A fundamental issue with measuring the exposure start time with the AP47p (and many other commercial
CCD's) is that the camera only records the exposure start time to whole seconds,
but NEO work can demand time to be recorded significantly more accurately than this.
To get around this limitation the PC's clock is set to better than one second
accuracy and then exposures are started at the moment the PC's clock changes from one second to the
next. If both are done correctly then the actual exposure mid-time can be
determined to significantly better than one second accuracy.
During an observing run the PC is connected to the Internet and the
freeware utility Dimension4 is used to
keep the PC accurately aligned to UT.
A Visual Basic program called CCDCamCtl is used to initiate one or more
exposures. Before each one it waits for the PC clock to change from one
integer second to the next. As soon as it does the program instructs Maxim DL/CCD
to open the camera shutter. The number of milliseconds between this command and the
program receiving control back again is measured. Generally just a
few hundredths of a second elapse from the PC clock second changing and control
returning to the program. The start time + the millisecond count is used to set
the Exposure start time FITS header keyword with a value to three decimal
From the fastest NEOs measured at Great Shefford with this setup the timing error
appears to be of the order of 1/3 second or better in absolute terms, the
actual amount being masked by a combination of the real clock inaccuracy and all other
components contributing errors (e.g. star catalogue positions, imperfect optics etc.)
an example, click
below to see a stack of images of 2003 DW10, taken during its close approach to
Earth on 2003 March 2.8 UT when it was moving at over 150 arcsec per minute:
a close approach like this is probably the easiest way to check in absolute
terms the precision of time measurement, which can be rather difficult to
The residuals for 2003 DW10 published by the MPC for
the 24 positions reported from J95 on 2003 March 2 were used to give some idea
of the maximum timing error by assuming each residual was entirely due to
errors in timing.
The standard deviation of this error works out to be
equivalent to a timing error of +/- 0.34 seconds.
Other errors not considered here (e.g. star catalogue errors) are likely to reduce the actual
error solely due to timing to slightly less than this value.