How much leakage current is tolerlable in a varactor stack?

I'm trying to design a varactor stack like figure 2 (d) here:

http://ectm.dimes.tudelft.nl/publica...ocument985.pdf

and I'm getting confused by a varactor's typical reverse leakage current rating.

I've found varactors with only 20nA leakage but it seems like even that would forward bias one of the diodes.

It seems like, no matter how small the leakage current is, it would still be enough to forward bias one of the anti parallel diodes, which would defeat the purpose of the diodes to provide a high impedance bias connection. Or am I wrong about that?

How do people make this work?

George

Is the purpose of the anti-parallel diodes, not to block small signal RF but rather to provide compensation for the non-linearities that occur from one or other of the varactors becoming forward biased by larger RF signals?

If yes, then maybe the leakage doesn't matter that much?

If yes then is it important for the anti parallel diodes to have a turn on voltage that matches that of the varactors?

There is a relatively stiff voltage source at the Vbi. So any leakage thru one diode will go out that voltage source instead of forward biasing the other diodes(s)