High Frequency Gate Driver

[Meri96]

Hello everyone, I am researching the gate driver IC for the resonant SEPIC LED driver circuit I designed. The switching frequency is 10 MHz. Input voltage 3.7V and output voltage 3V. Power is 3W. It can be with or without isolation, it doesn't matter. I have single MOSFET. I chose SiSS26LDN as MOSFET, considering Rds and dimensions. As a result of my research, I calculated the Ig value as 7.83A according to the mosfete. While selecting the gate driver, I found a few ICs that the typical output current value bigger than Ig.

These are the gate drivers I found for 10MHz: DEIC420, IXDD415SI, IXRFD615, 1EDN7550, 1EDBx275F

Would it be ok for me to use any of these? or do you have any other suggestions? Is there anything I should pay attention to when choosing? This is my first project i need your ideas, thank you in advance.

 

[dick_freebird]

I'd suggest you step away from MOSFETs and set up for GaN
devices instead. 10MHz and nF of Cgg are not going to be
pretty, efficiency-wise.

There are a few, really sporty GaN drivers out there. The
control loop and resonant tuning, that's a different
dimension of the problem.

 

[mtwieg]

Even with a low switching frequency, that would be a poor choice of a MOSFET. Your gate driver alone would be dissipating over a watt.

 

[Meri96]

I researched it again and I think I can use EPC2045 GaN FET to increase efficiency in line with your suggestions.

--- Updated Apr 22, 2021 ---

So what should I pay attention to when choosing a gate driver for this GaN FET?

 

[mtwieg]

Closer, but still way off.

If Vin=3.7V and Vout=3.0V, you'd be fine with a FET rated for 20V. And if your max Pout is 3W, you don't need an Rdson <10mohm. Have you actually done any calculations to estimate what your efficiency will be? Have you considered that the freewheeling diode alone will probably dissipate ~20% of your input power? What's your motivation for such a high operating frequency?

As for GaN drivers, TI produces some gate drivers dedicated to GaN. For highest speed, look at the LMG1020. They also make half bridge drivers like the LMG1210 and LMG1205.

 

[Meri96]

It is important to me that the circuit sizes are as small as possible, so I chose high frequency. I thought the smaller I picked the Rdson, the less the loss would be.

MOSFET loss calculate for EPC2045 and LMG1020:
Pon = (Irms) ^ 2 * Rdson = 1.77 ^ 2 * 5,6 m = 0,017 W
Pgate = Qg * Vgs * fsw = 6n * 5 * 10M = 0,3 W

I wonder am I thinking wrong?

Have you considered that the freewheeling diode alone will probably dissipate ~20% of your input power?

this is not fully understood.

 

[mtwieg]

It is important to me that the circuit sizes are as small as possible, so I chose high frequency.

Sure, the power circuitry for this thing could probably made to fit into a 1cm square footprint. Though I'm guessing that the controller will be at least that large. Diminishing returns for such a low power level.

I thought the smaller I picked the Rdson, the less the loss would be.

MOSFET loss calculated for EPC2045 and LMG1020:
Pon=(Irms)^2 * Rdson = 0.9^2 * 5.6m = 0.00453W
Pgate= Qg * Vgs * fsw = 6n * 5 * 10M = 0.3W

I wonder am I thinking wrong?

Ok, so if you picked a smaller geometry FET like the EPC2035 with Rdson=0.045 and Qg=0.88nC, you would get:
Pon=(Irms)^2 * Rdson = 0.9^2 * 45m = 0.0365W
Pgate= Qg * Vgs * fsw = 0.88n * 5 * 10M = 0.044W
So total losses are much lower.

I couldn't fully understand this.

I'm assuming your SEPIC will use a rectifier/freewheeling diode. It's in series with the output, so its dissipation will be roughly:
Pdiode=Iout*Vf
or another way of stating it,
Pdiode=Pout*Vf/Vout
So if Vout=3.0V and Vf=0.5V, then the diode alone will dissipate 16.7% of the output power. Getting very high efficiency at low output voltages requires synchronous rectification, but that's difficult to implement in a SEPIC converter.

 

[Meri96]

Ok, so if you picked a smaller geometry FET like the EPC2035 with Rdson=0.045 and Qg=0.88nC, you would get:
Pon=(Irms)^2 * Rdson = 0.9^2 * 45m = 0.0365W
Pgate= Qg * Vgs * fsw = 0.88n * 5 * 10M = 0.044W
So total losses are much lower.

I now understand the part that I need to pay attention to for the MOSFET selection, thank you very much.

I'm assuming your SEPIC will use a rectifier/freewheeling diode. It's in series with the output..

You are right about the diode loss, I chose the DFLS230L diode and it becomes PD = If * Vf = 1.12 * 0.31 = 0.347W. This means 11.56% of the output power is lost.