flicker (1/f) noise

Some noise is loudest at low frequencies and fades as the frequency climbs, a slow, wandering, drifting kind of hiss rather than a flat, steady one. This is flicker noise, also called 1/f noise because its power rises as you go down in frequency, roughly as 1 divided by frequency. Where thermal and shot noise are a flat background hum, flicker noise is the slow heave of the sea underneath them: the gentle drift you see in a meter reading that just will not settle.

Its power spectral density scales as 1/f, so there is more noise per octave the lower you look, and it grows without limit as the frequency approaches DC, which is exactly why very slow measurements drift. Every device has a corner frequency where its flicker noise has fallen to the level of its white noise. Below that corner flicker dominates; above it, thermal or shot noise dominates. That corner can be just a few hertz for a good bipolar transistor but hundreds of kilohertz for a MOSFET, because flicker noise comes largely from charge being trapped and released at imperfect surfaces and interfaces, and MOS transistors are surface devices.

Flicker noise is why DC and very-low-frequency precision is hard: offset and gain seem to wander over seconds and minutes. The classic cures move the signal away from DC where flicker lives, by chopping or auto-zeroing the amplifier, or simply by choosing bipolar parts over MOSFETs for slow precision work. Be honest about one difference: unlike thermal and shot noise, flicker noise has no clean universal formula. It depends on the specific device, its bias, and how it was fabricated, so you read it from the datasheet rather than calculate it from constants.