Direct mount antenna bad idea?

Gateways
1

IMPORTANT BEFORE POSTING:

Summary
Tuned antenna performs worse than untuned*.*

Logs
Please paste any error logs here and use the text format (< / >) option (No screenshots!)

Hardware
Older RAK7249, untuned 5.8 860-915 antenna vs McGill 5.8 915 antenna

I have a older RAK 7249 mounted externally 15 m high. I have the RAK 5.8 untuned 860 - 930 rod antenna screwed directly on top of the metal case. As I am in North America I thought I’d try a tuned antenna 5.8 antenna from McGill microwave to improve my reception in a heavily treed environment.

After several days of testing the McGill tuned performs worse than the untuned antenna. I have uploaded all the moment by moment data to several AI models and the answer is that the metal box (ground plane) is messing with the highly sensitive tuned antenna. Their answer is to only mount a tuned rod antenna to a length of LLM400 cable first , never to the box directly.

Any thoughts on the issue of direct mount? I am NOT looking for answers about raising the antenna higher, or doughnut effects. Just the issue of whether direct mounting is a bad idea.

Here is Chat GPT solution " If the antenna is mounted directly to the RAK7249 (or similar RAK outdoor gateway) without a pole, we are looking at a classic case of Near-Field Interference and Chassis Detuning.

When an antenna is screwed directly into the N-type bulkhead of a metal gateway enclosure, the gateway’s own body becomes part of the antenna system. Here is why your “Untuned” antenna is winning and how the “Tuned” one is being sabotaged:

1. The Chassis as a “Ground Plane”

A LoRaWAN antenna is designed to radiate into “free space.” When you mount it directly to the gateway:

  • The Metal Box Effect: The metal enclosure of the RAK gateway acts as a massive, asymmetrical ground plane. This pulls the “resonant frequency” of the antenna.

  • The “Tuned” vs. “Untuned” Performance: Because the McGill antenna is so precisely tuned to 915 MHz, this shift moves its “sweet spot” to perhaps 905 MHz or 925 MHz. It is now “de-tuned” for the actual LoRaWAN frequencies. The “Untuned” antenna has a much broader resonance (lower Q), so when its frequency shifts, it is still “good enough” to cover 915 MHz."

2

I’ve spent the last 6 months mapping out the breadth & depth of antenna design, propagation and deployment, I’d consider it to be a 10,000 hours endeavour and the rest. Everything is a compromise. Modelling works until it doesn’t.

As AI is fed the body of knowledge called web pages and anyone can publish, which they do, and because it is a dark art and radio amateurs like to dabble, the aggregate nonsense is going to dilute the actual facts quite considerably, so anything AI has to offer can only provoke some thought. Or despair.

At least AI did note that the ground plane of the heavy metal box may be messing with the McGill antenna, what we don’t know is if the RAK antenna is tuned for being on the box.

I have a number of RAK 434 & 868 antennas and none of them have a ground plane at the top of the 6m Aluminium mast I use. They are connected via WestFlex WF-103 (close in spec to LMR400 but 1/3rd the price but still expensive). They seem to work just fine. But if I’ve got a null spot, I’d install a supplementary gateway.

More general comments:

The first thing for antennas is that every situation is unique.

The shape of the terrain and soil composition figure up the top of the list for even the most basic of antenna setups. If you are in a wooded area, the type of trees, how wet the leaves are, both internally & externally, how close together they are will also be a major factor.

So all you can do is test using a tool, he says ironically, like TTN Mapper.

The second thing to consider is that RAK’s kit is well regarded, the antennas they bundle are not a PoS ( :poop:) but for sure, have to fit a balance between construction, efficacy and durability, so probably not the absolute best.

And then there are the frequencies in use: 902.3 to 914.9 MHz for the 64 channels and 923.3 to 927.5 MHz for downlinks. So the antenna HAS to be OK across that range. The 915 bit is middle ground and as such is a huge red herring (:blowfish:), aka over simplification, aka misleading aka shorthand for the ISM band of 902–928 MHz.

I can’t see the particular antenna from McGill, but they may be able to help you out with the right specific product. Given some of the data sheets don’t have the polar diagrams on, I’m not sure what to note. However, if you can afford some decent connectors plus a length of LMR400, it’s worth a try if the deployment is worth it.

Remember, you need to set the gateway for the dBi of the antenna so you don’t exceed local legal limits. Sometimes a big honking antenna is standing still or going backwards.

@Jeff-UK will go in to the details, but woods & UHF aren’t a great mix, other discussions over the years on how to do LW in woods (if a device transmits in the woods but no gateway hears, does it actually use any power?) are available by searching (top right).

TL;DR: Do some mapping then install more gateways.

1 Like
3

Any decent antenna available with gateways that allow pole mounting will be a dipole antenna. These create a virtual ground plane between the upper and lower pole and do not require an actual ground plane (which would be hard to come by on when mounting on a pole).

Not knowing the exact model of the new antenna you obtained it is hard to determine if it’s a dipole or monopole but probably it is a dipole. Mounting a dipole on top of a metal box shouldn’t be an issue if the box isn’t a couple of feet square. Antennas meant for car use will require a ground plane to be present and in that case the gateway will be too small.

In the 11 years I’m working with LoRaWAN I never found a third party antenna that improved upon the (decent) antennas available from the gateway vendor. Yes, there have been some gateways with token antennas, mostly indoor gateways, however these are easily recognizable as being sub standard (a 5 cm antenna included with a gateway is not the same quality as the 50 dollar one offered by the same vendor but it allows getting started without damaging the gateway)

2 Likes
4

Hi, my 2 cents:

Both 5.8 860-915 antenna and McGill 5.8 915 antenna are collineairs and their radiation patterns are independant from the enclosure it is mounted to. Because the enclosure is under the antenna, influence can be neglected. For as far as I know there is no “The Metal Box Effect” in the installation you describe.

I have a strong believe that the words “Tuned” and “Untuned” are incorrectly being used. Any antenna is resonant or tuned to the frequency it was made for. therefore the words “Verified” and “Unverified” are better in this context. I believe that McGill verifies the antennas they have manufactured in China. That sure helps gain confidence in the product they sell.

You may compare this with: why commercial guys sell antennas with a gain in dBi and RF engineers use dBd.

In all my years as RF engineer I have not observed significant difference between an antenna on the top of a mast or a antenna on a enclosure on the top of the mast. Here i explicitly state that no objects (antennas) are in close proximity to the antenna. This however may be the case with the RAK7249 enclosure.

afbeelding
afbeelding161×379 17.2 KB

3 Likes
5

Thank you for your expert opinion. I used Gemini for their AI analysis. I found it to be very good at looking at several hundred data points and putting into a chart the degradation of the signal when when moving to the McGill “tuned” antenna. This was backed up by the increasing frequency of missing packets as I watched in the real world what happened. So I like AI for arranging data, not so much for explaining why.

Which leaves me mystified why, with everything else being equal, the McGill performs less well than the RAK. Maybe it’s because the RAK gateway and antenna was bought as a system with an antenna matched to the gateway by a single manufacturer and as opposed to a system pieced together from disparate parts.

PS I am going to add something that I didn’t want to mention in the original post as I didn’t want to add factors that I didn’t think relevant. I am using Milesight EM400 MUD sensors to monitor waste bins in a proof of concept for our town resource management department. Our 1m x 1.5m x 1.5m bins are giant above ground fully enclosed metal containers (we have a bear foraging issue). So a Faraday cage. My gateway is about 15m above the ground and the bins are as much as 1.36k away in a treed, low density environment. The sensors report hourly. They have ADR turned off and are set to SF 10. They are fastened to the underside of the lid on a plate. The plate is attached to the lid with 6 strong rare earth magnets that are at least 10cm away from the sensor. I get transmissions 90% of the time which I think is an astounding success given the robustness of the metal containers. It’s good enough for a workable system.

I didn’t want to mention this as I thought people would focus on a solution of adding gateways etc. instead of focusing on my original question. I thought spending a few more dollars on a so-called tuned antenna might get me to 95% success.

For your amusement see the video at the bottom of the web link here. The bin pictured is a small version of what we use. https://www.haulall.com/trashcans

6

Did you verify that the McGill antenna connectors are matching to the gateway?

7

They are exactly the same as the RAK and are securely fastened.

8

Hi,

It is a bit unclear to me what you have compared? Could you give some further explanation before I mis interpret and we add more confusion to the equation?

Great story, I do like watching a good test and great, skilled and dedicated testers. :slight_smile:

1 Like
9

Intrigued by your case I studied again your introduction to the thread and stumbled over this:

I concluded myself that I am a bad reader too.

Bringing electronics close to the antenna is not always advisable, it largely depends on the quality of the electronics design regarding EMC/EMF compliance. Although I have had positive experiences with RAK gateways, I cannot comment on their EMC/EMF performance specifically. However, near-field interference could be a contributing factor to your degraded link performance. It can desence your gateway receiver.

Without a conclusive root cause analysis, which may not be straightforward in your case, a practical quick fix is to lower the gateway by a few meters, keep the antenna at the top of the mast, and use several meters of good-quality coaxial cable. This approach may yield satisfactory results and help identify whether the issue originates from equipment at the top of the mast. While it may not fully resolve the underlying problem, it is a reasonable step toward troubleshooting.

1 Like
10

I tested by downloading 70 minutes of gateway traffic log for 1 antenna and then 70 minutes for the 2nd and then asked ai to compare them based on the rssi snr per device. There are about 15 devices within range. They report at 20-60 minute intervals.

I have ordered some LMR400 cable and will try your advice. After seeing the video you’ll now understand why I prefer to focus my efforts 15m in the air rather than at ground level. :grinning_face:

11

We are all really starting to lose our collective skills of evaluating data..

12

Hmmm, what I did first is use my signal value charts where I could see the decline in efficiency. I also reviewed my sensor dashboard where the end results were obvious. Only then did I turn to ai to confirm what I suspected. It was able to chart hundreds of data points faster than me. So in my opinion it’s a useful tool. My building a spreadsheet to tease out a trend is subject to the likelihood I’ll introduce some errors or typos. Using ai eliminates some of drudgery of assembling data which means we can get to the point where we can use our collective ability to interpret the data more quickly

Where ai seems to fall short is in explaining why the decline happened. And that’s when I turned here for it’s wisdom.

13

Having it make a chart for your to evaluate is something different indeed :wink: that sure can be useful

14

Not all of us.

If you have a data set and import it in to a spreadsheet, it can chart any number of data points much faster at a significant lower environmental cost than anything else. You don’t have to type in the data, so no typos.

These tools look like they can help, but they are working in mysterious ways. Charting is fairly simple but if AI can’t explain the results, I’m seeing a trend where people assume that it is beyond explanation. Or the application of the scientific method - come up with a hypothesis, test it, review results.

In this case, as you are in a non-typical environment, some investment is indicated, as I, @pe1mew and yourself have mentioned, in trying ‘stuff’ out.

So even if you see improved results by moving the antenna away from the case, it may be worth trying at tree top height.

With regard to the units being inside the bins, any chance the manufacturers can put some form of case on the outside? Alternatively, look at BLE on the inside with a receiver that relays via LW, I find BLE works through fairly dense materials. Or an encapsulated BLE sensor on the outside. If it does get knocked off, eaten or damaged, they are less expensive that a LW device.

15

Using live traffic is a proven method. I have used a single node sending data and the gateways in my area to compare a unknown antenna (RAK 5,8 dBi if I remember well) with a known one (dipole). Using this method I was able to verify the gain of the RAK antenna within 0,1 dB. I have posted this years ago on this forum.

Its a flash-back in history but this post explains how I verified the antenna then: The BIG and SMALL ANTENNA topic part 2 - #267 by pe1mew

1 Like
16

In the last few days I’ve separated the “untuned” RAK antenna from the gateway and raised it 3 m higher than the gateway with a LMR400 6 meter cable. The antenna now sits about 22 m above the ground. It now has a slightly better line of sight through trees and over rooftops to 3 sensors sitting within metal waste bins at distances as much as 1.4 k away.

RSSI improved slightly or not at all. However SNR improved by as much as 2 - 5 db .This is just enough to almost eliminate 10-20% of the failed reception of hourly reports.

At some point I may try the “tuned” antenna but for now the simple improvement in line of sight has brought good results. This change may not resolve the idea that a direct mount antenna is any worse than a separated antenna because the improvement is likely attributable to the gain in height.

17

Hi,

This reminds me that when it comes to through-forest communication I have observed that a transmitter should be below 1 meter or above aproximately 6 meters to prevent “blocking” by bushes and low vegitation. I observed this at a HAM-camp where I delivered WiFi access to a large group of campers.