How to supress this resonance frequency in a LPF?

[LTCC]

ltcc resonant box

Hello, guys. I have designed a LPF. In the graph below, it has a resonance frequency in 15ghz (mark3). I want to find the reason and supress it.

Does anybody know what may cause this resonance and how to solve it?

Thank you so much.

 

[LTCC]

resonance frequency lpf

Any suggestion?

 

[Phytech]

Without information on the design parameters of your filter, how can one give you suggestions?

 

[RF-OM]

Without detailed information on your design it is hard to say something, but I can give a couple of hints. First of all it looks like you have a parasitic resonances in stop-band. The most probable reason is self-resonance of one or more components. If you use lumped component check them for their self-resonance frequencies (SRF). It also may be discontinuities inti ground and transmission lines. even soldering points with excess of solder may lead to parasitic resonances. Your plot also shows steady decline of attenuation for higher frequencies. It may parasitic capacitive coupling from input to output, series connected inductance from one or more elements to the ground (usually it is via) or something else. Try to check all these points. If you need additional assistance I can help you, but I need more information.

Best regards,
RF-OM

 

[LTCC]

Thank you, RF-OM.

Here is the filter model in HFSS. It's a microstripline LPF using LTCC substrate.
And vias are to connect the intermedia ground plane and bottoom ground plane.
The filter is housed in an Aluminum box using conductive epoxy.

 

[LTCC]

The stopband of this LPF is designed to be below 40dB from13-16GHz. The following graph is my simulated results(dot lines) vs. the measured results(solid lines). In the simulated results(mark3), the resonance frequecy is about 18Ghz, which is the resonance frequecy of the Aluminum box.

Now, I didn't know why this frequency shift to 15.1Ghz(mark2), or an unexpected resonance frequency has occured?

 

[Mazz]

In real filter, How did you solder the connectors ground to the filter? I suspect that the intermediate ground is not so "cold" as it has to be, especially close to the connectors.
Try to make a second prototype adding more vias close to the connectors, or simply avoid to use a 3 layer board.
Another possibility could be that, for some unknown reason, the resonant freq of the box is lower than the simulated one. Could you remove/modify it and repeat the measurement?
I hope it can help.
Mazz

 

[RF-OM]

Dear LTCC,

It looks like you design 9th order minimum inductors stepped LPF. From your picture of HFSS simulation box one can see that you placed three rows of via in upper area and two rows at the bottom. These upper via located very close to printed capacitive elements. The rule of thumb is to keep all grounds at least 3 trace widths away from filter elements. It looks in your design upper via placed closer. This may lead to some local parasitic resonances. I also do not understand why you use more than one ground. This is usually unacceptable in microwave design unless it is really necessary for some reason.

I did some simple design for your spec just to check possibility to move away reentry mode peaks that you have around marker m4. I personally dont like any Ansoft products and HFSS too. These programs are bulky, unfriendly and rarely provide really good results, at least results that are better than from other programs. Even old Genesys Empower EM simulator provided better and faster results than powerful HFSS. This is why I used new Genesys for my quick simulation. As you can see from included picture for 9th order stepped LPF we may move reentry mode peaks out of 25 MHz zone. Please keep in mind that this is circuit simulation, not EM, so actual response may be shifted down for about 10 to 20 %. It is easy to retune the filter back to the same characteristics. By the way this is answer to your question about why maker2 is shifted down from marker3. The reason that circuit simulators cannot count many of parasitic unless you include them manually into schematic. But to so you need to know all of them a priory, which is possible only when you have good models for all elements. EM simulator counts more parasitic effects, therefore frequency response shifted down. By the way, I used Rogers 4003 for substrate.

Next point is markers 2 and 3 peaks. It is possible that cavity resonance is responsible for it. Try to make your box smaller, especially for it smaller size. You can check it by removing box at all. If cavity resonance phenomenon is the reason parasitic peak will disappear. If it still there your box is not the reason of peak. You may also try to remove the second ground plane and check its effect.

Finally about via rows. It is rare occasion when you really need three rows. Two rows usually enough, but it is better to shift one by half the pitch. It will increase attenuation of via rows but preserve the ability to flow to other currents.

Best regards,
RF-OM

 

[Phytech]

Did you use the correct dielectric constant for your substrate in your simulation? When the cavity is loaded by the dielectric materials, the resonant frequency will move lower.

 

[flatulent]

Transmission line filters always have responses at harmonics of their cutoff frequency.

The usual trick is to cascade differrent types of filters.

 

[LTCC]

Thank you for your advice, especially RF-OM and Mazz.

For some reason, I have to use the multilayer ground. So the vias are unavoidable.

Added after 11 minutes:

In real filter, How did you solder the connectors ground to the filter? I suspect that the intermediate ground is not so "cold" as it has to be, especially close to the connectors.
Try to make a second prototype adding more vias close to the connectors, or simply avoid to use a 3 layer board.
Another possibility could be that, for some unknown reason, the resonant freq of the box is lower than the simulated one. Could you remove/modify it and repeat the measurement?
I hope it can help.
Mazz

I didn't solder the connectors ground to the filter. I have uses epoxy to connect the pin of the SMA connectors with the transmission lines of the filter. Then the SMA connectors are screwed to the box so that all the ground planes are connected. I do think vias close the connectors are very important. I am doing anther prototype by adding more vias.

 

[RF-OM]

To LTCC,

Check the possibility to use CPW instead of microstrip. I do not know all the details of your project, but it looks like CPW may be better for your application. There is one SMA connector which can be screwed and may work up to 40 GHz, I unfortunately do not remember the type and part #. The footprint for this connector is very important for you, so design it carefully. You also may try to use absorber material inside the cavity. Often it can help with parasitic resonances.

Good luck,
RF-OM

 

[LTCC]

Thank you so much, RF-OM, for all your suggestions and filter model.

I'll first use absorber material to see if the resonance frequency shift or remove. If it doesn't work, I will modify the size of the box, use 40GHz SMA connectors and try again.

After doing these changes, I'll update how it's going on with the fitler.

Added after 2 minutes:

Did you use the correct dielectric constant for your substrate in your simulation? When the cavity is loaded by the dielectric materials, the resonant frequency will move lower.

I have already considered this situation, which is that the box is loaded by the dielectric materials. Thank you for your advice.

 

[RF-OM]

To LTCC,

You are very welcome! Please let me know if you need something else.

BR,
RF-OM