Bias tees and DC blocks: the two parts that quietly set your amplifier's bandwidth
Most RF chains are designed around the amplifier, the filter and the mixer. Then a bias tee gets picked off a shelf in five minutes, and the low end of the band stops behaving.
This is a short note on the two passive parts that get the least attention and cause a surprising share of the debugging.
What a bias tee actually does to your return loss
A bias tee injects DC onto a line that also carries RF. The RF port sees a capacitor to the DC source and an inductor to the supply. At the frequency where that inductor's reactance is no longer much larger than 50 ohms, it starts loading the line.
Rule of thumb: keep the inductor impedance above roughly 500 ohms at the lowest frequency of interest. That is a factor of ten above the system impedance, which keeps the shunt loading under about 1 percent — small enough that it does not show up as a VSWR step.
Concretely, for operation down to 10 MHz you need roughly 8 µH. For 1 MHz it is 80 µH, and at that point the part is physically large and its self-resonance is a real constraint. If a datasheet quotes "10 MHz to 6 GHz" for a bias tee, check whether the low end is the -3 dB point or the point where return loss is still acceptable. They are usually different numbers.
The DC block is a high-pass filter you did not design
A DC block is a series capacitor, so it is a first-order high-pass. The corner is set by the capacitor and the 50 ohm system impedance:
f_c = 1 / (2 * pi * R * C)
With R = 50 ohms:
| Capacitance | Corner frequency |
|---|---|
| 100 pF | ~32 MHz |
| 1 nF | ~3.2 MHz |
| 10 nF | ~320 kHz |
| 100 nF | ~32 kHz |
Two things follow. First, "DC to 18 GHz" on a DC block almost always means the corner sits somewhere low enough that the vendor rounds it away — read the actual capacitance. Second, a large capacitor gets you a low corner but pushes the series resonance down, so the high end suffers. A 100 nF part may be fine to 2 GHz and useless at 18 GHz.
Where the two interact badly
The failure mode I see most often is a chain that is fine on paper and 2 dB down at the bottom of the band in the lab:
- Amplifier with a bias tee on the output
- DC block before the instrument
- Long cable with its own low-frequency loss
Each of these is a high-pass with its own corner. Three corners within a decade of each other multiply into a noticeably steeper roll-off than any single datasheet suggests. If you are chasing a low-frequency droop, measure the three separately before replacing anything.
Checking it in five minutes
- Sweep S21 from well below your lowest band edge, not from the band edge itself. The interesting behaviour is usually below.
- Look at S11 at the same time. A bias tee whose inductor is too small shows up as a return loss bump, not as insertion loss.
- Warm the part. Ferrite-cored inductors and Class II ceramics both drift; if the corner moves with temperature, you have found your instability.
Components referenced above are the ones we build at DB Design — bias tees and DC blocks covering 10 MHz to 40 GHz, in coaxial and waveguide form. The numbers in this post come from our own characterisation work, not from a marketing sheet.
Top comments (0)