Voltage at the Drain Terminal of a Switching MOSFET

Question

I am attempting to drive an STP55NF06 N-Channel MOSFET with a TC4427 gate driver, switching at ~100kHz. The desired outcome is a higher current PWM square wave at the MOSFET's drain (Circuit in Figure 1).

However, my actual implementation of the circuit doesn't behave as it does in simulations (where there is just a normal square wave at Vout). The Vout in the physical implementation is shown in Figure 2 as a pink trace, with the yellow being the driver's voltage level. The Vout has a difficult time reaching the rail's voltage. The problem seems to be the time the driver is taking to make the MOSFET conduct as a higher PWM duty-cycle or lower frequency switching speed always means a slightly better waveform. A smaller current-limiting resistor helps, but only the slightest bit, and 10 ohms is close to the lowest this driver can go at 12V.

My real confusion lies in the fact that switching the pull-up resistor R to the other side of the MOSFET and measuring the voltage between the (now pull-down) resistor and source terminal shows a nearly perfect square wave. The resemblance between that voltage waveform and the driver's voltage is nearly the same, making me think the problem is not the current delivered to the gate.

To give some context to this problem, I have been attempting to create a Class D amplifier, but I have run into a roadblock with the output stage. This question has abstracted out a bit, but the problem is essentially that I do not know why an n-channel MOSFET isn't working here. I anticipate that the answer is that I am misunderstanding the way that an n-channel MOSFET works, but I am not sure in what way.

Answer 1

Your results more or less match my simulation using a similar (to your actual) MOSFET (FDS5670) model.

The problem is that the rather large (real) 50A/60V MOSFET has a great deal of drain capacitance and the 10K resistor is far too high in value to pull up the drain quickly. It turns 'on' quite snappily, both in simulation and in reality. The default LTspice MOSFET, on the other hand, is very small, like in an IC and you are not putting any realistic capacitive load on it (such as a test probe would have).

With a datasheet drain capacitance of 300pF typical, the time constant with 10K is 3 microseconds and you have only 5 microseconds on-time. The MOSFET I simulated has closer to 700pF typical drain capacitance so it is even slower.

If you reduce the 10K resistor to a few ohms (and if it is non-inductive) you'll see a vast improvement. Of course the resistor will tend to get very hot, so make sure it is rated for the required dissipation.

When you switched it to a source follower, the gate driver was actually driving the load through the massive gate-drain capacitance and the MOSFET was not really doing much.