I'm considering 3-terminal linear regulators (LM317, LM350, TL783, etc.) for a power supply that regulates a 150V wobbly DC voltage down to 145V or 140V. My load is like 30-65mA. So, depending on parameters power dissipation might be 150mW or 650mW. I'd like the same circuit to handle either scenario.
A general phenomenon I've seen with the LM317: with a static load, the output voltage starts at 145.00V and quickly (~10s) drifts down to 144.93V, then over ~60s down to 144.91V, then over many more minutes drifts by some 10mV-order voltage.
I assume this is a negative temperature coefficient effect -- the regulator heating up. I can make Vout rise by absorbing some heat with my fingertip. I've been exclusively using DPAKs without a heatsink. Previously I've seen the opposite phenomenon: a slow rising voltage -- this was caused by high tempco carbon-composition resistors.
I've observed this effect under various loads, output voltages, and circuit layouts. The ultimate goal is to have a power supply which becomes quickly (seconds) stable to within 1mV. I don't want to come back an hour later and find it has dropped by 10mV.
A. How might one build the circuit to negate voltage drift due to temperature?
B. What is the difference between these? Some datasheets list many of them.
- Thermal Regulation (%/W). Mentioned in LM350
- Output voltage change with temperature. Mentioned in TL783.