Why is it usually the case that the negative rail for ICs needs more decoupling capacitance (has worse PSRR) than the positive rail?

Question

That the premise in the question seems to hold can be seen from various sources, including:

• comparing the datasheets of various clones of LM317 and LM337 (too many to list, but generally the datasheets for the latter recommend more decoupling on input, about an order of magnitude more than for the former, e.g. TI's datasheet for LM317 recommends 0.1uF input/supply bypass, whereas the one for LM337 recommends 1uF for the same.)
• related to the above, the TI datasheet for uA78xx has a split rail power supply schematic where the decoupling for the positive regulator is less than that of the negative one. This is reproduced below.

• the Analog appnote MT-101 shows worse PSRR for the negative pin than the positive pin:

So the question is why is this asymmetry usually present.

Answer 1

This is true because the LM7815 is stable with any output capacitance- the capacitor is just there to reduce the output impedance at high frequencies. Vout comes from the emitter of the NPN pass transistor.

The LM7915, on the other hand, is made with a similar semiconductor process but has to produce a negative output voltage. Vout comes from the collector of the NPN pass transistor. It's not stable without a largeish capacitor on the output. With only 100nF on the negative regulator it will likely oscillate under some conditions, whereas the positive regulator will be fine.

LM78xx

LM79xx

---

As far as the AD8099 goes, it probably has to do with the (internal) compensation capacitor being connected to the negative supply. Op-amps do not have ground pins generally.

So, any change of the negative supply pin relative to 'ground' is coupled to the amplifier.

What appears to be a pattern is actually from two quite different root causes.

Answer 2

It occurs because semiconductor devices themselves are not perfectly symmetrical. Devices that rely on "holes" as their primary charge carriers (PNP BJTs and P-channel FETs) generally have slightly lower performance than the corresponding devices that use electrons. This manifests itself as slightly slower switching times and higher resistances. This can be offset somewhat by increasing physical dimensions in certain ways, but then this leads to higher parasitic capacitances.

In the case of the 3-terminal regulators, the simple-minded approach would be to simply "invert" the circuit of the positive design in order to create the negative design, reversing all of the voltage polarities and swapping NPN and PNP transistors throughout, including for the main pass transistor. However, this works so badly that a completely different circuit topology (using mostly NPN transistors) had to be developed instead, and its stability characteristics are also quite different.

For the opamps, you'd have to look at the internal schematic of the specific device in order to understand the details.