Add support to set the clock rate and have sntp use it. #3236
Conversation
docs/modules/rtctime.md
Outdated
| @@ -139,8 +139,10 @@ It is highly recommended that the timestamp is obtained via NTP (see [SNTP modul | |||
|
|
|||
| Values very close to the epoch are not supported. This is a side effect of keeping the memory requirements as low as possible. Considering that it's no longer 1970, this is not considered a problem. | |||
|
|
|||
| You specify the seconds parameter with the optional microseconds (defaults to 0). Setting the rate is independant. | |||
There was a problem hiding this comment.
Generally speaking, docs don't use 2nd person like this. Separately, that's remarkably terse. How about "Times are specified across two arguments, seconds and microseconds. If microseconds are not specified, the default is 0 (i.e., time is exactly at the boundary between two seconds). In addition to the time, a third argument specifies the local clock's drift rate estimate, as documented in rtctime.get(). If this argument is not given or is nil, the current rate estimate will not be altered, and the rate estimate may be set without changing the time by leaving the seconds and microseconds arguments nil."
There was a problem hiding this comment.
Fixed.
As a side note -- this is the rate of a WEMOS D1 MINI over almost 24 hours:
I suspect that differences are temperature related caused by localized heating due to the amount of work being done by the CPU. I'm going to try modifying a board to use a TCXO instead and see what happens.
|
Delightful. I had done something similar for #2819 and hadn't cleaved it out as its own PR. |
|
I'm happy for this to land. |
Fixes #3235
devbranch rather than formaster.docs/*.This just adds support to
rtctime.setto allow it to be used with the first two args as nil. In this case, if the rate is supplied as argument 3, then it is set.The sntp module also initializes the clock PLL from the current rate.