| Re: label_date precision problem [message #9985] |
Wed, 01 October 1997 00:00 |
R. Bauer
Messages: 137
Registered: November 1996
|
Senior Member |
|
|
mh wrote:
>
> Hello All,
>
> I've got some timeseries data every 15 minutes for a period of a
> couple of
> days, and I'd like to use the label_date routine to do the time
> (x) axis. In order
> to use label_date, I am converting my time axis into absolute Julian
> Day -
> a real number when you convert the time as well. For example,
> 2450717.5
> would be 12:00 on 9/25/1997. If I want 12:15, though, I get
> 2450717.5104.
> But, it appears IDL isn't maintaining the precision of a double when I
> plot, and
> the .5104 is getting truncated to .5. This makes for an ugly plot,
> with 4 or 5
> y-values collapsing onto one x-value.
>
> Right now, I'm working around it by pretending I'm in year -4710,
> which is
> basically the start year for -4713 (1/1/-4713 = Julian Day 1) in order
> to keep
> enough precision to get down to hours. It works, but, it's a cludge,
> and requires
> some thinking around leap years.
>
> Anybody have a suggestion how to make this work? Or another easy way
> of
> doing a time axis in IDL?
>
> Please respond via email.
>
> Thanks,
> Mike
>
Hi Mike
We are using therefore Ray Sterners definition of julian seconds (js)
It's defined: seconds since 2000-1-1 00:00:00 UTC.
Now you have a double precision negative value which will become
positive after 2000-1-1 00:00:00 UTC.
Ray has written a lot of routines to handle these timeformat.
He has lots of routines to transform times in and out to this format and
he has written routines like timeaxis and jsplot.
timeaxis makes whith julian seconds a time axis on a plot. jsplot is a
plot routine to do timeseries plots.
Please look at this documentation for more details.
http://fermi.jhuapl.edu/s1r/idl/s1rlib/time/time.html
I think it's a very good definition for time and the most of our data
are using this js time!
--
R.Bauer
Institut fuer Stratosphaerische Chemie (ICG-1)
Forschungszentrum Juelich
email: R.Bauer@fz-juelich.de
|
|
|
|
| Re: label_date precision problem [message #10017 is a reply to message #9985] |
Thu, 25 September 1997 00:00  |
Robert.M.Candey
Messages: 23
Registered: June 1995
|
Junior Member |
|
|
In article <34299552.C000CFB9@seanet.com>, mh <mh@seanet.com> wrote:
> Hello All,
>
> I've got some timeseries data every 15 minutes for a period of a couple of
> days, and I'd like to use the label_date routine to do the time (x) axis. In
> order to use label_date, I am converting my time axis into absolute
Julian Day
> - a real number when you convert the time as well. For example, 2450717.5
> would be 12:00 on 9/25/1997. If I want 12:15, though, I get 2450717.5104.
> But, it appears IDL isn't maintaining the precision of a double when I
> plot, and the .5104 is getting truncated to .5. This makes for an ugly plot,
> with 4 or 5 y-values collapsing onto one x-value.
>
> Right now, I'm working around it by pretending I'm in year -4710, which is
> basically the start year for -4713 (1/1/-4713 = Julian Day 1) in order to keep
> enough precision to get down to hours. It works, but, it's a cludge, and
> requires some thinking around leap years.
>
> Anybody have a suggestion how to make this work? Or another easy way of
> doing a time axis in IDL?
>
> Please respond via email.
>
> Thanks,
> Mike
IDL uses 4 byte floats for the axes values, thus your truncation. We
generally subtract the time (in Julian values) at the beginning of the year
(or the beginning of the first day plotted) from the time array first and
added it as a parameter to the time_axis routine (we use the JHU APL date
routines). So the plot is made first with no time axis and then time_axis
is called. You could, I suppose, also pass the beginning date in a common
block to the routine called by the tickmarks.
--
Robert.M.Candey@gsfc.nasa.gov
NASA Goddard Space Flight Center, Code 632
Greenbelt, MD 20771 USA 1-301-286-6707
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
