Dear Patrick: Thank you so much for a good answer that makes sense.
>The effect of refraction of light in the earth's atmosphere is least when >the sun is highest but in practice is not a major factor when the sun's >altitude is more than 10 degs above the horizon. For this reason you should >use the band encompassing two or three hours before and after noon for >setting up. Estimation of refraction is possible but is rarely accurate >since it depends on humidity and atmospheric pressure along the whole line >of travel of the light through the atmosphere. Therefore I suggest you use >times of high sun altitudes. This seems like a good compromise , about halfway between high noon and sunrise and sunset thereby avoiding the atmospheric problems that occur when the sun is near the horizon and also avoiding the inaccuracy resulting from the difficulty of reading a sundial at noon beacause of the small distance between hour lines. > >It isn't automatically the case that the best setting up comes from using a >time within the above band where the angular difference between the hours >is greatest for the design that you have made, for example as you suggest >2-3 o'clock in the pm and 9-10 in the am. If the dial is truly delineated >and aligned then I think that to do what you suggest is the best. However, >if there might be errors then it will be better to use Noon. makes sense to me. > >In the Compendium (Journal of the North American Sundial Soc) a series of >articles recently discussed errors in timekeeping resulting from an error >in delineation or manufacture. In many of the errors considered the >maximum error was to be observed at Noon. For example a 1 degree Eastward >tilt of the dial plate - gnomon still vertical and aligned - results in a >3.25 min error at noon at Lat 45deg N on 1st Sept and only 2+ mins in the >morning hours. Also a 1 degree turn of the dial plate (again on its own as >an error) results in nearly 6 mins error - again at noon. A similar turn >but of the gnomon only has a maximum error of 4 mins - but again it's at >Noon. Thus one might conclude that by setting up at noon one might better >compensate for some of these effects. To avoid these possible errors, that is why I qualify my statement that the "time method for setting a dial only works on those dials that are properly designed, constructed and leveled." > >There is also an effect with Latitude (some errors seem to be less with >increasing latitude) and with time of year (the Summer Solstice has the >highest error at 4 mins for the dial tilt mentioned above compared with 2 >in the winter).. Oddly a twisting misalignment just of the dial plate has >a constant effect throughout the year whereas twisting misalignment just of >the gnomon does not. So maybe your clients should be advised to put a tad >more priority into the gnomon's alignment? This would be hard for the average sundial customer to do because he probably does not have access to an inclinometer for vertical alignments. He also has no easy way to align his sundial horizontally to true north. A compass is not reliable because a small error in reading a compass may result in a large error in setting the sundial because a sundial is so much larger than a compass. Not to mention additional errors that might be caused from interference from nearby iron objects or an error in correcting a compass for magnetic deviation. Polaris is not very accurate either as we have seen. From what I gather, setting a sundial using polaris may result in an error of up to 2 1/2 minutes. So it seems that we are in agreement that the "time method" is the prefered one if and only if the sundial is properly designed, constructed and leveled. >Hope this helps It does. Thanks Patrick John Carmichael
