I am using the 915 MHz RAK3172 on a 4 layer 61 mil FR4 PCB. I’m using the IPEX and not the RF pin (12)
The 4th layer (bottom layer) GND plane was not removed, the inner GND planes are removed. This has a dramatic affect on RF TX power and RX sensitivity. According to the capacitance calculator the stray capacitance is ~0.1pf. The pad size of 0.7mm x 2mm with a distance of 61 mil (1.6 mm) I’m seeing the RX sensitivity drop by 35 dB and the transmit power drop by 40 dB, this is very dramatic.
I resolved the issue by removing the part and not soldering the RF pin 12.
My question is why is the RF pin 12 so very sensitive to having a GND plane under it?
This is a warning in the RAK3172 specification:
WARNING
When using RF pin for antenna connection and not the IPEX connector variant, make sure there is no ground plane (in all layers of the PCB) under the RF trace path to eliminate the possible effects of unwanted stray capacitance which can cause degradation of the RF signal levels.
This warning should also that any ground plane under the RF pin 12 should be removed. Whether using the RF pin 12 or not. Indeed the ground plane clearance should be bigger than pin 12.
The warning should be updated ASAP.
I would suggest looking at the design of the RF output to reduce the sensitivity to RF pin 12 to the GND plane. The stray capacitance being so small there must other effects going on?
Can you please share me on what section you find that warning in the documentation?
It was already updated since ground plane optimization for non-IPEX variant of RAK3172 is dependent on other factors aside on stray capacitance. For example, on monopole antenna, the ground plane is really needed.
In your situation (of soldering the RF pin), that pin is isolated to the RF path via 0201 resistor in RAK3172 IPEX variant so it is really strange that it has effect on the IPEX connected antenna performance.
If you can share me some image of your setup and other relevant info, I can share it to our RF engineer so he can have a look.
I have attached RAK3172 Sept. 2023 spec sheet with page 4 about the warning the ground plane underRF pin pin 12.
I’m not using the RF pin 12. My U.FL cable connects directly to the IPEX on the module.
Even thought I have removed the GND plane (4 layer board) on the inner layers beneath pin 12, I left a ground plane on the bottom layer 61 mil away beneath the pin. Yet I’m seeing 40 dB of RF attenuation. I solved the problem by removing the RAK3172 and soldered a new part and not soldering pin 12 to the PCB.
What I suggest when using the IPEX on the RAK3172 nodule, that all ground planes be totally removed on all layers around pin 12. Indeed i would also suggest that the ground plane be removed beneath the rF section of the module around the IPEX connector, an area of 300 by 200 mils. I my design I removed pin 12 from the pad stack.
See attached layout
I am very keen to know what RAKWireless has determined on this report - I have a narrow time window to make changes to our own PCB using the same module.
If this has been determined to be an issue, we may have to retest an external antenna design that is being finished at this moment.
We do not have a dev board that uses RAK3172 with IPEX connector and with ground plane so we cannot validate. It is really strange issue and actually this is the first time this is reported to us. In my opinion, removing the ground plane on the RF pinout will not cause harm in overall design so adding that provision can be practical.
I am about to start testing for this issue to see if it can be confirmed.
I remain baffled that it might be an issue. From looking at the RAK3172 under magnification, it appears to me that the RF trace to pin 12 comes to a pad next to the IPEX connector’s centre pin pad.
I would presume that the difference between the two SKUs is the installation of either the IPEX connector or else a bridge resistor across to the Pin 12 RF trace.
This RF trace appears to be a “coplanar waveguide with ground” design due to the vias either side of the trace.
Assuming that these vias link directly with a ground plane within the RAK3172, this should effectively shield the RF trace from any host PCB ground plane in this area - up to the point the signal hits the pad for Pin 12 - after which it becomes a problem for the host PCB designer to handle RF transmission from that pad.
As the Pin 12 RF trace is not connected, it should have zero impact.
Now, if for some reason, the bridging resistor was installed as well as the IPEX connector, then this would create a stub that would interfere with the RF path. This might be possible, but it seems to me that the footprints for the resistor and the connector overlap so would be both difficult to do as well as visually obvious.
In short, I can’t see a reason why the reported problem might be occurring short of an actual product manufacturing issue.
Nevertheless, I am about to start testing a custom antenna and will be checking for the reported issue.
The modules I will be testing are roughly 18 months old so the fact they work (assuming they do so) doesn’t discount a more recent issue.
@carlrowan Have there been any board revisions to the RAK3172 module in the past year or two?
I am checking now with our hardware team if we have any significant change on the RAK3172 design in the past year.
As for the ground plane, could it be also small parasitic capacitance from the plane that affects the RF path? As far as I know @frank_r deeply work on this issue and found that only removing the plane under the RF section he got improved signal levels. But it can be that this findings is case to case basis as well.