High SWR protection method

Most high SWR protection circuits I saw are using a SWR bridge to sample the reflected power. But what will be the disadvantage to simply measure the RF voltage on the transistor drain or collector? Because in the end this is what kill the device in case of antenna mismatch. And is more precise to fix a threshold in volts rather than in SWR ratio.

I have done exactly that before and it works very well, but it does have drawbacks. One drawback that comes to mind is if the load (antenna) has a very low Zin. This will cause the output transistor to conduct much higher current than with a matched load and could cause failure of the PA transistor, without the protection ever kicking in.
In my application I was using the sampled RF output for ALC and the SWR protection came as a bonus. However, I suspect a directional coupler will provide better control and they're not difficult to make.
Dick

But what is the relation between the directional coupler output and the drain voltage?

If the load impedance is too high then the reflected wave arrives back at the PA and can (depending on the phase relationship) present a much higher voltage across the Drain-Source. From The PA's perspective the load impedance is higher than it should be. If the reflected wave arrives back at the PA in anti-phase then the load impedance as seen by the PA is lower than it should be.
The directional coupler just measures the magnitude of the reflected wave and when it exceeds a certain value it causes the PA to reduce its power.

Yes, I know that. But more specific, if I want that drain voltage stay bellow 50v where should I set the threshold of directional coupler (assuming worst case of reflected wave phase)?

I'm not sure.
I would connect a scope to the Drain of the PA transistor (Assuming it can cope with the additional capacitive load. This may not be advisable above 50Mhz or so) and adjust the mismatch until the maximum tolerable Vds voltage was reached and then calculate the reflection coefficient.