700 watt high-side MOSFET switch?

[Dave92F1]

I'm trying to design a high-side MOSFET switch for an electric RC airplane.

It needs to switch up to 60A at 11.7 volts (3S LiPo), so about 700 watts of DC power.

Is there any P-channel MOSFET that can handle that much power?

It's a simple on/off switch - it doesn't PWM or anything fancy like that.

I was going to use a Infineon IPB80P03P4L-04 (datasheet: http://www.infineon.com/dgdl/Infineon-I80P03P4L_04-DS-v01_01-en.pdf?fileId=db3a30431ddc9372011e07e95eb827d7).

That's rated to handle 80A with a Rds(on) of 4.4 milliohm (max), but I think it can't dissipate the 15 watts of power that passing 60A thru it would generate (at least, not for more than milliseconds).

The thermal resistance on that part is 62 K/Watt (junction to ambient). So it seems like it can't dissipate more than a couple of watts without cooking itself.

Am I confused about this?

Also, my schematic is below.  Do I need clamping or freewheel diodes? The load might be somewhat inductive (I'm not sure).

[Siwastaja]

You can put several in parallel if you can't find a more suitable part.

With that 100k resistor, turn-off will be so slow that you will probably exceed the SOA. The time in linear operation will be too long and the die may hotspot during that. If on-state power energy efficiency is not paramount, you can easily make this resistor smaller, say, down to 1k, which would dissipate 160 mW. (Then, you need to make R1 smaller, too.) Otherwise, you need to consider a push-pull driver.

No need to guess on PCB trace resistances, the web is full of calculators. If space is constrained and this is a one-off application, it may be easiest to enforce the PCB traces with some copper sheet (or wires). 60A is not that difficult. Make everything small. Connectors will be the hardest part.

Freewheeling diode is highly recommended, even when some loads may not need it. Contrary to common misbelief, the body diode in the FET doesn't help at all because it's in the wrong place. Freewheeling diode is connected in parallel with the load, but close to the FET (not close to the load) so that the inductive loop area is small between the bypass caps, FET and the diode.

[Jeroen3]

Is there any P-channel MOSFET that can handle that much power?
...
The thermal resistance on that part is 62 K/Watt (junction to ambient).

Yes, but not on standard board. Most RC speed controllers use more copper, and are often sandwiched between aluminium plates. The other alternative is a module. But that won't fit your purpose.

It's 40 k/w on 6cm^2 copper plane 70um.

Also, the Safe Operating Area in the datasheet says that you can't run more than 40A at 10V. And that is with 1ms pulses.
Their datasheet is a bit contradictory, since they both have 80A for continuous and Avalanche current, single pulse.
They measure this by testing the mosfet while it's being contidioned by a thermostreamer. Which virtually give you infinite thermal capacity.

[NiHaoMike]

What about an automotive relay?

[MarkL]

Take a look at some of Infineon's PROFET high-side switches.  You'll find most of them in the automotive category on their web site.  They have some great protection features built in.

Here's one example that might do it for you, BTS550:

  http://www.infineon.com/dgdl/Infineon-BTS550-DS-v01_00-en.pdf?fileId=db3a30432ba3fa6f012bd3e978183b76

EDIT: Sorry, I meant the BTS555 which has a better Rds(on) = 4mohm (max):

  http://www.infineon.com/dgdl/Infineon-BTS555-DS-v01_00-en.pdf?fileId=db3a30432ba3fa6f012bd3dfdd0b3b65

They can also be paralleled.

[calexanian]

Good lord. 60 amps? Is that like some kind of stalled rotor overload condition or something? Perhaps they should look at increasing the voltage to get the current down. What kind of R/C airplane is this?

[nctnico]

I'd go for a push-pull driver for the MOSFET to make the switching faster (and stay within the SOA limit). With N channel MOSFETs the choice will be much wider because the RDSon for the same die size is 2.5 times lower than for a P channel MOSFET. An automotive relay may be a good solution as well. It is probably cheaper and more robust in the long run.

[Dave92F1]

Thanks for all the replies!  Naturally, answers lead to more questions. 

Worse, consider the milliohms resistance involved in PCB tracking or wiring.  It doesn't take much wiring to add up to another few milliohms.

I was planning 12 GA solid copper wire for the high-current path; not more than 2 inches (5 cm) of it.

But any heating from the wire resistance will be on the wire, not inside the MOSFET, right? So am I correct to think that's not a factor re burning out the MOSFET?

As for ESD / transient protection, You hopefully will be switching into a softly starting load and not abruptly breaking the circuit with the FET at full 60A load current.  Or maybe you will.. Anyway if you are trying to switch from 60A to 0A in some nanoseconds or microseconds there can be significant stored inductive energy released so adding a snubber or freewheeling dioded is not a bad idea.

When it switches on/off the load will be less than 1A. The 60A is just when the airplane is at full throttle (I wouldn't be turning it off then).

In the modified schematic below I added D2 and D3 as freewheel diodes. Are these in the right places? Do I need both of them (or maybe just D2)?

[The airplane is turned on by closing REED_SW with a magnet. An onboard PIC powered by this circuit raises SUSTAIN to 3.3v, so power stays on once the magnet is withdrawn. The PIC monitors SENSE; when it sees REED_SW close again later, it drops SUSTAIN to 0v, so when the magnet is withdrawn power turns off.]

You will have to be sure the channel stays heavily enhanced so you can get down to milliohm level RDSOn ratings as well if you just use a giant low R FET.

I'm pulling the gate to ground (so 11.7v from the source). Vthreshold is only -2.5 volts, so I think I'm OK on that.

With that 100k resistor, turn-off will be so slow that you will probably exceed the SOA.

The 100k (R2) is pulling the gate high, so keeping the MOSFET turned off. Q1 pulls the gate to ground to turn it on. How does R2 affect the turn-on time?

Contrary to common misbelief, the body diode in the FET doesn't help at all because it's in the wrong place.

But D3 (in my new schematic above) is parallel with the body diode. So is D3 redundant? (Not sure if I need it or not.)

Also, the Safe Operating Area in the datasheet says that you can't run more than 40A at 10V. And that is with 1ms pulses.

That's the main thing that worries/confuses me. The MOSFET is rated 80A drain current at 10v but Figure 3 of the datasheet seems to show it can handle it for only 1 millisecond!

I don't know how to interpret that. Am I misreading the datasheet?

Get a proper charge pump based high side load switch driver and use a much better N MOSFET or eGaN GET.

Something like the BTS555 that @MarkL suggested? The best N channel MOSFETs I can find have a Rds(on) only maybe 1/2 of what this part does - I'm not sure that's enough to solve the problem.

What about an automotive relay?

It's a propeller airplane. There is a lot of vibration; I fear it'd shake the contacts loose. Plus big and heavy. (Maybe not compared to the huge heatsink I may need for this tho...I'm trying to avoid that!)

Good lord. 60 amps?

Totally normal for a medium-sized electric RC airplane (~ 1 meter wingspan) at full throttle. Have a look: https://hobbyking.com/en_us/speed-controllers-esc-1/40-79a.html

I'd go for a push-pull driver for the MOSFET to make the switching faster (and stay within the SOA limit).

I guess I'm confused - why is switching speed an issue? Switching will happen only when the current is low (< 1A). The 60A flows only when the airplane is in flight, not on the ground when it's turned on or off.

And what does switching speed have to do with the SOA? I thought (maybe wrongly) the SOA had to do with how much power the MOSFET can pass without cooking itself. How does switching speed relate to the SOA?

(As is probably obvious, I'm new to this power switching stuff.)

[Jeroen3]

Get an N-channel with high side driver please. Preferably with m4 mount holes.
SSR perhaps. Multiple challanges have shown up here:
- you can't run that much current through the mosfet continuously.
- you can't run that much current through standard 35um pcb.

[Siwastaja]

In the modified schematic below I added D2 and D3 as freewheel diodes. Are these in the right places? Do I need both of them (or maybe just D2)?

One parallel with the FET is really redundant. The body diode is most likely good enough.

I'm pulling the gate to ground (so 11.7v from the source). Vthreshold is only -2.5 volts, so I think I'm OK on that.

Vth is irrelevant (another catch for the young players: it doesn't tell you much anything about when the channel is fully enhanced - you need to go look at the Vds/Id curves drawn at different Vgs). Of course, 11.7V is enough for any typical switching MOSFET, with a good marging.

With that 100k resistor, turn-off will be so slow that you will probably exceed the SOA.

The 100k (R2) is pulling the gate high, so keeping the MOSFET turned off. Q1 pulls the gate to ground to turn it on. How does R2 affect the turn-on time?

Please reread more carefully. But if you are 100% sure that you are always turning off at about Id = 1A and not 60A, this won't be a problem.

That's the main thing that worries/confuses me. The MOSFET is rated 80A drain current at 10v but Figure 3 of the datasheet seems to show it can handle it for only 1 millisecond!

Only? 80A at 10V is 800W of dissipation, in a miniscule die. You are using a tea kettle power to heat up a grain of sugar. Or similarly, you are using full nuclear reactor power to heat up one cup of tea!

Of course it can handle it for only 1 ms. But, sadly, these SOA curves can be totally unrealistic, so you need to apply derating for robust operation. So it's going to be a lot less than 1 ms.

Why would you want to operate the FET at Id=80A and Vds=10V? It would leave no voltage for the plane - you are powering a short circuit. Do you realize this condition only happens during switching?

So, switch faster!

1ms is a loooong time. Don't use the FET in linear mode for that long, it has no advantages. You get the EMI reduction / ringing mitigation robustness advantage of slow switching easier, you don't need to go that slow. Think about the range of about 1-10 us. Dissipate that 800W only for 1us, and the die will be OK!

But with that 100k resistor of yours, your switching may actually take 1ms, or even longer. That's why you should use a smaller resistor or a gate driver. You can approximate the switching time from the RC time constant caused by that resistor and the Total Gate Charge of the MOSFET.

The best N channel MOSFETs I can find have a Rds(on) only maybe 1/2 of what this part does - I'm not sure that's enough to solve the problem.

Any proper distributor should have several N MOSFETs with Rds(on) < 1.5 mOhm (when fully turned on) in stock. Parts below 1 mOhm start get a bit rare. You can parallel two.

Good lord. 60 amps?

Nothing special or difficult. No need to up the voltage, IMO, although it would allow some saving in wirings, thus weigh, but wiring is probably very short to begin with. Additionally, it's probably not doing full throttle all the time, so some efficiency loss on wiring is allowed.

I guess I'm confused - why is switching speed an issue? Switching will happen only when the current is low (< 1A). The 60A flows only when the airplane is in flight, not on the ground when it's turned on or off.

You didn't say this earlier, in fact, you explicitly asked how to switch 60A.

Now, with the specification significantly changed; if you can guarantee no switching at 60A, then it's ok to design it not to handle switching the full current. If you try to switch anyway, your switch will blow up as a short circuit and can not turn it off anymore, so you cannot use it as a safety shutdown.

If you really decide to design a system that cannot switch the full load current, then it's a decision which will make the design easier, and you can forget most of our SOA discussion. Even that 100k would be OK, although it's so horribly big that I'd still reduce it to at most about 4.7k.

And what does switching speed have to do with the SOA? I thought (maybe wrongly) the SOA had to do with how much power the MOSFET can pass without cooking itself. How does switching speed relate to the SOA?

While you switch, you are quickly dissipating huge amounts of energy, because the MOSFET is partially conducting, acting as a resistor, dropping a lot of voltage, producing losses P = U*I. Only when it's fully on, the U over the FET is small and losses defined by Rds(on). When it's fully off, the U over FET is the full voltage, there is no voltage left for the load, and the I must be 0. With many types of loads, the losses are biggest when the voltage is shared 50%/50% between the FET and load.

[MarkL]

Something like the BTS555 that @MarkL suggested? The best N channel MOSFETs I can find have a Rds(on) only maybe 1/2 of what this part does - I'm not sure that's enough to solve the problem.

You can put multiple BTS555 in parallel.  You could do 2 or 4 if you really wanted to get the total dissipation down.  The thermal resistance junction to case is 0.35 K/W max, which is quite good, even if you have single device dissipating 14.4W (max).  Typical values will be better.

This and similar parts have high side charge pump, over-current protection, over-voltage protection, over-temperature protection, ESD protection, with low-current drive (1.5mA) all taken care of for you.  There's also a current sense output, if you're interested in what the motor is drawing.

I'm not sure I see the utility of doing it with discrete.

[Dave92F1]

That's the main thing that worries/confuses me. The MOSFET is rated 80A drain current at 10v but Figure 3 of the datasheet seems to show it can handle it for only 1 millisecond!

Only? 80A at 10V is 800W of dissipation, in a miniscule die. You are using a tea kettle power to heat up a grain of sugar. Or similarly, you are using full nuclear reactor power to heat up one cup of tea!

Of course it can handle it for only 1 ms. But, sadly, these SOA curves can be totally unrealistic, so you need to apply derating for robust operation. So it's going to be a lot less than 1 ms.

Why would you want to operate the FET at Id=80A and Vds=10V? It would leave no voltage for the plane - you are powering a short circuit. Do you realize this condition only happens during switching?

No, I didn't realize it. I thought Vds would be 11.7v, but now I realize that once the MOSFET is fully on, Vds is just the drop across the MOSFET, which is tiny because it's only about ~ 4 mOhms.

I guess I'm confused - why is switching speed an issue? Switching will happen only when the current is low (< 1A). The 60A flows only when the airplane is in flight, not on the ground when it's turned on or off.

You didn't say this earlier, in fact, you explicitly asked how to switch 60A.

Quite right, I did. (Sorry!) What I meant was the switch has to handle 60A when it's fully turned on, but the actual current when it switches is < 1A.

I think when I wrote that I didn't realize that it makes a huge difference. (So, I learned something here.)

And what does switching speed have to do with the SOA? I thought (maybe wrongly) the SOA had to do with how much power the MOSFET can pass without cooking itself. How does switching speed relate to the SOA?

While you switch, you are quickly dissipating huge amounts of energy, because the MOSFET is partially conducting, acting as a resistor, dropping a lot of voltage, producing losses P = U*I. Only when it's fully on, the U over the FET is small and losses defined by Rds(on). When it's fully off, the U over FET is the full voltage, there is no voltage left for the load, and the I must be 0. With many types of loads, the losses are biggest when the voltage is shared 50%/50% between the FET and load.

OK, I think I get it. When the MOSFET is fully on the current thru it is 60A (max). With 4 mOhm of Rds(on), that's 15 watts to dissipate still, so a pretty hefty heatsink, I think.

This has been very educational, but I think I'm going to take a hard look at the BTS555 solution suggested by @MarkL.

[Dave92F1]

You can put multiple BTS555 in parallel.  You could do 2 or 4 if you really wanted to get the total dissipation down.  The thermal resistance junction to case is 0.35 K/W max, which is quite good, even if you have single device dissipating 14.4W (max).  Typical values will be better.

This and similar parts have high side charge pump, over-current protection, over-voltage protection, over-temperature protection, ESD protection, with low-current drive (1.5mA) all taken care of for you.  There's also a current sense output, if you're interested in what the motor is drawing.

I'm not sure I see the utility of doing it with discrete.

You're very convincing.    I didn't realize such things existed (plus I wanted to learn how to do it myself).

But the BTS555 does sound more practical. Does it also need a freewheel diode, or does it have one internally?

[MarkL]

You can put multiple BTS555 in parallel.  You could do 2 or 4 if you really wanted to get the total dissipation down.  The thermal resistance junction to case is 0.35 K/W max, which is quite good, even if you have single device dissipating 14.4W (max).  Typical values will be better.

This and similar parts have high side charge pump, over-current protection, over-voltage protection, over-temperature protection, ESD protection, with low-current drive (1.5mA) all taken care of for you.  There's also a current sense output, if you're interested in what the motor is drawing.

I'm not sure I see the utility of doing it with discrete.

You're very convincing.    I didn't realize such things existed (plus I wanted to learn how to do it myself).

But the BTS555 does sound more practical. Does it also need a freewheel diode, or does it have one internally?

It has a body drain-source diode like other power MOSFETs.  The device has no path to ground, except through the low-current IN and IS control connections, so you will need an external snubber diode across your load.  I would put this diode in even if you think your load is not inductive.  The external diode is discussed in several areas in the datasheet.

But if part of your goal is to learn about high power MOSFET switching design, an integrated device doesn't do much for that!

[tatus1969]

more things to consider if that had not been posted beforr:
- increase of Rdson over temperature will cause rising losses, more temperature, more loss. can rhermally run away. i use to calculate with 2 to 2.5. times nominal here
- your load probably has significant capacitance. when charging this through the FET, this dumps significant amount of energy into the chip. this happens fast enough so that it cannot be removed through the case, and can crack it. make sure the transistor can handle this, look for single pulse avalanche energy rating.

[Dave92F1]

For whatever it's worth, I got tired of waiting for my BTS555s to arrive from China. I built this over the weekend, using parts I had around:






It's not ideal because it's a low-side switch. But it seems to work fine.

--Dave

[tatus1969]

why a 60v type to switch 11.7v? that has way too high Rdson for 60a without heatsink. Farnell spits out AUIRF1324S-7P with 24v / 800uOhm.

[Marco]

It's not ideal because it's a low-side switch. But it seems to work fine.

What makes a high side switch preferred in this application BTW?

[Benta]

For whatever it's worth, I got tired of waiting for my BTS555s to arrive from China. I built this over the weekend, using parts I had around:

Nice and simple solution. But you'll still need a heatsink, power loss here is around 10 W.

[NiHaoMike]

If you're looking for a junk box source of low voltage, high current MOSFETs, an old motherboard will have several.

[Siwastaja]

For whatever it's worth, I got tired of waiting for my BTS555s to arrive from China.

Another catch for the young players here - if by China you mean a random Ebay or Alibaba supplier (or similar), the chances of getting fakes is too high. It works out sometimes, but the times it won't will cost you so much time and frustration that you should just start buying semiconductors from proper sources such as Element14/Farnell, Digikey, Mouser, etc.

(And beware of Ebay Justice Warriors who will jump on every thread when they see this advice )

[tatus1969]

Ebay Justice Warriors

thanks for that expression, haven't heard that yet 

[mikerj]

I'll take the bait. The BTS555S is not a commonly used chip, so why would anyone to clone it?

Because they are expensive devices and it's easy and quick to re-mark plastic packages once you have the tools.  How do you know they are not commonly used BTW?  There are so many different industries using semiconductors that only the manufacturers or distributors would have that information.

[BravoV]

I got tired of waiting for my BTS555s to arrive from China ...

Interested on the BTS555 you procured, please update us once you've toyed with it.

[rx8pilot]

Get a proper charge pump based high side load switch driver and use a much better N MOSFET or eGaN GET.

Yes - that! But eGaN is not a good choice for a DC switch - much better for high speed switching and they have very delicate gate drive requirements. You can get high current sub-millohm N-FET's and drive them with a charge pump based driver. Paralleling can put a lot of load on a single FET pushing it through the SOA. To be reliable, you may find that is not as easy as it looks.