Previous: Sources of dead-time
Up: Time-resolved observations with ISIS
Next: Future developments
Previous Page: Sources of dead-time
Next Page: Future developments

Reducing dead-time

There are a number of ways in which dead-time can be reduced. Below we list the methods currently available and give some indication of how effective they each are.

Using MULTRUN -- This is a useful ICL command which should be used when it is necessary to take repeated exposures of a single object. The procedure does not do anything particularly clever -- it simply sets up an exposure loop which removes the dead-time involved in typing repeated RUN commands. The loop can be aborted by typing KILLMULT. The saving in dead-time is surprising, particularly towards the end of a night!

Windowing the chip -- By windowing a CCD, only a subset of the chip is read-out and transferred to the VAX. This significantly reduces read-out time by an amount which scales with the number of pixels in the window. It makes little difference if this window is chosen at the top, middle or bottom of the chip. The only disadvantages of this method are that one obtains a reduced field of view and reduces or removes entirely the bias strip. The former is not really a problem for spectroscopy and the latter can be resolved by allocating multiple windows, so that one window contains the spectrum and another the bias strip (although the multiple windows option will not be available until next year).

Binning the chip -- With the CCD controller it is possible to change the clocking routines to provide on-chip binning in which the summed charge from several adjacent pixels is read out as a single charge. This reduces the number of pixels which are read out and transferred and therefore also reduces the read-out time by a factor which depends on the number of pixels in the binned chip. An added advantage of binning is that the summed charge from the adjacent pixels is measured with the read-out noise corresponding to a single pixel. Hence the effective read-out noise is reduced by a factor equal to the square root of the number of pixels in the bin. The disadvantage of binning is that it reduces spatial resolution and the dynamic range of the chip, since the larger pixels still have the count limit of a single pixel. If neither of these is a problem then the use of on-chip binning on its own or in combination with windowing should be considered.

Selecting the read-out speed -- There are currently two read-out speeds available when using ISIS -- READOUT_SPEED FAST and READOUT_SPEED STANDARD; the speed must be typed in upper case. For an EEV chip, the full-frame read-out time in standard mode is approximately 110 s with a read-out noise of approximately 4 electrons. The equivalent figures for fast mode are 70 s and 5 electrons. The fast read-out speed also gives a slightly larger dynamic range. Clearly, if the observations are sky limited there is little to gain in terms of signal-to-noise by using STANDARD, although there is evidence that the charge transfer may be poorer; please consult your support astronomer for details.

Accumulating the data in the DMS -- It is possible to accumulate data in the DMS using datacubes. These are three-dimensional data structures in which individual frames form the planes. This removes the significant dead-time due to transferring data from the DMS to the VAX. Using datacubes it is possible to obtain dead-times of approximately 5 s with a CCD in fast read-out mode and windowed . The DMS can hold datacubes of up to 32 MBytes in size and hence approximately 800 frames of pixels. There are two drawbacks to this method: Firstly, no file headers and hence exposure start times are saved, and secondly, there is a large overhead when the datacube is transferred to the VAX. At the present time, the datacube option is not offered on a routine basis and any users who are interested in employing this method should first contact their support astronomer.

Using the IPCS -- In terms of signal-to-noise, there are few occasions when it is advisable to use the IPCS on the blue arm of ISIS instead of the TEK CCD. However, the IPCS does have the advantage of a lower dead-time. The IPCS is not an integrating detector. This means that as the individual photons are recorded, their positions are immediately passed to the DMS. At the end of an exposure, there is no chip read-out, only the normal overhead of writing the data to the VAX (and even this can be overcome by using datacubes in the DMS). It is even possible to time-tag each photon, so that one can choose the time-resolution and signal-to-noise per frame at the data reduction stage. However, this mode is still to be fully tested. If the IPCS can produce an acceptable signal-to-noise and the very-highest time resolution is required, it should be considered as the detector for the blue arm of ISIS.