ESP32-PICO-V3-02 main CPU crystal oscillator

[derricksenva]

Hi,

There's a lot of conflicting information from the datasheet for the ESP32-PICO-V3-02 (https://documentation.espressif.com/esp ... eet_en.pdf). We ultimately need to know how accurate the main clock driving the CPU is — whether or not it's running off an external crystal oscillator (and its ±ppm), and whether or not we can provide one externally on the board.

- Page 1 says "It integrates all peripheral components seamlessly in a single package, including a crystal oscillator"
- Page 2 shows a block diagram illustrating that a crystal is integrated.
- Page 13 shows pins 44 and 45 as no-connect.
- Page 30 shows a reference schematic with a 40 MHz crystal connected to pins 44 and 45.

What's the correct answer here?

[MicroController]

AFAIK, the crystal is actually included in the PICO's package, no external crystal needed or supported. Figures 7, 8, and 9 in "7 Schematics" seem to show the schematics of the in-package circuitry (notice e.g. the flash chip and the open-ended "LNA_IN"), while "8 Peripheral Schematics" with figures 10, 11, 12 shows the "typical application circuit", i.e. the external components needed.

[lbernstone]

See the hardware design guidelines. The on-module primary crystal will be +-10ppm from 40MHz, but the RTC is fractional off that source, so will drift significantly if you are doing serious timekeeping.
You can add a 32.768kHz crystal (at an accuracy of your choosing) on gpio 32/33 to provide a more accurate movement.

[Sprite]

See the hardware design guidelines. The on-module primary crystal will be +-10ppm from 40MHz, but the RTC is fractional off that source, so will drift significantly if you are doing serious timekeeping.
You can add a 32.768kHz crystal (at an accuracy of your choosing) on gpio 32/33 to provide a more accurate movement.

That's not the reason why the RTC (without an external 32KHz crystal) is known to drift. While powered on, the RTC should indeed be 10ppm stable as the 40MHz crystal is like that (the fact that for the timers this gets divided down does not affect the precision). The issue is the modes when that crystal is not powered on, i.e. deep sleep: when no 32KHz crystal is connected, the RTC is clocked off an internal 128KHz (iirc) clock source which has a much wider error tolerance, causing the clock to drift appreciably.

[MicroController]

I.o.w., when you divide (or multiply) a clock, the drift in Hertz also gets divided (or multiplied) by the same factor; the ppm stay the same.

[derricksenva]

See the hardware design guidelines. The on-module primary crystal will be +-10ppm from 40MHz, but the RTC is fractional off that source, so will drift significantly if you are doing serious timekeeping.
You can add a 32.768kHz crystal (at an accuracy of your choosing) on gpio 32/33 to provide a more accurate movement.

That's not the reason why the RTC (without an external 32KHz crystal) is known to drift. While powered on, the RTC should indeed be 10ppm stable as the 40MHz crystal is like that (the fact that for the timers this gets divided down does not affect the precision). The issue is the modes when that crystal is not powered on, i.e. deep sleep: when no 32KHz crystal is connected, the RTC is clocked off an internal 128KHz (iirc) clock source which has a much wider error tolerance, causing the clock to drift appreciably.

Can you provide hard confirmation for my original question — Does the ESP32-PICO-V3-02 integrate the 40 MHz crystal into the SiP? And is it spec'd at ±10 ppm?

[Sprite]

Can you provide hard confirmation for my original question — Does the ESP32-PICO-V3-02 integrate the 40 MHz crystal into the SiP? And is it spec'd at ±10 ppm?

Yes. See figure 9 of the datasheet, which covers the internal schematic of the module. The crystal is indicated as '40MHz+/-10ppm'.