Driving multiple MOSFET

[joesaliba]

Hi,

As per the title, I need to drive 7 N-channel MOSFET's in parallel with a PWM of around 1KHz.

I was intending to use a non-inverting optocoupler. Apart from doing all gate together, should I use a resistor to the gate, or direct from the optocoupler is enough?

And to turn OFF the MOSFET's, a single resistor to ground is enough?

I am switching pure resistive load.

Thank you

Joe

[ricardourio]

Hello Joe,

     It depends which opto you will use. You can use same opto for inverting or non-inverting, only need to choose between ground or positive gate drive. Isn't mandatory to use a gate resistor, but I recommend to protect opto, as the gate current can reach more than 01 amp peak, and you'll use a low frequency. Remember, enough normally isn't the better. Using opto to drive positive, a resistor is only what you need at 1 Khz.

Regards

Ricardo Urio

[midali]

When connecting in parallel, the capacitance of the FET Gate will increase, which will be 7x. Even if it is a decent frequency of 1 khz, you will have to use the fet-driver to control them. Why do you want to use 7 in parallel?

[trastikata]

If the intention is current sharing, you should use FET drivers, otherwise because of the increased gate capacitance and other parasitic effects, you will face slow rise times which would create on its turn power dissipation problems.

[ricardourio]

When connecting in parallel, the capacitance of the FET Gate will increase, which will be 7x. Even if it is a decent frequency of 1 khz, you will have to use the fet-driver to control them. Why do you want to use 7 in parallel?

Good point, I assumed that would be 1 opto to each mosfet.

Ricardo Urio

[GDeSantis]

It's not unusual to drive multiple MOSFETs from a common source; but precautions must be taken.

For example, the attached file shows four paralleled MOSFETS driven by a hex inverter with Schmitt trigger inputs.  The circuit can operate a 1KHz; but three hex inverters must be paralleled to provide sufficient gate drive to handle the ~10 nF of capacitance associated with the MOSFET circuitry. 

If the MOSFETs drains are not connected, then each should include a separate diode (RURP8100) to suppress reverse energy which could potentially be generated by an external inductive load.  Also, an optocoupler is not shown in the circuit but could very well be required in your application.

Although the circuit could be expanded with modifications, THIS IS ONLY AN EXAMPLE with no guarantee of applicability for your application.

[rick.curl]

Hi Joe-,There are some opto-isolated IKGBT/MOSFET drivers that incorporate the opto and the driver in one chip.  As an example. take a look at ONsemi's FOD3184. It will do 2.5 amps peak, which sounds like a  lot, but the high current is necessary for a few microseconds to get the MOSFETS turned on or off as quickly as possible.

-Rick

[joesaliba]

Thanks all for the replies.

The output of the MOSFET's will not be in parallel. Each MOSFET will have a resistive load of about 4A, practically a small heating element. I require these to be separated so I can monitor if any of these heaters went open.

The intention to use PWM will be to chop down the 8.4v supply voltage to about 2v for the heaters.

I just do not know if lowering the 1KHz will solve some problems.

@rick.curl Nice little device, but voltage must be between 15v to 30v, and I am using 8.4v

Regards

Joe

[John Lawton]

Why would you need 1kHz PWM to control resistive heating loads? Use a fraction of that frequency or indeed don't use PWM at all, some form of pulse density modulation should be fine.

[joesaliba]

Why would you need 1kHz PWM to control resistive heating loads? Use a fraction of that frequency or indeed don't use PWM at all, some form of pulse density modulation should be fine.

In fact I am going the lowest possible frequency PWM would handle with 4MHz crystal, i.e. around 145Hz.

If I manage to use that very low frequency, than it would be ok to just parallel the gates?

Regards

Joe

[midali]

Even if you do not tie the MOSFETs in parallel, the gate's driver is the problem due to gate capacitance. It is a bit more complicated to calculate after the datasheet, but from experience I tell you that at a load of 4A and 7 gates in parallel, drived from PIC, you will have big problems with heat dissipation. I would recommend you to analyze half bridge IFX007 from Infineon, where you can put in parallel as many pieces as you want , but in your case 2 pieces are enough, each can drive 3 or 4 loads .

https://www.infineon.com/cms/en/product/power/motor-control-ics/brushed-dc-motor-driver-ics/single-half-bridge-ics/ifx007t/

[david]

Hi Joe,
8.4V sounds like a 2S Lipo source.   Some time back I did a brushed motor controller that averaged about 15A running current and 60A starting.  It used a single high current FET with no heatsink and was driven at either 4kHz or 8kHz. It was driven directly off the little 8 pin micro.
I appreciate there can be significant capacitance at the gate but at these low frequencies the source impedance of the micro will drive a large amount of capacitance. 
I would choose an appropriate FET in a small package (low voltage, low RdsOn) and give it a try.  If it was 60-100kHz with high swings at the drain it would be a totally different story.
A perfect 125Hz PWM source via 300R (to represent a PIC pin) and driving 140nF won't have a lot of rounding and I doubt even 7 small FETs in parallel is going to present this much C.  Also, switching 0A-4A will happen over a fairly small change of gate input voltage on a logic level FET.
I'd attach a simulated plot of the waveform but can't seem to add pictures unless from a URL.....   Hmmmmm.


Cheers,
David

[david]

Finally.....
Simulated drive.

Cheers,
David

[joesaliba]

Correct David,

I am using an 2s LiPo. I ordered some MOSFET's and will try them running at 150Hz.

Regards

Joe

[david]

Hello Joe,
For a 4MHz clock I think about 244Hz may be the lowest PWM you can select but it's definitely worth keeping it as low as you can.
If you approximate the FET turn on/turn off as a linear ramp up/down then maximum device dissipation occurs at half volts, half current.  So when transitioning from 0V, 4A to 8.4V, 0A the power is maximum midway at 4.2V, 2A and is near zero at either end.  The power loss follows an inverted parabolic function with a maximum of 8.4W and an average of about 5W for the turn on/turn off period.
If you use 244Hz (preferably lower) you have a period of about 4000uS and if for example your turn on time was 25uS and the turn off time 25uS then your switching losses would amount to ~50/4000 x 5W or about 62mW average.  This does not include your losses due to the RdsOn of the FETs selected.  I think your switching times will be faster than 25uS.
I'd be tempted to drive each FET via a 100R resistor just for some isolation between devices and I would include a common 10k resistor from the PWM pin to ground to hold things low during any tristate condition at start up.
Good luck with your project and take care with the Lipo wiring.

Cheers,
David