Raspberry Pi clock trivia

You mean just 10 minutes, 43 seconds.

The hour is an inherent automatic daylight saving feature :-)

it is a computer not a clock the time is maintained by a counter that is periodically updated by an interrupt

There are many things that can affect the accuracy of this counter.

you would see the same effect on a PC left running for any length of time as they only check their RTC chip on start-up (or at least that was the case with MSDOS)

You could install a crystal based clock module! There are many on internet. Google 'raspberry rtc' to see! Regards

If you actually gave an RA*..yes.

Normal quartz oscillators have a deviation of +/- 100 ppm. That gives about +/- 9s per day in worst case. As frequency of a crystal is also influenced by temperature the measured drift looks rather normal to me. For better accuracy you need an external clock module or sync regularely with a time server if possible.

I thought it was fairly typical, but thanks for the rigorous analysis. I happened to have the data, it seemed a shame to keep it to myself. I rather like the idea of getting the time from a GPS receiver.

Either that or run an instance of ntpd, the Internet time server, which keeps your system clock in sync with UTC time as well as providing the time to its client. Its one of the Debian standard packages, so is dead easy to install and has a default configuration, so it mostly 'just works'.

See the bit at the top about "losing it's internet connection"?

Or something like a WWVB receiver. Very simple, low power and just as accurate.

These may help....

On 03/02/2016 13:59, rickman wrote: []

Far less accurate than GPS/PPS - most of the time!

When reconnected to internet it will update time.

???? How is that?

Exactly so. I also read 'losing its internet connection' to mean a temporary loss, rather than a permanent disconnection. If the latter was the case, why not say so?

GPS/PPS is typically within 1/10 microsecond, 100 ns, or better. Radio propagation changes introduce far greater uncertainties.

On Wed, 3 Feb 2016 10:33:11 -0500, rickman declaimed the following:

The typical GPS time server is, besides computing a lat/long/alt, also computing the time... One reason at least four birds are needed to get a good fix, as the receiver time is back-fit to the propagation delays of the multiple birds. That is: "based on these three birds I think I'm standing here; if I am here, then the time delay from the fourth should be /d/... It's not, so my clock must be off... Adjust the clock and recompute the location, check time delay..."

WWVB is a single time signal, with somewhat uncertain propagation delays (if really bad, one might be getting the long-path signal -- around the back side of the earth -- rather than the short "direct" path [which may still be different "bounces"])

I have worked with GPS PPS signals and they are seldom as accurate as theory allows. There are many realities in low end units so that the actual accuracy is in the 10's or even 100's of us. I can get that from a WWVB signal at 60 kHz.

I guess the real question is, can a rPi make any use of accuracy better than a ms or so? Both signals will have excellent long term accuracy. Both signals provide time of day and calendar, etc. The WWVB type signal may be hard to receive in some areas while the GPS signal requires an outdoor antenna in some cases.

The OP did say "One of my Raspberry Pis has been running continuously since losing its internet connection in April."

James

If I'd been the OP I'd have said 'permanently losing'.