jahboater wrote:The time from NTP is pretty accurate checked against MSF rugby radio clocks.
The time from NTP is going to be more accurate than just about anything other than a good GPS, since the protocol incorporates steps to account for and filter out the round-trip-time of the synchronizing packets over the network.
And, unlike a typical consumer-level "atomic" clock (which simply syncs to a radio signal once a day, or more often if you're lucky), NTP synchronizes continuously. And, if it needs to make adjustments, it doesn't simply add or subtract seconds - since this could cause time to repeat or go backwards from the viewpoint of processes on the system, and databases and such (which would wreak havoc when sorting transactions by timestamp to see the order in which they happened) - it actually makes the computer's notion of a second be a tiny fraction longer or shorter for a while, until it "catches up", so time is always increasing. This is vital to making networked systems of computers work well together (starting with "which system has the newer version of his file"). It's a lovely, elegant, design, really, by very smart people. No cumulative drift, and accuracy in the 50 millisecond range.
And precisely why one of my first Pi projects, with my first Pi model B, was to add a PiTFT display and make a super accurate clock for my living room. I don't actually need fraction-of-a-second accuracy, that's a side benefit, but the lack of drift and built-in handling of things like daylight savings time changes (coincidentally happening tonight in the USA) are a huge win.
Mine also has a TSL2561 luminosity sensor pointing out the side of the semi-transparent case, and it takes a reading every 15 seconds and publishes it to the in-house MQTT broker, then that Pi and several others that also have displays use the data from MQTT to continually adjust the brightness of their backlights. Works a treat.