Hi,
The BF996 is unprotected.
Actually, the datasheet says it's got zener diodes internally. Curiously, they don't give an ESD rating, so it's not clear if they're good for anything at all...
But yeah, even if they gave a rating, it would probably be a paltry 2kV (HBM), if that. An ESD strike to the antenna will easily deliver that, and it will be at a lower impedance (because of the coax and filtering components), so, more destructive than an HBM condition.
It would need to be clamped to something like plus or minus 5V and just putting some zeners back to back, or just two diodes back to back (for a 0.5V clamp) is probably going to have way too much capacitance.
The reason for C3 - a trivial 3.3pF - is unclear, given the signal passes through say 10m of 50 ohm coax... The operating frequency is about 300MHz.
Well, coax is a transmission line, it's not a capacitance. If it's properly matched, then it's 50 ohms, period -- no capacitance at any frequency!
3.3pF at 300MHz is hardly trivial, it's 160 ohms.
The response of the input network is approximately:
As you can see, the output voltage (the voltage at the gate) peaks around, well, low UHF, let's say -- it has a bandpass characteristic. The above is plotted with a load capacitance representative of the gate capacitance plus "select" capacitor. It's not obvious if L5 is providing resistance or inductive reactance. Probably, it's a bit of both, which has a damping effect, and the gain and bandwidth at 300MHz will be more useful.
I have done this sort of thing before but for low leakage (precision ADC input amp) rather than low capacitance. That was done with low leakage reverse biased diodes.
The tricky thing here is not just the magnitude of capacitance: as you note, zeners (typically 100pF+) are right out, here. But if you used ~3pF diodes (say, replacing C3 with junction capacitance), you have a more subtle problem: the capacitance varies with voltage, so large signals will experience a different frequency response than small signals. It's a knock-on effect, so signals will get
mixed together under this condition, and now you have weird spurs all over: IMD. This is especially tricky in an RF front end, where you want maximal dynamic range and linearity.
The suggestions others have made are about as good as you can get. Haven't looked, but I'd expect the Mini-Circuits limiters have IMD data, and show good (linear) operation up to the limit. There are commercially available ESD diodes for radio and high speed digital applications, which have the downside that, although they have little capacitance (under 1pF, and the change in capacitance versus voltage is a fraction of that), they're simply small to begin with, and can't withstand too many ESD strikes before failing. (A replaceable inline module would indeed be handy.
)
Tim