# RF Via fences/stitching spacing

I'm working on a 4-layer PCB with a U-Blox module and I'm trying to calculate the space between the fencing vias next to the Antenna trace and for the stitching vias.

According to the datasheet we have the following possible frequencies:

The case which would cause the closest vias is the last one.

According to the calculations suggested here https://electronics.stackexchange.com/a/42028 I should put them ideally with the following spacing: $$\frac{\lambda}{20} = \frac{C/2170Mhz}{20} = \frac{138.25mm}{20} = 6.9mm$$ but if we look at this image from the datasheet it seems way too much compared to what they did; consider that the side of the module (the white outline) shown in the above figure is 16mm, so according to calculation there should be a dot at the beginning and at the end of the track and that's it.

My best guess is they are basing the fence spacing on the Greatest Common Divisor for the mean value of each frequency reported in the table to cover each operational mode.

$$\lambda = \frac{300}{F\times\sqrt{\varepsilon_{R}}} = \frac{300}{2170Mhz\times2.097} = 60mm$$ now $\lambda/8$ is 7.5mm and that should be the necessary spacing for the ground stitching (Er = 4.4 for typical FR-4 PCB material).

To sum up:

1) How much space there should be between each fencing via in my case?

2) How much space there should be between each stitching via in my case?

3) Does the frequency actually relates to the fencing/stitching space, or placing the vias closer than the smallest $\lambda/20$ (for fencing) and $\lambda/8$ (for stitching) is all that matters?

4) Is all that stitching shown in figure really necessary from an RF point of view?

• Off-topic: For a moment I thought this was going to be about using a chain-link fence as an antenna. ("RF via fences") Mar 4, 2018 at 23:37

Spacing your vias by 1/20 wavelength is very unlikely to cause any problems.

Does the frequency actually relates to the fencing/stitching space, or placing the vias closer than the smallest λ/20 (for fencing) and λ/8 (for stitching) is all that matters?

The frequency and the wavelength are intimately related, so the operating frequency has just as much to do with the via spacing as the wavelength does.

Is all that stitching shown in figure really necessary from an RF point of view?

Probably not. But for a low-volume board like a demo board, adding a few extra vias is cheap insurance against having to re-do the layout in the unlikely event there's a problem caused by spacing the vias too widely.

• thanks for the answer, let me just clarify what I was trying to ask in the second point: If I want to make them closer to each other, should I consider that the 3rd harmonic is 8680mhz so I could place the dots every 1.725mm or placing them less than 6.9mm is all that matters? To my understanding your answer cover this affirmatively. Mar 6, 2018 at 21:17
• @Evil, 1. 8680 would be the 4th harmonic of 2170 MHz. 2. Do you expect the source to produce a large 4th harmonic component? Are you especially worried about emissions from this harmonic? Can you clean up the SMA footprint to give reasonable shielding at 8680 MHz? Mar 6, 2018 at 21:31
• Not really, just checking if I should base my calculations on harmonics to make the vias closer. Thanks for answering, I'm going to mark this as answer Mar 7, 2018 at 20:40
• Definitely if you're dealing with a digital signal, and not very cost-sensitive, then place the vias based on the 3rd or 5th harmonic of the bit rate. RF is not my area, but I'd expect if the module you're using is any good, it will not be producing much harmonic content. Mar 7, 2018 at 21:08