Lead current limit for different MOSFET packages?

I am trying to put together a power stage for the Lebowski controller that can output a short duration maximum of ~18 kW from a ~60 V nominal 16s high capacity battery pack. The resulting peak (as in occasional, short duration) current is going to be about 70% of the maximum battery discharge rate.

This is going to drive a Heinzman PMS120 motor rated at 7 kW continuous S1 duty and, according to the Heizmann datasheet, able to deliver 18 kW and 45 Nm of torque for a 'short duration'. The motor is an axial flux motor with center rotor and coils heat-sinking against the case halves. Weighing in at 27 lbs this motor certainly feels like it has sufficient thermal mass for short bursts of the maximum 18 kW power. It was originally used in the Brammo Enertia motorcyle.

There are plenty of MOSFETs advertising currents sufficient to drive this motor to its maximum output. However, most of the datasheets mention in the fine print that the package leads can sustain a significantly lower current than the silicon can handle. For the TO-247 case 75 A RMS continuous is mentioned and for the SOT-227 somewhere around 200 A.

Do I have to be concerned that the few seconds of maximum current the motor can handle may be too much for the leads with their very low thermal mass?

On paper, 8 FET per phase in the TO-247 case should carry more lead limited current than 2 FET per phase in the SOT-227 packaging. How does this coincide with your experience?

I remember 50A RMS was recommended for TO247. But it's not as bad as it sounds, because the phase current is divided between the high and low side FETs. With 1 high and 1 low side that's 100A RMS output, and that would be 141A phase current for the sine wave peak. As far as I know the sine wave peak is specified for the commercial controllers, but the manufacturers don't explain it.
But making a half bridge with 1+1 TO247 for 141A is already a big challenge. Clean transient waveforms, fast switching and low dissipation are the requirements, but I think they can't be all satisfied for 141A. 100A is more feasible, that is 35Arms/FET, but it also depends on the FET type. Paralleling many of them is another difficult story, the output current is somewhat less than n*100A. (n is the number of FET pairs)
I didn't use SOT-227 so I can't talk about them. Generally for hard switching applications (inductive load) I think the limiting factor is not the RMS current capability of the leads, but the issues I mentioned.
I also never understand what people and the manufacturers mean by "peak power" for "short duration" about controllers. By your explanation, you need a 18kW controller with a heatsink/cooling designed only for 7kW. This would reduce the weight and size, but does it make sense?

peters said:

I remember 50A RMS was recommended for TO247. But it's not as bad as it sounds, because the phase current is divided between the high and low side FETs. With 1 high and 1 low side that's 100A RMS output, and that would be 141A phase current for the sine wave peak. As far as I know the sine wave peak is specified for the commercial controllers, but the manufacturers don't explain it.
But making a half bridge with 1+1 TO247 for 141A is already a big challenge. Clean transient waveforms, fast switching and low dissipation are the requirements, but I think they can't be all satisfied for 141A. 100A is more feasible, that is 35Arms/FET, but it also depends on the FET type. Paralleling many of them is another difficult story, the output current is somewhat less than n*100A. (n is the number of FET pairs)
I didn't use SOT-227 so I can't talk about them. Generally for hard switching applications (inductive load) I think the limiting factor is not the RMS current capability of the leads, but the issues I mentioned.
I also never understand what people and the manufacturers mean by "peak power" for "short duration" about controllers. By your explanation, you need a 18kW controller with a heatsink/cooling designed only for 7kW. This would reduce the weight and size, but does it make sense?

Click to expand...

Thanks for reminding me that we are dealing with a sine wave crated by TWO switches. That makes the whole scenario more promising. Your explanations also make it more understandable why FET manufactures advertise silicon currents that you could not possibly get in and out of the packaging.

The motor mentioned above will be used in an ATV. The duty cycle there is "Full throttle", "1/3 throttle", "Full throttle" and so on. Breaking the wheels free with burst of power is more fun in the turns. That's where the unusual ratio of peak to rated power comes from. The heat generated during any 18kW burst still needs to be dissipated before the next burst so skimping on the cooling is not an option.

In fact, I am currently working on the design of an 8 FET array (per phase) that allows excellent cooling and has low parasitic inductance. I will start a separate thread about that soon.

Thanks again for your helpful comments.

peters said:

But it's not as bad as it sounds, because the phase current is divided between the high and low side FETs. With 1 high and 1 low side that's 100A RMS output,

Click to expand...

Since they're in series, how does that work? :?

Current is only distributed among parallel circuits.

Current is always the same in series connections.


Within the 3 phases you have the high side of one bridge sourcing current, out thru the windings, then back into the low side of one of the other two bridges sinking it to ground.

The FETs in parallel in the same bridge half do share current with each other...but they dont' share current with the others in the other bridge half of the same phase (or the one they are in series with thru the windings).


So if you have a six FET controller, only two FETs active at any one time, in series, a 50A limit per FET means a 50A limit per phase. Not 100A.


If it was a 12 FET controller...then four FETs would be active, two parallel two series, so *then* you'd get 100A, because of the parallel pairs. (well, not exactly because current flow is never evenly distributed, so something a bit less, but close enough for this discussion).

amberwolf said:

Since they're in series, how does that work? :?

Click to expand...

Hmm, I missed something, I remembered the RMS of the square wave wrongly...
My point was that the phase current is distributed between the high and low side FETs in 'time division', but it is false.
In fact it's not the current, but the power that is distributed between the high and the low sides in time, due to the PWM switching. The half of the power passes through the high side FET and the other half through the low side, if the voltages and currents are balanced.
When the PWM is high, then Irms_phase^2*R is dissipated on the high side FET lead and 0 on the low side, and vice versa. (R is the lead resistance, but it is true also for Rdson). But the average power for a pwm period is proportional to the duty cycle, and the average duty cycle is 50% for a sine wave. So the average power on one FET is Irms_phase^2*R*0.5. Then the RMS current of a single FET: Irms_fet=Irms_phase/sqrt(2) (= the continuous current producing the same power on the same resistance).

For sine wave this could be verified by simulation or numerically with a program. (measuring RMS current of one FET, the current and the pwm are modulated by sine waves. The phase between the current and the pwm (voltage) shouldn't make any difference.)

So if the allowed continuous current (RMS) on a FET lead is 50A, then the phase output of the half bridge can be 70.7Arms (=50*sqrt(2)), and that is 100Apeak for the sine wave (instead of 141Apeak). Some available controllers do more than that.

But if the motor stalls then you are fully right, and that is a problem. If it's blocked in a position where the current of a phase is close to the peak (100% or 0%), then the full phase current passes on either the low side or the high side FET. As per the example, it is 100A, that dissipates 4 times the power on the leads compared to the allowed 50A. Controllers could detect stalling and reduce the current after 0.5sec. My controller does this, but originally the purpose of this function was to protect the motor, not the controller.

Stick to 45A rms per device for to-247 and 35A to-220 devices for continuous operation and it "should" be ok based on my own testing. The leg temps tend to be about 10C hotter than when the case temp is 80C (which you don't want to exceed). Keep in mind that when I'm talking about continuous, i really mean continuous, like running for a year in varying environment conditions.

Ideally you would not let your continuous operation case temp exceed around 65C so you still have thermal head room to absorb high current transients. After 80C on the case, the device tends to go into thermal runaway with very little additional current.

This info is based off tests I've performed to find the real limits, not bs spec sheet info. Done correctly, your devices can handle short term high current transients, sometimes to 2x your continuous limits for a few seconds. Really depends on how well you handle the thermal issues.

My testing was done with a DC power supply to study the thermal limits of packages and was done with several devices from various manufactures.