Bias JFET with a diode in the source leg

Why do you like this? A JFET has a high input Z for small signal excursions even at zero volts gate-source---it's not as if the gate-channel diode begins to conduct a lot for a small positive voltage. Reverse-biasing is only necessary for positive gate-source voltages of more than a few hundred millivolts, depending on temperature and drain-gate voltage. In fact there is an operating point, what Van der Ziel describes as floating gate operation, where the net gate current is zero, with the forward gate-channel diode current is equal to the drain-gate leakage current. The impedance is still quite high under those conditions, and an added benefit is that transconductance is even greater than that at Idss. And the author says nothing about the overall temperature coefficient of the operating point. For some FETs it may be made more stable with the diode, for others less so--it depends on the pinchoff voltage. For example the 2SK170/LSK170 has about a zero tempco at Vgs = 0 for roughly 12mA Idss bracket parts. A diode would just foul that up.

A look at the J309 data sheet shows the well known JFET design problem, the large spread in drain current between devices. The drain current at -0.7V bias can be anything between 10mA and 45mA depending on which part you happen to pick up. This is somewhat mitigated by the reduction in drain voltage brought about by the drain resistor but not a design for production quantities.