Designing a Compact 5 GHz Band-Pass Filter for 5G

The rollout of 5G is happening faster than most of us can keep up with, and every new base‑station or handset needs a filter that is small, sharp, and cheap. If your design team is still using a bulky lumped‑element filter from a decade ago, you are probably losing out on board space, power efficiency, and even signal quality. In this post I walk you through a practical, step‑by‑step method to create a compact band‑pass filter that meets the 5 GHz 5G specifications most often required today.

Understanding the Requirements

What does “band‑pass” really mean?

A band‑pass filter (BPF) lets frequencies inside a chosen window pass through while rejecting everything outside that window. For 5 GHz 5G we typically target a center frequency of 5.0 GHz with a 3 dB bandwidth of about 100 MHz (±5 %). That means the filter should pass signals from 4.95 GHz to 5.05 GHz and attenuate anything lower or higher.

Key specs to nail down

ParameterTypical value for 5 GHz 5G
Center frequency (f0)5.0 GHz
Bandwidth (BW)100 MHz
Insertion loss (IL)< 1 dB
Return loss (RL)> 15 dB
Size< 5 mm × 5 mm (board area)
Power handling0 dBm typical

These numbers are not set in stone, but they give you a realistic target to work toward.

Step 1 – Choose the Filter Topology

For a compact design at microwave frequencies, the most common choices are:

  • Microstrip coupled‑line resonators – easy to fabricate on a standard FR‑4 or Rogers board, and they naturally lend themselves to a small footprint.
  • Hairpin resonators – essentially a folded half‑wave resonator that saves space.
  • Stepped‑impedance resonators (SIR) – provide a tighter bandwidth control with fewer sections.

In my own lab, I start with a hairpin because it folds the resonator length into a U‑shape, cutting the area roughly in half without sacrificing Q‑factor. If you need even tighter control, you can switch to an SIR later.

Step 2 – Calculate the Physical Length

A half‑wave line at 5 GHz on a typical 0.8 mm thick Rogers RO4003C (εr ≈ 3.38) has a guided wavelength (λg) given by:

λg = c / (f0 * sqrt(εeff))

where c is the speed of light (3×10⁸ m/s) and εeff is the effective dielectric constant (≈ 2.9 for this stack‑up). Plugging the numbers:

λg ≈ 3e8 / (5e9 * sqrt(2.9)) ≈ 30 mm / 1.7 ≈ 17.6 mm

A hairpin uses a half‑wave resonator, so the total line length is about λg/2 ≈ 8.8 mm. Because the hairpin folds back on itself, the footprint can be roughly a square of side 4.5 mm, which meets our size goal.

Step 3 – Determine the Coupling Gap

The bandwidth of a coupled‑line filter is primarily set by the gap between the two parallel sections of the hairpin. A smaller gap gives tighter coupling and a wider bandwidth; a larger gap narrows the band.

A good starting point is a gap of 0.2 mm for a 100 MHz bandwidth on the RO4003C stack‑up. Use an EM simulator (ADS, HFSS, or the free Sonnet Lite) to sweep the gap from 0.15 mm to 0.3 mm and watch the S‑parameters. Aim for a -3 dB points at 4.95 GHz and 5.05 GHz while keeping the return loss above 15 dB.

Step 4 – Add Input/Output Matching

Even a perfect resonator will look like a high‑impedance stub to a 50 Ω source if you do not provide matching. The simplest method is to use a tapered microstrip line (a quarter‑wave transformer) on each side of the hairpin.

Calculate the required characteristic impedance (Z0t) of the transformer using:

Z0t = sqrt(Zsource * Zload)

Since both source and load are 50 Ω, Z0t = 50 Ω, which means you can simply use a straight 50 Ω microstrip of length λg/4 ≈ 4.4 mm. In practice, a slight taper (e.g., 30 Ω to 70 Ω) smooths the transition and reduces reflection.

Step 5 – Simulate, Optimize, and Validate

Run a full‑wave EM simulation of the entire layout, including the feed lines, hairpin, and any via or ground connections. Look for:

  • Insertion loss – should stay below 1 dB across the passband.
  • Return loss – aim for better than 15 dB at the center frequency.
  • Spurious response – make sure higher‑order modes appear well above 6 GHz.

If the insertion loss is high, check the copper thickness and surface roughness; a 35 µm copper with a smooth finish can shave off a few tenths of a dB. If the return loss is poor, tweak the taper length or the gap.

Step 6 – Layout Tips for a Small Footprint

  1. Keep the ground plane solid – any slots or stitching vias near the resonator will disturb the fields.
  2. Use blind or buried vias for the feed pads if you need to route other signals underneath.
  3. Avoid right‑angle bends – use mitered or curved bends to reduce radiation loss.
  4. Place the filter away from high‑current traces – strong currents can induce unwanted coupling.

Step 7 – Prototyping and Testing

Once the layout is finalized, order a small batch of boards. For a first prototype, a standard 0.8 mm thickness works fine. Use a vector network analyzer (VNA) with a frequency range covering 4–6 GHz. Calibrate with a SOLT kit, then measure S21 (insertion loss) and S11 (return loss). Compare the measured curves to the simulation; small deviations are normal, but if you see a shift of more than 10 MHz in the center frequency, revisit the dielectric constant value used in the simulation.

If the filter is a bit off, a quick fix is to trim the gap with a fine tip or add a tiny piece of solder to adjust the coupling. This is a handy trick I learned during my PhD when we didn’t have a precise laser cutter.

Step 8 – Finalize for Production

When the prototype meets the specs, generate the Gerber files and a bill of materials (BOM). For volume production, consider:

  • Copper thickness – 70 µm (2 oz) can improve Q and lower loss.
  • Surface finish – ENIG or immersion gold reduces oxidation and keeps the insertion loss low over time.
  • Panelization – place several filters on a single panel to reduce per‑unit cost.

Wrap‑Up

Designing a compact band‑pass filter for 5 GHz 5G does not have to be a mystery. By picking a hairpin topology, calculating the guided wavelength, setting the coupling gap, and adding simple matching sections, you can hit the required bandwidth and loss while staying under a 5 mm square. The key is to let the EM simulator do the heavy lifting and then fine‑tune with a quick prototype. I’ve used this exact flow on several projects at the lab, and it consistently delivers filters that fit inside the tight spaces of modern RF modules.

Happy filtering!

Reactions