Inconsistent sleeping with SDK 2.1.0 uint64_t sleep times

[torntrousers]

I'm trying the extended deepSleep times available with the latest ESP8266 SDK 2.1.0. Some sleep time values work fine but some don't seem to work and the ESP wakes up almost immediately. Bit strange.

For example, calling system_deep_sleep with these works as expected:
uint64_t sleepTime = 10000000; // WORKS: 10 seconds
uint64_t sleepTime = 7200000000; // WORKS: 2 hours
uint64_t sleepTime = 9000000000; // WORKS: 2.5 hours
uint64_t sleepTime = 86400000000; // WORKS: 24 hours

But calling system_deep_sleep with values like these doesn't work and the ESP8266 wakes up almost immediately:
uint64_t sleepTime = 21600000000; // FAILS: 6 hours
uint64_t sleepTime = 74049000000; // FAILS: bit over 20 hours

In all cases system_deep_sleep is returning true, and the ESP8266 wakes up with rst cause:2, boot mode:(3,6).

What could be wrong, is this a bug?

[pratik]

Reset cause 2 means that you are getting the reboot due to reset pin activation.
Ensure that this has nothing to do with loose connections, if you are working on a prototype. Meanwhile I will let you know if I can reproduce this exact problem.

[torntrousers]

Were you able to recreate this? Should be pretty easy - just try a deepSleep for 6 hours and it should wake up right away instead of sleeping.

[Her Mary]

May it be the maximum limitation of sleep time?
It is mentioned in the non OS API guide.

[torntrousers]

I'm having trouble with the maths. In the API guide in system_deep_sleep in 3.3.9 it says:

The theoretical maximum value of time_in_us can be calculated by formula:
(time_in_us / cali) << 12 = 2^32 - 1
• cali = system_rtc_clock_cali_proc(), the cali is the RTC clock period (in us)

In system_rtc_clock_cali_proc in 3.3.22 it says:

RTC clock period (in us); bit11 ~ bit0 are decimal.
Example os_printf("clk cal : %d \r\n",system_rtc_clock_cali_proc()>>12);

When I try that printf on an ESP8266 it outputs "clk cal : 6"

So (time_in_us / cali) << 12 = 2^32 - 1
So time_in-us = ( (2^32 - 1) >> 12 ) * 6
= 6291450

Which is just 6 seconds! Clearly wrong as it sleeps for longer than that fine. Can anyone do the maths better?

[Her Mary]

I think the "<< 12" means "bit11 ~ bit0 are decimal"... So the "cali" should not be "6", but 6xxxx...
It is "cali = system_rtc_clock_cali_proc()", not "cali = system_rtc_clock_cali_proc()>>12".

[torntrousers]

If thats the case then the longer sleep times should be fine, right? So why aren't they working for me? Could someone else here just try a deep sleep for 21600000000 (six hours) and see if it works for them or not?

[Her Mary]

The latest doc changed to (time_in_us / cali) << 12 = 2^31 - 1
Mine is 3.5 hours.

[Bremer]

I found another problem with system_deep_sleep(0)

[ESP_Faye]

Hi Bremer,

The deep sleep 0 issue has been solved, please have a try with the latest SDK. https://github.com/espressif/ESP8266_NONOS_SDK

Thanks for your interest in ESP8266!