ESP32C3 Custom Board

[jackinditec]

My friend and I are developing a custom board for BT/Wifi using the ESP32C3 and an onboard antenna based off: [TI 2.4 GHz Antenna][/https://www.ti.com/lit/an/swra117d/swra117d.pdf].
In our first version, the BT/Wifi was not working properly. We used the tx test tone in the [ESP Certification and Test Tool][/https://www.espressif.com/en/support/do ... ools?keys=] to check the emitted signal and compared it to an official ESP dev kit:

PCB v21 signal measure.jpeg (213.11 KiB) Viewed 2138 times

Marker 2 is our dev kit and Marker 3 is our board. We can see that there is not only a difference in power (not significant in this picture because we might have placed our board closer the receiver) but more importantly there is a difference in frequency at max power. I'd also add that we later changed the clc circuit on the antenna and the signal strenght greatly increased.
Occasionally if we stopped and restarted the test tone while the c3 was hot, the signal would reduce but be the right frequency...
Reading that this might be caused by problems with the oscillator or antenna circuits, we tried various CL values for both and managed to increase power somewhat however the frequency issue persisted... Could this be related to auto calibration by the ESP?
We have no VNA at our disposal unfortunately.

We have since also added the inductor L2 to oscillator, better separated oscillator and antenna signals and created a thicker ground place isolating the antenna:

PCBv21 Osc.PNG (18.16 KiB) Viewed 2138 times

The proximity to the antenna has also been optimized in our last version:

PCB v21 layout osc&antenna.PNG (198.19 KiB) Viewed 2138 times

Before we order the new boards, we were wondering if anyone had any insight on what our last error might have been?
Also, if 1mm thickness is sufficient for a 4 layer board? Lastly if the choice of antenna is significant as long as they have the same characteristics? (50 Ohm impedance at 2.4Ghz)

Thank you.
Jack

[Sprite]

24pf seems a bit high for a crystal with 10pf load capacitance... how did you arrive at that value?

[jackinditec]

My apologies, the picture is outdated. As I indicated we used several values for the oscillator capacitor. The theoretical best value is 16pF but we settled on 13pF through experimenting with the oscillator. The behaviour did not change and the problem persisted regardless.

[jackinditec]

Additional hypothesis:
The frequency seems to fit into a different (the fourth) wifi channel while the dev board is in the . Is it possible, that the ESP32 antenna calibration software automatically recognises that our antenna has a resonant frequency that better suits a different wifi channel and automatically chooses this frequency for our signal? We have tested it on all classic examples (ie: SoftAP, WifiScan, etc) and have not seen any way to configure this.

wifi.png (168.68 KiB) Viewed 2003 times

[Sprite]

I'm not sure. You may have better luck collecting your findings and sending an email to sales@espressif.com; they tend to have a bit more experience with RF tuning than the people here on the forum. If you find out what it is, please do post your findings here, btw.