This page is part of the ING document
manual for UltraDAS
The detailed handling of integration timing is quite subtle.
This section tells you what to expect in the various records for the different
kinds of observation that UltraDAS can do.
Integration vs. exposure
For a CCD, the integration starts when the CCD is cleared, and the integration
is counted from the end of clearing. The integration is deemed to end when
the CCD starts to read out. The exposure starts when the shutter opens
and ends when the shutter closes. The exposure starts after the integration
and finishes before the integration; the exposed time is slightly shorter
than the integrated time. The actual exposure and the effective exposure
are identical for this case.
An integration is defined as the interval during which the camera's
clocks are set to trap and retain charge. The integration is delimited
by the behaviour of the readout electronics, and is entirely independent
of the exposure made by the shutter.
An exposure is the interval during which the detector is exposed
to incident light by the shutter. The exposure is entirely independent
of the integration. If the camera has no shutter (e.g. INGRID), then there
is one continuous exposure.
The effective exposure is the interval in which the camera is both
exposing and integrating.
The effective integration for a particular readout is the time from
the start of integration to the start of readout; this distinction is important
for IR cameras.
For cameras that don't have a shutter, the actual exposure is continuous
through all integrations. Effective exposure times are identical to the
matching integration times. This mode applies to cameras that have a shutter
if the shutter is opened explicitly before the observation: for this case,
UltraDAS treats the camera as shutterless.
For CCDs, readout destroys the charge pattern on the detector by moving
the charge into the output amplifier. This means that readout ends the
integration. IR devices (e.g. INGRID) typically do non-destructive readout
where the stored charge is measured in situ. This means that the
effective exposure stops at readout, but that the actual exposure goes
on through readout and afterwards.
"Idling" vs. integrating
Charge from light leaks and thermal noise builds up on the detector between
observations. If this charge is integrated by the detector, it may not
be completely cleared away by the clear cycle of the next observation.
In that case, the observation will be contaminated by extra counts. (Often,
this appears as a ramp in the background leading up to a saturated region
in the low-numbered rows.)
To avoid this problem, the detector is made to clear itself continuously
between observations. This is called "idling", and is reported as such
on the mimic.
In the mimic:
This detail affects both the start times and the durations. One result
of this arrangement is that the displayed integration for INGRID and similar
IR cameras will continue to count up throughout the observation. The final
integration time at the end of observation will be higher than the effective
integration time of the last readout in that observation.
The actual integration is shown.
The effective exposure is shown.
In the FITS header:
The effective integration length is shown, under keywords ELAPSED
The effective exposure length is shown, under keywords EXPOSED and
The value with the UTSTART keyword is either the start of the effective
exposure (if the detector was exposed in this observation or the
start of the effective integration (if there was no exposure).
The date associated with UTSTART is recorded under DATE-OBS.
Source and accuracy of timing
Exposures delimited by shutter movements are timed by the clock in the
SDSU detector-controller. This clock only measures the length of the exposure,
not the absolute start-time. The exposure is considered to start when the
shutter starts to open and to end when the shutter finishes closing.
Start-times of exposures and integrations and lengths of integrations
(and by extension, lengths of shutterless exposures) are measured from
the clock of the DAS computer. This clock gives absolute time. The computer
clock is regulated by network time protocol (NTP) and its accuracy is limited
by the length of the network path to the primary time-server.
All time measurements are precise to the nearest millisecond.
Accuracy of timing is difficult to test. We believe that these are the
If funds are made available, it is possible to add a GPS time-service to
each DAS computer such that the absolute timings are accurate to better
than a microsecond.
The response time of the shutter makes shutter-timed exposures inaccurate
at the level of tens of milliseconds.
The NTP regulation of the computer clock is quite poor at present: individual
times are likely to be inaccurate at the level of ~100ms.
NTP should prevent any long-term drift of the computer clock.