An RF inductor cannot be selected from its nominal inductance alone. The useful comparison is its impedance and loss across the circuit's operating band, measured under conditions that represent the intended assembly. Start with the function, then decide which Q, self-resonant frequency (SRF) and current data actually constrain it.
1. Define the circuit task before setting limits
An impedance-matching network or selective filter needs a predictable reactive element. A bias-feed choke needs sufficient impedance to isolate the RF signal from the supply; its useful band can include a resonance-related impedance peak. Those tasks do not justify the same SRF rule. Coilcraft explicitly distinguishes them in its RF selection note [R1].
For matching and higher-order filtering, that note gives an SRF at least ten times the operating frequency as a rule of thumb. Treat this as a starting point for investigation, not a performance guarantee or a requirement for every choke. Final limits should come from the circuit's allowable mismatch, insertion loss and bandwidth. A blanket “SRF above three times the operating frequency” does not establish those limits.
Record the operating band, target impedance or inductance, permitted variation, signal level, DC bias and thermal environment. Packaging and assembly then become part of the electrical evaluation rather than late mechanical checks.
2. Read Q at the frequency you will use
For an inductive impedance, Q is the ratio of its imaginary to real part [R1]. In a series small-signal representation, the same ratio relates series reactance to effective series resistance. The following is an assumed calculation example, not a part specification or measurement:
Q = Im(Z) / Re(Z) = X_s / R_s
f = 100 MHz; L_s = 10 nH; Q = 50
X_s = 2πfL_s = 6.283185… Ω
R_s = X_s / Q = 0.125664… Ω
The assumed Q of 50 gives an effective series resistance of about 0.126 Ω at 100 MHz. That resistance includes the loss represented by this frequency-specific model; it is not the winding's DCR. It also does not establish circuit efficiency. Another part's Q measured at a different frequency, bias or fixture cannot be ranked directly against it.
Set a Q requirement from the actual loss and bandwidth budget. There is no universal Q threshold that makes an inductor suitable for every RF or microwave circuit. Read the manufacturer's frequency curves or validated model alongside the single table value [R2].
3. Use SRF to examine the model's limits
A lossless inductor in parallel with a parasitic capacitance illustrates why apparent inductance changes before resonance [R3]. This is a separate hypothetical example from the preceding Q calculation:
L = 10 nH; Cp = 1 pF; ω = 2πf
f0 = 1 / (2π√(L Cp)) = 1.591549… GHz
Z_LC = jωL / (1 − ω²L Cp)
L_eff / L = 1 / (1 − (f/f0)²) (0 < f < f0)
Below f0, the model gives these reactances. Deviation is relative to the ideal inductor at the same frequency; values are rounded.
| Frequency | Ideal L reactance | Parallel-LC reactance | Deviation |
|---|---|---|---|
| 100 MHz | 6.283 Ω | 6.308 Ω | 0.396% |
| 1 GHz | 62.832 Ω | 103.817 Ω | 65.230% |
Writing the ratio explicitly also tests the proposed threefold SRF margin:
f = f0/3
L_eff / L = 1 / (1 − 1/9) = 1.125
(L_eff / L − 1) × 100% = 12.5%
Thus, even this simple model predicts a 12.5% increase in effective inductance at one third of resonance. This is not a universal tolerance prediction for real parts. At f0 the ideal model diverges; real loss and additional parasitics limit the impedance peak. The lossless model cannot supply a real component's Q or loss at resonance.
SRF is also affected by the fixture and mounting environment. Coilcraft's measurement note discusses substrate capacitance, fixture effects and the difficulty of comparing results from different setups [R3]. An isolated catalog number is not a complete mounted-circuit model.
4. Make the measurement conditions traceable
For each candidate, record the exact part number and document version, test frequency or sweep, excitation level, DC bias, temperature, instrument, fixture and calibration or de-embedding reference plane. Preserve the manufacturer's inductance tolerance and the conditions under which it applies [R2].
When using S-parameters, check the port impedance, fixture removal and model validity band before converting them to impedance. Include the intended pads, ground return and nearby conductors in the assembly evaluation. A discrepancy near resonance can reflect those boundaries rather than a change in the component itself [R3].
5. Compare candidates, then verify the assembly
- Screen nominal inductance, tolerance and package against the circuit requirements.
- Compare impedance and Q across the whole operating band using compatible measurement conditions. Separate guaranteed limits from typical curves.
- Check DC bias, signal current, DCR, temperature rise and the manufacturer's current-rating definition. Q or SRF alone cannot establish current capability.
- Simulate the actual topology with an appropriate manufacturer model and layout parasitics; examine mismatch, insertion loss or choke isolation as applicable.
- Measure the assembled circuit over the intended bias, signal and temperature range. Preserve both the setup and the acceptance criteria with the result.
The inductor parameter guide explains why current ratings need their test conditions. The magnetic core materials guide shows how to read loss and bias evidence without turning a material label into a device rating.
6. Sources and revision
- [R1] Coilcraft: Key Parameters for Selecting RF Inductors — Document 671-C, “Revised 01/07/21” as printed; p.1: circuit task; p.2: Q and conditional SRF guidance.
- [R2] Coilcraft: Testing Inductors at Application Frequencies — Doc119, sections “Inductor parameters,” “Application frequency testing” and “Specifying Inductor Tests.”
- [R3] Coilcraft: Measuring Self Resonant Frequency — Doc363, sections “Self Resonance of Inductors,” “Fixture Effects” and “Comparing SRF.”
Sources checked 2026-10-30. The two HTML notes did not expose an explicit revision in the retrieved text; the access date is not their publication date.
Revised 2026-10-30: replaced blanket Q and SRF thresholds and unsupported performance claims with source-based selection steps and reproducible hypothetical examples. No product tests or independent engineering approval are claimed; application-specific assembly validation remains necessary.
