Could not get flat response for microstrip

[ahelmi]

Hi all,
Hope someone can help me with this problem. I'm designing a simple microstrip but when I made an S21 measurement, I could not get a flat frequency response up to 10 GHz. Microstrip design and S21 measurement are attached. The microstrip dimensions are as below:

72 mil width
30 um thick copper
2 cm long
FR4 board (dielectric constant = 4.5)
1 mm board

any idea why we cant get a flat S21 for the microstrip and how to get a flat response?

thanks in advance

 

[vfone]

Should be better than what you get. Are you sure the VNA calibration is ok up to 12GHz?
Also I don't understand why the ground of the SMA connectors is isolated from the top ground.

 

[ahelmi]

the vector network analyzer was calibrated up to 20 GHz with the cables used in the S21 measurement.

hmm, about the ground being detached from the top plane, it was not because of any logical reasonreally. just put things without much thought. should the top plane be ground as well?

thanks for the help!

 

[VSWR]

In addition to vfone's posting above:

FR4 is unreliable for use above 3 GHz. How do you know that er is 4.5 @ 5-10 GHz for your FR4-board?

The horizontal distance between the microstrip transmission line and the top ground ground plane is very small, about 0.2 times the width. Should be >3 times the transmission line width to avoid coupling/mismatch.

 

[vfone]

The top shall be ground as well, and the distance between gnd and microstrip as was mention above, to avoid CPW mode.
Anyway, floating ground on the top side for sure is the reason of those resonances at high frequencies.
Is true, FR4 cannot go for very high frequencies, but whatever bad it is, cannot give great discontinues vs frequencies (what you get now). Generally the losses vs frequency on FR4 are monotonically.

 

[FvM]

When attaching an edge-mount PCB connector to a microstrip, the bottom ground leads have to be wired primarly (without any thermal relief). You didn't show their connection style. If the pads also have thin wires to the ground plane, the resonances would be pretty understandable.

The top pins should be soldered additionally, at least for stronger mechanical retention. As another point, if you have copper pour at the top side, it should be connected to the ground plane with a tight via grid to avoid resonances. Then a hybrid MS/CPW design would be no problem, apart from achieving the correct impedance. In your design, the impedance is decreased to about 45 ohms due to the low ground separation.

 

[ahelmi]

thank you for all your replies.
the bottom pad of the SMAs do not have relief, hence are soldered directly to ground plane

i will redesign the microstrip and test it again. hopefully it'll be better this time round.

many thanks again for all the helpful comments. appreciate it