Clock spread / precision RC_FAST_CLK and RTC_SLOW_CLK

[ShackShawn]

Hello Espressif,

I could not find any documentation mentioning the spread / precision of the RTC_SLOW_CLK and RC_FAST_CLK on the ESP32-S3

Could you please let me know? This is very important to us.

The only mention of it was in this video, where somebody did their own research on it:
https://www.youtube.com/watch?v=fZAR8WTKiSg

Thank you.

[ESP_Sprite]

Sorry, as far as I know we do not have this information at this moment.

[ShackShawn]

Thank you for the quick reply.

You do not have numbers on the drift for either clocks? I saw from an other post

In light sleep and deep sleep, RTC clock is used for timekeeping. This is either the internal 150kHz oscillator, or external 32kHz XTAL. The internal 150kHz oscillator is +/-5%.

Perhaps if I rephrase what I am looking for

- What is the dependence of the drift to the temperature? Linear, quadratic? How many ppm / C or ppm / C² ?
- What is the over-all deviation?

Thank you for your help.

[ESP_Sprite]

Those numbers are for the original ESP32, it's very likely that they still hold but I don't have a confirmation on that. I do not have information on the other values you want.

By the way, out of curiosity, why would you need those values? Those clocks generally are only relevant when you don't have a main crystal running, which would normally be in deep sleep... but in deep sleep, you generally also are not awake to measure anything the drift may depend on.

[ShackShawn]

We are taking samples during the deep sleep phase using the ULP. In order to have reliable timestamps on those samples, we are using a combination of the clock and the NTP update whenever we upload them.

To increase reliability we would need to have a precise clock source, so as to not solely count on the NTP update.

In this case we will have to make a board revision with an external oscillator it seems.

[ShackShawn]

Would you be able to provide info on what the product name or the type of oscillator is used? We could do our own research to estimate the variance.

[ESP_Sprite]

The internal oscillators are on-die; they're not separate components.