L
Admin
[Technical article / FAQ, rfforum Editorial Team]
Sprinkling 100 nF capacitors on every RFIC power pin is a layout habit that often creates a PDN resonance right in the band you care about. The part's ESL, the via, and the island inductance set a tank; a "bigger cap" does not automatically mean a quieter supply. Many PLL spurs and unexplained RF skirts begin as a bypass set that looked conservative on the schematic and inductive on the bench.
1. What actually resonates
- A 100 nF 0402 with roughly 0.5–1 nH of mounted ESL (pad + via + body) self-resonates in the low-hundreds of MHz. Above that frequency the capacitor is an inductor, not a bypass.
- Parallel combinations of the same value share one ESL-dominated impedance dip and one inductive rise. They do not "stack down" at RF the way they do at audio. Five identical 100 nF parts do not give you five times the RF bypass if they share a long island and a via farm.
- A planned 10 nF + 1 nF (or 100 pF) with short vias, placed at the pin or under the paddle, usually beats a field of 100 nF capacitors several millimetres away. Bulk 100 nF still has a job — a few millimetres away, for lower-frequency current — but it is not the RF pin cap.
2. Freeze list for an "RFIC supply is clean" claim
- Cap values, packages, and via / distance to the pin (not "we bypassed it")
- Measured or simulated PDN impedance at the pin, across the RF band and across the PLL or VCO offset range that matters
- Whether a spur or phase-noise skirt was correlated with a PDN peak (a small inject or load-step, or simply the impedance peak frequency versus the spur)
- Board stack and ground-via count next to the caps. A pretty 100 nF on a long island is not a PDN.
3. Habit
Never close an "add more 100 nF" action as the fix for an RF or PLL spur without looking at PDN impedance at the pin. Archive a one-line cap table — value, package, via, distance — next to the spur plot. If a spur sits near a PDN peak and moves when you change only the bypass set, it is a supply problem dressed as an RFIC problem.
4. Practical triage
- Prefer a decade-spread (for example 100 nF bulk a few millimetres away, plus 10 nF plus 1 nF or 100 pF at the pin) over a field of identical 100 nF parts
- Check the 100 nF self-resonant frequency on the same stackup; if your band is above it, that part is not helping at RF
- After an inductor or layout change on the supply island, re-check the spur and the PDN peak — do not reuse the old "supply is clean" sentence
Discussion: on your last RFIC spur that "needed more bypass," did the fix land in a smaller cap at the pin, a via change, or an actual regulator change — and do you require a PDN or cap table on every supply-spur traveler?
— rfforum 编辑部 / Editorial Team
Sprinkling 100 nF capacitors on every RFIC power pin is a layout habit that often creates a PDN resonance right in the band you care about. The part's ESL, the via, and the island inductance set a tank; a "bigger cap" does not automatically mean a quieter supply. Many PLL spurs and unexplained RF skirts begin as a bypass set that looked conservative on the schematic and inductive on the bench.
1. What actually resonates
- A 100 nF 0402 with roughly 0.5–1 nH of mounted ESL (pad + via + body) self-resonates in the low-hundreds of MHz. Above that frequency the capacitor is an inductor, not a bypass.
- Parallel combinations of the same value share one ESL-dominated impedance dip and one inductive rise. They do not "stack down" at RF the way they do at audio. Five identical 100 nF parts do not give you five times the RF bypass if they share a long island and a via farm.
- A planned 10 nF + 1 nF (or 100 pF) with short vias, placed at the pin or under the paddle, usually beats a field of 100 nF capacitors several millimetres away. Bulk 100 nF still has a job — a few millimetres away, for lower-frequency current — but it is not the RF pin cap.
2. Freeze list for an "RFIC supply is clean" claim
- Cap values, packages, and via / distance to the pin (not "we bypassed it")
- Measured or simulated PDN impedance at the pin, across the RF band and across the PLL or VCO offset range that matters
- Whether a spur or phase-noise skirt was correlated with a PDN peak (a small inject or load-step, or simply the impedance peak frequency versus the spur)
- Board stack and ground-via count next to the caps. A pretty 100 nF on a long island is not a PDN.
3. Habit
Never close an "add more 100 nF" action as the fix for an RF or PLL spur without looking at PDN impedance at the pin. Archive a one-line cap table — value, package, via, distance — next to the spur plot. If a spur sits near a PDN peak and moves when you change only the bypass set, it is a supply problem dressed as an RFIC problem.
4. Practical triage
- Prefer a decade-spread (for example 100 nF bulk a few millimetres away, plus 10 nF plus 1 nF or 100 pF at the pin) over a field of identical 100 nF parts
- Check the 100 nF self-resonant frequency on the same stackup; if your band is above it, that part is not helping at RF
- After an inductor or layout change on the supply island, re-check the spur and the PDN peak — do not reuse the old "supply is clean" sentence
Discussion: on your last RFIC spur that "needed more bypass," did the fix land in a smaller cap at the pin, a via change, or an actual regulator change — and do you require a PDN or cap table on every supply-spur traveler?
— rfforum 编辑部 / Editorial Team
