Continue to Site

Problem with LoRa antenna trace

L

laszlo.penzes

Newbie
Newbie level 4
Joined
Mar 7, 2025
Messages
5
Helped
0
Reputation
0
Reaction score
0
Trophy points
1
Activity points
61
Hello everyone,

I have been struggling with my LoRa project for some time. I tired so many things without success, I feel that now I need help from someone...
I'm using LoRa-E5-HF in a tiny LiPo-powered device. The problem is that radio transmission is very poor, I can't get better RSSI than -90dBm and SNR 1. It works by a distance of a few metres, then the signal is lost.

The antenna trace can be seen at the bottom-left corner, highligted white. It leads from RF pin of LoRa module to a U.FL connector.

1741347666419.png

  • L1 is top signal layer with partial ground plane (red)
  • L2 is a ground plane, covering basically the whole board (except vias)
  • L3 is 3.3V power plane
  • L4 is bottom signal layer
The PCB was made by JLCPCB. I used one of their 4-layer controlled impedance stackups and calculated microstrip width based on that:
1741347878656.png


There is no ground plane around microstrip on L1 (so it is a microstrip, not a coplanar waveguide).
Cleareance around RF trace is ~2mm. There are no components around RF trace or U.FL connector on bottom side, and no trace crosses it on any layer.

I measured the RF trace characteristics with a NanoVNA:
  • Soldered a 50ohm resistor to the pads of LoRa-E5, between GND and RF pins. LoRa module is not present on the PCB.
  • Connected an antenna cable (ending in a U.FL plug) to NanoVNA and calibrated it.
  • Then connected NanoVNA through the U.FL connector on the PCB.
  • S11 logmag shows -25dB and according to Smith chart, the impedance is 45 +j2.5 Ohm at 873 MHz, which is acceptable
I tried with two types of antennas (on both the transmitter and receiver): a flat pach (sticker) and a ~4cm long whip antenna. Both have 3dBi gain and are optimized for LoRa frequencies.

Do you happen to have any idea on what might be wrong with my setup?

Thank you!
 
Sort by date Sort by votes
You didn't mention about conducted measurements.
What is the measured conducted output power (according datasheet, should be about +20dBm) and what is the conducted sensitivity (should be about -120dBm).
 
Upvote 0 Downvote
I unfortunately don't have tools for that. My hope was that someone would point out the obvious problem and I won't have to spend more money on buying a power meter and/or spectrum analyzer. Now I'm looking for some cost effective tools to carry out those tests.
 
Upvote 0 Downvote
I would use your VNA to measure both antennae * cables for s11,s22 and s21 at 1m gap. Set the Tx to a suitable level and wide video bandwidth.
Then use the Tx antenna on a LORA and attempt to capture the peak pulse power on any of the chirps for possible correlation from the same distance and use Friis loss attempt to estimate Tx levels. Similarily for RSSI from VNA Tx. Room reflections may interfere but don't risk VNA Rx overload with direct connections. Capturing the sweep rates of both LORA and the VNA will lead to random levels but using the peaks may be useful. Your Tx BW and sweep rates will also have some effect on RSSI threshold but not RSSI levels at a fixed distance.

What else can reduce your excessive path losss shown by the RSSI?

Can you scope Vdd ripple with a diff probe or tip/ring probe? Could this cause VCO Tx dither and reduce RSSI? PC audio scope may work..

Use a commercial LORA card & Antenna to compare results.
 
Last edited:
Upvote 0 Downvote
Unfortunately cannot do RF design without a minimum of instruments. You cannot assume that the conducted measurements are as should be. There are many factors that could make conducted parameters to not be as in the datasheet.

A cheap SA for about $150 gives +/- 1dB accuracy measuring output power. And who knows, maybe you will use it again for further projects.
 
Upvote 0 Downvote
vfone said:
Unfortunately cannot do RF design without a minimum of instruments. You cannot assume that the conducted measurements are as should be. There are many factors that could make conducted parameters to not be as in the datasheet.

A cheap SA for about $150 gives +/- 1dB accuracy measuring output power. And who knows, maybe you will use it again for further projects.
Click to expand...
I'm about to order a tinySA Ultra+ ZS-406. Do you guys think that this is sufficient? If you can recommend any other purchase in the same price range which would help to investigate the problem, please fire away!
Also, what other supplementary suff do I need for an SA? Some articles suggested that using an attenuator is a must, as a too high signal might damage the SA.

One quite cheap stuff which came across is the logarithmic power meter based on AD8317:
https://www.amazon.de/-/en/gp/product/B083W1NC27/ref=ox_sc_act_title_1?smid=A34J6P0EZGOP86&psc=1
Does it make sense to buy it or does a spectrum analyser cover all my needs?
 
Upvote 0 Downvote
Other than presumed in the thread title, I don't expect an "antenna trace" problem. Trace impedance is at least near to 50 ohms according to PCB parameters. It's more likely an antenna or software configuration problem. Using a pair of devices under test is possible for verfication, but no good way to debug the design. Spectrum analyzer measurements would be my first step to verify transmitter operation, checking output power, frequency, modulation and spurious emissions. It can be also used for qualitative check of radiated field strength in combination with an antenna of known antenna factor, e.g a lambda/2 dipole. Consider that attenuators (20 - 40 dB) are required for measurements at antenna output.
 
Upvote 1 Downvote
If the receiver cannot get lower signals than -90dBm (spec -120dB) you could have "RF desense".
For more information see the link below:

Hardware and Systems Engineering Design - RF Desense

Created; April/2020 Last Updated: July/2024 Introduction: Have you ever wondered why the connectivity of your electronic product can sometimes be unreliable or flaky, even when it's designed with a wireless module from a reputable vendor using the latest standards? While it might be tempting to
www.hwe.design

Reasons could be: digital noise (coupled conducted and/or radiated), layout, grounding, filtering, etc.

I do have a tinySA which is working fine. Is specified at +/-2dB accuracy but actually it has +/-1dB accuracy. Maybe I was lucky.
Sweep time is a bit slower for wide bandwidth, but for the money is fine..
 
Upvote 0 Downvote
I don't know yet about said tinySA, it looks to me like a kind of software defined radio (SDR). If so, it should have additional features like linking a demodulation software that can decode LoRa protocol, e.g. HackRF package.
 
Upvote 0 Downvote
vfone said:
Unfortunately cannot do RF design without a minimum of instruments. You cannot assume that the conducted measurements are as should be. There are many factors that could make conducted parameters to not be as in the datasheet.

A cheap SA for about $150 gives +/- 1dB accuracy measuring output power. And who knows, maybe you will use it again for further projects.
Click to expand...


The cheapest solution is a load resistor on a plug to a low C RF diode and small Cap to measure with a DC meter.


Aliexpress de-listed that SA for some (good) reason. UPDATE: It was my script blocker... Sorry.

Note the fake suspicious prices.
1741539143015.png

--- Updated ---

laszlo.penzes said:
I'm about to order a tinySA Ultra+ ZS-406. Do you guys think that this is sufficient? If you can recommend any other purchase in the same price range which would help to investigate the problem, please fire away!
Also, what other supplementary suff do I need for an SA? Some articles suggested that using an attenuator is a must, as a too high signal might damage the SA.

One quite cheap stuff which came across is the logarithmic power meter based on AD8317:
https://www.amazon.de/-/en/gp/product/B083W1NC27/ref=ox_sc_act_title_1?smid=A34J6P0EZGOP86&psc=1
Does it make sense to buy it or does a spectrum analyser cover all my needs?
Click to expand...
Did you try what I suggested ?

It is cheaper to buy a known good device (KGD as we Test Engineers call it) to compare (cheap on Ebay)

You might have a software bug :rolleyes:or test with a DMM using a matched load, diode + cap on a plug.
 
Last edited:
Upvote 0 Downvote
Yeah. There are many options to measure the RF power. You can even touch the output trace with your finger and feel the RF power heat. But this power meter is hard to calibrate. :)
$150 I paid for groceries few hours ago, so buying a 5.3GHz spectrum analyzer for the same money I think is totally worth.
 
Upvote 0 Downvote
vfone said:
Yeah. There are many options to measure the RF power. You can even touch the output trace with your finger and feel the RF power heat. But this power meter is hard to calibrate. :)
$150 I paid for groceries few hours ago, so buying a 5.3GHz spectrum analyzer for the same money I think is totally worth.
Click to expand...
2 bags worth,,... don't forget to get a pair of commercial PCB's that work.
 
Upvote 0 Downvote
Thank you for all the help I got from you, it was truly life-saving!
As multiple people suggested (thank you, FvM and Tony!), it was indeed a sw bug, not PCB design flaw. On sender side, RF switch table has to be initialized in a slightly different way depending on the actual Lora module model number, so E5-HF and E5-LE-HF models are slightly different.

When initializing the radio module, the respective line has to be used:

C++: 
static const Module::RfSwitchMode_t rfswitch_table[] = {
  {STM32WLx::MODE_IDLE,  {LOW,  LOW}},
  {STM32WLx::MODE_RX,    {HIGH, LOW}},
  {STM32WLx::MODE_TX_HP, {LOW, HIGH}},  // for LoRa-E5-HF
//  {STM32WLx::MODE_TX_LP, {HIGH, HIGH}},   // for LoRa-E5-LE-HF
  END_OF_MODE_TABLE,
};

MODE_TX_HP (High Power) setting instructs the module to use another, amplified RF path.
The range is way better now, 15 metres indoors; RSSI: -80, SNR: 8 at 14 dBm output power.
 
Upvote 0 Downvote
laszlo.penzes said:
I'm about to order a tinySA Ultra+ ZS-406. Do you guys think that this is sufficient? If
Click to expand...

I double checked with a friend who uses this exact device, and he is quite happy with accuracy and capabilities.

But don't forget an attenuator, I would recommend 30 dB with at least 1W power rating.
 
Upvote 0 Downvote


Write your reply...

LaTeX Commands Quick-Menu:

Similar threads

Part and Inventory Search

Welcome to EDABoard.com

Sponsor

Back
Top