3。7 x 10-**cm but take an energy level of 0。44 eV。 Banghart

et  al  [9] give vyNp  —  4。11      ]  25  (T/77)*cm‘2 $-1 and  with

these data we calculate zq = l4qs at l5°C (assuming no field enhancement)。 This value is much less than the time for reading one line (= 850qs)。 Hence during readout the charge will be deferred into the trailing pixel only。 On the  other hand, the line move time during frame transfer (and for fast dumping of unwanted  parts of the image) is 2ps - much  less

that Tax So during these transfers charge is deferred over many pixels。

These effects are illustrated in figure 2 which shows the charge deferred into the next trailing pixel in the column direction for a clocking scheme which has a frame transfer (576 lines) plus three normal readouts of 64 lines interspersed with two fast dumps of 192 lines each。 The slope of the plot does not give the same CTE as figure 1 since the horizontal axis shows the total number of transfers (slow readouts plus fast dumps)。 If this is plotted in terms of just the slow readouts the CTE is tlle same as in figure 1。 Hence we see that the effects of radiation-induced traps depends on the clocking sequence used。 It is interesting to note that the first data value in figure 2 shows a deferred charge of =2%。 This can only be because there are also a significant number of traps which can defer charge into just one or two pixels during the frame transfer time。 In fact significant deferred charge  is also seen  in the  second  trailing  pixel,  w'hich can

only happen in this way。 For a time constant •e to be comparable or less than 2jis requires an energy level of W。3 eV or enhanced emission due to an electric field of order 104 V/cm [14]。

The x-ray technique was extended in order to measure the serial CTE。 The readout register was operated as a linear CCD by using a timing waveform that reverse clocked the image and storage regions and allowed the readout register to

Column Number

Fig。 3 Peak x-ray signals for the readout register operated as a linear CCD and with charge shuffling for columns 400 to 800 so as to give additional pixel transfers and hence CTE loss。

integrate charge before readout。 An additional feature was to readout the first 400 pixels and then repeatedly shuffle the remainder sideways by 400 pixels (i。e。 move them towards the output amplifier and then back again。 X-ray events were recorded as above。 Results for the peak signal (in single pixel events) are given in figure 3 for 8 shuffles (corresponding to an extra 18 x 400 transfers)。 It can be seen that there is a pronounced discontinuity at column 400 - corresponding to the charge loss for the additional transfers。 The discontinuity was 8。2%o of the average peak signal, corresponding to a CTI of 0。000011 per pixel。 It was found that the charge loss was proportional to the number of shu8les, as expected。 The dose is somewhat uncertain Since, for part Of the irradiation, the proton beam did not cover the whole of the readout register。 However it is estimated that the dose was =l krad so we have a CTI/krad of W。00001 per pixel which is a factor 5 less than for the parallel direction。论文网

The  values  obtained  for  CTE  damage  are  less   than

previously  reported  for  low  temperature,  slower   readout

conditions。 For example Holland et at [1] give a damage constant of 2x10-'3 c 2 for parallel CTE compared with our value of 3。4x10 14    °。  Hosvevcr the effect of traps on CTE

depends on the ratio of the capture time constant to the time available for trap filling [5) - assuming (as in our case) that the traps are initially empty before the ncxt signal 'event' is transferred。 Robbins ct at [5] have calculated that for EEV devices  the  capture  time  constant  is  of the  order  of a few