2n3904/2n3906 driving mosfet gate

I'm building a MOSFET driver like this. The signal source is 7V peak-to-peak. Vcc is 12V. The push-pull stage is 2n3904/2n3906.

I've tested the push-pull output before adding the transformer, and using a 50 ohm load, the output voltage is reduced by half, so I concluded that the output resistance of the push-pull stage is 50 ohm.

Then according to the MOSFET spec (IRF740), C_iss = 1400pF, at 13.5MHz I calculated the impedance to be 8 ohm. So I used a 2:1 transformer (therefore 4:1 impedance transformation) to match the load at the gate.

But now I get a flat signal at the Q3 gate, AC is almost non-existent. If I disconnect the MOSFET Q3, I get a good sine-wave. Can anybody help me find out what's the problem with this circuit?

C2 is a 104 capacitor which is practically a short at RF. The 2n3904 has a spec of peak 200mA current capacity, which I translate to be 2.5V for 1/4 period, which should get doubled by the transformer and is supposed to be enough to charge the gate to a observable voltage.

• For completeness you might want to mention what C1 is and the frequency of your sine signal. Dec 16, 2014 at 13:59
• drain and source connected to ground. Dec 16, 2014 at 13:59
• What happens to the drive signal at the bases when you connect the MOSFET gate? The current gain of those transistors won't be much at 13MHz, maybe 20, so you're loading the oscillator output with < 1K. Dec 16, 2014 at 14:11
• For digital (square wave drive), a gate driver chip such as Micrel MIC4420, for a video amplifier such as MAX9650. Both are cheap. An Apex power amplifier would be better again, but \$. Dec 16, 2014 at 14:37
• Start simple. Leave out T1 and Q3. Load the output at C2 with the equivalent RC of Q3 gate. That way you don't worry about T1 and Q3 now. Expect to need a class AB bias for Q1/Q2. With the bases tied together there is a dead spot in the emitter drive will be killer at 13 MHz. Don't do any of this on a protoboard. Physical layout of all the parts and routes will be very important ... keep all loop areas as small as possible. Dec 18, 2014 at 4:11