gate to ground? im assuming Gate to Source is -11V from the zener diode. Im wondering what this voltage tells us?
The measured temperature went above 170C in just 4mins (with the 25A load), we stopped the test at this point so we dont damage the board.
We also tested with 13A load and temperature was <70C. We are assuming that if we have 2 parallel mosfets it would half the current running through them, hence the 13A.

 

If you added a heatsink and saw no significant temperature change your choice of heatsink was poor.

If you want to use purely passive cooling (no fan, &c.) your heatsink will need to be a lot bigger. But you might try putting a fan on the one you are currently using, as a test, to see if you get a reasonable thermal equilibrium.

Keep in mind, too, that on a PCB you can add copper floods to increase the heat dissipation and protect surrounding components. Using multiple devices could work, but for it to be effective you will be using a very overrated array of MOSFETs to keep the temperature down.

If space is a big problem, the multi-device route might make sense but if you have the real estate, a larger heatsink can do wonders. Adding a fan, even more.

 

I calculate 9W dissipation at 25A. A reasonable heat sink should be able to handle that.

 

I calculate 9W dissipation at 25A. A reasonable heat sink should be able to handle that.

Yes, I agree... between 6 and 12.5W depending on junction temp...

@dr3iz What is the model # of your heatsink?

 

"" assuming Gate to Source is -11V from the Zener-Diode. ""

The Zener simply insures that the Gate Voltage
won't be able to "Exceed" -11-Volts.

There may be some "slight" advantage to changing the Zener to ~18-Volts,
but this may do nothing at all if the driving Voltage from other Circuitry is not adequate.

Assumptions won't do,
You must verify the Gate-Voltage with a Meter, or an Oscilloscope.

For a start, change the 10K Resistor to 100K.
.
.
.

 

For a start, change the 10K Resistor to 100K.

That will, of course, significantly reduce the turn-off time of the MOSFET(s).

 

What is the model # of your heatsink?

the heatsink we tried looked similar to the attached image. not sure what the part number is as they are extra components from previous projects. We are planning to use this similar heatsink https://au.mouser.com/datasheet/2/670/hss_b20_01-1777955.pdf
it is twice the height of the one we tested.

If space is a big problem, the multi-device route might make sense but if you have the real estate, a larger heatsink can do wonders. Adding a fan, even more

Space is indeed an issue so we cant use those massive heatsinks. The electrical side is also enclosed in a case, so no airflow. if this parallel fets with this small heatsink will not solve our issue then we will investigate modifying the design to allow a bigger heatsink and/or airflow.

The Zener simply insures that the Gate Voltage
won't be able to "Exceed" -11-Volts.

There may be some "slight" advantage to changing the Zener to ~18-Volts,

Could you explain why the Gate voltage is interesting. i thought as long as i keep Vgs>Vth and Vgs<Vgsmax then the mosfet would turn ON no problem.

 

The Gate-Voltage determines the "Rds/On" Resistance of the FET.
( Look at the graph in the Data-Sheet )
The Rds/On-Resistance determines how much Power will be lost
in the form of HEAT that must be dissipated by the FET.

A FET is NOT an "On-Off" Switch.
.
.
.

 

Your heatsink is way too small.

Look at the Safe Operating Area chart. On the DC line at 25A the Vds will be around 0.4v giving a dissipation of 25 X 0.4 = 10W approx at a nominal junction temperature of 25C, but of course that's impossible without an infinite heatsink. However that gives a value for Rds(on) of 0.4/25 = 16mOhm at 25C.

As a rule of thumb, case temperature should be no more than 75C which would give a junction temperature of Tc + P * Tjc = 75 + 10 * 0.6 = 81C
From the Rds v temperature chart the actual Rds(on) will be approx 1.3 higher giving dissipation as 25 * 25 * 0.016 *1.3 =13W approx which will lead to a slightly higher junction temperature.

Keeping with a case temperature of 75C and a Max dissipation of 15W to give some wiggle room and assuming an ambient of 30C your heat-sink needs to be better than (75 - 30)/15 or <= 3C/W, about 6 times more efficient than the one you propose... And that's in free air. No wonder it was getting hot with an even smaller sink!

Basically you need to look at forced air cooling or some other approach.

Putting multiple MOSFETs in parallel will help. Let's say 5 in parallel, each handling 5A. The Rds(on) is around 0.02ohm suggesting each MOSFET dissipates around 0.5W. In free air with no heat-sink that would give a junction temperature of around 32C above ambient. That's much more promising. Assuming the ambient inside your case is around 40C then thats more viable. Ideally you want a finned metal case with the devices bolted to the inside to try and get heat out as even 40C will have an adverse impact on the other components and that will impact overall reliability. If you can, put an external heat-sink on there; even a few sq cm will pay dividends in the long run.

 

You may be able to find a device with a lower Rds(on). If not, parallelling would work, or if a big redesign is needed even to consider using an NMOS which would need a small DC-DC converter to drive the gate to 10V above the output rail (i.e. around 58-60V.) Plus protections etc. That is not particularly difficult, but is more complication for sure. But an efficient converter (if it's only an on-off switch) does not have to provide much power (a swtiched capacitor charge pump may suffice) as once the gate voltage reaches 10V drive virtually no current is needed) and a low Rds(on) FET may be cheaper than several FETs in parallel. The FET you chose had a max RDs(on) of .02 ohms at 175C and I suggest you should be looking at a fifth to a tenth - equivalent of 5 in parallel at minimum as said above.

 

A low-side NMOS switch in the -ve battery rail is also an option, and NMOS MOSFETs with Rds(on) of 1 - 2mOhm are readily available. @neonstrobe is right that a simple capacitive charge pump for a high-side NMOS switch is a viable option. The PMOS MOSFET chosen is about as good as it gets for PMOS esp in a TO-220 case. There are some better ones (sub 10mOhm) in a SMT DPAK or similar case but heatsinking those is more problematic as needs large PCB copper real-estate.

 

the heatsink we tried looked similar to the attached image. not sure what the part number is as they are extra components from previous projects. We are planning to use this similar heatsink https://au.mouser.com/datasheet/2/670/hss_b20_01-1777955.pdf
it is twice the height of the one we tested.
View attachment 296077

Space is indeed an issue so we cant use those massive heatsinks. The electrical side is also enclosed in a case, so no airflow. if this parallel fets with this small heatsink will not solve our issue then we will investigate modifying the design to allow a bigger heatsink and/or airflow.



Could you explain why the Gate voltage is interesting. i thought as long as i keep Vgs>Vth and Vgs<Vgsmax then the mosfet would turn ON no problem.

Gate voltage [Vgs] is critically important - the higher it is the lower the drain-source resistance, as you will discover by re-reading the data sheet. N.B. Vgs is measured from the gate pin to the source pin, NOT to ground!

Google "heat sink design" for many useful guides. See : https://www.allaboutcircuits.com/tools/heat-sink-calculator/

As others have said, your chosen heatsink is far too small, especially in a small case without active airflow. Any chance your enclosure is metal? It might have enough surface area to do the job? My elderly [1975] National Semiconductor Voltage Regulator Handbook has a worked-out design example showing that a single black-anodized aluminum 1/16" [1.6mm] thick vertical fin 4.25" [108mm] square would have an effective thermal resistance of 4 degC/watt. Your ~10 watts of dissipation would cause that fin to rise ~ 40 degC above ambient IN FREE AIR. Take it from there; just understand that higher temperatures will also make the FET's Rds(on) go UP, i.e. the part will get even hotter. Re-read the data sheets carefully, work out some examples.

Just for fun, check out DigiKey's J115F11A5VDCSH.6 mechanical thru-hole power relay, fully sealed, which would occupy ~ same space as your MOSFET circuit & dissipate far, far less power in its drive coil [5V * 120 mA = 0.60 watts]. You'll need to add a diode across the coil to clamp the turn-off inductive spike, plus a similar arc suppression device across the contacts of this relay [is your load highly inductive?]. Can your 5 volt logic source provide 120 mA? If not, use your existing NMOS driver circuit to power the relay, but now from your 48 volt rail. Relay part # J115F11A48VDCSH.6 . Same 0.60 watt coil power. Just a thought. Enjoy!

 

Also,
I don't think anyone has mentioned the possibility of using a Low-Side N-FET.
.
.
.