Via design for 24GHz with JLPCB JLC7628 Stackup

[veeti]

Hey

I'm trying to homebrew my first 24GHz design, is there someone familiar with microwave office - i would need help to confirm my via design. This is my first time when I go so high with freq that i think i should allso give a thought with via dimensions.

Screenshot for Microwave office
https://drive.google.com/file/d/1rrPvnnGdo2i_PZYPcYvvoviSJcWO2Df2/view?usp=sharing

PCB stack dimensions
https://cart.jlcpcb.com/impedance

Mirowave office tutorial for Vias
https://awrcorp.com/download/faq/english/docs/Elements/viam.htm

BR: veeti

[Marsupilami]

Hey veeti,

Your model configuration looks correct from the topology perspective assuming that you feed the line via microstrip and not coplanar waveguide but the top and bottom capture pads look very chunky. I have never used this model and honestly I'm a bit baffled by how is it even supposed to work in practice. Normally your return current should flow beneath the signal trace  and have ground vias around to provide a path for it between the two inner gnd planes. These gnd vias placed close enough to the center one would provide some extra capacitance to counteract the inductance from the long and narrow signal via thus alleviating the need for the chungus pads on the top and bottom.
Also run the simulation in a wide enough band to make sure you're not accidentally catching a resonance peak.



Are you sure though that the FR4 dielectric will do whatever you want it at 24GHz? Dielectric constant can be really far from 4.6 not to mention losses.

[neilhao]

Hi,

According to my experience with infineon 24ghz radar and FR4 board. I did not successfully design the VIA for 24GHz with acceptably insertion loss. Thus, I finally did not use VIA.
https://uniteng.com/index.php/2020/04/24/infineon-bgt24ltr11n16-24ghz-radar-mmic-impedance-matching/

The attachment was the via structure used. According to the VNA measurement, for 4layers FR4 board with L1 to L4 VIA, the VIA could work up to 6GHz with less than 0.5dB insertion loss. However, the EM simulation model could work up to 18GHz. I GUESS the reasons for the mismatch were:
Fiber Weave Effect
Er were different along XY plane and YZ/XZ plane. I should not use the Er for XY plane for VIA design but measure the Er of YZ/XZ plane for the VIA.

This is just my result, you may be able to find a better design.

[veeti]

Hey

Thank you for both for your answer - i have to be honest, i totally forgot that FR4 material - i just went straight on words impedance matched pcb ant 4.6

I allso doubt that i would find any "cheap" fab for 24ghz pcb - so i may put this design to side and start new one with 6 ghz - I'm trying to do FMCW radar, its just somekind mental milestone for me to make radar at microwave freq.

 

[Marsupilami]

In case you want to try a more suitable material I recommend Eurocircuits. They have a pooling service for RF laminates. The Isola I-Tera in this configuration is a lot more affordable then fully custom RF stackups. It's still going to be more expensive than JLPCB.
FR4 might work to an extent just make sure that when you simulate, run variations over a wide range of Er values. E.g. between 4 to 5 maybe and check that you hit your target metrics regardless. Another issue with FR4 the higher dissipation factor, put that in the simulator too as it will increase your insertion loss significantly.
Another thing you can try with FR4 is to use a coplanar waveguide instead of the microstrip and a thicker board. This way the field strength and overall contribution will be stronger above the copper in air and lesser in the sketchy FR4 dielectric. The ground gap size gives you an extra parameter to tune your line impedance other than the dielectric thickness and the line width.

The via, I would EM simulate that. The planar simulator should be good for that in AWR or the full 3D if you want to go fancy. Alternatively you can experiment with the VIAMC element to simulate coupling between your signal via and surrounding gnd vias but it's really hard to verify if you set it up good.

In general if it's just for random experimentation I recommend the 6GHz circuit. It's a lot more forgiving and gives you a lot more opportunity to learn and understand what exactly is happening. It's definitely not without its challenges but compared to 24GHz it makes a big difference.

HTH.
Good luck!