High frequency sinks to MOSFET's drain?

Question

I have the following setup on a custom PCB. The WT32-ETH01 module can be normally booted into download/flash mode by holding the IO0 to the GND during the start. The IO0 is also used as a clock between the ESP32 and the LAN8720A ethernet PHY chip on the same module.

I had the IO0 trace leading from the WT32-ETH01 across the entire board to the boot button. Everything worked except it made a huge antenna for the 50MHz frequency, which was a problem during EMC testing.

I hoped to overcome this issue by "isolating" the frequency physically near the WT32-ETH01 module with this simple circuit. However, the 50MHz signal leading from LAN8720A ended up partially sinking into the Q1 MOSFET even in its fully closed state. I suppose that as a result of this, the communication between ESP32 and LAN8720A is not working. Though I am not entirely sure as the signal should be ok for clock purposes even with the Q1 connected (see measurements).

I never dealt with such a high frequency, maybe someone can give me a lead here?

Below are also the results of measurements on IO0.

^{simulate this circuit – Schematic created using CircuitLab}

IO0 not connected to the isolation circuit:

IO0 connected to the isolation circuit:

IO0 connected to the isolation circuit and boot button pressed:

Answer 1

Consider the capacitance of AO3400A. Roughly speaking, C_rss forms a capacitive divider between drain and gate. At 50MHz, this is 64Ω, plenty to carry the signal through wires and radiate.

It also severely loads the signal even when off. It's not clear if the waveform is actually as slow as shown (is that a 100MHz scope?) but this massive transistor isn't doing any favors.

Solution? Use a modest size transistor.

You can't really buy signal MOSFETs anymore. RUM001L02T2CL is the last one I looked up with a fairly small junction; you will still need filtering to pass EMI (or better layout, or probably a wide range of other options).

Even better: use an analog switch. The logic input is well isolated from the switch path, and probably won't need filtering. Better still, an SPDT switch (AKA 2:1 mux/demux) can switch between input and GND so no signals are shorted out in the process.

Better still, just solve it with digital logic, it's a digital signal to begin with, right?

^{simulate this circuit – Schematic created using CircuitLab}

A 74LVC1G08 will do trivially.

Note: it's not clear to me if this signal is routed onboard a module. I can't find a "WT32-ETH01" schematic offhand. So I'm going to assume this is connected by the user and so can be wired as above. If not, an analog switch or MOSFET shorting to GND, with some RC filtering to the gate if MOS, is the next closest thing.