Proposal for a change in the design of the

octagon multiplicity detector

Open for discussion !!

Rudi Ganz, August 13^th 1997

Prompted by the concern about common noise problems in the OCTAGON multiplicity array both for the alternative electronics but also (but probably not so significant) for the current PHOBOS electronics, we propose the following change in the octagon sensor design:

Current design:

92 Octagon sensors with 2 x 32 ch ea. 64 channel chip read out

(see the present design)

Proposed new design:

92 sensors (same geometry and mounting) with 4 x 32 ch ea. with 2x64 channel bonding pad.

( 2 x 64 ch read out with the current electronic, and 1x128 ch VA chip with the alternative electronic design; fan out to 64 channel bonding pads on the substrate.)

___________________________________________________________________

Several considerations:

1) Common noise

(See the discussion about empty channels)

Since we double the number of channels the occupancy will be cut in half. In the current design the probability that we have (only !) one or more empty channels for all the 44 sensors (high occupancy region, this includes the central sensor octagon ring and 3 rings on each side with respect to the vertex) in a central Au+Au collision is 18 %. When you require two or more empty channel on each chip this drops to 0.034 %. Especially in the multiplicity array, where we are looking for fluctuations, we cannot afford this!

In case of the two 64 channel chips on all 44 of the now 128 channel wafers, the probability, that all the 88 chips have at least 4 empty channels is 99 % (but drops when you require more than that)

In the most favorable case, where we use 128 channel chips on 44 of the 128 channel sensors, the probability is 99.5%, even if one requires more than 12 empty channel on all 128 ch chips.

Remark: The assumption, that 44 sensors all have an occupancy of 3 (in the current design) is truly a worst case one. (see table below)

2) Pulse sharing

Since we propose to keep the ETA segmentation but increase the PHI segmentation, the pulse sharing won't change significantly since the particle traverse the detector in the z-axis component.

The detector is at a distance of 45 mm from the beam axis; the width of ea. pad along z is (and will be) 2.6 mm. The active sensor has a thickness of about 0.3 mm.

The estimate of the h - and Z (z-distance of the pad from the primary vertex) dependence of the occupancy of primary particles, assuming dN/dh = 1000 = const

(then the occupancy is given by 1000 * Dh /16) all based on the 2 x 32 segmentation of the current design. Secondary particles will add another 5- 20% to the occupancy.

Remark1: in the central region there are only 4 sensors in each octagon ring due to the spectrometer and the vertex detector opening.

Remark 2: The sensors (bold font) at 0 mm and 80 mm and 160 mm and 240 mm are a total of 44 sensor as used as an estimate for the high occupancy region sensors in the section about the occupancy.

Pulse sharing -besides increasing the occupancy- will produce hits with energy deposition only fractions of a MIP. These then will be misidentified as "Empty Channel" and goes into the common noise subtraction. One might think of suppression of those by requiring no hit on the neighboring channels. In any case this will be rare, since particles which penetrate the detector at an angle will also deposit more energy.

And again: Pulse sharing is the same for both the current and the proposed solution when increasing the phi segmentation.

3) Data size increase

In the current design the whole multiplicity has 8960 channel. 5888 ch of those are in the octagon. In the proposed design this would go up to 14848 channel (65%) (due to the increase on the octagon to 11776). We suggest reading out all the channels until a proper Common noise suppression is established off-line. This would mean a 35 % increase in the event size of a central collision. In case we read all of them out for events with the alternative electronic this would mean an even larger increase for the minimum bias events. Which we may be able to afford only during the low luminosity phase of the startup of RHIC and then implement the zero suppression on the TDR board.