[SOLVED] Basic LC High Pass filter simulation in ADS is not the same with real PCB

[new_rf_engineer]

Hi,

I am new in RF electronics. I am learning ADS and other theoretical stuff about RF. I thought that i can create an Elliptic LC high pass filter on real PCB board to exercise. Before i create the board, firstly, i have chosen a high pass cut off frequency from RF-tools website. The specs, circuit schematic and graph can be seen in Fig. 1, Fig. 2, and Fig. 3 respectively. The cut off frequency which i have chosen is 1050 MHz. Then, to see the same figure in ADS, I simulated the fig2 schematic in ADS. In the ADS simulation, i used the standard component values for capacitors and inductors by rounding the exact values from RF Tools schematic. This circuit schematic and graph can be seen in Fig. 4 and Fig. 5. respectively. In the next step, Because I am going to use my real components on board, I had to make a realistic simulation. I have the capacitors and inductors with SMD 603. I added them to ADS as C_pad1 and L_pad1. Also, I attached the microstrip lines to the ADS.

Finally, the microstrip lines and 3d components(SMD 603 capacitor and inductor) are ready and same with the board. This circuit schematic can be seen in Fig. 6 and corresponding graph in Fig. 7. As it can be seen in Fig7, the cut off frequency is not the same which i selected at first and i think that this is due to the lines. The graph in fig.7 is my expectation on Real PCB.

I created the same circuit with the same lines on PCB. The lines and components are the same with the ADS in PCB. After i observe the graph of this PCB in spectrum analyzer, i got the graph in fig. 8. From now on, I told you the step by step implementation of a high pass LC filter by me. The graph is totally different and cut off frequency is about 954 MHz instead of 1050 Mhz which i have chosen at the first. PCB is very small and it contains only the necessary components in the circuit from fig. 6. The PCB has FR-4 material.

Why ads simulation and pcb implementation is too different ? Do i make something wrong in ADS? What points do i need to check to distinguish failure?

Thanks in advance

 

[crutschow]

Those capacitances are very small and are likely significantly affected by parasitic capacitance on a real PCB.

 

[new_rf_engineer]

Those capacitances are very small and are likely significantly affected by parasitic capacitance on a real PCB.

Rf tools choose the right values for a specific cut off frequency, I do not. If capacitances are the main cause of these parasitic effect and cut off frequency shift, It means that i have to design a different circuit for the same cut off frequency. Do you know any other reliable tools to design a LC filter with the high capacitances ?

 

[BigBoss]

Because you newbies are doing ( like you ) the same mistakes when you design something in RF.
You're using ideal elements and then you expect the same response from a implemented pcb.
The real world is not ideal.
You have to consider components' tolerances too. Layout should be simulated with components and Monte Carlo statistical analysis must be done to predict the filter response. Pareto analysis will give you the sensitivity and it'd better to do.

 

[andre_luis]

It means that i have to design a different circuit for the same cut off frequency. Do you know any other reliable tools to design a LC filter ...

As general rule, we have to add extra devices in order to surpass intrinsic unwanted behaviours, a tradeoff on final stability vs. performance, the guarantee that things will work predictably. Doing this, the simulation will converge to similar results regardless of the tool you choose.

 

[BradtheRad]

I made a simulation similar to yours. The scope trace results in a high-pass filter admitting frequencies above 1.2 GHz.

The 1 ohm resistors are added to provide some amount of realism.

 

[volker@muehlhaus]

Because you newbies are doing ( like you ) the same mistakes when you design something in RF.
You're using ideal elements and then you expect the same response from a implemented pcb.
The real world is not ideal.
You have to consider components' tolerances too.

100% agreed. It is not a tool issue, you just ignored the importance of parasitics in components and the importance of layout/routing/interconnect.

This is very typical for unexperienced users who do their first filter design, impressed by the nice curves with ideal elements. For better results, use component libraries based on real data (Murata etc have ADS component libraries, Modelithics offers substrate-scalable libraries for many vendors) and learn about circuit-EM-cosimulation. This is the only reliable way to get accurate results for demanding filter specs.