Comparing a FM Coax Notch Filter vs Lumped Elements FM Bandstop Filter
Over on his YouTube channel Adam 9A4QV has uploaded a video that compares a coax notch filter and lumped elements filter band stop for the FM band. Bandstop filters are useful as they can be used to block out extremely strong signals that can overload an SDR dongle (or any radio).
A coax notch is a very simple band stop filter that is made from a length of coax cable at 1/4 wavelength of the frequency that you want to block. Just connect the 1/4 wavelength coax with a T-junction connector and you’ll get a notch at the frequency you want to block. A lumped elements filter is one made out of inductors and capacitors. Designing this type of filter generally requires a few more calculations, and ideally simulation. Then building it is a bit more difficult as you either need to buy or make the inductors, and then solder them together.
But as Adam shows in the video coax notch filters have a problem in that the notch is not only at the exact frequency that you want to block. Instead there will be multiple odd spaced harmonics of the blocking frequency as well. For example if your desired notch is at 100 MHz, you’ll also get notches at 300 MHz, 500 MHz, 700 MHz and so on. So a coax notch filter still needs to be carefully designed to not block out your frequency of interest.
I added “Another means to reduce the width of a 1/4 Lambda (wavelength) transmission line notch at the cost of a decreasing the notch depth, is by loosely couple the 1/4 Lambda circuit via a small capacitor.” because I remembered, that a report on a VHF-repeater in DL, where despite the use of several meter long cavity filters with high Q, one signal was not sufficiently suppressed. They operator experimented with different coax types directly coupled to notch this unwanted signal, but the notch width was to wide for them. Only when they they tried the loose coupling approach could they notch the undesired signal.
For some time there was also the proposal by one HAM’s to add an additional HPF path for e.g. harmonics of a Tx and terminate them in a resistor load to dissipate harmonics in heat. Except for diplexer using a series resonant circuit for the IF output path and a terminated parallel circuit with a 50 Ohm resistor to achieve wideband matching at for the IF-output, I have not seen this approach of using resistors in filter in any other design.
Vy 73
_.. .._. ….. .__. _.__
PS.: while VNA are more available everyone relies to much on the assumption that they are under all conditions unfailable. Settings differ unfortunately from manufacturer to manufacturer that e.g. due to incorrect/not applicable settings results are invalid to start with.
Using a diplexer with a 50 ohms termination to dissipate the harmonics is a smart way how to improve the amplifier linearity and to decrease distortion and other unwanted effects. Even better option is to use two 90deg hybrid couplers and two bandpass filters. Both 90deg couplers should be terminated with the 50 ohms where all products other than the filtered one will be dissipated in the 50 ohm loads and the amplifier will see output impedance all the time.
All notch-filter is imho not a filter, since it just provides a short at the resonance frequency. Also Notch filters are normally not used to replace High Pass Filter (HPF) Low Pass Filter (LPF) or Band Pass Filters (BPF), but as addon to reduce very strong signals, that cannot be suppressed sufficiently by a LPF, HPF or BPF. When tuned to such very strong signals the additional 15 to 30 dB in notch depth can make the difference.
Independent if a simple LC resonant circuit or the equivalent using some form of transmission line, e.g. coax, strip line-circuit, slot or 2 wire line, funtions at as short. Additional resonances occur at odd multiples of it’s frequency, e.g. for a 100 MHz 1/4 Lamda this will perform as 3/4-Lambda at 300 MHz and so on.
The BPF or FM Trap as used by 9A4QV consists of a notch filter, as can be seen in his diagram in the middle, with the addition of 2 parallel resonant circuits before and after the Notch/short, which act as an open circuit at the resonance frequency.
The width of the notch and the peak attenuation of notch vary with the Q of the LC-circuit/transmission line and can be improved with increasing Q within limits. Another means to reduce the width of a ¼ Lambda transmission line notch at the cost of a decreasing the notch depth, is by loosely couple the ¼ Lambda circuit via a small capacitor. Use of this method will produce a lower resonance frequency and requires shortening of the transmission line length. The required length will as always depend on the transmission line properties.
@snn47, Your reply is on-point in my opinion.
However I do take exception with your statement: “All notch-filter is imho not a filter, since it just provides a short at the resonance frequency.”
A transmission line is simply an infinite number of series linear time invariant lumped RLC elements. For a fixed excitation frequency, fixed source, and fixed load impedance, when you change the length of a transmission line, you are changing the values of the lumped RLC elements that represent the piece of coax. This resuts in an S11 (input return loss) and S21 (forward insertion loss) response that is identical to a physical lumped element filter (one made with components, not a transmission line).
So mathematically, whether you make a “notch” or “bandpass” filter using identical lumped elements with respect to the input and output nodes, the two are identical in the time and frequency domain.
In realization however (if you actually build something), A transmission line filter will likely behave quite different compared with a filter built with lumped physical components (e.g. parts). The difference between the two is the proverbial “Devil in the Details”. These differences are usually referred to as “Parasitic” parameters.
I like that you said: “Another means to reduce the width of a 1/4 Lambda (wavelength) transmission line notch at the cost of a decreasing the notch depth, is by loosely couple the 1/4 Lambda circuit via a small capacitor.”
I suggest this statement can be taken further. One example is using small inductance and/or capacitance values at the source and or load of a transmission line fiter to “tune” the filter. These added physical small inductance and capacitance parts are typically called “Gimmick” parts. See here:
https://en.wikipedia.org/wiki/Gimmick
However, the use of Gimmicks is bad practice unless they are used carefullY. Especially in production quantity designs. But Gimmicks are quiet useful during the design/prototyping phase.
At VHF through SHF frequencies, coaxial notch or bandpass filters are quite effective, especially compared with physical passive lumped element filters (Q fator is the key). But there are down sides to using tuned lengths of coax, especially if the filter is being mounted outdoors – temperature: The characteristics of coaxial cable varies with temperature. This can wreak havoc with a high-Q coaxial notch filter (it gets even worse when you replace pieces of coax with physical resonant cavities).
It is possible to manually tune coaxial stub filters without cutting the coax length and without adding Gimmicks. Research devices like coaxial “Trombone Lines” or “Sliding Lines”. But without clever design, the likes of a sliding line device is usally relegated to the laboratory enviornment. As Vector Network Analyzers (VNAs) become more affordable these days, they present a vastly better solution in my opinion.
Regards, David