RobG
|
Just to expand a bit. The bandgap output voltage is (ideally) VOUT = UT*ln(IBE/IS) + K*UT where UT is our friend kT/q and K is a gain ~8
K is determined by ratios so it is process independent. UT is determined by your favorite God, and he/she doesn't appear to have plans for changing it, so count that as process independent.
IBE is the current in the relevant PN junction -- it is process dependent. Generally it is limited by the sheet resistance of your process: +/-30% or so, plus the temp co of the resistors. The temp co can increase or decrease the curvature of the temp response.
IS is the saturation current (reverse bias leakage) and is +/- 30% or so over process if you have a lazy fab. (IMO it is very well controlled, much better than +/-30%, like the Nwell resistor).
So your process dependent part of the equation is UT*ln(IBE/IS) and the variables are tempered by the ln function.
There are other non-ideal effects, like base current drop across the base resistance. These can be strongly process dependent and need to be designed around.
Power supply rejection is just a matter of shielding the pn junction and its current from power supply variations.
|