Re: [ferret_users] time axis definition and precision

[Martin Schmidt <martin.schmidt@xxxxxxxxxxxxxxxxx>]

Hi Peter,

> that works fine for daily files. But how to deal with the data, when they 
> cover -lets day a whole year or more?
> The problem is, that you need the full information in seconds for a year, but 
> the variable to create the time axis is modified (see below).
>
> Any idea?
>
>  
You need to keep the span of times within the 4-byte limit. If this is 
not possible - I think no chance.

However, possibly data with a period of seconds are not needed for a 
year or so. You will have to condense this huge amount of information 
somehow anyway. I suggest, to do this as soon as possible to get time 
units, which are more handsome.

Greetings,
Martin

> let xmpl=DAYS1900(1900,12,31)
> let secs={1,2,3,4,5,6,7,8,9}
> list/format="(f14.2)" xmpl*86400+secs
> list/format="(f14.2)" xmpl*86400+secs
>             VARIABLE : XMPL*86400+SECS
>             BAD FLAG : -9.9999998E+33
>             SUBSET   : 9 points (X)
>             X        : 0.5 to 9.5
>   31449600.00
>   31449602.00
>   31449604.00
>   31449604.00
>   31449604.00
>   31449606.00
>   31449608.00
>   31449608.00
>   31449608.00
>
>
>
> Am Dienstag, 20. Mai 2008 14:50 schrieb Martin Schmidt:
>  
>
> > Peter,
> > I had the same problem several times.
> >
> > To read daily data with a period of 1 s from a file, where times have
> > the form
> >
> > 2008-02-19 00:00:00
> > 2008-02-19 00:00:01
> > ...
> >
> > I do the following:
> > let offset = DAYS1900 ( 2008, 1, 1)
> > define axis/x/x=1:86400:1 xax
> > define grid/x=xax xg
> > COLUMNS/deli=" "/grid=xg $1
> > let v1o=v1-offset
> > define axis/t/from_data/t0=1-jan-2008/unit=days time=v1o+v2/24 + 1e-5
> > define grid/t=time tgrid
> >
> > From now on things work well.
> >
> > Two things are needed, the offset to keep the time values limited and
> > the 1.e-5. Otherwise I get the same error message as you have got.
> >
> > There are several show-ups of single precision limitations:
> > - long time axes with small time periods
> > - strong zooming of other axes, which gives occasionally good looking
> > but wrong plots
> > - seawater thermodynamics, i.e. relations between salinity, temperature
> > pressure and conductivity and so on ...
> >
> > Greetings,
> > Martin
> >    
>
>