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.
I was mistaken on the RAK3372. The WisBlock core board has external IPEX and I thought that’s the one we use. But I confirm that (1) it uses the built-in IPEX connector on RAK3172 and (2) there is a copper plane under it.
It would be useful to see the top copper layer as a ground plane there will have much stronger coupling to the module.
I am really trying to understand as much as possible about the potential issue as we are sitting on several thousand dollars of boards that may get scrapped.
Unless your measuring the TX RF power with an RF power meter or a spectrum analyzer your assumption are rather presumptive. If your setting the TX RF power to +22 dBm then you should see this on the instrument. You can also conclude that if the TX chain is as expected then the RX chain will work as expected.
Another way without instruments is to use the path loss equation and plugging-in the variable, such as antenna gain (Gt, Gr), distance (D), TX RF power.
20 LOG(D) + 20 LOG(f) + 20 LOG (4 PI / c) - Gt - Gr.
If the RSSI reading is what the equation predicts then you know it’s 100% its working.
This graphs shows you the expected RSSI.
Antenna gain = 2dBi
F = 915MHz
TX PWR = +22 dBm
I agree that my preliminary tests are presumptive. Their purpose was a first-pass evaluation of some antennas.
But anything resembling the 40dB reduction in transmit power that was originally reported should have been clearly evident, presuming the NUCLEO board wasn’t similarly affected.
It would have been useful to test my board with the NUCLEO’s antenna as well - something that will happen when I find the appropriate pigtail.
At this point, I am satisfied that there is not an issue with the stock of modules I have on hand.
However, this doesn’t discount that @frank_r actually has a problem.
I would like to pose a thought which might lead to another conclusion. I won’t defend this thought as I simply don’t know.
There are two versions of the RAK3172 in respect to range (high freq., low freq.). The RUI3 firmware determines which version based on whether the pin PB12 is tied high or low.
I don’t use RUI3 and I can only guess what is happening under the hood with the RAK3172 itself.
If the low frequency module version uses the “low power” Tx pin (max +14dBm) while the high frequency version uses the “high power” Tx pin on the STM32WL chip, then this scenario may play out…
Pin 27 of the module is PB12. If this were accidentally tied to ground, this might force the RUI3 firmware to treat the module as the wrong type and not enable the PA for the “high power” Tx pin.
I avoid using this pin completely and leave it floating in my design.
However, Frank’s original PCB snippet shows a net connected to pin 27.
Is it possible that this connection to pin 27/PB12 is overriding the hardware type to the firmware and resulting in incorrect transmitter configuration?
The idea could be complete rubbish.
Feel free to ignore. As I said - I won’t spend time defending the suggestion.
I was the having the same TX RF power on two batches 2 years apart. It seems unlike the RAK would have changed the layout of the RAK3172 module since introduction?
Yes indeed I was having some issue with PB12 (pin 27) and tied it high with a jumper. Some modules worked when left floating and some didn’t. Really I don’t understand the point of this pin. The firmware should recognize what frequency is sent and adjust its itself accordingly.
Another user reported a problem with TX RF power, I’m not the only one.
When I first detected this problem, the PCB had copper removed under pin 12 except the bottom layer 61 mils away (as recommended in the spec). By removing the module and reinstalled a new module without pin 12 connected the TX RF power returned to normal. On another board with 2 RAK3172 modules I completely removed all copper under all the layers under the RF output section. The board never had any problems.
WARNING
