Bandpass filter design - wideband

[deki]

I'm trying to design a bandpass filter with a BW of 50MHz in LTSpice, however my simulations aren't coming out as I'd like. I'm using the frequency transformation method, with a centre frequency of 2GHz., fractional BW 2.5%. I'd like a wider passband, however I can't seem to get away from the following type of response (images attached):

And the circuit:

Any ideas on how to approach this?

 

[meBigGuy]

How did you come up with that circuit?

You probably need a 50 ohm source also.

http://www.electronicproducts.com/Bandpass_Filter_Design_Calculator.aspx

 

[the_emi_guy]

You are intending to build this, or is this a paper exercise?

 

[deki]

There is the possibility of building it, though it's not confirmed. As of now I removed the second resonator, so it's a 3rd order filter instead of the 4th order I begun with. It has helped, but my passband isn't as flat as I'd like.

 

[Jeff Rosenbury]

I'm having trouble with the posted response curve. The region of interest seems to be the pass band, but there looks to be only two data points near there. Is it possible to get a closer picture of the pass band?

Also, what don't you like about it? Is it just gain fluctuations in the pass band you object to, or do other concerns come into play like the group delay?

Are you wedded to the Caur topology? There are some nice distributed element filters that start to work around that frequency.

BTW, I'm no filter guru, so don't expect answers from me. Just questions.

 

[deki]

I've managed to alter it closer towards what I want. I want a more flat passband, though I'm still getting these ripples even though my frequency transformations didn't account for them (I'm using the freq transformation technique to get the BPF). And if possible to get it to resemble more of an ideal BPF with steeper roll off. Or would I have to go into the microstrip realm?

 

[the_emi_guy]

Using discrete Ls and Cs at 2GHz is impractical, are you planning to transform your schematic into a stripline filter?

 

[meBigGuy]

Am I wrong about the need to drive it from a 50 ohm source?

 

[the_emi_guy]

Am I wrong about the need to drive it from a 50 ohm source?

No, you were right.

Looks like OP included source term in post #6 with Rser card.

 

[Jeff Rosenbury]

As I understand it, filter design is about trade offs. It's possible to build a flatter pass band, but at the cost of less roll off -- for an equal number of poles.

Also, there's something wrong with your simulation. You have actual gain in that bump at the top of your pass band. That should never happen in a passive network. Something about perpetual motion I think.

 

[LvW]

For a flat passband and - at the same time - a relatively large bandwidth it is best to use a series connection of a highpass (lower corner) and a lowpass (higher corner) . The roll-off characteristic is determined by the filter order of both sections. To avoid loading errors both sections should be isolated with a buffer amplifier in between.