Dear ST,
I developed a RF PCB with STM32WL Microcontroller. The goal was to create RF Transceiver for 868 MHz transmission. Before I started developing PCB I read thoroughly through documents: DS13105, AN5407 and AN5457. As I am familiar with STM32 microcontrollers and its peripherals I only struggled with RF design, so let HP RF Path only be a subject to the following analysis. I developed a simple board on a following stackup:
(Prepreg’s permeability is 4,1, RF signals are on Top layer, solid GND layer was on layer L2. Traces have matched 50 Ohm impedance. Again, GND plane is solid without any discontinuities on this layer)
On PCB I put LC impedance matching filter for 50-Ohm U.FL connector. After it was assembled, I programmed it to emit a continous carrier frequency signal. I connected it through U.FL-BNC adapter and then connected it to IFR2398 spectrum analyzer. The result was a beautiful 20,5 dBm carrier frequency and a satisfactory range. The prototype was a success and we decided to develop it further.
In the next step I combined LP/HP/RX as it was developed on Nucleo board, you can see a schematic in attachments, and RF section only on picture below:
The schematic was developed in accordance to AN5457 guide. I used smith chart tool to obtain LC values. You can check my „simulation” for HP circuitry below, starting with components’ values selection:
and a resulting smith chart:
(DP8 is a final output impedance, it is near the center of a chart)
I also present you a layout on my PCB:
New prototype was designed on different stackup:
When device was assembled, I ran the same carrier test and the strength of the RF signal was only 9,5 dBm. It is way less than expected. I started searching for sources of power loss and here is what I have found:
a. a filter between RF Switch (U4) and antenna connector (CON4) is unnecessary, there was almost no power gain when I desoldered all components and soldered 0R 0402 resistors in series to maintain connection
b. I removed all filtering components except of first LC matching impedance filter. From PCB I removed C60, L18, C63, L19 and C64 and shorted L18 and L19 with 0R 0402 resistors. The power output increased from 9,5 dBm to 10,5 dBm. C61 is there to provide a required DC-Block for U4 RF switch, without it output power drops to around -25 dBm and according to Smith Chart, it introduces almost no impedance change.
c. When carrier signal is constantly generated, VDDPA voltage drops from 3,3V to 2,9V. In my opinion, series resistance of 74LVC1G3157 analog switch (U1) is too high for this design. R_on in range from 7 to 12 Ohms or other switch limitations are unacceptable. I desoldered this switch and shorted VDDPA to 3,3V and observed power gain from 10,5 dBm to 13,,5 dBm. (In case of any questions – yes, 3,3V is stable, no significant voltage drop is observed when carrier is generated)
d. Knowing that analog switch has high series resistance I desoldered U2 switch and shorted VR_PA to L13 (HP_PWR) inductor. This resulted in further increase of output power from 13,5 to 15 dBm
e. I also desoldered RF switch ans shorted OUT_HP with ANT_FLT with a kynar wire. I know that the short may introduce additional impedance discontinuity and signal reflection, but with Spectrum Analyzer I observed power output increase from 15 dBm to 15,5 dBm.
I modified the HP RF path to following circuit:
line represents short, „X” represents removed component.
So in the end, Prototype 1 and Prototype 2 have similar RF circuitry, yet there is a 5 dB power difference between designs. There are 2 major layout differences between prototypes:
A. Prototype 2 has all metal layers removed below antenna connector, while Prototype 1 has GND plane on L2 layer, below the antenna connector.
B. Prototype 2 has RF paths impedance equal to 56 Ohms – 6 Ohm mismatch.
Although 56 Ohm impedance mismatch introduces low to little reflection losses the lack of GND layer below antenna switch may be concerning. My questions are:
1. Can you suggest any other Analog Switch IC for U1, U2? Is there any you can recommend?
2. Do you have any idea why there is a 5 dB power difference between my designs? I do know it is because of PCB design, as by the end there was identical RF circuitry on both prototypes. If you can spot any critical errors in my design, I would gladly read them and correct the layout.
3. As both prototypes were built on different stackups, are there any laminate properties that favour thinner stackups? Let’s assume that on both stackups there are impedance matched traces.
I do realize that this is a complex problem, I don’t expect an easy answer to my problem, but any kind of feedback will be appreciated at any time. If you need more information regarding the prototypes or measuring setup, I will gladly provide additional data.
Kind regards
Marek Jaworski.
