Hi all,

I am trying to design a Microstrip bandpass filter using edge Coupled Tapped line structure. Ater designing and simulating it in ADS i found variations in the actual performance and simulation result. I found that there is a shift in the band and poor return loss.

Can anyone please suggest me
1.why i am getting shift in the band??
2. Why there is mismatch in the simulation and actual response even though i am feeding the substrate information in the simulation?
3. How to control this shift in the band?
4. How to improve the return loss perfomance?

Please help me to solve this problem.

 

2. Why there is mismatch in the simulation and actual response even though i am feeding the substrate information in the simulation?

How accurate is your substrate information-- could that be off? Also did you simulate the parasitics with the vias used in a tapped line? How accurate/precise is the built circuit-- build tolerances, copper thickness....

 

What frequency range are you working with?

Is your substrate dry? A bit of moisture can throw things off considerably. Try putting it in a vaccuum oven for several hours.

Did you account for parasitic capacitance of the filter's enclosure?

How far was the shift in frequency and bandwidth?

Can you post your schematic, the simulation plot of insertion loss and return loss, as well as your actual results?

 

Hi Nanovate,

I am using the substrate information provided by the supplier only.

There are no vias used in this structure. Please see the attachment for the structure used.

I don't know how to simulate for parasitics (in case of microstrip filters) in ADS. Please tell me if you know.


Hi SgtWookie,

Currently I am working on two filters: 1) Centre freq 3 GHz
2) Centre freq 11GHz

The substrate I am using is dry. I am not using any enclosure for these filters.

The shift in the freq is more than half of the bandwidth.

Please see the attachment for the structure used in both the designs.

 

OK, I reviewed your design.

You're probably experiencing a downwards shift (ie: lower in frequency), correct?

You could simulate parasitics by adding small values of capacitance to ground along your L's. How much depends upon how thick your substrate is, what the electrical parameters (particularly, dielectric constant) of the substrate are, and what the surface area of your L's are. However, the distance from your substrate to the plane of your enclosure could vary significantly due to how it's attached (ie: adhesive thickness). You might consider leaving a portion of ground plane on the substrate, and simulate it - easier to compensate for a known evil than an unknown one. Something else to consider is the coefficient of expansion for the substrate vs the enclosure materials. Ceramics have a very low coefficient of expansion, where aluminum has a very high coefficient of expansion. While this is more of a reliability concern, it can also cause strange things to happen due to the stresses.

I don't care what the manufacturers of the substrate tell you, your ceramic substrate WILL absorb a certain amount of humidity. After all, it got washed after the etching process - some of that still has to be in there. If you don't have a vacuum oven, just put it in a 110°C-125°C dry oven overnight, and try again right after it cools off. Finger oils, other contaminants like flux, cleaning chemicals, etc. can also throw your filter out of kilter. Make your last soak before bake be in isopropyl alcohol.

Toroids aren't supposed to change values with humidity, either. I had a heck of a time with them in a 15mhz narrow bandpass filter; just a few PPM would throw me out of the ballpark. I wound up having to bake them overnight every few days; and inbetween store them in a dry gaseous nitrogen-filled cabinet.

Check your connections over very carefully! It could be that your connections to the substrate traces are too close to the case, the shielding or similar.

I assume you're using semi-rigid coax, 50Ω impedance, correct? Did you calibrate the network analyzer properly, using the same cables you're looking at the filter with? Did your "short" look exactly like your filter's connections? (same fittings, etc) Inspect your semirigid cables really closely for cracks, and especially for contaminates on the mating surfaces. When you're in the GHz range, even seemingly small bits can cause big errors.

Do you have any hanging poles? Look well outside the bandwidth. They might be quite small. Little lost hanging poles will ruin your day, along with your return loss. Once you spot a hanging pole, you can find out where it is in your filter by adding just a bit of capacitance using your finger (don't forget finger cots so you don't muck up the substrate; finger oils can be hard to get out.) Look at both S11 and S22. If your network analyzer won't display them simultaneously with S21 or S12, that's fine - just turn it around.

We used Eagleware for our LC and crystal filter simulations, at a company I used to work at, so other than suggesting to put in caps that aren't physically there, I don't know what to tell you.

Oh, I didn't see any "physical" caps - but just so you know, SMT multilayer caps go up in value a tremendous amount as frequency increases over, say, 200MHz. By 500MHz, they can easily be triple their value (the higher the pF value, the more drastically they increase). Single-layer caps don't exhibit this phenomenon nearly as much.