making his sundial as accurate as possible. Would the morning and
evening errors that he mentions be due to refraction, or to one of the
other factors that he mentions ?
Here are some extracts from his emails :-
Dear Richard,
The hour lines are offset to allow for the difference between my
longitude and that of Greenwich. The twelve-o'clock line therefore
doesn't coincide with the local meridian. Local noon is marked by two
lines exactly matching the thickness of the gnomon. This means that it
is far easier to judge the moment of local noon than any other time. As
you rightly point out, the fuzziness of the gnomon introduces a
considerable degree of uncertainty into dial readings, because it is
extremely difficult to judge when the ill-defined shadow is right on an
hour line. However, it is far easier to judge when the shadow of the
gnomon is exactly filling the space between the two lines marking local
noon. By taking many observations, plotting the results on a graph and
taking means, I think I've managed to get quite a good picture of the
setting of the dial.
It was commissioned for me by my wife as a present for my 50th birthday
last year. As we were then in the process of house-hunting, she wisely
deferred the purchase until we were certain of our new address. It was
made by a chap from the Bristol area who trades as Merlin Sundials. It
was delivered last September or October, but for various reasons I have
only just got around to setting it up.
It's a smallish horizontal dial of conventional design, eight inches or
so in diameter. As I think I mentioned in an earlier message, the hour
lines are offset so as to enable one to read Greenwich time direct from
the dial (subject, of course, to the EOT). An attractive graphical
representation of the EOT is reproduced on the dial plate, so that one
can make a rough estimate for any day of the year to within a minute or
so. I was also able to choose my own motto (and I believe I have an
original one, in Latin, which I'm fairly confident has not been used on
a sundial before - if you're interested, I'll tell you about it). It is
chemically etched and is graduated at 1/4 hour intervals.
Anyway, my sons and I have levelled the dial plate pretty accurately, if
one can trust spirit levels, and I think we are probably within about 6
arc minutes (20 seconds of time) in azimuth. That's what I'd really
like your advice about. Is that a good adjustment, or should I try to
improve it? To be honest, I'd expected a larger error, because I'm
pretty hamfisted at DIY, and I thought that errors would inevitably
creep in with drilling holes in the top of the plinth and so on. If I
got the holes even a small fraction of a millimetre out, it would
obviously affect the accuracy of the dial. In fact, through luck more
than design, I fixed the dial in azimuth in exactly the position I
intended. The only slight problem was that I was using a
not-very-accurate value (1 m 15 s) for the EOT when setting the dial in
the meridian. I actually set it about 15 or 20 seconds 'fast' on the
correct value, which was more like 1 m 34 s.
The position seems to be that at local noon the dial is more or less
bang-on the intended setting, except that I have a hunch that
(fortunately) I may have set it a few seconds slower than I meant to do,
thus fortuitously reducing the error with respect to the correct EOT
value from 18 to about 15 seconds. I hope I've made this reasonably
clear. What it all boils down to is whether an error of 15-20 seconds
of time in azimuth at local noon is acceptable. What do you think? How
does it compare with normal practice? This is the only sundial I've
ever set up, so I don't really have anything to compare it with.
There are other errors, which I imagine arise from slight errors in the
construction of the dial - possibly in the angle of the gnomon, lack of
parallelism of the two edges or 'styles', the etching of the hour lines
and so on. The result is that when it first catches the sun (about 0830
near the summer solstice), it is about a minute fast. This diminishes
quite quickly, so that from about 10 a.m. to 1 p.m. it is pretty much
exactly right, within half a minute or less. After 1 p.m., it gradually
gets fast again, until by about 5.15 p.m., when the sun leaves the dial
in June, the error is of the order of 2.5 minutes. I don't find this a
problem, but have no idea whether it represents acceptable performance
to an experienced dialist. I know this error doesn't flow from an error
in the azimuth setting, because the dial is bang-on at local noon and,
indeed, for at least an hour either side.
I believe I have now solved the problem of computing the EOT. I have a
copy of Peter Duffett-Smith's 'Easy PC Astronomy', with its accompanying
Astroscript software. I've used this to calculate the EOT and now get
figures closely compatible with the predictions in Whitaker's Almanack
and the NASA ephemeris generator, and reasonably close to those provided
by the calculator on the BSS website. In other words, I was right to
criticise the 'standard' tables, which on certain dates can differ from
the true value by well over 20 seconds.
Best wishes,
Tommy
-
