Designing a circuit with a microcontroller and a crystal or resonator there should some measures done to ensure the oscillator runs stable and reliable in mass production and under climatic conditions.
For STM there exist an "Oscillator design guide"". Similar for Espressif is not available. Wrong capacitors in an ESP32 design led to malfunction in Bluetooth but not with the rest of the MCU.
What measures are recommended, especially in cases like Espressif, where no detailed specifications are available?
ST even mention in their guide that usually not much attention is given to this. But Espressif does give some guidelines here:
https://www.espressif.com/sites/default/files/docu... in section 3.14.
of 8pF in its datasheet (Got it approved by Espressif for ESP32S3). We assembled it without external capacitors for experimenting. It did run at 40MHz and also the application software did fine. The strange thing was, that a Bluetooth connection was only possible with the capacitors assembled.
There must be something like unstable frequency without the external capacitors on the crystal. We like to have some measurements or test to ensure proper oscillator functionality in mass production. Therefore we ask for qualification methods for the oscillator.
Did you respect the recommended layout guidelines?
Yes, but it can be a sensitive zone. Therefore my questions here how to test.
Somewhere in the data sheet is stated that for BT max. +-10ppm deviation is allowed.
Why no external load capacitors on the crystal? Any reason?
As I assume that you understand German take a look at the Quarzkochbuch, available online here:
Chapter 6 page 17 on the load capacitor.
Does Bluetooth require a specific crystal frequency or multiple thereof ?
Program an output to reflect a fraction the clock frequency, then measure that output for jitter.
From experience, even small capacitance variation on the crystal terminals can prevent it from even starting (ICT test fixture).
