filters · 2026-09-21

Sysmocom CF868 cavity filter

An 868 MHz band-pass cavity filter for LoRa and MeshCore nodes, measured with a NanoVNA-H4 from 800 to 1000 MHz. The question: how much does it cost in the 868 MHz band, and how well does it keep out the mobile networks next door?

868mhzcavityvnananovnalorameshcoresysmocom

Key figures

Figure	Measured
Insertion loss, 863–870 MHz	0.42–0.82 dB
Insertion loss at 869.5 MHz	0.82 dB
Lowest loss	0.42 dB at 863.75 MHz
−3 dB passband	857.25–874.25 MHz (17.0 MHz)
Return loss, 863–870 MHz	13.4–19.2 dB
800–821 MHz (LTE 800 downlink)	≤ −53.8 dB
925–960 MHz (GSM/LTE 900 downlink)	≤ −50.7 dB

The stopband values are at the NanoVNA's noise floor, about −55 to −60 dB. The real rejection is at least this, probably more. The sweep starts at 800 MHz, so the lowest part of the LTE 800 downlink (791–800 MHz) is not covered.

Frequency response

S21 transmission, normalized S11 reflection

Move over the plot to read values.

How it was measured

NanoVNA-H4, 800–1000 MHz, 801 points (250 kHz steps).
The screenshot below is the live S21 trace with four markers.
The exported CSV was not through-normalized: its S21 reads 2.3 dB lower than the screen. I scaled S21 by +2.342 dB, the mean offset at the two passband markers (863 and 870 MHz). After this, all four markers match the screen within 0.09 dB.
S11 is as exported. The screenshot (12:02:40) and the CSV (12:04:00) are two sweeps a minute apart, so the noise in the stopband differs a little.

NanoVNA S21 trace from 800 MHz to 1 GHz with markers at 833 MHz −52.40 dB, 863 MHz −0.428 dB, 870 MHz −0.798 dB and 890 MHz −47.31 dB

Marker	Frequency	Screen	Normalized CSV
1	833 MHz	−52.40 dB	−52.47 dB
2	863 MHz	−0.428 dB	−0.441 dB
3	870 MHz	−0.798 dB	−0.785 dB
4	890 MHz	−47.31 dB	−47.23 dB

Files

Normalized data (CSV)Comma-separated, dot decimals. Columns: frequency_hz, s21_db, s21_re, s21_im, s21_db_raw, s11_db, s11_re, s11_im

CSV · 79 kB

Original NanoVNA export (CSV)Semicolon-separated, decimal commas. Frequency, then real and imaginary parts of S11, S21, S12, S22. Not normalized.

CSV · 139 kB

NanoVNA screenshotS21 with markers, 2026-09-21 12:02:40

PNG · 99 kB

Manufacturer datasheetrf_filter_cf866.5-kt30.pdf

PDF · 387 kB

Manufacturer: sysmocom – s.f.m.c. GmbH, Berlin.