Previous: Reducing dead-time
Up: Time-resolved observations with ISIS
Next: A note on accurate time and ISIS
Previous Page: Reducing dead-time
Next Page: A note on accurate time and ISIS
By reprogramming the CCD controller it should be possible to reduce dead-times
still further. For example, at present the CCDs are cleared prior to all
exposures. The charge is transferred to the output but not digitised. This
operation takes approximately 10 s per pixel, and hence 15 seconds for a
full EEV chip. There are alternatives. It is possible to flush the charge out
almost instantaneously by using all of the CCD clocks simultaneously instead of
one-by-one, as in the normal horizontal clocking process. Alternatively, since
each full-frame read-out essentially clears the CCD anyway, we could avoid
clearing the chip altogether, assuming that the previous frame was not
over-exposed and there is not much light leakage or dark current since the
last read-out. In addition, since the read-out noise is known to be a function
of read-out speed, it should be possible to offer a greater variety of read-out
speeds and noise levels.
For even greater reductions in dead-time, it should be possible to store a
series of short exposures on-chip. The interval between exposures can be very
short, especially if the image-window is almost one-dimensional, ie.
spectroscopic. A small box, defined by a dekker, consisting of rows at the
top of the CCD is illuminated. On completion, the box is moved quickly down the
chip and the clear area can be exposed again. This can be repeated until the
chip is full. The whole frame can then be read-out from the bottom of the CCD.
The exposure time
per box can be in the range
ms. For an EEV chip
it takes about 0.04 ms to shift down one row of data; thus the total time per
box
ms. For a box of size 10 rows, and a 60 s exposure, the
total time interval between boxes would be 60.0004 s. Even if such a system
were available now, it would still be impaired by the relatively slow DMS-VAX
link. However, work is in progress to replace this next year with a system
which saves the data directly onto the observing SPARCstation. This should
provide an order of magnitude improvement in the dead-time due to archiving
data on disk.
It should be noted that most of the above enhancements are still at the drawing-board stage and it is by no means clear when, if ever, they will become available to the community; this will depend on demand from users and the availability of engineering effort.