vfone, here is layout of passive part (DR + FET transistor as a varactor):
I did some stub tuning, especially drain stub. Now it works much better, current from drain to ground is less than 0.01mA (10uA) now.
Maybe it is because DRAIN and SOURCE pins being not perfectly grounded at RF. It could be optimized for even lower current, but then oscillation stops. I think it is because "gate stub" line is coupled too tightly to DR.
Ideally FETs source and drain are grounded using vias to ground plane, and then negative bias is applied to gate. I am trying to avoid negative voltage circuity, so using quarterwave stub instead.
biff44, about using voltage source, maybe it will work, but there is another problem: when I decrease series resistor value on a biasing line (default is 1kOhm), then I get more and more huge spikes on biasing waveform. I think it is because waveform changes voltage rapidly, so current starts to flow somewhere between FET and bias voltage source.
My further idea is to reduce "varactor" line coupling to DR, and make better RF grounding on drain using radial stub.
Should I ground source using quarterwave stub also?