I have been struggling with design that is a modification of a National Semiconductor application (DS009062-12). This application uses a LM117HV regulator and multiple LM195 power transistors in parallel to create a high voltage high current adjustable regulator. I modified the design by replacing the multiple LM195's with a single MJ15001 power transistor and an LM317HVH. I don't require the protection that the LM195's provide and space is limited. I ran some Pspice simulations with this modification and all looked well so I had a prototype CCA built. For the prototype however I used an NTE60 (direct replacement for the MJ15001) since the Motorola parts were back ordered. The prototype seemed to work very well under both static and dynamic load conditons. However when I replaced the NTE device with the MJ15001, regulation became erratic when going from a no load to a loaded condition. If I reduce the value of the base emitter across the MJ15001 the circuit will begin to work properly until the power transistor begins to heat up at which point the regulation becomes erratic again. The circuit requires that a minimum load current of 30mA flows through the regulator. This provides sufficient current for the regulator to work properly and produces approximately a diode drop across the 2N2905. When the circuit goes out of regulation it appears that the regulator is being current starved by the pass transistor during the transistion from no load to load. I did some bench testing on the NTE60 and MJ15001 and found that the Motorola parts had much more gain than their spec. sheet defines while the NTE's had much less but were still within specification. Need help !