Power Supply Rejection Ratio

Power supply rejection ratio measures how well a circuit keeps noise and ripple on its own supply rail from leaking through to its output — a specification every analog block needs precisely because no real power rail is the clean, noise-free voltage source a textbook schematic implicitly assumes it to be. Expressed in decibels, PSRR is the ratio between a disturbance injected on the supply and however much of that disturbance actually shows up at the output, and like most rejection specifications, the number that matters is rarely a single value — it depends heavily on the frequency of the disturbance being rejected.

PSRR is typically excellent at DC and low frequency, then degrades steadily as the disturbance frequency rises, which is exactly the opposite of where a designer would prefer the weakness to sit. A well-designed amplifier or regulator can reject supply noise by sixty, eighty, even a hundred decibels at DC, but that rejection falls off with frequency as the circuit's own internal loop gain and bypass capacitors lose effectiveness — the same gain-bandwidth tradeoff this series has already covered for amplifier design generally applies here too, since whatever feedback mechanism provides supply rejection at DC is itself bandwidth-limited. This degradation lands precisely in the frequency range a switching regulator's own ripple occupies, which is why PSRR at the actual switching frequency, not just the DC specification, is what determines whether a downstream circuit sees real supply noise coupling through or not.

This is a direct, practical consequence of exactly the ripple voltage this series has already covered as an unavoidable byproduct of switched-mode power conversion. A buck converter's output ripple sits at its switching frequency and harmonics, typically in the hundreds-of-kilohertz to low-megahertz range — precisely the band where most analog circuits' own PSRR has already begun rolling off from its DC peak. A precision amplifier powered from a switching regulator can therefore see meaningfully more supply-coupled noise at the ripple frequency than its impressive DC PSRR specification alone would suggest, which is exactly why real systems pair switching regulators with a downstream linear regulator or dedicated filtering specifically to clean up ripple before it reaches a PSRR-limited precision stage, rather than trusting PSRR alone to reject it.

PSRR is frequently confused with CMRR because both are expressed the same way, as a decibel rejection ratio, but the disturbance each one rejects enters the circuit through a completely different path. Common-mode rejection ratio describes how well a differential circuit rejects a disturbance appearing identically on both of its signal input terminals, a rejection mechanism that depends entirely on how well-matched the differential pair itself is — the same device-mismatch concerns common-centroid layout addresses elsewhere in this series. Power supply rejection ratio instead describes rejection of a disturbance on the supply rail, a path that never touches the signal inputs at all and couples into the circuit instead through internal device parasitics, substrate coupling, and whatever finite power-supply-independence the biasing circuitry actually achieves.

Because these are genuinely independent coupling paths, a circuit's CMRR and PSRR numbers can diverge significantly, and neither one predicts the other. A differential amplifier with an exceptionally well-matched input pair can deliver outstanding CMRR while still coupling supply noise through its tail current source or output stage biasing, producing comparatively weak PSRR despite excellent common-mode behavior. The reverse combination is equally possible — a circuit engineered with careful supply-rejection techniques in its biasing and output stages can achieve strong PSRR even with input-pair matching that leaves CMRR only adequate. This is exactly why both specifications are published separately on any serious analog datasheet rather than one being inferred from the other.

Charge pumps deserve specific mention here, since their own output is itself a significant source of the high-frequency noise downstream circuits need good PSRR to survive. A charge pump's switched-capacitor operation produces its own characteristic ripple at the pumping clock frequency, riding on top of whatever boosted DC level it delivers — any circuit powered directly from a charge pump output, rather than from a filtered or regulated derivative of it, is exposed to exactly the kind of high-frequency supply disturbance this entry has already identified as the hardest band for PSRR to reject. This is part of why charge-pump outputs intended for sensitive analog circuits are typically filtered or post-regulated rather than used directly, even in applications where the charge pump's raw current capability would otherwise be sufficient.

Power supply rejection ratio is the specification that determines whether a circuit's supply rail is actually clean enough once it reaches the circuit, regardless of how clean the regulator generating that rail claims to be on its own datasheet. It completes a triangle this series has now covered from every angle — ripple voltage describes how dirty a switching supply really is, charge pumps and power sequencing describe how that supply gets generated and brought online, and PSRR describes how much of whatever noise survives actually reaches a circuit that was never designed to tolerate a perfectly clean input in the first place.