openEMS · CadQuery · Unterhaching, DE

Designing a 21 cm H I-line preselector that survives a Munich suburb.

Four cavity-filter topologies, 63 FDTD simulations, three tuning campaigns, and the physics of why you can't always have low insertion loss and strong GPS rejection in the same box.

1420.4 MHz
target — HI line
63
openEMS FDTD runs
4
topologies explored
< 0.6 dB
insertion-loss target
−30 dB
min GPS-L1 rejection

01 The mission

The 21 cm hydrogen line sits in the ITU-protected 1400 – 1427 MHz band, but at a Munich-suburb site like Unterhaching the spectrum around it is anything but quiet. The filter's job is to admit the 10 MHz around 1420.4 MHz with negligible loss, and reject everything else — especially GPS L1 (155 MHz above) and Iridium (200 MHz above), both always-on satellite downlinks the receiver can never point away from.

FM broadcast
87.5 – 108 MHz
Olympiaturm, Wendelstein — multi-kW ERP, line-of-sight.
DAB+ Band III
174 – 240 MHz
Regional multiplex networks.
GSM 900 / LTE 800
790 – 960 MHz
Dense suburban cellular; WR-650 cutoff is 908 MHz, rejecting most of this structurally.
Mode-S / ADS-B
1030 / 1090 MHz
Munich Airport (EDDM) 40 km NE; high pulse power but low duty.
L-band ATC radar
1250 – 1350 MHz
Primary surveillance radars.
GPS L1 / Galileo E1
1575.42 MHz
Always on, every sat in view (≈ 10 at Unterhaching horizon).
Iridium downlink
1616 – 1626 MHz
Canonical HI-observer pest; burst mode.
GSM / LTE 1800
1710 – 1880 MHz
Dense.

02 Five design milestones

Each card is the final geometry of its column in the campaign log — click through to see the iteration journey that got there.

03 3D geometry viewer

Spin, pan and zoom the machined aluminium housing for each of the five designs. Drag to rotate, scroll to zoom, right-click-drag to pan. STL files are the same ones the HI_Filters/runs/ pipeline writes after each simulation — geometry in the viewer is exactly the geometry the FDTD solver saw.

Loading 3D viewer …
drag · scroll · right-drag

04 Interactive S-parameter comparison

Overlay any combination of the five designs. Toggle S11 or S21, zoom into the passband, and watch how the 1553 MHz spurious in the as-designed interdigital (run 035) sits right where GPS lives. The dashed vertical lines mark the key RFI frequencies.

Designs
Trace
View
Markers

05 Out-of-band rejection matrix

Simulated S21 (dB) at each interferer frequency, interpolated from the saved sweeps. The HI-line row is the passband; every other row is rejection — more negative is better. Red cells fail the ≥ 30 dB threshold typical for dense-RFI sites.

06 The iteration log

Click any campaign to see its full tuning trajectory — rod length and spacing for the coaxial designs, cavity length and iris gaps for WR-650, iris size for the interdigital+septa fixes. The scatter plot shows where each run landed in (f_peak, bandwidth) space; the table is the per-run parameter change.

07 Insertion loss — the 0.6 dB pinch point

Each filter's in-band IL is set by the product of unloaded Q and bandwidth. The iris-coupled interdigital is trapped: the aperture that suppresses the 1553 MHz spurious caps BW near 5 MHz, and bare-aluminium Qu then forces IL above 0.6 dB. The combline escapes this because its 10 MHz BW comes from rod spacing, not iris aperture.

The physics

IL ≈ 4.343 × Σgk × f0 / ( Qu × BW )

For a 3-pole 0.1 dB-ripple Chebyshev, Σgk = 3.21. With f0 = 1420 MHz, meeting IL < 0.6 dB requires Qu × BW ≥ 33,000 MHz.

Realistic unloaded Q for a machined coaxial cavity at 1.4 GHz:

FinishQu range
Bare 6061 aluminium3,500 – 4,500
Silver plate / Cu strike (~2 µm)4,700 – 6,000
Copper cladding or solid Cu6,000 – 7,500
WR-650 waveguide (inherent)8,000 – 10,000

Interactive calculator

10.0 MHz
4000
3
0.00 dB
estimated insertion loss
— / —

08 Verdict

Under the IL < 0.6 dB hard spec, the combline wins in bare aluminium. The interdigital + iris rescue is feasible but requires silver plating or copper cladding. WR-650 is the clear choice when size and budget aren't constraints.

PRIMARY RECOMMENDATION

Combline

Meets all specs in bare aluminium — IL ≈ 0.46 dB, GPS rejection −47 dB, post-fab tunable.

  • 10 MHz BW native, no iris-coupling trap
  • 3 cap-screws for post-fab centering
  • €110 bill of materials, 0.6 kg
  • 1495 MHz spurious (−8 dB) in an empty band
COMPACT ALTERNATIVE

Interdigital + iris

Smaller footprint (0.5 kg), but the spurious-suppression iris forces BW down to 4.5–5.5 MHz. Plating required to meet IL spec.

  • Run 047 (Cu-clad): IL 0.60 dB, GPS −30 dB
  • Run 063 (Ag-plated): IL 0.61 dB, GPS −27 dB
  • +€50–100 plating cost
  • No tuning screws — tolerance-sensitive
RESEARCH-GRADE

WR-650 iris

IL ≈ 0.05 dB, structural cutoff at 908 MHz rejects FM, DAB, GSM 900 and Mode-S infinitely. 10 kg / 77 cm.

  • Fixed-pier mount only
  • €850–1100 inc. CPR-650 + coax transitions
  • Highest unloaded Q of the bunch
  • Spurious-free to ~1.8 GHz (next WG mode)
CriterionWinnerNotes
IL < 0.6 dB in bare ALcomblineOnly bare-AL coaxial design that meets it.
f-peak centering accuracyinterdigital+iris (Δ = +0.1 MHz)FDTD sampling resolution-limited.
Lowest achievable ILWR-650 (0.05 dB)10× better than any coaxial.
Below-1-GHz rejectionWR-650Infinite below cutoff; coax drops monotonically.
GPS L1 rejectioncombline ≈ WR-650Both adequate; interdigital+iris −30 dB.
Compactnessinterdigital+iris (run 047)50 × 70 × 130 mm, 0.5 kg.
Post-fab tunabilitycombline3 cap screws, ±40 MHz trim range.
Cost in bare ALcombline (€110)Interdigital+plating ≈ €200; WR-650 > €850.
Overall for Unterhaching hobby usecomblineBest IL/spec/€ balance in bare AL.
Research / observatory installWR-650When the TSYS budget justifies the mass.