Help selecting SMT component selection for a bandstop filter

[uski]

Hi

TL;DR: making a PCB band stop filter with SMT parts, please help me make an educated choice for my capacitors and inductors, dielectric type, and wire wound vs multilayer

I am venturing into making a homemade 88-108MHz FM bandstop filter, with -30dB cut in this band. Just for fun and learning. This type of filter is to be used for reception, not transmission, so it is not expected to handle any serious power.

The goal is for this filter to have good performance until 1.5GHz.
After this frequencies, I expect parasitic phenomenon from the layout, and from the PCB material itself (FR4) to start becoming problematic.
Is this reasonable?

I used this website to calculate components :
https://rf-tools.com/lc-filter/



I routed a small PCB :



I calculated the trace width with a coplanar waveguide calculator with ground plane :
https://chemandy.com/calculators/coplanar-waveguide-with-ground-calculator.htm



I think this is not super critical anyway since most of the traces will not be actually that width due to the size of the components.

I am at the stage where I am selecting components.

For the capacitors, I plan to use regular SMT capacitors. Since the values are small, I plan to go with NP0. But would X7R be problematic? Why?

For the inductors however, I don't know what to do. Should I use wire-wound inductors, or multi-layer inductors?
What are the selection criteria?
I expect the DC resistance is going to create losses, but that's about all I know, and I am not even sure it is true.
What about the self-resonance frequency, does it matter? What about the Q factor? Those things are a bit blurry to me and I would appreciate some pointers to make a more educated choice

Thanks!

[yl3akb]

Non-tuneable lumped element filter for up to 1.5GHz is pushing the limits but still is feasable.

A few quick comments:
Dont use coplanar microstrip, but plain microstrip instead, because GND on top layer close to SMT components increases the parasitic capacitance to ground.
Use simple microstrip and use two individual vias for each grounding point. But if coplanar waveguide is really necessary, at least add some clearance between SMT parts and top GND layer.

Self resonance frequency and Q is very important for both inductors and capacitors. You want to have SFR af all elements to be above the max. operating frequency and Q as high as possible to reduce the insertion loss and maintain the steep filter roll off. Dont worry about dielectric materials and physical construction, etc., of SMT part so much, but instead look for parts which has actual RF perofrmance charts available (Q, SFR vs. freq.) and judge the performance using that data.

For example, I personally have a good experiance with these capacitors: https://www.johansontechnology.com/s-series
And inductors: https://www.johansontechnology.com/ceramic-inductors

[eb4fbz]

Don’t expect the parasitics to come exclusively from the PCB. You need to include, at least, the inductors parasitics and capacitors Q in the simulation. Most manufacturers provide S-Parameters files.

[BigBoss]

Refer to Murata for ceramic 0805 or 0603 package C0G feature capacitors. GRM series is appropriate for that frequency.
For coils, I recommend you to use adjustable wirewound coils to tune the filter.
https://www.coilcraft.com/en-us/products/rf/tunable-rf-inductors/vertical-mount/uni5/

You can also use fixed wirewound coils but tuning will be a bit constrained.
https://www.coilcraft.com/en-us/products/rf/air-core-inductors/#/

You can wound your own inductors too. It's simple to calculate and they will be adjustable by spreading the length.

Why I'm saying to use adjustable inductor because fixed SMT type inductors have large tolerances and PCB stary effect impact the filter response. Therefore adjustable inductors are better option.
Tolerances for capacitors are less dangerous and it's pretty possible to find very low tolerance capacitor.

[radiolistener]

You can use high Q porcelain RF capacitors for better results.
Manual winding coils can be tuned by spreading the length with ceramic tweezers.
Use perpendicular layout for the coils between adjacent sections, it allows to reduce parasitic coupling between inductors.

But don't expect ideal response like you show in the simulator.
In my case I get about 1-2 dB loss for bandpass frequency region.
And it's hard to tune such filter which consists a lot of sections.

[uski]

Thanks so much everyone for all the answers. Several good tips in there. I will redo the PCB without the top ground plane.

I do have experience with Johanson Technologies passives. Years ago, I used them in the RF chain of a product I designed, for impedance matching.
Can confirm they did the job just fine. I see that they have kits available on DigiKey, might just get one.

Interestingly, I took a quick peek at the specs and it seems like the bigger components have better specs than the smaller ones, at least in terms of SRF and Q factor. Intuitively, I would have expected the smaller components to be better.