Field notes from European Microwave Week 2026, London.
One of the quieter demos at EuMW 2026 was on the Smiths Interconnect stand: an aluminium block with a row of tuning screws, three SMA cables, and an Anritsu ShockLine four-port VNA. The DUT was a Ku-band filter and directional coupler built into a single housing for a space application. What made the screen worth photographing was the frequency axis. The sweep ran from 50 MHz all the way to 43.5 GHz, far beyond the passband, and that is where the more important information about a waveguide filter shows up.
The DUT: one housing instead of two
According to the exhibitor's demonstration sheet, the assembly was developed as a multi-function assembly for a spaceborne telemetry, tracking and command application. The filter and coupler were not procured as separate units. The sheet gives the reasons as fewer interfaces, lower insertion loss, and a mechanically robust package suited to environmental qualification. The main signal path runs through a waveguide bandpass filter, and a coupled monitoring port provides around 20 dB of coupling across the operating band.
The bench used three of the VNA's four ports. Port 1 fed the input, and Ports 2 and 4 went to the output and coupled ports. The fourth cable was left unconnected. A three-port DUT on a four-port VNA can be fully characterised in one connection.
What the screen showed
Both plots share the same 50 MHz to 43.5 GHz axis. The marker readouts are too small to read in the photo, so all values below are estimated from the plot grid.
- Passband: roughly 10.5 to 12.9 GHz. The markers sit near the band edges, at positions consistent with the 10.7–12.75 GHz Ku downlink range. S11 in band is around −21 to −25 dB, with one null near −33 dB. S21 sits close to 0 dB at the 10 dB/div scale.
- Below the passband: S21 falls almost vertically towards −100 dB.
- Above the passband: S21 shows spurious transmission responses. They reach about −50 dB around 15–16 GHz, about −38 dB around 22 GHz, and about −37 dB in the 34–43 GHz range.
- Out-of-band S11: close to 0 dB, as expected for a reflective filter, but with many narrow dips above the passband. The deepest, near the top of the sweep, reaches about −13 dB.
Reading 1: waveguide filters are asymmetric by nature
The steep lower skirt is not a feat of filter design. A waveguide simply does not propagate below its cutoff frequency, so the low side of the response gets very high rejection without any design effort.
The upper side is different. Cavity resonators have higher-order resonances, waveguides support higher-order modes, and coax-to-waveguide transitions add modes of their own. Together these create re-entrant passbands above the design band. Upper-stopband rejection is something the designer has to achieve, and therefore something the test has to verify.
That is why this sweep goes to 43.5 GHz. For a filter centred near 12 GHz, the second and third harmonic regions sit at roughly 24 and 36 GHz. A sweep confined to the passband, or even to 20 GHz, would never show the responses in those regions. For space hardware, the specification should state upper-stopband rejection up to a defined frequency, typically at least two to three times the passband, and the test has to cover it.
Reading 2: S11 dips that do not match S21
The most useful habit on this screen is to compare the two plots at the same frequency.
Near the top of the sweep, S11 dips to about −13 dB. That means only about 5% of the incident power is reflected, so roughly 95% goes somewhere else. Yet S21 at those frequencies peaks at about −37 dB, which is around 0.02% of the power. For a lossless three-port, |S11|² + |S21|² + |S31|² = 1. The missing power therefore has two possible destinations: the coupled port, or dissipation, for example in a lossy resonance or a higher-order mode trapped inside the structure.
The practical conclusion: an out-of-band S11 dip that is much deeper than the matching S21 spur is a reason to measure S31 and the loss term, not something to ignore. This matters most when the coupled port drives a power detector, because out-of-band energy arriving at the monitor port goes straight into the power reading.
Reading 3: integration changes where the mismatch comes from
Each SMA interface pair adds mismatch, insertion loss and repeatability uncertainty. Combining filter and coupler in one housing removes the interface between them. The in-band S11 of around −21 to −25 dB then mostly reflects the coax launches at the ends, which are now the main remaining contributors. In a qualification context, those launches are also what to re-check after vibration and thermal cycling. So are the tuning screws, because a screw that shifts moves both the passband and the spurious responses.
A short checklist for waveguide and cavity filter verification
- Sweep from below cutoff to at least 3× the upper passband edge, or to the limit stated in the specification.
- Record the worst spurious response, both its frequency and its level, not just in-band insertion and return loss.
- Measure every port of an integrated assembly: coupling, directivity and isolation, and the coupled port's out-of-band behaviour.
- Use |S11|² + |S21|² + |S31|² to track where the missing power goes at out-of-band S11 dips.
- Handle unused VNA ports correctly when a four-port VNA measures a three-port DUT.
- Repeat the wideband sweep after environmental testing.
Takeaway
The screen showed a clean Ku-band passband, and the more useful information sat to the right of it. A waveguide filter gets its low-side rejection from the waveguide cutoff, while its upper stopband has to be designed and then verified. Sweeping to 43.5 GHz and comparing S11 with S21 at the same frequency shows whether out-of-band power is reflected, transmitted, coupled or lost. For hardware that will spend fifteen years in orbit, those spurious responses are part of the specification.
Notes: Observations and interpretations are my own, based on what was running at the stand during the show. Screen values are estimated from the plot grid in my photo. DUT information is paraphrased from the exhibitor's demonstration sheet. Passband specification, marker values and calibration state were not confirmed with the exhibitor.