High power MOSFET question

Circuitsoft · 2012-01-15 13:48

I'm planning to make a brushless DC motor driver out of the prop. The motor I intend to drive has a 150A maximum current, and 60V maximum drive voltage. I plan to make a 3-phase driver with three SiR882DP mosfets per leg per phase; That's 18 MOSFETs per motor. I plan to use an integrated half-bridge gate driver IC, but I'm not sure how much gate drive current I need.

According to the SiR882DP datasheet, with a 4.5v gate voltage, it has a worst-case gate charge of 27nC. If I'm driving the bridges via differential DUTY mode on the counter, and I want to switch the MOSFET within 1/2 clock, then it seems I need 27nC*160Mhz=4.3Amps gate drive Per MOSFET. Am I going way overboard on this? Did I do some of my math wrong?

Circuitsoft · 2012-01-15 13:55

BTW, I'm thinking of using an LM5101A as the gate driver, pending the results of this thread.

Phil Pilgrim (PhiPi) · 2012-01-15 16:32

Driving MOSFET gates with a DUTY-mode signal is not a good idea, due to the very high frequencies involved. Some pulses can be as short as 12.5 ns, and your MOSFETs will be working in their linear region for a much-too-high percentage of the time. It would be way better to use one of the PWM objects available in the OBEX.

-Phil

T Chap · 2012-01-15 16:41

Wouldn't it be a lot easier to use a BLDC controller/driver? I use a moto MC33035 and HP4086 to drive the mosfets(IRF540). I don't think this will work at the voltages you are needing, but there should be something similar that would. If the goal is purely a DIY quest, then I apologize for the suggestion.

Phil Pilgrim (PhiPi) · 2012-01-15 18:17

Another thing that concerns me about this proposal is the plan to parallel three MOSFETs per leg. Although MOSFETs are capable of being paralleled, load-sharing at these current levels will be extremely tricky to pull off. The reason is that the extremely low "on" resistance of each MOSFET has to be surpassed (by an order of magnitude) by the connections between the individual transistors, in order that the loads be shared equally. At 150A, this will not be accomplished by soldering the MOSFETs to a PCB, but by attaching them to copper buss bars in a way that accomplishes a very low-resistance connection.

-Phil

Circuitsoft · 2012-01-15 19:58

The plan was to have the MOSFETs attached to large copper pours on the circuit board, and to see if I could get 2oz copper on the board. I am planning to do current monitoring as well, with 0.5 mOhm resistors on each transistor.

Phil Pilgrim (PhiPi) · 2012-01-15 20:07

Have you researched and analyzed the width/thickness/shape of the copper pours required to accommodate 150A? Or are you just hoping for the best? I haven't researched it, BTW, but my gut feeling is that you're being over-optimistic. I could be wrong.

-Phil

Circuitsoft · 2012-01-15 20:21

When it gets to that point, I'll be asking a coworker for input on the layout. I have another motor to test the design on, rated at 7.2v,28A. I'm hoping I can just swap out my current sense resistors and run the driver at a lower voltage.

On another topic of current flow, I've found inductors usable to 84 amps. Can I stack two on opposite sides of the board as long as I wire them in phase? If the bobbins stack in line, will I keep the inductor's rating and double the saturation current, or will the inductor value be halved?

jmg · 2012-01-16 12:32

Circuitsoft wrote: »

According to the SiR882DP datasheet, with a 4.5v gate voltage, it has a worst-case gate charge of 27nC. If I'm driving the bridges via differential DUTY mode on the counter, and I want to switch the MOSFET within 1/2 clock, then it seems I need 27nC*160Mhz=4.3Amps gate drive Per MOSFET. Am I going way overboard on this? Did I do some of my math wrong?

The math is correct, and gate drivers of multiple peak-amps are easily available.
I'd say your assumption of needing to drive this in ~6ns is a little over-cooked though.
Your operating PWM frequency will be much lower, and slower FET edges are often chosen for better EMC.
So long as tr ~ tf, your drive delays do not greatly affect duty cycle.

Another reality check, 150A is not easy to manage on PCBs. If you take a nominal 10 milliohm sheet/path resistance, you get 225 watts of loss.
That's one fried assembly

Even your 0.5mOhm sense resistor, is 11 watts.

kwinn · 2012-01-16 16:41

For 150A the suggested conductor size is 42.4 square mm. If this is your peak current and the continuous current is lower you may be ok with a smaller conductor area. The difficult part will be designing the circuit so that each transistor is carrying it's share of the load, and the circuit layout will be the critical part of that.

The 2KW RF generators of the ICP's I service use 2 sets of 4 fets to generate 1200 Watts required for the plasma, and the mismatch in load sharing is the greatest weakness of the design. As the components age the some transistors will carry less of the load and others more. Eventually this leads to a cascade failure.

Circuitsoft · 2012-01-16 19:05

jmg wrote: »

The math is correct, and gate drivers of multiple peak-amps are easily available.

I'm only finding 2A or so peak drive in this voltage range. LM5101A is an example.

jmg wrote: »

I'd say your assumption of needing to drive this in ~6ns is a little over-cooked though.
Your operating PWM frequency will be much lower, and slower FET edges are often chosen for better EMC.

I figured if I had the transition time within 1/2 of the step size in the PWM, then I wouldn't need to worry about how the driver will act near the ends of the PWM range.

jmg wrote: »

So long as tr ~ tf, your drive delays do not greatly affect duty cycle.

What are tr and tf?

jmg wrote: »

Another reality check, 150A is not easy to manage on PCBs. If you take a nominal 10 milliohm sheet/path resistance, you get 225 watts of loss.
That's one fried assembly

So I'll be trying to keep the resistance under .25milliohm.

jmg wrote: »

Even your 0.5mOhm sense resistor, is 11 watts.

The intent was one resistor per MOSFET, so 50A max per resistor. Since they're 2W resistors, and 150A is only a peak value, I think I'm okay.

kwinn wrote: »

For 150A the suggested conductor size is 42.4 square mm. If this is your peak current and the continuous current is lower you may be ok with a smaller conductor area.

150A is certainly a peak value. The motor, I think has ~10AWG or 8AWG wires going into it, so I don't plan on spending too much time at 150A.

kwinn wrote: »

The difficult part will be designing the circuit so that each transistor is carrying it's share of the load, and the circuit layout will be the critical part of that.

So, with 2oz copper, I need a trace 60cm (not 60mm) wide? I am sure that width is unnecessary since I've seen computer power supplies capable of 50A on the 12V rail, and they're less than 20cm wide in total, not to mention that I've seen PSUs good to 120A, so I know that can't be right. Finally, I've seen PCBs designed to take 10-20A and the traces aren't more than a few mm wide, and it didn't seem like even 2oz copper.

Should I build the output portion of this circuit out of buss bar? The MOSFETs are made to put heat into the PCB, not to an external heat sink, yet they're rated at 53W power dissipation at 70C.

kwin wrote: »

The 2KW RF generators of the ICP's I service use 2 sets of 4 fets to generate 1200 Watts required for the plasma, and the mismatch in load sharing is the greatest weakness of the design. As the components age the some transistors will carry less of the load and others more. Eventually this leads to a cascade failure.

My plan was to put one current-sense resistor on each transistor. It wouldn't be too hard to check the balance between the resistors in addition to adding them.

kwinn · 2012-01-17 21:43

@Circuitsoft, I'm not trying to discourage or dissuade you from building this, only to warn you of the problems I have encountered with equipment operating at these power levels. I am very interested and contemplating a project that may need something similar so I hope you come up with a good workable design.

150A is certainly a peak value. The motor, I think has ~10AWG or 8AWG wires going into it, so I don't plan on spending too much time at 150A.

The 42.4 sq mm recommendation is the wire size for conductors carrying 150 amps continuously so it would be overkill for a brief peak current over a short path. You do however need to size the traces for the continuous current the motor draws at full load.

Should I build the output portion of this circuit out of buss bar? The MOSFETs are made to put heat into the PCB, not to an external heat sink, yet they're rated at 53W power dissipation at 70C.

Components dissipating that kind of power have a tendency to turn PCB's into charcoal after a while so it would be a good idea to have something to dissipate the heat as rapidly as possible. True, the maximum rating is 53W at 70C, but the trick is keeping them at 70C when they are dissipating 53W. Mounting the output portion on a buss bar would certainly help with that. Fortunately if you can drive the gate hard enough to get the Rds on down to the .0073 ohms in the data sheet you will be under 20 watts dissipation at 50 amps current.

My plan was to put one current-sense resistor on each transistor.

That will help a little with the load balancing but the Rds on of the transistor is approximately 15 times greater than that. From the data sheet the typical/max Rds on at 20A is 0.0071/0.0087 ohms. If that is extrapolated to 2 parallel transistors with 0.0005 ohm sense resistors providing 100A current the currents through the individual transistors would be approximately 55A for transistor 1 and 45A for transistor 2. Not a big difference, but one that must be taken in to account when operating close to the maximum ratings.

High power MOSFET question

Comments