A resistive divider is six resistors in a box. A Wilkinson divider is two quarter-wave lines and one resistor. They cost about the same, they both split power, and picking the wrong one is a mistake you will not notice until you are chasing an error budget three months later.
Here is where the two actually diverge, and how to decide in about five minutes.
What each one is doing
Resistive (lossy) divider. Every port sees 50 ohms, power is split by resistor network, and the split is flat from DC to as high as the resistors and the package will behave. The penalty is that the resistor network absorbs most of the input power before it reaches the outputs.
Wilkinson (reactive) divider. Quarter-wave transmission lines perform the split, and a single resistor bridges the two output ports. In principle the network is lossless — all the input power appears at the outputs, minus conductor and dielectric loss.
The number that separates them is the ideal insertion loss:
| Divider | Ideal insertion loss (2-way) |
|---|---|
| Resistive | 6 dB |
| Wilkinson | 3 dB |
A 2-way resistive splitter throws away half the power in the resistor. In a receive path that costs you 3 dB of noise figure for nothing. In a level-setting application where you were going to pad 3 dB anyway, it costs you nothing at all.
The isolation resistor is the whole point
The single resistor across the outputs of a Wilkinson does something the resistive network cannot: it absorbs power that reflects back from one output port, so it never appears at the other output.
That is what "isolation" means physically. A 2-way Wilkinson gives 20 to 30 dB of isolation between outputs over an octave, and better over a narrower band. A resistive divider gives 6 dB, because the resistor network is symmetric and a signal entering output 1 has a clear resistive path to output 2.
Practical consequence: if the two output ports feed anything that can reflect, you need a Wilkinson. Two antennas on a roof, two amplifier inputs, two channels of a phased array — all of those present mismatches that a resistive divider will happily hand to the neighbouring port.
If the outputs go to two independent receivers that are always well matched, and the isolation requirement is modest, the resistive part is smaller, cheaper and works down to DC.
Bandwidth is the trade
A Wilkinson's quarter-wave sections are quarter-wave at one frequency. Away from that frequency the lines are electrically wrong, isolation collapses, and return loss degrades. A single-section Wilkinson typically holds its specifications over 10 to 20 percent bandwidth. Two sections extend it to an octave at the cost of size and loss.
A resistive divider has no such limit in theory — it is flat from DC until the package parasitics take over, which for a good broadband part can be 40 GHz.
So the question inverts:
- Broadband, and you can afford 3 dB → resistive.
- Narrow to moderate band, and you need the power and the isolation → Wilkinson.
A 2-way Wilkinson specified 2 to 18 GHz exists, but it is a multi-section design, it is physically large, and its loss at the band edges is not 3 dB.
Three things that get missed on the datasheet
1. Amplitude and phase balance, not just insertion loss. For a phased array or a balanced amplifier, the number that matters is how closely the two outputs track. Ask for amplitude balance in dB and phase balance in degrees, at the band edges, not at centre frequency. A part quoting 0.3 dB balance at mid-band may be 0.8 dB at 18 GHz.
2. Power handling with an unbalanced load. A Wilkinson rated 30 W assumes both outputs are terminated. If one output goes open, the full input power divides between the resistor and the remaining port, and the resistor sees roughly twice its balanced-mode dissipation. A resistive divider is more forgiving here because it is already dissipating.
3. Reverse power. In a Wilkinson, reverse power entering an output port is split between the isolated port (absorbed by the resistor) and the input. If the resistor is only rated for the forward split, hot reverse power will destroy it. This is a real failure mode in TDD systems that share an antenna.
Where the isolation resistor physically goes
Inside the package, the resistor sits between the two output arms, and its parasitic inductance sets the upper frequency limit. Above about 18 GHz, thin-film resistors on a ceramic substrate are used with tight geometry control, and above 30 GHz the mechanical tolerances of the housing start to dominate the electrical performance.
This is also why a Wilkinson that has been dropped often still measures fine at 1 GHz and fails at 12 GHz: the resistor has micro-cracked, and the crack is invisible until the parasitic path it creates resonates.
A decision rule
- Is the band more than one octave? Resistive, unless you can afford a multi-section Wilkinson and the size.
- Do the output ports see mismatches? Wilkinson.
- Is this a receive front end? Wilkinson — the 3 dB is noise figure.
- Am I just combining two sources at similar level into a load? Resistive works fine, and it is more robust.
- Do I need amplitude and phase balance under 0.2 dB / 2 degrees over temperature? Ask for measured data on the specific unit, not a typical curve.
Where I sit
I work on RF and microwave passive components at DB Design / MeiXun in Wuxi, so this is written from the manufacturing side. We build power dividers — resistive and Wilkinson, 2-way through 16-way — alongside directional couplers and coaxial loads and terminations for the ports that end up unused.
If a divider in your system is behaving oddly, the first thing worth checking is whether the output ports are actually terminated the way the datasheet assumed. Most of the "bad divider" reports I see are unterminated ports and an unbalanced load.
Top comments (0)