Question about S11 value of microstrip bandpass filter

[DimaA]

S11 of filter

I'm designing microstrip bandpass filter to be used in 3.1-3.8Ghz
I'm getting about 18dB in S11
Is it good or bad? I'm using FR4 with Er=3.7

 

[g579]

Re: S11 of filter

Is it a design result on a paper or from a soft, or a measured one ?
If it is a realized and measured one you should be satisfied very much.

 

[DimaA]

Re: S11 of filter

This is measured result
Here is the picture, what do you think? I know that insertion loss is pretty big 7.6dB, but this is the best I could get
Someone told me that S11 of filter should be less than 10dB, is this correct?

 

[g579]

Re: S11 of filter

There should be some design problems with the filter. Seems that almost all input energy is dissipated (low out of band S11, and vey high insertion loss).
Give some details of the filter construction and design, perhaps someone can help you here.

 

[BigBoss]

S11 of filter

This filter is quite lossy. I can not say anything without information about filter design methodology but it's lossy.
Input reflection coefficient is almost very good.If S11<-18dB but S21>7.6dB, there are mistaken components such as substrate and/or components if it's done lumped elements..

In additional to, if PCB has been lacked some chemicals, this could be a loss factor too..
It should be very deep analyzed.

 

[lguancho]

Re: S11 of filter

Hi. Perhaps you can post a picture of your physical filter design? There are many factor to high insertion loss, such as poor implementation (such as discontinuity at the RF connectors).

You should also provide a broadband plot on S21 of your filter. This can show problems such as additional waveguide modes which result in poor out-of-band attenuation.

I don't see nice sharp nulls in your measured S11 plots. The nulls are related to the order of the filters.

 

[estradasphere]

Re: S11 of filter

i also designed a filter for Wimax applications with microstrip lines almost at these frequencies and the insertion loss was much less than yours (3.5 dB). if you can, try to use the pcb "rogers 4003" (Er=3.3) , is more suitable for rf applications. there are better pcbs even, such as duroid, but they are too thin and cause mechanical problems during soldering and etching.

on the other hand, your reflection loss is good.

 

[cwjcwjcwj]

Re: S11 of filter

I have experienced of designing a microstrip filter by using FR4....I think >7 dB of insertion loss is a bit too much.........u may check again with ur scheme.......

 

[BigBoss]

S11 of filter

And also, your filter got a little bit inconsistency. Did you calibrate your VNA well?? Because S11 is quite good but S21 ???
Check this out...

 

[DimaA]

Re: S11 of filter

Hi all, tnx for reply
the pcb himself is quit lossy, tan=0.01 at 2mhz, must bi higher at higher frequencies
I got newer picture from the network analyzer, with much accurate calibration kit ( the first one was pretty old kit)
the s21 graph didn't change, but the s11 is little different
this the filter layout and the simulated result in Momentum ADS
the loss is only about 3db
also, seems that s11 is better in reality than in simulation???

 

[estradasphere]

Re: S11 of filter

it's normal that the reflection loss is less than simulated, because some power will be radiated (somehow not well predicted by momentum) and won't be reflected/transmitted to your input/output (lower S11/S21). I've also experienced that effect. The simulated S21 was for my filter about 1.2 dB and the measured 3.5 dB.

 

[g579]

Re: S11 of filter

1. Try a new simulation with loss of 0.025 and 0.03. This may explain the lower than simulated reflection.
2. Are the real realized gaps between the coupled lines identical to the designed gaps in the simulation ?
This may cause higher insertion loss than expected.

 

[VSWR]

Re: S11 of filter

Diel const of FR4 can vary from 4.3-4.8 from batch to batch. And is temperature sensitive as well. Loss Tangent can vary between 0.018-0.030, depending on number of layers, board thickness, etc.

Using FR4 as substrate for microwave circuits, requires accurate tolerance analyses of the FR4 parameters in a microwave design program, to predict spread in performance of the finished design. Using e.g. Monte Carlo analysis here can avoid many traps.