L
Admin
A bias tee looks trivial: an inductor to feed DC, a capacitor to block it. Over a narrow band it is. Over a decade or more, the bias tee often sets the low and high frequency limits of the whole amplifier.
1. The inductor has two jobs that fight
- Low end: the choke reactance must be large compared with the line impedance. A rough rule is X_L at least 5 to 10 times Z0, so for 50 ohm at 10 MHz you need roughly 4 to 8 uH.
- High end: any real inductor self-resonates. Above its SRF it becomes capacitive and can create a sharp notch or a ripple in S21. Big inductance means low SRF.
2. Ways around the conflict
- Stack two or three inductors in series: a small air-core or ceramic chip inductor (tens of nH) close to the line for the high band, then a large ferrite-core inductor further away for the low band. Put a damping resistor (tens to a few hundred ohm) across or in series with the junction to kill the resonance between them.
- Conical (tapered) inductors wound on a ferrite cone give a single part that behaves well from tens of MHz to tens of GHz. They are expensive and orientation-sensitive; mount them at the angle the datasheet shows.
- A ferrite bead in series with the DC feed adds loss rather than reactance and helps flatten residual resonances.
3. The blocking capacitor
- Low end: its reactance must be small compared with Z0 at fmin.
- High end: series resonance and the pad discontinuity matter. Broadband single-layer or 0201 multilayer capacitors behave best; larger MLCCs show parallel resonances in the GHz range.
- Watch the DC bias derating of high-K dielectrics; a 100 nF X7R at its rated voltage may be much less than 100 nF.
4. Layout
- Tap the choke onto the line with the smallest possible stub; a few millimetres of track is already an open stub at mmWave.
- Keep the choke body away from the ground plane edge and other inductors to avoid coupling.
- Decouple the DC side with a ladder of capacitors so the supply impedance stays low and the RF does not see a resonant supply.
5. Verify
Measure S21 and S11 with the bias tee alone, and again with DC current flowing: ferrite permeability drops with current, so a choke that is fine at 10 mA may lose much of its inductance at 500 mA.
How do you build bias tees above 20 GHz: conical inductors, printed quarter-wave lines with radial stubs, or both?
1. The inductor has two jobs that fight
- Low end: the choke reactance must be large compared with the line impedance. A rough rule is X_L at least 5 to 10 times Z0, so for 50 ohm at 10 MHz you need roughly 4 to 8 uH.
- High end: any real inductor self-resonates. Above its SRF it becomes capacitive and can create a sharp notch or a ripple in S21. Big inductance means low SRF.
2. Ways around the conflict
- Stack two or three inductors in series: a small air-core or ceramic chip inductor (tens of nH) close to the line for the high band, then a large ferrite-core inductor further away for the low band. Put a damping resistor (tens to a few hundred ohm) across or in series with the junction to kill the resonance between them.
- Conical (tapered) inductors wound on a ferrite cone give a single part that behaves well from tens of MHz to tens of GHz. They are expensive and orientation-sensitive; mount them at the angle the datasheet shows.
- A ferrite bead in series with the DC feed adds loss rather than reactance and helps flatten residual resonances.
3. The blocking capacitor
- Low end: its reactance must be small compared with Z0 at fmin.
- High end: series resonance and the pad discontinuity matter. Broadband single-layer or 0201 multilayer capacitors behave best; larger MLCCs show parallel resonances in the GHz range.
- Watch the DC bias derating of high-K dielectrics; a 100 nF X7R at its rated voltage may be much less than 100 nF.
4. Layout
- Tap the choke onto the line with the smallest possible stub; a few millimetres of track is already an open stub at mmWave.
- Keep the choke body away from the ground plane edge and other inductors to avoid coupling.
- Decouple the DC side with a ladder of capacitors so the supply impedance stays low and the RF does not see a resonant supply.
5. Verify
Measure S21 and S11 with the bias tee alone, and again with DC current flowing: ferrite permeability drops with current, so a choke that is fine at 10 mA may lose much of its inductance at 500 mA.
How do you build bias tees above 20 GHz: conical inductors, printed quarter-wave lines with radial stubs, or both?