Python epoch and phase computation

[Chtululu]

Hello,

Epoch question :
In the solve function the default epoch is defined as "The default corresponds to the Python standard library datetime proleptic Gregorian calendar, starting with 1 at 00:00 on January 1 of year 1"
On the other hand, in the notebook ["utide_real_data_example.ipynb"](url), time is defined as time = mdates.date2num(obs.index.to_pydatetime()) which epoch is correspond to calendar starting with 1 at 00:00 oon January 1 of year 1970.
Finally which epoch is used in the solve function?

Phase question :
I'm having hard time to recover the phase of my signal. I use real data (sea surface elevation in the Bay of biscay)
With t = mdates.date2num(obs.index.to_pydatetime())
and where obs.index=pd.date_range(start='2017-07-09 12:00:00',end='2017-07-30 12:30:00',freq='1min')

I recover the main correct component and their corresponding amplitude, but the phase is different then expected. May this be related to my first question?

I don't know if it's related but I then started by constructing a simple signal with a semi diurnal frequency such as : Acos(wt-p).
I recover my amplitude and pulsation (w) but I couldn't recover the phase p. When I computed the phase of the signal with p= 0, meaning : Acos(wt). The phase was around 114. I was then able to recover p when using a signal: Acos(wt-(p-114)). I thought I might use this trick for my real signal but I could'nt recover a correct phase, even with a signal made of sum of Aicos(wi*t), with i the indices of the main constituent of my signal.

[Chtululu]

The phase given by the code consider the epoch starting with 1 at 00:00 on January 1 of year 1 when python date2num starts with 0 at 00:00 on January 1 of year 1970.

As I use date2num I also need to use the keyword 'epoch='1970-01-01'' in coeff calculation to recover the correct phase. (This should also be the case in the notebook example)