Recorder Drain
conclusion from the future: voltage buffering OpAmp works flawlessly!
Problem
harvest source is weak, so it should be connected with low leakage
current drain is a Mosfet, that can pull the source to the desired voltage
mosfet is controlled by an inverted opAmp-voltage-buffer, resulting in a PWM-Controller that is smoothed out by a slow RC-Filter
RC-Lowpass is delaying (big) voltage-changes by 1 to 1.7 ms
second problem: opAmp is oscillating, there must be a better way
Ideas / Approaches
JFET
pro: larger resistive range than mosfet
con: negative voltage required, very little variety
PID controlled Mosfet
controller can hopefully be reduced to just one OpAmp
con: very complex, tuning required
TODO: maybe there is just some integrator and a small proportional part needed
Bipolar Transistor
TODO: can a BJT go into very high impedance? Seems not (small sample check)
drain directly into output of OpAmp (voltage buffer)
pro: fast, clean, simple
con: no high impedance, could feed into a harvest source
Switched OpAmp-Sink
Mosfet above OpAmp ⇾ would need voltage offset to control mosfet
analog switch would be a nicer solution than mosfet
con: could still feed into harvest source
oneway OpAmp-Sink
Diode above OpAmp to stop reverse flow
could be included in opAmp-return-path (compensate diode offset)
high impedance (over opAmp voltage range) with a low leakage diode
con: still no (real) high impedance, needs dual supplied opAmp to reach 0 V
Conclusion
combined switched oneway OpAmp-Sink is a proper solution
diode-solution should be fine
WARNING for diode solution
watch out with power-levels mixed with enabled-states
if OpAmps VOut settles at V-, the diode will drain harvester below GND
V- could be higher than V+ in disabled state (LDO for Pos, BoostInverter for Neg)