What keeps frying my RF switches?

Question

Short version: What might be breaking the first sensitive component connected to an ungrounded 14cm UHF dipole antenna mounted 30' above the ground? Could it be related to static electricity?

---

I have designed a built a few simple weather stations that I have installed in my region. The system consists of an atmega MCU that monitors wind and temperature, and sends a message over LoRa every few seconds to a receiver at my house.

All good, except after a few months in the field the signal cuts out. I take a replacement out and bring the broken unit home to find out what went wrong. What I have found is that the most of the circuitry is still working fine, but the radio has stopped transmitting. Further investigation revealed that RF switch (PE4259) of the broken units is inserting 30dB+ of attenuation (This was determined by soldering some thin coax cable at various places along the radio signal path and measuring with a spectrum analyser).

The units are mounted 30' above the ground, solar powered, transmitting +22dBm at 424.8MHz (w/ ham license).

The radio units I use are small devices based on the SX1268 chip, and encompass all of the hard RF stuff into a simple package that I just attach to my board: http://www.dorji.com/products-detail.php?ProId=64. I have opened them up on broken units (necessary to troubleshoot) and they pretty much follow the reference design from Semtech:

The signal path is: Tranceiver RFO -> Filter & Impedance matching -> RF Switch -> More impedance matching -> U.fl connector -> Coax -> SMA Connector -> Coax -> Antenna.

The antenna is a home-made dipole antenna from ~13cm long aluminium bar on the end of a plastic arm. It is mounted about 1' below sensor units, which are at the top of the pole (See this image of an assembled unit, not yet mounted on a pole).

The station has no connection to any sort of real ground - the entire thing is floating.

While I am not completely familiar with the internals of the radio unit, I assume the manufacturer has done a decent job of impedance matching along the signal path. I have tested the part of the signal path that I am responsible for (from where the signal leaves the radio units to the antenna) and measured a SWR of 2.2 - not ideal, but low enough that I wouldn't expect it to damage the components.

So - getting back to the problem - I am at a bit of a loss as to what is causing the RF switch to fail, and how to prevent it. A recent thought that occurred to me is that the switch is the first 'sensitive' component any signal from the antenna will see. Perhaps this is caused by large voltages from the antenna? This article / video and its comments have me wondering if it may be caused by static electricity developing on the antenna. The stations are mounted at ridge-top locations, sometimes in snow, sometimes in cloud, sometimes in dust. I assume that lightning is not to blame because I would expect much more significant and widespread damage.

If it is static, what would be a way I could confirm that? What would be a suitable fix? A sufficiently large inductor or resistor between the antenna elements?

Sorry about the rambling details, but my problem here is I'm not even sure if I'm looking at the right factors.

Answer 1

There are two ways you can kill that RF switch:

1. Too much input power. This seems unlikely, as you would need to exceed 33dBm (2 watts) to damage it that way. This would be an issue with the environment and out of your control. But if your weather station is able to pick up more than 2W of 933MHz (I assume from your dipole length) RF, something has gone terribly wrong near the weather station. Perhaps some mad scientist is frying the entire area with space-based RF weaponry (933MHz because even mad scientists still fear the FCC and stick to ISM bands)

2. Too much voltage blows a hole through the thin oxide layer of the CMOS die. This kills the switch. Lightning on the other hand would vaporize the chip (and probably some capacitors as well) entirely so the existence of something to even test rules that out.

The answer is, of course, number 2 but with the only remaining mechanism left to cause it: a build up of static electricity. No confirmation needed, it is the only plausible mechanism that could damage that RF switch.

Your schematic appears to be missing a bleeder resistor between the SMA center pin (from the dipole antenna) and ground. Pollutants and other particulates like dust will get blown over the antenna and deposit charge over time. Factors like humidity, air quality, and the size of the antenna will all play a role in what sort of peak voltage might be seen and the time scales it might occur over.

I would note that the RF switch you're using only has 100V ESD withstand for the machine model, which is the model one would use for, say, a discharge originating from a dipole antenna rather than a meat resistor (finger/human body discharge model).

Fortunately, this is a very easy problem to fix.

Just put a 1MΩ resistor across the SMA connector's center pin and its ground shield. You want to provide a path for charge to dissipate slowly from the antenna so it never gets a chance to build up very much. 1MΩ is the standard 'rule of thumb' value for a bleeder resistor, it is usually enough of a path to discharge any buildup without meaningfully compromising the antenna's performance.

Answer 2

I have made two changes and the problem has not occurred again:

• as recommended, I put in a bleeder resistor between the two elements of the antenna.
• I connected the station ground to earth.

I do not know which of these was the necessary fix.

I realise I left out what may have been an important detail: most stations are mounted on metal poles. I have one station with the same design but mounted on a wooden 4x4 post. The wood mounted station had never blown the switch in 4 years (most stations lasted less than 6 months)

It’s a bit of conjecture, but i could imagine static charge building up on the station until it reached a voltage sufficient to breakdown the insulation between the antenna coax and the metal pole (the apparent weakest link), at which point the charge would flow to earth - only problem is the charge on the ungrounded antenna element must flow through the switch, which obviously didn’t like it.