What manufacturing defect can cause lack of WiFi functionality?

[PanicanWhyasker]

Hi,
we are manufacturing dozens of ESP32-based devices, and most of them work perfectly.
We have a few (e.g. 2% of the production run) which otherwise seem to work fine, as long as the other hardware is related, but the WiFi seems to be dead; the chip WiFi network scan results in 0 networks visible, even if the device is held 1m from the router. The whole PCB is our design, we don't use ESP modules.

Our antenna config is very simple; pin#2 is connected through a 5.6pF cap to a PCB trace antenna.
We've verified that the pin is soldered correctly and is not shorted to the nearby pins. We have continuity from pin #2 to the cap one side, and from cap's other side to GND.

Our suspicion is that somewhere along the waveguide from the cap to the PCB trace antenna there may be a tiny copper whisker that shorts to the GND plane. Since the waveguide is connected to GND through the antenna anyway, there's no way to detect it with a DMM (at DC everything is connected). Our PCBs also have electrical test, but again, at DC this test is useless for such a defect. Visually we also cannot see anything odd.

Is there any other possibility that something's bugging us? Some troubleshooting tips?

Kind regards!

[PanicanWhyasker]

Any ideas?

[ESP_Angus]

Hi Panican,

Is it possible you have a cold joint under the ESP32 QFN pad for the antenna?

Is your crystal 40MHz and its frequency tolerance is good? May be worth comparing XTAL frequency on working vs non-working boards if you can.

Does reworking the antenna components fix the problem on any of the PCBs?

[PanicanWhyasker]

Hi @ESP_Angus!

On the non-working boards we tried reflowing pin #2 with a fine solder tip. As I said we see continuity between pin 2 and the respective side of the 5.6pF cap. It is possible of course that the connection is only made when you press the DMM probe to the pin.

The crystal is 26 MHz ±30ppm. I don't have any equipment to measure 26 MHz, but I guess I can use LEDC to output a square signal that should be a round fraction of XTAL, e.g. 100.00kHz if the XTAL is 26.000 MHz.

We haven't tried reworking the 5.6pF cap, might as well try that! We were able to get one board working by replacing the entire ESP32 chip with a new one. But I'm not sure how to interpret this result and what was the culprit.

Kind regards!

[PanicanWhyasker]

An quick update:

I implemented the LEDC thing and ESP_Angus's suspicion on XTAL was spot-on correct. I made an unused pin to output 10.000 kHz exactly, and on working devices the frequency was accurately held. On problematic devices it varied, staying mostly around 6.6-8.6 kHz but behaved erratically.
Our suspicion is now on phase shifting caps around the XTAL, we'll reflow/replace them and report if this fixed it (other XTAL troubleshooting tips are very welcome).

[ESP_Angus]

Glad you narrowed down the root cause.

The ESP32 datasheet was updated (I think in mid 2017) to specify a 40MHz crystal if Wi-Fi and BT functionality are required.

I believe this is because there were some reports of stability problems locking the PLL on some 26MHz crystals. Some older devices do still use 26MHz with minimal problems, but migrating to 40MHz is recommended.

EDIT: Sorry I didn't reply to the previous post, could have saved you some messing around with the LEDC peripheral.

[PanicanWhyasker]

Cool, many thanks, glad found out what's the issue with these devices.

[AdiXPS]

Hi
Sorry for pumping this thread. But unfortunately we are facing the same behavior, everything works correctly, even UART/Serial but when canning wifi shows zero APs. We are using 40mhz crystal.

We've reflowed the PCBs also. Moreover this is not the case for all our production, only few of them, shows this behavior.. Does that mean its from the crystal? keeping in main that UART is working @ 115200..

[ESP_Sprite]

Hard to say; perhaps you can swap crystals to see if that fixes the issue? The UART working is not really a good indication; an UART has a lot more tolerance to frequency offsets than WiFi.

[AdiXPS]

Hi,
Thanks for the reply. God bless, we've found the issue it was from the crystal and its capacitors. we've replaced it along with its caps and its working now. Thanks.