I have the circuit shown below: MOSFET Circuit I want this circuit to switch between 2.5V and 3.6V at the drain of the PFET when I input a square wave into the gate of the PFET. I would expect that this circuit would do this - when I put 0V into the gate, the PFET should become a low-impedance resistor, causing the voltage to equal 5V - 2 0.7V diode drops = 3.6 V. When I put 5V into the gate, the PFET should act as an open circuit, so the drain voltage can be found using the voltage divider to be 2.5 V.

In reality, when I input a 10 kHz square wave into the gate of the PFET, I get the following voltage at the drain of the circuit (measured with an oscilloscope in a lab): Voltage plot at gate and drain of circuit

As you can see, there are massive voltage spikes that happen every falling edge. These voltage spikes take about 10 µs to dissipate.

I have been able to trace some information about these spikes. The voltage at the source node of the PFET also spikes, indicating that these spikes are not entirely due to my MOSFET selection: Voltage plot at gate and source of circuit

A plot of the source-drain voltage of the PFET and of the voltage of the diodes also reveals a few things. The source-drain voltage appears to have an incredibly long 10 µs rise time for some reason. In addition, the diodes appear to be having an incredibly long 10 µs reverse recovery period: Plot showing MOSFET Source-Drain voltage and Diode voltage

When I have tested the diodes individually, I found that their reverse recovery time was orders of magnitude shorter than what I have seen here (the time on the datasheet is 8 ns). In addition, I found that the rise and fall times of my PFET, the ZVP3306A, was significantly faster when I tested it alone with a simple MOSFET test setup. On the datasheet, it has a rise time of 7 ns. So, individually the diodes and MOSFET can rise and fall incredibly quickly; however, when I am putting them together, I am getting massive rise times and reverse recovery times.

I will note that I performed a simulation of this circuit using Cadence PSpice. In the simulation, the large voltage spikes disappeared; However there were still very large 10 µs fall times at the drain of the PFET at every falling edge.

Does anyone know what may be causing these large rise times/reverse recovery times to appear? Am I correct that these are even rise times and reverse recovery times? Finally, is there any way I can adjust this circuit to speed these times up or remove this massive falling time at the drain entirely? I am trying to make this circuit switch quite fast, so any tips would be appreciated.


1 Answer 1


You can expect Miller capacitance between gate and drain to react in both directions. Thus with a high impedance load, the gate voltage is being seen at the drain. Conversely, with a very low impedance load and higher impedance source, you can expect negative feedback to the gate.

Being aware of the FET model capacitance helps explain its behaviours. Capacitance tends to increase as RdsOn rating decreases such that for a given family the product is relatively constant.

enter image description here Ref

A similar package in a PNP BJT could be less than 3 pF vs above FET is 50 pF for Ciss and 8 pF for Crss. This is the 2N3906-G. it also has a lower Rce=Vce(sat)/Io ~ 4 Ohms vs 14 ohms for your FET. Similarly, if you examine off-state leakage for the BJT and feedback capacitance Cob, as well as On impedance this BJT has a lower Rce*Cob=T than the FETs such as yours. This gives a hint as to why FET input & BJT output devices called IGBT's are popular for high slew rate and kV voltage switching.

enter image description here

The 1N4148 has even lower capacitance and leakage resistance so it is expected to see more of the transient across the diodes.

But since you have no tolerance specs for feed-thru voltage and load impedance, I cannot comment on a better solution. What is the purpose of this experiment?

  • \$\begingroup\$ Thank you so much for responding. I will have to look more at the devices you mentioned, and may try using the 2N3906-G after further testing and simulation. This circuit was a subcircuit I found to be working incorrectly inside a larger differential signalling circuit. I will also have to look into resistor sizes - now that I look at it, if I decrease the resistor sizes, the rise times and recovery times will be faster. Due to some considerations in the larger circuit, I may not be able to decrease resistances too much though. Once again, thank you so much for the help. I really appreciate it \$\endgroup\$ Jan 27, 2022 at 21:36

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.