Timing information in Gemini Instruments

Summary

Here is a summary of the accuracy of the various instrument with respect to timing. Please read the section on each instrument below for further details. Only GMOS-N/S have been measured directly by looking at satellites and the other instruments have been analyzed by looking at the timestamps and their origin, more here

Instrument Name	Timing Accuracy	FITS Keyword
F2	see instrument section below for details	UTSTART and UTEND
GHOST	TBD	TBD
GMOS-N	0.207±0.02s	UTSTART and UTEND
GMOS-S	-0.164±0.03s	UTSTART and UTEND
GNIRS	see instrument section below for details	UTSTART and UTEND
GPI	see instrument section below for details	UTSTART and UTEND
GRACES	TBD	SHUTOPEN and SHUTCLOSE
GSAOI	see instrument section below for details	UTSTART and UTEND
NIFS	see instrument section below for details	UT
NIRI	see instrument section below for details	UTSTART and UTEND

Timing Basics

Accurate timing is critical for many science goals with Gemini instruments. The level of required accuracy may go from tens of seconds down millisecond accuracy. Here we have aimed to state what are the accuracies in the timing information in the Gemini instrument FITS headers, and most importantly state what FITS header keyword that should be used. The usage of the wrong keyword and not taking into account the event which the keyword refers to is the most common cause for an apparently wrong timing.

For historical reasons, the time stored in the Gemini FITS headers have many notation standards (UTC, ST, TT, local time) and it's important to understand this diversity in the notation to derive the most accurate time.
Also for historical reasons, it is not always clearly described in the FITS header what is the notation used; this page is intended to help with this.
Finally, it's important to understand the event the FITS keyword is referring to. For example, a keyword that refers to the start of an observation does not refer to the start of the actual exposure. Again, this page is intended to help you figure out which keywords mean what.

Known Facts

These are the known facts relevant to the timing information available for all instruments.

Millisecond accuracy in the timestamp itself.
Subsecond accuracy in mapping timestamp to the actual start of the exposure. Known limits are:
- In optical the Shutter opening and closing speed
- In IR we have a known unknown (the READ cycle) delay, this can be upto several seconds depending on the detector controller configuration.
The FITS header contains all necessary keywords
- Most keywords except UTSTART and UTEND (for most instruments) are NOT directly related to the exposure and set by the Seqexec
- FITS header help comments are not explicit in many cases on the timing standard being used. See the FITS header section for each instrument below.
- Most keywords set by the Seqexec (the TCL software controlling the instruments and telescope), these have variable delays from a few to several seconds.
  - Seqexec is not realtime
  - Set by the start of the event
  - Timestamp from shared memory channels, these have delays depending on the update frequency of the shared memory and query process.
Known hardware issues:
- Clock drifts in detector controllers
  - Controllers are old and not able to synchronize at enough frequency with the NTP. In some cases the
    synchronization done through a secondary computer that synchronizes with the local NTP server.
  - Misconfiguration of NTP server causing errors
    - iFor example the GPS misconfiguration in 2015-2016 affecting principally GMOS-S.

Timing Standards

There are many timing standards being used in the FITS headers and understanding their relations are critical.

International Atomic Time (TAI): A time scale that combines the output of some 400 highly precise atomic clocks worldwide.
Coordinated Universal Time (UTC): TAI corrected by leap seconds to correct for the slow down of Earth's rotation. Corrections are done to keep it within 1s from UT1. These corrections are planned 6 months in advance and happen either 23:59:59 June 30th or December 31st as determined by International Earth Rotation and Reference Systems Service (IERS) . Defined as:
- UTC=TAI-10s- “leap seconds” (applied since 1972). Today (2021) UTC≅TAI-37s
UT1: Computed from observations of distant quasars using VLBI, laser ranging of the Moon and artificial satellites, as well as the determination of GPS satellite orbits.
DUT1 = UT1 - UTC: UTC is maintained via leap seconds, such that DUT1 remains within the range −0.9 s < DUT1 < +0.9 s. The reason for this correction is partly that the rate of rotation of the Earth is not constant, due to tidal braking and the redistribution of mass within the Earth, including its oceans and atmosphere, and partly because the SI second (as now used for UTC) was already, when adopted, a little shorter than the current value of the second of mean solar time. Forecast values of DUT1 are published by IERS Bulletin A. The historical values for DUT1 if you want to convert from UT1 to UTC can be found here
Terrestial Time (TT): TT ≅ TAI + 32.184 seconds and thus
- TT ≅ UTC + 69.184s (2019)
Sidereal Time (ST)
Julian Date(JD) and Modified Julian Date (MJD): The Julian Day Number (JDN) is the integer assigned to a whole solar day in the Julian day count starting from noon Universal time, with Julian day number 0 assigned to the day starting at noon on Monday, January 1, 4713 BC. The Julian date (JD) of any instant is the Julian day number plus the fraction of a day since the preceding noon in Universal Time. The MJD has the epoch 0 on 0h Nov 17, 1858 and thus MJD = JD − 2400000.5. User should be that these epochs are given in the smoothly running TT reference frame.
GPS = TAI − 19.000 seconds.
Network Time protocol (NTP): Network synchronization that updates the clock and also corrects for computer clock drifts by synchronizing with a server.

The Gemini Timing reference and its accuracy

This is a short summary of the Time reference hardware being used at Gemini and its accuracy.

At Gemini the time reference is a local Network Time Protocol (NTP) server linked to a GPS at each site. With a recent upgrade to the NTP protocol, the accuracy is at the order of a few milliseconds.
All computers are synchronized to the NTP server on a regular basis. Offsets to the time standard and drifts are of the order of a few tenths of a millisecond. Thus the underlying accuracy is of that same order.

Events and timestamps

Each event gets a time-stamp at the moment the software/seqexec sends the trigger for the start of the event. Delays to the actual event may range from a second to several seconds, thus understanding what event and what process sets the time-stamp is cirtical in obtaining an accurate timing. In general this is the sequence on how a timestamp is set:

The SEQEXEC (a Tcl/Tk application running on a non real time system) starts an observation through an observer action.
The application obtains the timing value from the Telescope Control System (TCS) through an shared memor channel.
Sends this value to the Data Handling System (DHS).
Seqexec then sends the OBSERVE command to the instrument.

This pertains to ALL FITS keywords in the headers except UTSTART and UTEND for many instruments, and thus all keywords except these latter two are inherently including variable delays between the time-stamp and the start of the actual event.

At present, the UTSTART and UTEND keywords are written by the respective detector controllers for GMOS and GPI. Unless explicitly stated in the FITS Header details (below) for each instrument then these are two keywords that should be used as they are the best representations of the start and end of an exposure. See the separate section for each instrument on the pertinent details. It should be noted that:

The detector controllers are not real time systems
These controllers are mostly synched to a computer that in turn synchronizes to the main NTP Server.

Warning on the MJDOBS keyword

One of the most misinterpreted keywords is the FITS keyword MJDOBS. Contrary to the general assumption that it is UTC this is given in the TT (Terrestial Time) notation. JD and MJD actually correspond to the smoothly-running TT, not the UTC which is affected by leap seconds. Currently (Feb 2019), TT - UTC is about +69.2 s, which is very nearly the offset between the UT and MJD times in the headers. The definite link on JD and TT is https://www.iers.org/IERS/EN/Science/Recommendations/resolutionB1.html

F2

The F2 headers are typical in that they contain various FITS keywords with times in different notations and referring to different events. As for many other instruments we highly recommend using the UTSTART and UTEND even though the keywords are obtained through the TCS. Accuracy of TCS keywords are several seconds as there is a significant delay between TCS and DC starting.

FITS Keyword	Comment in header	Origin of Timestamp
LT	/ Local time at start of exposure	UTC + time zone correction
UT	/ UTC at start of exposure	UTC from TCS
DATE	/ UTC Date of observation (YYYY-MM-DD)	UTC from TCS
ST	/ Sidereal time at the start of the exposure	ST from TCS
TIME-OBS	/ Time of observation	UTC from TCS
DATE-OBS	/ UTC start date of exposure	UTC from TCS
UTSTART	/ UTC at observation start	UTC from TCS
UTEND	/ UTC at observation end	UTC from TCS
MJD-OBS	/ MJD of start of obseration	TT from DC, but large errors (AVOID)

The calculation of MJD time at the detector controller computer has a significant error and the time is off by as much as 2.5 hours.

GMOS

The GMOS headers are typical in that they contain various FITS keywords with times in different notations and referring to different events. As for many other instruments we highly recommend using the UTSTART and UTEND keywords for the highest accuracy.

FITS Keyword	Comment in header	Origin of Timestamp
LT	/ Local time at start of observation	UTC + time zone correction
UT	/ UT at observation start	UTC from TCS
DATE	/ UT Date of observation (YYYY-MM-DD)
ST	/ Sidereal time at the start of the exposure	ST from TCS
TIME-OBS	/ Time of observation	UTC from TCS
OBSEPOCH	/ Epoch at start of exposure
TIMESYS	/ Time system used
DATE-OBS	/ UT Date of observation (YYYY-MM-DD)
UTSTART	/ UT at observation start	UT1 from detector controller
UTEND	/ UT at observation end	UT1 from detector controller
EXPTIME	/ Exposure time in seconds
ELAPSED	/ Elapsed observation time in seconds
DARKTIME	/ Dark current integration in seconds
MJD-OBS	/ MJD of start of obseration	TT from TCS

GMOS UTSTART / UTEND. Before June 2 10:00 (GMT-3) 2016, these values came from the Seqexec, same as UT and TIME-OBS. The time is now read directly from GMOS DC, no TCS connection.
Shutter closure speed is approximately 10-15ms.
Observations of satellities indicate that the GMOS UTSTART keywords are ahead by -0.164±0.03s for GMOS-S, and 0.207±0.02s for GMOS-N.

GNIRS

The GNIRS headers are typical in that they contain various FITS keywords with times in different notations and referring to different events. Only MJD-OBS is set directly by the detector controller, see note below, UTSTART and UTEND keywords are set by the TCS so there are larger delays (see note below the table).

FITS Keyword	Comment in header	Origin of Timestamp
LT	/ Local time at start of observation	UTC + time zone correction
UT	/ UTC at start of exposure	UTC from TCS
DATE	/ UTC Date of observation (YYYY-MM-DD)	UTC from TCS
ST	/ Sidereal time at the start of the exposure	ST from TCS
TIME-OBS	/ Time of observation	UTC from TCS
DATE-OBS	/ UTC start date of exposure	UTC from TCS
UTSTART	/ UTC at observation start	UTC from TCS
UTEND	/ UTC at observation end	UTC from TCS
MJD-OBS	/ MJD of start of obseration	TT from GNIRS detector controller

LT, UT, TIME-OBS and UTSTART agree to within 5 sec.
MJD_OBS is calculated at the same time as the WCS just before the start of the observation. The value is saved in the header while saving the dataset. Note that even though this is set by the DC, the origin of the time input to the MJD calculation is currently uncertain. It drifts quickly (~3%) and is reset via a DC reboot. The instrument team is investigating the cause of this and it is recommended not to use this parameter.
UTEND - UTSTART = exposure time + detector readout time + communication time + response lag from the seqexec's point of view. One of the steps as the seqexec starts the exposure is to populate "before" keywords. One of the steps while processing the end of an observation is to populate the "after" keywords.

GPI

The GPI headers are typical in that they contain various FITS keywords with times in different notations and referring to different events. As for many other instruments we highly recommend using the UTSTART and UTEND keywords for the highest accuracy. The GPI pipeline is automatically correcting for the known errors in timing as related to the astrometric accuracy. Full details on timing corrections can be found in the paper by R. DeRosa et al. (2019) in Section 3.1 ., please follow those guidelines to achieve the best possible timing accuracy.

FITS Keyword	Comment in header	Origin of Timestamp
LT	/ Local time at start of exposure	UTC + time zone correction
UT	/ UTC at start of exposure	UTC from TCS
DATE	/ UTC Date of observation (YYYY-MM-DD)	UTC from TCS
ST	/ Sidereal time at the start of the exposure	ST from TCS
TIME-OBS	/ Time of observation	UTC from TCS
DATE-OBS	/ UTC start date of exposure	UTC from TCS
UTSTART	/ UTC at observation start	UT1 from detector controller
UTEND	/ UTC at observation end	UT1 from detector controller
MJD-OBS	/ MJD of start of obseration	TT from TCS

GRACES

The GRACES headers are typical in that they contain various FITS keywords with times in different notations and referring to different events. We strongly recommend the SHUTOPEN and SHUTOPEN keywords as they are are set directly by the shutter blade controller.

FITS Keyword	Comment in header	Origin of Timestamp
DATE	/ UTC date and time of file creation	NTP server with a connected GPS at CFHT
HSTTIME	/ Local time in Hawaii	NTP server with a connected GPS at CFHT
SHUTOPEN	/ UTC Shutter blade opening time	NTP server with a connected GPS at CFHT
SHUTCLOSE	/ TC Shutter blade closing time	NTP server with a connected GPS at CFHT
MJDDATE	/ Modified Julian date at the start of observation	TT calculated from SHUTOPEN

GSAOI

The GSAOI headers are typical in that they contain various FITS keywords with times in different notations and referring to different events. As for many other instruments we highly recommend using the UTSTART and UTEND even though the keywords are obtained through the TCS. Accuracy of TCS keywords are several seconds as there is a significant delay between TCS and DC starting.

FITS Keyword	Comment in header	Origin of Timestamp
LT	/ Local time at start of observation	UTC + time zone correction
UT	/ Beginning of Observation (UT)	UTC from TCS
DATE	/ UTC Date of observation (YYYY-MM-DD)	UTC from TCS
ST	/ Sidereal time at the start of the exposure	ST from TCS
TIME-OBS	/ Beginning of Observation (UT)	UTC from TCS
TIME	/ Beginning of Observation (UT)	UTC from TCS
DATE-OBS	/ UTC start date of exposure	UTC from TCS
UTSTART	/ UTC at observation start	UTC from TCS
UTEND	/ UTC at observation end	UTC from TCS
MJD-OBS	/ MJD of start of observation	TT from TCS
OBSEPOCH	/ Epoch at start of exposure	From DC in unknown standard (AVOID)

NIFS

The NIFS headers are typical in that they contain various FITS keywords with times in different notations and referring to different events. NO keywords are set directly by the detector controller, and there are no UTSTART and UTEND keywords. We recommend using the UT keyword that is set by the TCS so there are larger delays.

FITS Keyword	Comment in header	Origin of Timestamp
LT	/ Local time at start of observation	UTC + time zone correction
UT	/ Beginning of Observation (UT)	UTC from TCS
DATE	/ UTC Date of observation (YYYY-MM-DD)	UTC from TCS
ST	/ Sidereal time at the start of the exposure	ST from TCS
TIME-OBS	/ Time of observation	UTC from TCS
DATE-OBS	/ UTC start date of exposure	UTC from TCS
MJD-OBS	/ MJD of start of observation	TT from TCS

NIRI

The NIRI headers are typical in that they contain various FITS keywords with times in different notations and referring to different events. As for many other instruments we highly recommend using the UTSTART and UTEND even though the keywords are obtained through the TCS. Accuracy of TCS keywords are several seconds as there is a significant delay between TCS and DC starting.

FITS Keyword	Comment in header	Origin of Timestamp
LT	/ Local time at start of observation	UTC + time zone correction
UT	/ Beginning of Observation (UT)	UTC from TCS
DATE	/ UTC Date of observation (YYYY-MM-DD)	UTC from TCS
ST	/ Sidereal time at the start of the exposure	ST from TCS
TIME-OBS	/ Beginning of Observation (UT)	UTC from TCS
TIME	/ Beginning of Observation (UT)	UTC from TCS
DATE-OBS	/ UTC start date of exposure	UTC from TCS
UTSTART	/ UTC at observation start	UTC from TCS
UTEND	/ UTC at observation end	UTC from TCS
MJD-OBS	/ MJD of start of observation	TT from NIRI detector controller
OBSEPOCH	/ Epoch at start of exposure	From DC in unknown standard (AVOID)