what remains to be done

5.

It would be good if we could refine our knowledge of the parameters in Table 2.3 for which the method is listed ``guess''. It would be possible to measure these if we had a large-statistics sample of minimum-biased events (i.e. with a trigger that required only the RPD OR). Unfortunately the run log does not record us having taken any such data during the June run, in particular, without the tagger OR in the trigger. We should think of how these numbers can be extracted from the data we have.

6.

Whether it is applied online or offline, an efficient CP veto is extremely important to this experiment. A subset of the phototubes on the CPV counters showed evidence of saturation at beam intensities that still lower than what is foreseen for 1999 running. Perhaps the advent of a higher energy beam will decrease the overall rates in the CPV, but whether there will be a substantial decrease we cannot be certain. We should be sure that the counters that cover the region close to the beam axis are equipped with bases that are capable of efficient operation at 5MHz.

7.

At present the UPV counters have been mounted with a 10cm x 10cm inner square hole for the beam. If the CPV is removed as an online veto then it is incumbent on the UPV signal to remove as much as possible of the background in the RPD OR associated with photon beam halo. The aperture of the UPV should be decreased to intercept as much of the beam halo as possible, without going so far that the UPV rate begins to introduce random vetoes as a significant level. Monte Carlo indicates that with the current shielding configuration the UPV OR rate would not exceed 20KHz/nA if the aperture were decreased to match that of the CPV. This would provide a better shield from upstream backgrounds than we presently have, and at the same time not introduce significant additional losses. Background from upstream conversions is especially important when CLAS is running, and the UPV helps to reduce our sensitivity to the presence of their target.

8.

With a minor adjustment it should be possible to reduce the adc gate width from its present value of 250ns. At some point, reducing the gate width further results in a loss of energy resolution as one cuts into the tail of the pulse. On the other hand, anything from an interaction within the duration of the gate, other than what caused the trigger, that creates a signal in the lead glass will corrupt the event and in most cases effectively remove it from our sample. This amounts to another dead-time effect in the electronics, and we need to understand its impact at the same level we have understood the trigger. This will require a Monte Carlo study with Gradphi[2].

9.

During 1998 running there was an unexpected failure mode in the adc modules which caused clusters of adjacent channels to generate anomalously large adc values. These blocks were effectively removed from active use by encoding their row,column addresses to values outside the range subtended by the LGD. However they add considerable overhead to the level 2 processing time, which had an adverse effect on dead-time. We would like to understand this problem, or at least have sufficient spares available to replace failed channels during 1999.

10.

During 1998 running, the zero-suppression feature of the adc modules was disabled so that every adc in the crate produced a data word in the event for every event. This generates unnecessary overhead both in DAQ dead-time and disk space. Anticipating a much higher event rate for 1999 running, we should switch over to use zero suppression.

11.

In order to have the possibility of running at design intensity in 1999 we must have the new level 1 trigger installed and working.