Coaxial Stub Notch Filter Designer

High Power Filters ... just made of Coaxialcable

This calculator is suiteable to calculate the length of a piece of coaxial cable to produce a notch at a desired frequency. Unlike to others, this notch is then matched to the source/load at the passband frequency. Matching may be achieved with an L-C-Lowpass or two coaxial cable stubs and an inductor. Or only coaxial cables. Few material & work, much gain in performance !!!

Motivation and Verification

This method of notching out unwanted frequencies is not new. It is a trick, well known among radio - television - technicians. Unfortunately this notch is usually very broad. We therefore run some simulations to see, how to improve this situation. Adding a second stub of the same length, together with an inductor, improves the situation dramatically. We therefore hacked together this calculator to share our experience. The two pictures below show the situation when "just" using a coaxial cable stub as a notch. Insertion Loss was 1.02 dB at 100 MHz and 32.8 dB at 200 MHz. With this additional matching components, Insertion loss lowers to approx. 0.4 dB while the notch is approx. 40 dB.




Example (not from practice, as we don't run a FM broadcast station - at least not often :-) Assume we transmit at 100 MHz. We would therefore want to attenuate the harmonic frequency at 200 MHz. Depending on the coaxial cable which is used, the stub is approximately 0.25 m long. (λ/4 * vf at 200 MHz, Stub open) Connecting this to a network analyser (Multiflex 86 for the test), we see, that the notching action repeats at 600 MHz, 1000 MHz and so on. We also see, that the notch is broad.

PROPOSED NETWORK I

The calculated value for the inductor was 79.6 nH, we used a Coilcraft 1206 CS (82 nH). For the capacitor, a Mica 10 pF and a trimmer (17pF) were used. The insertion loss could be trimmed to 0.39 dB at 100 MHz, while the notch depth was > 40 dB. As this combination is in fact a Lowpass, C - L - open Coax, this attenuates higher frequencies as well.

PROPOSED NETWORK II

The calculated value for the inductor was still 79.6 nH, we (still) used a Coilcraft 1206 CS (82 nH). For the capacitor, we used a piece of coaxial cable. As it turned out, that for matching to 50 Ω a capacitor of the same value as the notch is necessary (both are 31.8 pF). (In the calculation, the length is modified to 98% and 102% in order to get a broader notch. The insertion loss could be optimised to 0.35 dB at 100 MHz with the use of SMA connectors (quick hack), while the notch depth was > 40 dB. As this combination is in fact a Lowpass, open Coax - L - open Coax, this attenuates higher frequencies as well.

PROPOSED NETWORK III

In this third composition, we even replaced the inductor with a piece of coaxial cable. In all simulations (with different cables), we saw, that the length of the series coaxial cable must be longer than the notch by a factor of sqrt(2). Obviously this has nothing to do with the fact, that the notch frequency is twice the frequency of the passband :-(
The attenuation at higher frequencies is poor, compared to the other networks, as only the coaxial cable is the amplitude / frequency determining factor.

✈ Share your thoughts