boundary conditions detector

Design document description:

The barrel consists of 3 layers of double sided silicon strip detectors, also for the forward/rear wheel detectors double sided silicon detectors are foreseen.
The detectors are about 300 micron thick.
Detector dimensions are for the forward region: widgets of 70 mm height and 30 and 60 mm side dimensions, for the barrel : rectangular detectors of 42*84 mm²

Implications of design specifications

Contents of this page:

number of channels
data volume
stereo angle
p vs n
single vs double sided
input signal
noise level
temperature and cooling
other documentation

Number of channels:

In the forward region the number of detectors in 4 wheels of 16 double sided detectors is 64. For a pitch of 100 micron this gives 48 * 2 * 600 = 77 k channels.
In the barrel the number of detectors is (11 + 15+22)*8 = 384.
1. With a pitch of 100 micron the number of channels is: 384* (420+840) = 484 k channels.
2. If the pitch would scale with the radius the number of channels is: (11 + 15*0.68 + 22*0.46)*1260 = 395 k channels.
3. With a pitch of 100 micron and detectors bonded two by two together the number of channels is: 484/2= 242 k channels.
In the barrel alone the identification number for 242 k channels uses up to 18 bits (to be studied further).
In the following the total number of channels will be assumed to be : 77 + 242 = 319 k channels.

Data volume (see also slt-readout requirements):

Binary:
- an average occupancy of 1% (?)
- 319 k channels in total
- include data for nearest neighbours
- 3-bit data word for present, previous and next 'bunch'
- an overhead of 6-bits per packet and 8-bits address per dataword
The estimated datavolume is then: 319 k * 0.01 * (6 + 3 * (8 + 3))/8 = 15 kbyte/event or if we don't include data of neighbouring strips : 7 kbyte/event.
Analog: under the same assumptions but with a 8-bit data word per strip: 41 kbyte/event., not including neighbouring strips: 15 kbyte/event.
Analog: if we assume less channels because of a larger pitch and we assume the same ratio for analog to binary pitch as was chosen for ATLAS, the number of channels is a factor 0.66 of that of the binary option. The data volume is then: 27 kbyte/event, excluding neighbours: 10 kbyte/event.

The angle for the double-sided readout is of importance for how to bond the detector to the electronics. In the barrel one expects an angle of 0 and 90 degrees and for the widgets of +12 and -12 degrees. In the barrel strips will be bonded together to reduce the number of electronic channels. Bonding is a delicate procedure, to be studied in more detail. However, nothing more specific can be added at the moment.

P vs N:

A variety of possibilities is available:

p-strips on n-type silicon
n-strips on p-type silicon
n-strips on n-type silicon
p-strips on p-type silicon

Option 3 has been selected for ATLAS in combination with binary readout as a baseline solution. The CAFE chip is available at the moment for n- and p-strip readout. But as a consequence of the ATLAS choice, the CAFE preamp will be developed further for the ABC readout for option 3. If ZEUS chooses differently, the preamp probably has to be redesigned, especially if a different peaking time is required than for ATLAS.

Single vs double sided:

Comments (see ATLAS NOTE): With single sided detectors all readout electronics can operate at ground potential. For double sided readout the electronics on the two sides have to operate on different ground potentials separated by the bias voltage of the detector. Then the electronics output information has to be level shifted. This has been done without problems in the ALEPH experiment with capacitive coupling of the amplified signals. With optical coupling the level shifting occurs naturally.

Input signal and detector characteristics:

BaBar specifies in note 213:

Nominal input charge: 3.84 fC (1 MIP)
Minimum input signal charge: 0.94 fC (0.25 MIP)
Maximum input signal charge: 38 fC (10 MIP)

The value of the nominal charge represents the most probable signal from a minimum-ionizing particle traversing 300 micron of silicon. The system must be efficient for MIPs passing through detectors at angles as large as 73 degrees from perpendicular, in which case the particle traverses, for each of as many as nine of the 100 micron pitch z strips of the first layer, only 100/cos(0.3)=105 micron of silicon. Such strips see only a most-probable signal of 1.34 fC. Due to Landau fluctuations, 1 % of the signals would be below 0.7 times the most-probable value, so the minimal signal of interest in the system is 0.94 fC. Pulses larger than the listed maximum value of 10 MIPS are expected, but this is the interesting range.

Noise level:

BaBar specifies (our numbers will differ presumably because of a different time constant and background rate):

Maximum rate of noise hits per channel: 500 Hz
Nominal threshold: 0.8 fC
Maximum threshold: 4 fC
Maximum rms variation of threshold about the minimal value: channel-to-channel: 6% (with no tuning) and chip to chip: 6% (may be achieved by tuning)

The rate of noise per channel must be << hit rate from tracks. At a background rate of 5 kHz and time constant of 100 ns, the rate of threshold crossings would be 500 Hz if the threshold level were 4 times the rms noise level. To allow for 6% threshold variation, we take the average threshold to be 4.25 times the rms value. Then, assuming a minimum signal of 0.94 fC and setting the threshold 15% below that value, we conclude that the rms noise level, referred to the input, should be no greater than 0.94/1.15/4.25=0.19fC=1200 electrons equivalent noise charge. In standard terminology the signal-to-noise should be at least 24000/1200=20 in order to ensure good efficiency and good timing for all tracks of interest.

ATLAS specifies in the technical proposal (see below):

The total noise after 10 years of operation should be less than 1500 electrons equivalent noise charge (ENC), giving a signal-to-noise ratio (S/N) above 12:1, in order to maintain high efficiency and low noise levels compared to the hit rate from particles. Results obtained in the test beam, using prototype LHC readout indicate that a pulse-height cut on single strips would be viable if these specifications are met.

Temperature and cooling:

The operating temperature for the silicon detectors has to be stable and should not show a large temperature gradient. On the readout chips there can be larger temperature variations, although a limited temperature range of 30 till 40 degrees seems to be advisable. In ATLAS one quotes 4 mW/ch for the front-end chips, so in total the heat produced in ATLAS from this source is 20.5 kW alone (3 times that of the detectors!). If we use the same assumption the thermal load in ZEUS is about : 319.000 *4mW=1.3 kW.

For the ABC chip the power consumption is specified to be 64 mW, when operating at a 1% occupancy and an 100 kHz FLT rate (so 0.5 mW/ch). The CAFE chip is specified to dissipate 1.2-1.8 mW/ch.