From s.dalla at manchester.ac.uk Thu Mar 2 04:52:39 2006 From: s.dalla at manchester.ac.uk (Silvia Dalla) Date: Thu, 2 Mar 2006 12:52:39 +0000 (GMT) Subject: Request for modification of UCD: time.obs Message-ID: hi all, I propose modifying 'time.obs'. This UCD is currently defined as: 'Observation on-time, duration' However we are missing a UCD that identifies the time stamp associated with an observation: for example if data are collected between a 'time.obs.start' and a 'time.obs.end', frequently the mid point of this time interval is used to define the time of the observation. I propose that 'time.obs' should be 'time stamp associated to the observation'. A new UCD could be used to identify the duration of the observation: the obvious choice would be 'time.obs.duration'. Best wishes, Silvia ------------------------------------ Silvia Dalla School of Physics and Astronomy University of Manchester PO Box 88 Manchester M60 1QD UK Tel +44-161-306 8705 Fax +44-161-306 3941 ------------------------------------ From sla at ucolick.org Thu Mar 2 15:13:14 2006 From: sla at ucolick.org (Steve Allen) Date: Thu, 2 Mar 2006 15:13:14 -0800 Subject: Request for modification of UCD: time.obs In-Reply-To: References: Message-ID: <20060302231314.GA32136@ucolick.org> On Thu 2006-03-02T12:52:39 +0000, Silvia Dalla hath writ: > I propose modifying 'time.obs'. > > This UCD is currently defined as: 'Observation on-time, duration' > > However we are missing a UCD that identifies the time stamp associated > with an observation: for example if data are collected between a > 'time.obs.start' and a 'time.obs.end', frequently the mid point of > this time interval is used to define the time of the observation. > > I propose that 'time.obs' should be 'time stamp associated to the > observation'. > > A new UCD could be used to identify the duration of the observation: > the obvious choice would be 'time.obs.duration'. There are several FITS reserved keywords with definitions which are closely related to these UCDs. It would be dismaying if the UCDs and the FITS keywords became dissonant rather than consonant. First a quick review... >From the original Wells et al. paper there is DATE-OBS The current FITS standard asserts that this should be assume to refer to the start of an observation unless another interpretation is clearly explained in the comment field. In the original paper by Wells et al. this keyword was defined simply as "date of data acquisition". As a result not all instances of its use conform to the suggestion (added nearly two decades afterwards) that it refer to the start of an observation. FITS WCS Paper II (celestial coordinates) reserved MJD-OBS with semantics identical to DATE-OBS except for being expressed as a floating point count of days rather than as a character string indicating a calendar date. FITS WCS Paper III (spectral coordinates) reserved MJD-AVG and DATE-AVG with semantics indicating that the value should refer to an average date (of a presumably extended observation). Now I switch back to UCDs. DATE-OBS seems to correspond best to Silvia Dalla's suggestion for a change to "time.obs". DATE-AVG seems to correspond best to "time.obs;stat.mean". But that calls up a question of whether the current semantics for "time.obs", which indicate a duration, are not begging for a new UCD entry in the stat group -- should there be a UCD something like "stat.range" or "stat.span" to indicate the difference between the maximum and minimum of a quantity? I wonder why there is "time.obs.start" instead of "time.obs;stat.min" and similarly "time.obs.end" instead of "time.obs;stat.max" But upon reading closely through http://www.ivoa.net/Documents/latest/UCDlist.html I am left mystified by the distinction in semantics between the UCDs "time.obs" and "time.expo" Is there a more detailed explanation of the difference in the semantics of these two quantities? -- Steve Allen WGS-84 (GPS) UCO/Lick Observatory Natural Sciences II, Room 165 Lat +36.99858 University of California Voice: +1 831 459 3046 Lng -122.06014 Santa Cruz, CA 95064 http://www.ucolick.org/~sla/ Hgt +250 m From shaw at noao.edu Thu Mar 2 15:37:51 2006 From: shaw at noao.edu (Dick Shaw) Date: Thu, 02 Mar 2006 16:37:51 -0700 Subject: Request for modification of UCD: time.obs In-Reply-To: References: Message-ID: There is certainly merit in making a distinction among quantities such as these. For example, one needs to be able to describe the interval of time over which the data were obtained, as well as the effective duration of the exposure. For example: an integration may be started, interrupted, and then resumed. Or, multiple exposures of a region of sky may be obtained separately and later combined. This is common in ground-based IR, where sky brightness is an issue, or in the optical if the consecutive exposures are obtained to track a time-variable phenomenon, or merely to dither images spatially to remove artifacts such as seams in a large, focal plane array. In these cases the interval over which photons were collected is shorter than the difference between the start of the first exposure and the end of the last. Somehow this information needs to be recorded for high-level (i.e., combined) data products. Depending upon the degree to which one wishes to obsess about the details, a richer semantic description may be in order. Regards, Dick Shaw On Thu, 2 Mar 2006 12:52:39 +0000 (GMT) Silvia Dalla wrote: > > hi all, > > I propose modifying 'time.obs'. > > This UCD is currently defined as: 'Observation on-time, duration' > > However we are missing a UCD that identifies the time stamp associated > with an observation: for example if data are collected between a > 'time.obs.start' and a 'time.obs.end', frequently the mid point of > this time interval is used to define the time of the observation. > > I propose that 'time.obs' should be 'time stamp associated to the > observation'. > > A new UCD could be used to identify the duration of the observation: > the obvious choice would be 'time.obs.duration'. > > Best wishes, > Silvia > > ------------------------------------ > Silvia Dalla > School of Physics and Astronomy > University of Manchester > PO Box 88 > Manchester M60 1QD > UK > > Tel +44-161-306 8705 >Fax +44-161-306 3941 > ------------------------------------ #--- Dr. Richard A. Shaw, Scientist National Optical Astronomy Observatory 950 N. Cherry Avenue Tucson, AZ 85719 USA 520-318-8398 From seaman at noao.edu Thu Mar 2 15:54:57 2006 From: seaman at noao.edu (Rob Seaman) Date: Thu, 2 Mar 2006 16:54:57 -0700 Subject: Request for modification of UCD: time.obs In-Reply-To: <20060302231314.GA32136@ucolick.org> References: <20060302231314.GA32136@ucolick.org> Message-ID: <3029B3F6-05D8-443B-A272-5862A9E439EE@noao.edu> On Mar 2, 2006, at 4:13 PM, Steve Allen wrote: > It would be dismaying if the UCDs and the FITS keywords became > dissonant rather than consonant. Most certainly agree! Wouldn't mind, however, if the FITS standards process were to be informed by VO standards, not just the other way around :-) > DATE-OBS > The current FITS standard asserts that this should be assume to refer > to the start of an observation unless another interpretation is > clearly explained in the comment field. Have also seen headers with all variations of beginning, middle, and ending timestamps (and not all provided via DATE-OBS) and even both beginning and ending time stamps with no explicit exposure duration per se. Not clear that one size can, or should, fit all. Should also point out that the syntax and semantics of DATE-OBS changed significantly due to Y2K. The original usage was to specify only a calendar date. Current usage can combine date and time into an atomic whole. > DATE-AVG seems to correspond best to "time.obs;stat.mean". Just as there are various statistical measures of central tendency, an "average" representative timestamp for an observation may reasonably mean many different things. Various instruments, for instance, allow pausing and resuming an exposure for arbitrary periods. The camera shutter (if any) may not even be open during the claimed midpoint of an observation. Other instruments, for instance drift scans, have significantly different representative timestamps for different pixels. A shuttered camera on a space platform may have one (scalar) timestamp for captured "science" photons and another (vector) timestamp for radiation events that continue to accumulate during readout. Even simple (ground-based, shuttered, etc.) observations might benefit from correcting the overt midpoint timestamp for varying airmass during the observation. (Alternately, an "effective" airmass might be computed.) Hesitate to refer folks, yet again, to STC. Rob Seaman NOAO From andrea.preitemartinez at iasf-roma.inaf.it Fri Mar 3 02:52:27 2006 From: andrea.preitemartinez at iasf-roma.inaf.it (Andrea Preite Martinez) Date: Fri, 03 Mar 2006 11:52:27 +0100 Subject: UCD time.obs Message-ID: <200603042244.k24MiScC004217@mercury.hq.eso.org> Hi Silvia, Steve, Dick, Rob, let me extract from the discussion on the UCD time.obs three points: We should not mix up a "duration" (interval of time) with an "instant" (currently date/time), or different instants during the "duration". This semantic difference is reflected in the present list of UCDs. 1. time-instant: Astronomer are used to characterize their observations by a time-stamp (a UNIQUE time-stamp) that is reasonably to consider some sort of average between the beginning and the end of an observation. This stamp/label is called EPOCH of the observation. The corresponding UCD is time.epoch . 2. time-duration: First of all we need to consider that modern observations are often (I dare say always in the case of observations from space) the result of multiple exposures. While 30 years ago there was no semantic difference between observation and exposure, today we need to be precise, one observation being the result of more than one exposure, leading to an observation-duration usually greater than a single exposure-duration The UCDs time.obs and time.expo reflect this difference. 3. start/end, and the correspondance between FITS keywords and UCDs. I'm aware of the importance that FITS keywords still have in describing our data, but I'm also aware of the fact that (expecially in the "time" domain!!) there is a tendency to force (to forge?) the initial meaning and to proliferate keywords (Rob is right!). I think that we already have the UCDs corresponding to the quoted keywords: DATE-OBS = MJD-OBS : time.obs.start (time-stamp of the beginning of the observation) DATE-AVG = MJD-AVG : time.epoch[;obs] (epoch of the observation) I also agree with Steve sentence /> It would be dismaying if the UCDs/ /> and the FITS keywords became dissonant rather than consonant/ and Rob's comment to it, but I interpret his warning as: we should not make the same mistakes! Why start/end and not min/max: First of all for esthetic reasons. Most seriously, because in my opinion min/max whould indicate a min/max DURATION, while start/end are certainly referring, correctly, to two INSTANTs of time. Finally, start/end are the words that you find in the description of the data, not min/max. Andrea ------------------------------------------------------------------------ Andrea Preite Martinez andrea.preitemartinez-at-iasf-roma.inaf.it IASF Tel.IASF:+39.06.4993.4651 Via del Fosso del Cavaliere 100 Tel.CDS :+33.3.90242473 I-00133 Roma Cell.1 :+39.320.43.15.383 Cell.2 :+39.335.38.17.355 =================================================================================== From sla at ucolick.org Sat Mar 4 22:57:15 2006 From: sla at ucolick.org (Steve Allen) Date: Sat, 4 Mar 2006 22:57:15 -0800 Subject: UCD time.obs In-Reply-To: <200603042244.k24MiScC004217@mercury.hq.eso.org> References: <200603042244.k24MiScC004217@mercury.hq.eso.org> Message-ID: <20060305065715.GA13916@ucolick.org> On Fri 2006-03-03T11:52:27 +0100, Andrea Preite Martinez hath writ: > We should not mix up a "duration" (interval of time) with an "instant" > (currently date/time), or different instants during the "duration". This > semantic difference is reflected in the present list of UCDs. I agree. My confusion stems from the fact that the semantic definition for all of the UCD words consists of an incomplete sentence. In the absence of detailed descriptions such as those provided in this note I did not understand which ones were instant or duration or that observation is a superset of exposure. This may be true of other, non-temporal, UCD words. For the sake of averting subsequent misinterpretations, is there a way to include richer semantic definitions as part of the document itself? Or alternatively, to include an explicit, and obvious, reference to a separate "usage" document with examples, correspondences with FITS keywords, etc? -- Steve Allen WGS-84 (GPS) UCO/Lick Observatory Natural Sciences II, Room 165 Lat +36.99858 University of California Voice: +1 831 459 3046 Lng -122.06014 Santa Cruz, CA 95064 http://www.ucolick.org/~sla/ Hgt +250 m From s.dalla at manchester.ac.uk Thu Mar 2 04:52:39 2006 From: s.dalla at manchester.ac.uk (Silvia Dalla) Date: Thu, 2 Mar 2006 12:52:39 +0000 (GMT) Subject: Request for modification of UCD: time.obs Message-ID: hi all, I propose modifying 'time.obs'. This UCD is currently defined as: 'Observation on-time, duration' However we are missing a UCD that identifies the time stamp associated with an observation: for example if data are collected between a 'time.obs.start' and a 'time.obs.end', frequently the mid point of this time interval is used to define the time of the observation. I propose that 'time.obs' should be 'time stamp associated to the observation'. A new UCD could be used to identify the duration of the observation: the obvious choice would be 'time.obs.duration'. Best wishes, Silvia ------------------------------------ Silvia Dalla School of Physics and Astronomy University of Manchester PO Box 88 Manchester M60 1QD UK Tel +44-161-306 8705 Fax +44-161-306 3941 ------------------------------------ From sla at ucolick.org Thu Mar 2 15:13:14 2006 From: sla at ucolick.org (Steve Allen) Date: Thu, 2 Mar 2006 15:13:14 -0800 Subject: Request for modification of UCD: time.obs In-Reply-To: References: Message-ID: <20060302231314.GA32136@ucolick.org> On Thu 2006-03-02T12:52:39 +0000, Silvia Dalla hath writ: > I propose modifying 'time.obs'. > > This UCD is currently defined as: 'Observation on-time, duration' > > However we are missing a UCD that identifies the time stamp associated > with an observation: for example if data are collected between a > 'time.obs.start' and a 'time.obs.end', frequently the mid point of > this time interval is used to define the time of the observation. > > I propose that 'time.obs' should be 'time stamp associated to the > observation'. > > A new UCD could be used to identify the duration of the observation: > the obvious choice would be 'time.obs.duration'. There are several FITS reserved keywords with definitions which are closely related to these UCDs. It would be dismaying if the UCDs and the FITS keywords became dissonant rather than consonant. First a quick review... >From the original Wells et al. paper there is DATE-OBS The current FITS standard asserts that this should be assume to refer to the start of an observation unless another interpretation is clearly explained in the comment field. In the original paper by Wells et al. this keyword was defined simply as "date of data acquisition". As a result not all instances of its use conform to the suggestion (added nearly two decades afterwards) that it refer to the start of an observation. FITS WCS Paper II (celestial coordinates) reserved MJD-OBS with semantics identical to DATE-OBS except for being expressed as a floating point count of days rather than as a character string indicating a calendar date. FITS WCS Paper III (spectral coordinates) reserved MJD-AVG and DATE-AVG with semantics indicating that the value should refer to an average date (of a presumably extended observation). Now I switch back to UCDs. DATE-OBS seems to correspond best to Silvia Dalla's suggestion for a change to "time.obs". DATE-AVG seems to correspond best to "time.obs;stat.mean". But that calls up a question of whether the current semantics for "time.obs", which indicate a duration, are not begging for a new UCD entry in the stat group -- should there be a UCD something like "stat.range" or "stat.span" to indicate the difference between the maximum and minimum of a quantity? I wonder why there is "time.obs.start" instead of "time.obs;stat.min" and similarly "time.obs.end" instead of "time.obs;stat.max" But upon reading closely through http://www.ivoa.net/Documents/latest/UCDlist.html I am left mystified by the distinction in semantics between the UCDs "time.obs" and "time.expo" Is there a more detailed explanation of the difference in the semantics of these two quantities? -- Steve Allen WGS-84 (GPS) UCO/Lick Observatory Natural Sciences II, Room 165 Lat +36.99858 University of California Voice: +1 831 459 3046 Lng -122.06014 Santa Cruz, CA 95064 http://www.ucolick.org/~sla/ Hgt +250 m From shaw at noao.edu Thu Mar 2 15:37:51 2006 From: shaw at noao.edu (Dick Shaw) Date: Thu, 02 Mar 2006 16:37:51 -0700 Subject: Request for modification of UCD: time.obs In-Reply-To: References: Message-ID: There is certainly merit in making a distinction among quantities such as these. For example, one needs to be able to describe the interval of time over which the data were obtained, as well as the effective duration of the exposure. For example: an integration may be started, interrupted, and then resumed. Or, multiple exposures of a region of sky may be obtained separately and later combined. This is common in ground-based IR, where sky brightness is an issue, or in the optical if the consecutive exposures are obtained to track a time-variable phenomenon, or merely to dither images spatially to remove artifacts such as seams in a large, focal plane array. In these cases the interval over which photons were collected is shorter than the difference between the start of the first exposure and the end of the last. Somehow this information needs to be recorded for high-level (i.e., combined) data products. Depending upon the degree to which one wishes to obsess about the details, a richer semantic description may be in order. Regards, Dick Shaw On Thu, 2 Mar 2006 12:52:39 +0000 (GMT) Silvia Dalla wrote: > > hi all, > > I propose modifying 'time.obs'. > > This UCD is currently defined as: 'Observation on-time, duration' > > However we are missing a UCD that identifies the time stamp associated > with an observation: for example if data are collected between a > 'time.obs.start' and a 'time.obs.end', frequently the mid point of > this time interval is used to define the time of the observation. > > I propose that 'time.obs' should be 'time stamp associated to the > observation'. > > A new UCD could be used to identify the duration of the observation: > the obvious choice would be 'time.obs.duration'. > > Best wishes, > Silvia > > ------------------------------------ > Silvia Dalla > School of Physics and Astronomy > University of Manchester > PO Box 88 > Manchester M60 1QD > UK > > Tel +44-161-306 8705 >Fax +44-161-306 3941 > ------------------------------------ #--- Dr. Richard A. Shaw, Scientist National Optical Astronomy Observatory 950 N. Cherry Avenue Tucson, AZ 85719 USA 520-318-8398 From seaman at noao.edu Thu Mar 2 15:54:57 2006 From: seaman at noao.edu (Rob Seaman) Date: Thu, 2 Mar 2006 16:54:57 -0700 Subject: Request for modification of UCD: time.obs In-Reply-To: <20060302231314.GA32136@ucolick.org> References: <20060302231314.GA32136@ucolick.org> Message-ID: <3029B3F6-05D8-443B-A272-5862A9E439EE@noao.edu> On Mar 2, 2006, at 4:13 PM, Steve Allen wrote: > It would be dismaying if the UCDs and the FITS keywords became > dissonant rather than consonant. Most certainly agree! Wouldn't mind, however, if the FITS standards process were to be informed by VO standards, not just the other way around :-) > DATE-OBS > The current FITS standard asserts that this should be assume to refer > to the start of an observation unless another interpretation is > clearly explained in the comment field. Have also seen headers with all variations of beginning, middle, and ending timestamps (and not all provided via DATE-OBS) and even both beginning and ending time stamps with no explicit exposure duration per se. Not clear that one size can, or should, fit all. Should also point out that the syntax and semantics of DATE-OBS changed significantly due to Y2K. The original usage was to specify only a calendar date. Current usage can combine date and time into an atomic whole. > DATE-AVG seems to correspond best to "time.obs;stat.mean". Just as there are various statistical measures of central tendency, an "average" representative timestamp for an observation may reasonably mean many different things. Various instruments, for instance, allow pausing and resuming an exposure for arbitrary periods. The camera shutter (if any) may not even be open during the claimed midpoint of an observation. Other instruments, for instance drift scans, have significantly different representative timestamps for different pixels. A shuttered camera on a space platform may have one (scalar) timestamp for captured "science" photons and another (vector) timestamp for radiation events that continue to accumulate during readout. Even simple (ground-based, shuttered, etc.) observations might benefit from correcting the overt midpoint timestamp for varying airmass during the observation. (Alternately, an "effective" airmass might be computed.) Hesitate to refer folks, yet again, to STC. Rob Seaman NOAO From andrea.preitemartinez at iasf-roma.inaf.it Fri Mar 3 02:52:27 2006 From: andrea.preitemartinez at iasf-roma.inaf.it (Andrea Preite Martinez) Date: Fri, 03 Mar 2006 11:52:27 +0100 Subject: UCD time.obs Message-ID: <200603042244.k24MiScC004217@mercury.hq.eso.org> Hi Silvia, Steve, Dick, Rob, let me extract from the discussion on the UCD time.obs three points: We should not mix up a "duration" (interval of time) with an "instant" (currently date/time), or different instants during the "duration". This semantic difference is reflected in the present list of UCDs. 1. time-instant: Astronomer are used to characterize their observations by a time-stamp (a UNIQUE time-stamp) that is reasonably to consider some sort of average between the beginning and the end of an observation. This stamp/label is called EPOCH of the observation. The corresponding UCD is time.epoch . 2. time-duration: First of all we need to consider that modern observations are often (I dare say always in the case of observations from space) the result of multiple exposures. While 30 years ago there was no semantic difference between observation and exposure, today we need to be precise, one observation being the result of more than one exposure, leading to an observation-duration usually greater than a single exposure-duration The UCDs time.obs and time.expo reflect this difference. 3. start/end, and the correspondance between FITS keywords and UCDs. I'm aware of the importance that FITS keywords still have in describing our data, but I'm also aware of the fact that (expecially in the "time" domain!!) there is a tendency to force (to forge?) the initial meaning and to proliferate keywords (Rob is right!). I think that we already have the UCDs corresponding to the quoted keywords: DATE-OBS = MJD-OBS : time.obs.start (time-stamp of the beginning of the observation) DATE-AVG = MJD-AVG : time.epoch[;obs] (epoch of the observation) I also agree with Steve sentence /> It would be dismaying if the UCDs/ /> and the FITS keywords became dissonant rather than consonant/ and Rob's comment to it, but I interpret his warning as: we should not make the same mistakes! Why start/end and not min/max: First of all for esthetic reasons. Most seriously, because in my opinion min/max whould indicate a min/max DURATION, while start/end are certainly referring, correctly, to two INSTANTs of time. Finally, start/end are the words that you find in the description of the data, not min/max. Andrea ------------------------------------------------------------------------ Andrea Preite Martinez andrea.preitemartinez-at-iasf-roma.inaf.it IASF Tel.IASF:+39.06.4993.4651 Via del Fosso del Cavaliere 100 Tel.CDS :+33.3.90242473 I-00133 Roma Cell.1 :+39.320.43.15.383 Cell.2 :+39.335.38.17.355 =================================================================================== From sla at ucolick.org Sat Mar 4 22:57:15 2006 From: sla at ucolick.org (Steve Allen) Date: Sat, 4 Mar 2006 22:57:15 -0800 Subject: UCD time.obs In-Reply-To: <200603042244.k24MiScC004217@mercury.hq.eso.org> References: <200603042244.k24MiScC004217@mercury.hq.eso.org> Message-ID: <20060305065715.GA13916@ucolick.org> On Fri 2006-03-03T11:52:27 +0100, Andrea Preite Martinez hath writ: > We should not mix up a "duration" (interval of time) with an "instant" > (currently date/time), or different instants during the "duration". This > semantic difference is reflected in the present list of UCDs. I agree. My confusion stems from the fact that the semantic definition for all of the UCD words consists of an incomplete sentence. In the absence of detailed descriptions such as those provided in this note I did not understand which ones were instant or duration or that observation is a superset of exposure. This may be true of other, non-temporal, UCD words. For the sake of averting subsequent misinterpretations, is there a way to include richer semantic definitions as part of the document itself? Or alternatively, to include an explicit, and obvious, reference to a separate "usage" document with examples, correspondences with FITS keywords, etc? -- Steve Allen WGS-84 (GPS) UCO/Lick Observatory Natural Sciences II, Room 165 Lat +36.99858 University of California Voice: +1 831 459 3046 Lng -122.06014 Santa Cruz, CA 95064 http://www.ucolick.org/~sla/ Hgt +250 m