120 mhz bandpass filter 1

[davicente]

I'm working in a 120 mhz bandpass filter, i've designed it through a high-pass after a low-pass filter.The problem is that although i`ve simulated it with and without the pads and microstrip lines, whenever i make the PCB, i never obtain the desired results.I`ve tried to imulate it with RFSIM99, LIBRA and MICROWAVE-OFFICEE.The specification for my filter are:

BW1dB=14 MHz
BW3dB=16 MHz
bandpass s21>-2dB

I'm using SMD lumped elements on a miscrostrip structure PCB.
Can anybody help my?any idea?what am i doing wrong?What can i do to obtain more or less the same resaults as in the simulations?
Thanks in advance

 

[buding]

Hi!

Analysing of microstrip structure in software like LIBRA or others has always a little error connetcted to the analytical model of analysed structer.
To have more accurate results You should use some 2.5D simulators like MOMENTUM, SONNET etc. And also You have to remember about very detailed and accurate defining of dielectric material. In most cases it is the reson of all errors. Good luck.

 

[biff44]

It is generally best to do an actual bandpass, than a lp/hp combination, unless we are talking about huge passbands. Usually the filter departs from expected performance because 1) there is excess capacitance in the board layout acting in parallel with the lumped capacitors, 2) there is extra via hole or ground loop (you did use a good ground plane, didn't you) that changes the apparent series inductance used, 3) The component tolerances for values are inadequate 4) the component Q is too poor to achieve the desired roll offs.

 

[Higgler]

I made a series of these using fixed inductors and tunable capacitors for Cable TV. Without the tuning capability of capacitors, it's difficult to achieve perfect results.
We used Touchstone for analysis, results were pretty close to the Touchstone outputs.
kch
PS:biff44 is correct, do an actual bandpass design, fewer parts and better results.

 

[biff44]

Theoretically, a bandpass filter has twice the roll-off as a lowpass or highpass filter. So you can get away with fewer filter poles.

 

[krytar]

Hi,

Hipass-Lowpass approach is used when you have a very broadband or steep rejection on both skirts of the filter.

For the spec that is given, your approach is correct. If you use a bandpass design, then the ratio between L&C in series will be very hard to realize (e.g. 1pf/1000nH).

Use Lowpass & hipass with one finite transmission zero (finite pole) on each side of the BPF will solve the problem.

Krytar