NavList:

Antoine, you wrote:
"Since the objective is to compute position with just plain LinInt, we would need to restrict the tabular interval well below 1 hour (actually 1 hour gives about 6" accuracy with just linear interpolation)."

I don't think this is the case. Maybe try it again? Use your favorite astronomy software, e.g. Solex, to output the Moon's ecliptic latitude and longitude at half-hour intervals for some extended period of time. Then run that through some software or a spreadsheet and calculate the values for the times ending in half-hours from the data for integral hours by simple linear interpolation and compare against the actual values. What is the mean error? What is the maximum error? Try the same thing calculating the positions at odd hours by linear interpolation from the data at even hours.

You also wrote:
"You must have some QUITE CLEVER approach, which I would perfectly understand you might prefer to keep un published.
However, I am very curious here simply because your claim here-above has made my mouth water in the sense that I see an opportunity here to learn more on a subject of interest."

No really, there's nothing clever about it, and I haven't simply published my numbers because there's a benefit from trying different approaches, as independently developed as possible --at minimum, it saves us from the evils of "group think" masquerading as "consensus". One amusing quirk in my data files dates from a limitation in the web technology that I had available in 2004 when I first set this up. The "asp" software had to be tricked into working with binary files so in fact the data are saved as sequential bytes, like a big text file, with a zero byte to mark the end of the file. There are 111 bytes per day BUT zero bytes are not allowed in the data! That's got nothing to do with the method. Just a funny technical problem I had to deal with six years ago.

By the way, I should say here there's no reason to be purist when it comes to algorithms. An interesting compromise would use a very short analytic model to generate the almanac data to the nearest degree or tenth of a degree and then use tabulated data for the remainder. Off the top of my head, I would guess the total data size could be reduced by another factor of ten with only a couple of dozen extra lines of code. It's a best of both worlds approach. :) And folks who have heard a legend about "Thomson's Lunar Tables" may recognize an analogy here...

-FER

----------------------------------------------------------------
NavList message boards and member settings: www.fer3.com/NavList
Members may optionally receive posts by email.
To cancel email delivery, send a message to NoMail[at]fer3.com
----------------------------------------------------------------