[Jimbee (OP)]

About twenty years ago I was thinking of getting a sundial for my backyard.

To set a sundial, you need an almanac. I used The Old Farmer's Almanac. I noticed something interesting. Each solar day is a couple minutes faster or slower than one before and after. This is called sunfast.

Why is that? Is it because the earth spins faster when it is closer to the Sun?

[Halc]

Is it because the earth spins faster when it is closer to the Sun?

The sidereal spin rate (23:56:04) is fixed* to an awful lot of digits, but it is indeed due to Earth distance from sun.
The effect is so great on planets with high eccentricity like Mercury that the sun actually stops in the sky and goes backwards for quite a while every 'day'. Where it does this in the sky depends on where you are on the planet, but it's the same place each time.

When the sun is close, the Earth needs to rotate more degrees to get the sun back beyond the horizon, and that lengthens the day. The sun also moves faster relative to Earth, and that shortens the day, but not as much as the former effect lengthens it.


* The sidereal spin rate slows over very long times, due to tidal drag. The day used to be 10 hours or less, and will be again if the sun doesn't swallow Earth first, but it will.

[Janus]

Just to clarify on Halc's post at bit.  As stated, the physical rotation of the Earth remains pretty constant.  The Solar day, however is a result of this rotation and the Earth's orbit around the Sun. In a Sidereal day, the Earth has traveled a bit under 1 degree of it orbit. So, in order to Bring the Sun back to it noon position, the Earth has to rotate a bit more. Thus the Solar day, is roughly 4 min longer than the Sidereal day. 
The Earth's orbit is not perfectly circular, and due to orbital mechanics, varies its orbital speed over its course.  This alters slightly how much of its orbit it travels in  a sidereal day, and thus the length of the Solar day.   The Earth's axial tilt also plays a role in the apparent solar day length during the year, since the Earth's axial rotation is not aligned with its orbital one.

[Bored chemist]

If you have one of these you can give the almanac back to the farmer. :-)
https://en.wikipedia.org/wiki/Equation_clock

[alancalverd]

Fascinating stuff - I'm indebted to the experts!

I wonder if this explains the complexity of henges? Clearly the oldest single stone circles synchronise a calendar that can be recognised and used for trade over most of the globe, but I wonder if subsequent additions add corrections for longitude and seasonal variation of noon? 

[vhfpmr]

How long before the slowing of the earth's rotation will force us to choose between retaining 86400 seconds in a day and retaining the current definition of the second?

[Janus]

How long before the slowing of the earth's rotation will force us to choose between retaining 86400 seconds in a day and retaining the current definition of the second?

At a rate of 1.7 ms per century,  58,000-59,000 years for the Earth's rotation to increase by 1 sec. The definition of a sec is unlikely to change, as it it tied to so many things in physics.  What will happen is that, as time goes on, leap seconds will be added more and more frequently.  But by the time it becomes necessary to add one a day, I'm sure our clocks will be able to do it pretty much automatically as needed.

[alancalverd]

They already do! A cheap (?20) wall clock or a medium-price (?300) wristwatch now incorporates automatic synchronisation to national radio time standards every day.

Which brings me to a further question. Not all states in the USA or Australia adopt the stupid "summer time"  clock shift, but AFAIK there is only one national time standard. What happens in Arizona, for instance, when all the other radio clocks resynchronise?   

And when will the world wake up and adopt UTC everywhere? No need for a wobbly dateline, incompatible timetables,  faffing about with a manual watch or wondering whether you can actually receive the local signal when you travel: as I write, it's 1750 everywhere in the universe!

[paul cotter]

Quick question, Alan: there was a standard time transmission on a low frequency from Rugby, does that still exist?

[alancalverd]

It's still occasionally called Rugby time but the 60 kHz transmitter is now at Anthorn, Cumbria, and encodes the date and GMT/BST, referenced to NPL, Teddington.

[Bored chemist]

How long before the slowing of the earth's rotation will force us to choose between retaining 86400 seconds in a day and retaining the current definition of the second?

It already does. That's why we have leap seconds.