Recently a reader asked me about the shape of the curve of an op amp's PSRR (power-supply rejection ratio) versus frequency. He observed that any curve of CMRR (common-mode rejection ratio) versus frequency which is the same as the gain versus frequency curve is probably an error or a foolish piece of bad data taking.
How about PSRR curves? Is a PSRR versus frequency curve in error if it's the same curve as the gain curve? In general, the answer is no. If the curves appear the same, they probably really are the same. Here's why: Every op amp has one (or more) capacitors that roll off the amplifier's gain versus frequency. At one end, the main "Miller" capacitor's voltage moves as far as the output voltage. At the other end, the capacitor is referenced to one of the power supplies - sometimes the minus supply, sometimes the plus supply. In the old days of vacuum-tube amplifiers, the main roll-off capacitor was sometimes actually referred to ground (refer to the old Philbrick K2-W, etc.). So if there was some motion on the minus or plus 300-V supply, the output had no direct path to cause it to move.
But in the solid-state era, very few op amps are referenced to ground - the minus or the plus supply is the place where the compensation capacitor is referred to.
Please look into the 1976 article by James Solomon in the IEEE Journal of Solid-State Physics, VOL SC-9, No. 6 (also reprinted as Appendix 1 in the NSC Linear Applications Data Book, 1986-1990). Mr. Solomon confirms that most op amps will indeed have a PSRR that's similar to the Av versus frequency, at least for one of the supplies.
Are there any exceptions, any amplifiers for which the PSRR is better than the gain? If you take an LM301A and connect it with 30 pF from pin 1 to pin 8, and then add 30 pF from pin 5 to ground, that can help cancel out or neutralize the ac PSRR.
If you take an LM308, you might damp it with 30 pF from pin 1 to pin 8. OR, you might connect a 100-pf capacitor from pin 8 to ground. That would yield a much better PSRR versus frequency curve than the normal Miller-integrator scheme with capacitance from pin 8 to pin 1.
There are also some other amplifiers where the PSRR can be made very large at high frequencies. Noise gain damping can often be advantageous. So, in a case where you have problems, thinking is the order of the day, followed by measuring.
All for now./ Comments invited! RAP/Robert A. Pease/Engineer
Address: Mail Stop C2500A National Semiconductor, P.O. Box 58090, Santa Clara, CA 95052-8090