Matching network considerations for STM32WB55CEU6
Hi,
I am designing a system using STM32WB55CEU6 and I want to use the RF functionality. Looking at evaluation board I see that there is a matching network for the output pin. The datasheet does not specify the output impedance, but an ST employee posted their empirical measurements:
For the package I use the impedance is 42+j19. Using an online Smith chart tool it looks like I would only need a small 0.6pF capacitor:
However, the evaluation board uses a non-symmetric pi filter:
It was my understanding that the advantage of a pi filter would be the symmetrical nature, where signal direction does not matter.
Therefore, it seems it would make sense to use symmetrical values as such:
Using the values provided by evaluation design:
This lands pretty far off. I see that in evaluation design it is mentioned component values will be updated after evaluation, so is it the case that these values are more or less arbitrary and tuned with a VNA during proto testing? What would be a reason to decide to use non-symmetrical pi filter for initial design?
Regarding PCB design, most common thickness seems to be 1.6mm, however, with a 10% tolerance on the height could it be the case that the distance to return path changing from batch to batch could negatively impact performance of antenna, and therefore reducing overall thickness of the board prove beneficial to design?
2 answers
I see that in evaluation design it is mentioned component values will be updated after evaluation, so is it the case that these values are more or less arbitrary and tuned with a VNA during proto testing?
That was my first thought when I read that disclaimer.
Your original analysis suggested that you only needed to add a 600 fF capacitor. That's tiny, and easily swamped by the parasitic capacitance of the PCB traces, connectors, etc.
Your analysis also used a specific impedance for the STM pin which apparently didn't come from a datasheet but someone's measurements. Was that from a single unit? Did they measure many units and then average? What should you expect for unit to unit variation?
When fractions of a pF matter, then calculating everything up front when using ordinary PCBs is impossible. You start someplace plausible, then tweak by experimentation.
On a separate topic, the purpose of the C-L-C pi filter may not only be impedance matching. Attenuating out of band frequencies may be important for regulator compliance. Also, the asymmetrical capacitance should not be surprising when different impedances are connected to each end of the pi filter.
0 comment threads
It was my understanding that the advantage of a pi filter would be the symmetrical nature, where signal direction does not matter.
and
What would be a reason to decide to use non-symmetrical pi filter for initial design?
An asymmetrical PI filter will be used to match different impedances. I mean, it's already stated that the output isn't exactly 50 Ω so, an asymmetrical filter will be required to match the chip output to the transmission line resistive impedance. Yes, there is a t-line because that's why you match impedances (to prevent reflections).
Therefore, it seems it would make sense to use symmetrical values as such
No, that doesn't make sense to me for the reasons above.
reducing overall thickness of the board prove beneficial to design?
Unlikely if it's optimal performance is at the nominal thickness.
0 comment threads
0 comment threads