L
Admin
Capacitor class is not a BOM footnote in RF. The same 10 pF in C0G versus X7R can mean a stable match or a temperature-and-bias walking filter.
1. Dielectric behavior that matters
- C0G/NP0 (Class I): tiny voltage and temperature coefficients, low loss, predictable. First choice for resonators, matching, timing, and tight filters.
- X7R/X5R (Class II): high volumetric efficiency, large capacitance change with DC bias and temperature, higher loss. Fine for bulk bypass far from the RF node; risky in the RF voltage divider.
- Film / mica / porcelain RF caps: used where Q and stability beat size.
2. DC bias derating (Class II)
A 1 uF X7R at 16 V rated may deliver a fraction of that at 12 V DC bias. In a bias tee or supply rail next to an RFIC, that can still be OK for bypass. In a DC-blocked matching stub that assumed 10 pF, an X7R 10 pF part under RF voltage swing is the wrong tool.
3. ESR, ESL and SRF
Above self-resonance the part looks inductive. Datasheet impedance curves beat the printed capacitance. Parallel two values (e.g. 100 pF + 10 nF) only after you check mutual inductance and land pattern; sometimes one correct C0G wins.
4. Placement
- Matching / coupling in the RF line: C0G (or RF porcelain), short loops, tight GND vias.
- Local RFIC bypass: small C0G near pins plus bulk Class II further away is a common pattern.
- Do not share a lossy Class II part between an RF match and a digital rail "to save a line item".
5. Quick selection flow
1) Is the node RF-sensitive (match, filter, VCO loop, PA envelope)? → Class I.
2) Need bulk energy storage on a supply? → Class II OK with bias derating checked.
3) Need precision timing / TCXO load? → C0G and verify board stray C.
6. Verification
- Sweep S11 of the match over temperature if the product sees outdoors
- Re-measure after soldering; hand rework flux and tombstoning change small pF networks
- Compare two vendors' impedance curves at your band, not only capacitance tolerance
If the network is part of the resonance, pay for C0G. X7R belongs on the supply, not in the tank.
1. Dielectric behavior that matters
- C0G/NP0 (Class I): tiny voltage and temperature coefficients, low loss, predictable. First choice for resonators, matching, timing, and tight filters.
- X7R/X5R (Class II): high volumetric efficiency, large capacitance change with DC bias and temperature, higher loss. Fine for bulk bypass far from the RF node; risky in the RF voltage divider.
- Film / mica / porcelain RF caps: used where Q and stability beat size.
2. DC bias derating (Class II)
A 1 uF X7R at 16 V rated may deliver a fraction of that at 12 V DC bias. In a bias tee or supply rail next to an RFIC, that can still be OK for bypass. In a DC-blocked matching stub that assumed 10 pF, an X7R 10 pF part under RF voltage swing is the wrong tool.
3. ESR, ESL and SRF
Above self-resonance the part looks inductive. Datasheet impedance curves beat the printed capacitance. Parallel two values (e.g. 100 pF + 10 nF) only after you check mutual inductance and land pattern; sometimes one correct C0G wins.
4. Placement
- Matching / coupling in the RF line: C0G (or RF porcelain), short loops, tight GND vias.
- Local RFIC bypass: small C0G near pins plus bulk Class II further away is a common pattern.
- Do not share a lossy Class II part between an RF match and a digital rail "to save a line item".
5. Quick selection flow
1) Is the node RF-sensitive (match, filter, VCO loop, PA envelope)? → Class I.
2) Need bulk energy storage on a supply? → Class II OK with bias derating checked.
3) Need precision timing / TCXO load? → C0G and verify board stray C.
6. Verification
- Sweep S11 of the match over temperature if the product sees outdoors
- Re-measure after soldering; hand rework flux and tombstoning change small pF networks
- Compare two vendors' impedance curves at your band, not only capacitance tolerance
If the network is part of the resonance, pay for C0G. X7R belongs on the supply, not in the tank.