Dana DeLouis shared this with us in microsoft.public.excel.misc:

Hi. I'm not sure what a good suggestion would be. As far as the
Fourier idea goes, yes, you would be limited to 2^12, or 4096 data
points. I'm not sure of this idea, so I'll just throw it out. Since
your data has a period of 1 day, how about breaking the data up into
daily groups. Take the average of the daily highs and lows. Half
way between the high and low would be your offset, (or dc component),
The value of average high-offset would be your amplitude, Perhaps
take the average of the daily starting values to use as your phase.
A pivot table may be able to organize your data for you. Anyway, not
the best solution, but maybe a workaround. HTH. :)

That sounds like a lot of hand-hacking and manual work. I was hoping
for a "simple" solution where I could just enter the data points (or in
this case: import them from a text file or some kind of sql-ish data
source) and have instant results.
The idea was to have 2 sets of data in my chart: one with the original
data, unconnected, with a lot of "white noise", and another with the
calculated sine wave, points connected with a smooth line.


And now that I examine the raw data more closely, I get the impression
that not only there is a daily period, but also a weekly period - but
with a much smaller amplitude. Uh-oh... I think I'll have to dig up
some math books...

--
Amedee Van Gasse