start time problem on data requests?

[kaspervanwijk]

Hi All,

Am I doing something wrong here? I am getting significant differences between requested start time and delivered start time. IRIS server does not have this problem. Simple code for one example where the difference is more than 6 seconds:

from obspy.clients.fdsn import Client
from obspy import UTCDateTime

geonetclient=Client('http://service.geonet.org.nz')

t0 = UTCDateTime("2015-02-27T06:30:00.000") 
st = geonetclient.get_waveforms("NZ","WIAZ","10","EHZ",t0,t0+10)
print('requested time {}, received starttime {} and difference {}'.format(t0, st[0].stats.starttime,t0-st[0].stats.starttime)

[rumachan]

Hi Kasper,
Isn't it just how the data buffers boundaries fall relative to your request? If you grab the same month/day/time for 2017, the difference is only 0.05s.

[kaspervanwijk]

Dear Steve, Thanks for your reply. Not sure what is reasonable to expect! But every instance I've run on the IRIS FDSN service (pick your station and network) gets me within 0.1s. kasper Kasper van Wijk Department of physics University of Auckland https://www.physics.auckland.ac.nz/people/kvan637

…

On 9 February 2018 at 14:15, Steven Sherburn ***@***.***> wrote: Hi Kasper, Isn't it just how the data buffers boundaries fall relative to your request? If you grab the same month/day/time for 2017, the difference is only 0.05s. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#144 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AE253lvFMoQ1MDgSpsoecUWRoGaBQ5j-ks5tS5wVgaJpZM4R_M_W> .

[ozym]

Hi Kasper,

What you are seeing with these queries is a difference between the IRIS implementation of the FDSN waveform service and that of GeoNet.

In both cases the raw data is stored in miniSEED format blocks in an archive. The IRIS mechanics is to gather all the raw blocks that would match a request, then repackage them together and, at this point, trim the edges of the recovered data to match the request. This trimmed set is what's returned.

GeoNet, on the otherhand, simply gathers together all the raw miniSEED blocks that contain "any" data that has been requested and will return these blocks unchanged. The benefit, from the service point of view, is that it is much more efficient and results in a quicker response -- as the data itself doesn't need to be unpacked and examined. Another benefit is that the data returned is exactly what is currently stored in the archive. This may be useful in terms of time corrections, or data consistency checks, which tend to be stored in the block headers and would need to be interpolated.

The downside, however, is that the user is then required to extract the time window themselves from the returned data (along the lines of what IRIS does). Whether this is tricky or not will depend on the application the data is being used for and the software. There are others who use ObsPy more than I do who would know a good way of doing this, but the obspy.core.stream.Stream.trim & obspy.core.trace.Trace.trim functions look handy.

Mark