Calculating Pd and thermal req for hi-current MOSFET switch (SUM60020E)

[yan]

Hi, I'm trying a find a MOSFET to use as a hi-current switch for a 16V/100A supply and could use two or more in parallel.
I've found the Vishay SUM60020E part which has an Id=150A and could be controlled from a 5V Vgs signal, but I'm just trying to calculate the thermal req...

Rds is 0.0025ohms
So Pdiss = IxIxR = 100 x 100 x 0.0025 = 25W
if Tamb is 40deg does this mean Tj=Pdiss x Rthja = 25 x 40 = 1000degC?

Any help would be most appreciated.
Thanks

[Wolfram]

Temperature depends on Pdiss, which depends on Rdson, which depends on temperature again, so some iteration is needed to get realistic figures.

Your calculation doesn't account for the fact that Rdson increases with temperature, but it still nicely shows that the device cannot be used under the specifications that Rja are specified. Let's say we want to keep the junction temperature at 125 celcius, with a 25 celcius ambient. This leaves some margin, and limits the Rdson increase compared to running it closer to Tjmax. At this temperature, Rdson is 1.6 times the room temperature value. Also running the device at 5 V Vgs roughly doubles Rdson compared to the value at 10 V (page 3 in the datasheet). This gives us a realistic Rdson of around 60 milliohms, and a dissipation of 60 W at 100 A. Since Rth_jc is 0.4 k/w, and we want the die to not exceed 125 celcius, the case temperature needs to be kept below (125 C - 60 W * 0.4 k/W) = 100 C. With an ambient of 25 celcius, and a heasink temperature of 100 celcius, the heatsink needs to have a thermal resistance Rth_sa of ((100 C - 25 C) / 60 W) = 1.25 k/w. This is just an example based on some arbitrary but not unrealistic numbers.

Using a MOSFET with a lower Rdson and a lower Vgs_th would lower the losses and also simplify the cooling requirements. Using a package like a TO-220 or a TO-247 also makes it easier to heatsink. Lower voltage devices will also have lower Rdson for a given device size, so if you don't need the device to withstand 80 V in the off-state, using a 40 or 60 V device instead could also help.

[yan]

Hi, many thanks for taking the time to reply...
I have a few ques.
Where did you get the figure for Rdson being approx double for Vgs of 5V as the graphs on pg3 show just 7.5V & 10V & also the figure of 60mohms?

Would you have any recommendations for a device that would be more suited to this application as I thought Vgsth of 2V was quite low?
Also maybe it would help to use a level shifter on the gate to increase the gate voltage from the uProc and so reduce the Rdson?
Many thanks...

[Le_Bassiste]

quite an oldie, but still worth a thorough read:
https://usermanual.wiki/Document/1985SiliconixMOSPOWERApplications.64667194/help
chapter "4.2 HP-41CV Power MOSFET Thermal Analysis Program" might give you some insights on how to calculate heatsink requirements.

hth

[Wolfram]

Hi, many thanks for taking the time to reply...
I have a few ques.
Where did you get the figure for Rdson being approx double for Vgs of 5V as the graphs on pg3 show just 7.5V & 10V & also the figure of 60mohms?

Would you have any recommendations for a device that would be more suited to this application as I thought Vgsth of 2V was quite low?
Also maybe it would help to use a level shifter on the gate to increase the gate voltage from the uProc and so reduce the Rdson?
Many thanks...

Just a quick reply as I'm on the way out. I estimated the Rdson doubling from the output curves on the upper right graph on page 3. The voltage drop is roughly double at 100 A when Vgs is 5 V compared to 10 V. With less overdrive (Vgs - Vgsth), the transistor starts to saturate at lower currents, note that the curves start to flatten out at higher currents And lower Vgs. In this operating region, the transistor stops looking like a resistor and starts looking like a current sink, so the concept of an on-resistance makes less sense. We can still calculate an equivalent Rdson at a given operating point for use in power dissipation calculations, just note that Rdson depends on current here.

I don't have a particular device in mind, but have a look around using for example the parametric search at Mouser to find suitable devices. Vgs_th of 2 V is pretty low, but for a high current application with 5 V gate drive (worst case? Or can the voltage be lower in practice?) I would look at devices with a Vgs_th in the 1 - 1.5 V range. What matters ultimately is the total dissipation with the load current and Vgs you want to run.

A level shifter to get higher gate voltage can also help. Be careful about the switching times. Large MOSFETs tend to have high gate capacitance, and if you turn them on or off too slow with a load current of 100 A, they won't even outlive the first switching event. For more frequent switching, the accumulated dissipation of repeated switching events can also eat into the thermal budget, but this usually starts becoming a concern at hundreds of hertz to kilohertz.

[tooki]

Wolfram and Yan: Why all the guesstimating? The third graph on page 4 plots Rds(on) vs Vgs, at 25° and 125°.

5V is right in the middle of the elbow of those curves. That means that if your 5V signal sagged for whatever reason, your Rds(on) would rise sharply.

I’d either choose a logic-level mosfet, or use a mosfet driver to drive this mosfet cleanly.

[Wolfram]

Wolfram and Yan: Why all the guesstimating? The third graph on page 4 plots Rds(on) vs Vgs, at 25° and 125°.

5V is right in the middle of the elbow of those curves. That means that if your 5V signal sagged for whatever reason, your Rds(on) would rise sharply.

I’d either choose a logic-level mosfet, or use a mosfet driver to drive this mosfet cleanly.

Good point, I completely missed that plot, it's not one I'm too used to seeing in MOSFET datasheets. Agreed that 5 V Vgs is very marginal for 100 A Id with this device, graphs are most likely given for a typical device as well, but the Vgs_th specification for this device is pretty wide, 2 - 4 V, so worst-case values could be significantly worse than the typical curves.

[yan]

Many thanks for all the feedback.
Yes I see your point about increasing Vgs to reduce Rdson and maybe using a different device with  a 3.3V logic level gate turn-on.

I've found the BTS50010-1TAE which although rated at 40A, could possibly be used if three devices were connected in parallel.
Just trying to work through the thermal requirements for the device on pg13 for an amb of 40deg and a 100A load, but its a little confusing.
It looks like as long as VS is approx >10V Rdson should be around 0.1mOhms and with three in parallel would this be feasible?
Assume Rdson total could be taken as 0.1mOhms in parallel?
Thanks

[yan]

Hi, For the BTS50010-1TAE part and with three devices in parallel this does look more feasible to get the 100A and thermal rating:-

Rdson @ Tj=120degC is approx 1.5 times RT => 1.5 x 0.001 = 0.0015ohms
Rdson @ Tj=120degC for three devices in parallel = 0.0005ohms
Pdiss (for three devices) = 100 x 100 x 0.0005 = 5W
Max case temp @ Tj=120degC => 120 - (Pdiss * Rthjc) = 120 - (5x0.5) = 117.5 degC
Heatsink thermal Resistance Rthsa => (case temp - Tamb)/Pdiss = 117.5-25/5 = 18.5K/W

These devices will be soldered direct to the PCB so the brd copper area will likely be the main cooling which will be another calculation...
Do these figures look sensible?
Thanks for all the help....

[yan]

Hi, Just realised that for three devices in parallel the max current will be 100/3 so 33.3A in each device so Pdiss will drop to 0.5W.

Rdson @ Tj=120degC is approx 1.5 times RT => 1.5 x 0.001 = 0.0015ohms
Rdson @ Tj=120degC for three devices in parallel = 0.0005ohms
Pdiss (for three devices) = 33.3 x 33.3 x 0.0005 = 0.5W
Max case temp @ Tj=120degC => 120 - (Pdiss * Rthjc) = 120 - (0.5x0.5) = 119.7 degC
Heatsink thermal Resistance Rthsa => (case temp - Tamb)/Pdiss = 119.7-25/0.5 = 189.4K/W

As they will be soldered direct to the PCB, the brd copper area should be adequate for cooling now...
Do these figures look sensible?
Thanks for all the help....

[TopQuark]

Hi, I'm trying a find a MOSFET to use as a hi-current switch for a 16V/100A supply

What role does the MOSFET serve in your design? Is it the high freq switching element of the SMPS, or just a switch to turn the output off and on infrequently?

If you are calculating power losses of a switching element, losses in the MOSFET contributed by the switching process (MOSFET is ohmic when the gate is being charged up or discharged) can easily dominate the power losses, and static RDSon might not even matter that much, especially if you are switching at high speed with a weak gate driver. All depends what the MOSFET is doing in your circuit.

[yan]

Hi, the MOSFET is just being used to replace a mech relay and is only being used as hi-current switch.
There is no high speed switching.
Thanks