The circuit is very simple. The depletion MOSFET pair M1 and M2 (LND150) forms an CCS.
This CCS provides a stable current to develop a reference voltage across R4 for the gyrator. This voltage will bias the cascoded pair J1-3 and M3 through R6. D1, D2 and D3 protect J1-3.
R5 is simply a test resistor to measure the anode current.
R7 is the mu resistor which is optimized for each stage.
C1 provides the bootstrapping needed for AC operation and achieve the low output impedance in the mu output.
The top depletion MOSFET (M3) does all the heavy lifting. For low currents (e.g. 1-10mA) it doesn’t need a heatsink, however when the gyrator is used in a driver or when currents are greater than 10mA you’d expect to put a small TO-220 clip-on heatsink or bolt it to chassis if needed.
The low frequency response is primarily driven by the RC pair R6 and C1. Typically, I’d use the following combinations: 4.7MΩ and 220nF or 10M and 100nF.
The high frequency response is driven by the parasitic capacitances of the lower FETs. Therefore, you want to use a low-noise jFET in J1-3 instead of another depletion FET (unless it has low reverse capacitances – see the BOM and appendix section for more details). The jFET on this position is operating in very unfavorable conditions (i.e. low drain-source voltage) so best use a jFET here for best results.
The circuit has minimum protection and if you short accidentally any output you will kill M3 and J1-3 for sure. The voltage reference is pretty resilient, though, however it can be damaged as well. For the lower FET protection, you will need to add three 15-18V Zener diodes (D1-D3). This is covered later in more detail in the build section.
The value of R4 is determined by the optimal CCS current needed by the LND150 devices for best temperature compensation. This is about 500uA, so the value of R4 is roughly the output voltage divided this reference current.