- Email: [email protected]
Silicon pixel detector research and development
Nuclear Instruments and Methods in Physics Research A 368 (1995) 213-216
NUCLEAR
INSTMJMENTS L METHODS IN PHYSICS RESEARCH
Se&on A
ELSEVIER
Silicon pixel detector research and development T. Mouthuy * CNRS-IN2P3KPPM. 163 LIV.de Luminy, Case 907, 13288 Marseille Cede 09, France
Abstract Pixel detectors will be extremely useful in many future experiments. Firstly, due to their very small size, they have a very low intrinsic noise and are thus intrinsically radiation resistant. Secondly, their true 3-D capability improves the tracking efficiency and ghost rejection. They can then be close to the beam axis at the very high planned accelerator luminosities for improved vertexing. Several smart data-driven architectures are being considered for use at LHC and will be tested soon.
1. Silicon pixel detectors In many existing and future experiments, the high event rate and multiplicity requires sophisticated detectors. Silicon pixel detectors are particularly well suited, as will be shown for the CERN Omega [ 1 ] experiment, the DELPHI [ 21 detector at LEP and the foreseen ATLAS [ 31 and CMS [ 51 detectors at the LHC. Silicon pixel detectors have several advantages compared to other tracking devices: - Excellent timing resolution (with the largest part of the signal collected within 10 ns) . This is essential for detectors at the LHC, where beam cross-overs (BCOs) occur every 25 ns. - A large signal (typically 25 000 e- are collected for normal incidence in a 300 pm depletion depth). This requires a relatively simple amplifier to obtain a reasonable signal. - A much smaller surface than, for example, silicon microstrips. Thus their active volume and capacitance are lower and their intrinsic noise is reduced. A typical noise figure is less than 100 e-, compared to typically 1500 e- in silicon microstrips with a similar (50 ns) shaping time. As a consequence of the large signal over background ratio, a thinner detector can be used, even at the cost of signal reduction. - Intrinsically, pixel detectors will be more radiation hard. This comes naturally from the good signal to noise ratio. Radiation effects increase the leakage current, which in turn increases the detector noise by an amount proportional to the detector volume. Simultaneously, the depletion voltage also increases and could even reach the breakdown voltage. Working with thinner or under-depleted detectors helps at the expense of an extra loss of the signal. For pixel detectors the signal to noise ratio remains excellent.
* E-mail
[email protected]
0168~9002/95/$09.50 @ 1995 Elsevier Science B.V. All rights reserved SSDIO168-9002(95)00938-8
- Finally, because of their small size (50 ,um x 300 pm pixels are being considered for ATLAS), they have a low occupancy ( <10m4 per bunch crossing at maximum LHC luminosity of 10% cm-*s-l) and provide space-point information which is very useful for pattern recognition in the high multiplicity environment. Pixel detectors are typically built by bonding electronic readout chips (M 1 cm*) onto a detector chip (typically 8 x 2 cm*). The mechanical and electrical connections are made by a “flip chip” or “bump” bonding technique. At present, contacts between the connection pads of the detector and readout chips are made with solder or indium. 2. Pixel detectors in heavy ion physics Pixel detectors are currently being used in the WA97 [ 1] experiment at the CERN Omega spectrometer. The collision of Pb (33 TeV) on a Pb target produces a large number of tracks. The telescope, following the target, contains 4 pixel planes. At present, a 120 cm* system is operational, consisting of 3 x lo5 pixels of dimension 75 x 500,pm*. The readout scheme is based on a delay line, allowing time for the trigger decision. Binary information (the pixel is hit or not) is read out after a trigger. The decision is based on the comparison between the integrated charge and a threshold of around 5000 f 750 e-. The performance of the detector is excellent, with less than 3% dead pixels. The noise level is estimated with a Cd source by counting the number of hits above a varying threshold. By differentiating the curve, one clearly sees the two lines from Cd (at 22 keV and 25 keV) , which allows an estimate of the noise around 170 eAfter reconstruction, the probability of a noisy hit for a given pixel is less than low6 of the rate of hit pixels associated with identified tracks. This allows an easy reconstruction of the event, even at high multiplicity (around 35 tracks).
V. DETECTOR
AND TRIGGER
PROJECXS
214
i? Mouthuy/Nucl.
Instr. and Meth. in Phys. Res. A 368 (1995) 213-216
3. Pixels for LEP detectors The DELPHI [ 21 microvertex detector will be upgraded by the end of 1995. It will include two pairs of pixel cones to cover the forward region (angular ranges 15.6’-25.6” and 12.10-21”). The cones are tiled with (2 x 7 cm’) detector substrates onto which are bonded 16 readout chips: 10 chips reading 24 x 24 pixels, and 6 smaller chips reading 16 x 24 pixels. The pixels (330 x 330 pm*) are read out by a selective “sparse scan” scheme [4]. A hit pixel records its hit in an individual one-bit buffer. The readout scheme searches lines with hit pixels, and then hit pixels in the given line. A clock at 10 MHz is used to transfer out pixel information, giving a maximum readout time of 200 ns per hit pixel. A prototype of such a detector was beam-tested at CERN in June 1995. Pions of 100 GeV were sent through a microstrip telescope and the pixel detector, which had an angle of 40” with respect to the beam. This represents the worst case for LEP operation, in which the charge will often be split between adjacent pixels. Fig. 1 shows the measured efficiency as a function of the applied threshold for the 16 electronic chips of the detector. Good agreement is seen with a Monte Carlo simulation including only geometrical charge sharing. It should be pointed out that the chips bonded to the prototype detector were not pre-sorted before bonding. As the working points of the chips are controlled by currents (thresholds etc.) sent along a common substrate bus to the input connections which exhibit chip-to-chip impedance variations, the performance of the chips are not identical. Hence some chips behave less efficiently. Also one of the 16 chips was found to have a non-working column (due to a
short-circuit on a transistor), which gave a 4% inefficiency. For the final production rnn, all chips will be pre-tested and all chips having more than 1% noisy pixels at the working threshold will be rejected before bonding. Also, the detector substrate will be populated with chips that have been grouped into batches with similar control-line impedances. This should improve the behaviour of the detectors to be installed in DELPHI. 4. Pixels for LHC physics A large amount of activity is taking place at various laboratories relating to ATLAS [ 31 and CMS [ 51 microvertex detector R&D. This paper will concentrate on the development work for the ATLAS pixel detectors. Current CMS R&D is described in Ref. [ 61. Many LHC physics topics require a good inner detector to complement the calorimeter. For example, the measurement of B-decay vertices sets more stringent constraints and requires precision points to be measured very close to the beam axis. The barrel part of the ATLAS detector consists of two silicon pixel cylinders located at radii of 11.5 and 16.5 cm from the beam axis. The forward part is made of four silicon pixel disks [ 71. At high luminosity, a measurement of the top quark mass can be made quite precisely at high &“. In order to reduce the background, B-tagging can be used. An algorithm based on silicon pixel layers as a starting point for track finding shows an efficiency above 97% for tracks with pr above 1 GeV/c. The number of fake tracks crucially depends on the pixel layers. In the ATLAS barrel detector, with crossed strips only, the algorithm finds 160% fake tracks; 35% are found including strips at small stereo angle. The fake track fraction reduces to 4% with the help of the two pixel layers. The impact parameter resolution is usually expressed in the form B VIP = A c3 ~ p&m
Fig. I. Efficiency curves as a function of threshold for the 16 electronic chips of a DELPHI pixel detector. The curve represents the Monte Carlo prediction at 40° incidence angle.
.
(1)
The asymptotic term, A, depends mainly on the spatial resolution of the innermost layer. The second term, B, is the multiple scattering contribution which depends explicitly on the track momentum and on the crossing angle with respect to detector plane, 8. Within jets, a resolution ~QPof 28 @ 272/(pr&?) pm is found with the first pixel layer at 11 cm. With an extra “low luminosity” layer at 4 cm, the resolution becomes 17 @ 113/ (pi&&?) ,um. Fig. 2 shows the efficiency to tag a bjet versus the rejection rate against u-jets. The algorithm tags a jet as a b-jet if this jet contains more than n tracks with an impact parameter above 3c+. For a rejection of 50, the tagging efficiency is around 45%, well above specification [ 31. At the LHC, at a mean luminosity of 10” cm-*se’, the expected dose is 22.5 kGy/year [ 81. The total 1 MeV neutron-equivalent displacement per high luminosity year is
T Mouthuy/Nucl.
Instr. and Meth. in Phys. Res. A 368 (1995) 1
v
t 0
02
“4
Real
(MC)
track
“b
parameters
Eff~ckmcy
b-,&s
1
Fig. 2. B-jet tagging efficiency wcsus u-jet rejection.
estimated to be 6.4 x 1OL3cmm2. The ionising dose will deposit charge in the oxide layers and thus, for example, disturb the behaviour of MOS components. Incident particles also provoke atomic displacements in the bulk silicon crystal. These will degrade bipolar transistors by changing the effective doping concentration of the silicon. For silicon detectors, the consequence of a high radiation environment is a leakage current increase, depletion voltage increase and signal deficit [9,10]. Typically, the depletion voltage could reach 800 V at O’C after 10 years for a 250 pm thick silicon diode at a radius of 11 cm. Charge collection degrades, but remains above 50% at O’C with a working voltage of 200 V, independent of the required depletion voltage. Silicon pixel detectors are less sensitive to radiation effects than silicon microstrips. Due to their small size (50 pm x 300 pm), they have a lower intrinsic noise than strip detectors. The signal to noise ratio (S/N) is quite favourable, about 100. Pixel detectors could then work on a thinner detector substrate for example, 150 pm. This would reduce the required depletion voltage, as it is proportional to the square of the silicon thickness. The signal reduces linearly with the change in thickness. Furthermore, if the working voltage is lower than the depletion voltage, the detector is not fully depleted and the charge collection efficiency decreases. An extra loss of about 50% of the signal could be tolerated as the S/N ratio remains excellent. Different readout schemes are under study [ 111 for the high luminosity environment at LHC. The LBL [ 121 group propose a column-based architecture with peripheral storage FIFOs. A column fast OR is used to store the beam crossing number into peripheral memory. The pointer to the stored BCO is sent back to the hit pixels to be stored in electronic registers. Crossing time and time over threshold are measured. This serves to compensate the jitter caused by the
215
213-216
different amplitudes collected. It also provides an analogue readout of pixel charges. A structure made of one column of 64 pixels should be tested in a beam in 1996. The CERN-Genoa group in the RD19 collaboration proposes a digital readout scheme based on the Omega-2 chip. A fixed delay and a programmable delay line is used to compensate the trigger latency (around 2 ps). The “poor” resolution in the delay line (around 100 ns) is compensated by a fast OR in the column, used to time-stamp the hit with good time resolution (25 ns) . Furthermore, the chip has test input, pixel mask and status check, allowing easy debugging of the chip. A test cell with 16 columns of 128 pixels should be ready by the end of the year and beam tests are foreseen in the near future. The CPPM group is studying a digital readout scheme. This is also a column-based architecture but with a column shift register. The storage and coincidence are peripheral. The address of a hit pixel shifts down the column at the LHC clock frequency. When reaching the peripheral logic, a counter is started, which increments at the same frequency. The trigger latency (a fixed number) is compared with the sum of the counter and the pixel address. A positive comparison indicates a hit pixel in the given beam crossing. There are currently two designs under way in radiation-hard technologies: HSOI3-HD and DMILL. Cells with 8 columns of 15 pixels were tested in September 1995. New chips with 12 columns of 63 pixels are expected by the end of the year. Finally the PSI group is studying an alternative readout scheme for square pixels ( 125 x 125 pm2) for CMS. A column fast OR is used to store the beam crossing number into peripheral memory. The readout of hit pixels immediately starts and the charge is stored in analogue registers. On receipt of a Level 1 trigger, the registers are read out for digitisation, or are reset if no trigger arrives. Chips made in standard and radiation-hard technology (DMILL) are expected by the end of the year. Tests will then be made at CERN and PSI. 5. Conclusions The excellence of pixel detectors in pattern recognition has already been demonstrated in heavy ion experiments and is expected for LEP-200 and LHC applications. Pixels are required for survivability at small radii at LHC. They are essential for good impact parameter resolution, which allow efficient b-tagging. The technology is now well advanced. Pixels are operational at Omega. DELPHI pixels are currently under test and should be integrated into the detector by the end of 1995. Some prototype LHC readout schemes will be tested in a beam before the end of 1995. A lot of work is still required. In particular, the final readout scheme for ATLAS is yet to be decided. It is not yet known whether analogue readout is required. This will determine the layout of pixel diodes on the detector substrate, in order to improve the position resolution using charge shar-
v.
DETECTOR
AND TRlGGER
PROJECXS
216
7: Mouthuy/Nucl.
Instr. and Meth. in Phys. Rrs. A 368 (19951 213-216
ing between pixels. Mechanical structure studies, including cooling, must be pursued to achieve as thin a detector as possible with acceptable temperature stability. The optimum working temperature has yet to be decided. Possible alternative pixel scenarios could be studied. For example the use of CVD diamond as a detector is attractive since it appears to be stable under irradiation and could provide a reasonable signal for pixel electronics [ lo].
I41 P. Delpierre and J.J. Jaeger, Nucl. Instr. and Meth. A 305 ( 1991) 627; J.J. Jaeger et al., IEEE Trans. Nucl. Sci. NS-40 (1993) 400. [ 5 I CMS Collaboration, The Cornpa-t Muon Solenoid Technical Proposal, CERN/LHCC/94-38.
[ 61 171 I81 [9]
I IO] References
I II ]
1I]
F. Anghinolfi et al., IEEE Trans. Nucl. Sci. NS-39 (1992) 650. 121 DELPHI Collaboration, Proposal for the Upgrade of DELPHI in tbe Forward Direction, CERN/LEPC/92-13. DELPHI 92-142 GEN 135. 131 ATLAS Collaboration, ATLAS Technical Proposal for a GeneralPurpose pp Experiment at the Large Hadron Collider at CERN, CERNILHCC/94-43.
[ 121
D. Kotlinski and C. Racca, these Proceedings (3rd Int. Workshop on B-Physics at Hadron Machines, Oxford, UK, 1995) Nucl. Ins& and Meth. A 368 (1995) 115. A. Nisati, ibid., p. 109. Th. Moudmy, Radiation doses expected in LHC inner detectors: an update, ATLAS Internal Note INDET-NO-009 (1992). G. Hall, Ref. [6], p. 199. P. Weilhammer. presented at this Workshop (3rd Int. Workshop on B-Physics at Hadron Machines, Oxford, UK, 1995). P. Fischer et al., Pixel detector back-up document to support the ATLAS Technical Proposal. ATLAS Internal Note, INDET-NO-086 (1994). M. Wright, J. Millaud and D. Nygren, A pixel unit cell targeting 16 ns time resolution and radiation hardness in a column readout particle vertex detector, LBL 32912 and Proc. Int. Conf. on High Energy Physics and Technology, Como. Italy, June 1992).
NUCLEAR
INSTMJMENTS L METHODS IN PHYSICS RESEARCH
Se&on A
ELSEVIER
Silicon pixel detector research and development T. Mouthuy * CNRS-IN2P3KPPM. 163 LIV.de Luminy, Case 907, 13288 Marseille Cede 09, France
Abstract Pixel detectors will be extremely useful in many future experiments. Firstly, due to their very small size, they have a very low intrinsic noise and are thus intrinsically radiation resistant. Secondly, their true 3-D capability improves the tracking efficiency and ghost rejection. They can then be close to the beam axis at the very high planned accelerator luminosities for improved vertexing. Several smart data-driven architectures are being considered for use at LHC and will be tested soon.
1. Silicon pixel detectors In many existing and future experiments, the high event rate and multiplicity requires sophisticated detectors. Silicon pixel detectors are particularly well suited, as will be shown for the CERN Omega [ 1 ] experiment, the DELPHI [ 21 detector at LEP and the foreseen ATLAS [ 31 and CMS [ 51 detectors at the LHC. Silicon pixel detectors have several advantages compared to other tracking devices: - Excellent timing resolution (with the largest part of the signal collected within 10 ns) . This is essential for detectors at the LHC, where beam cross-overs (BCOs) occur every 25 ns. - A large signal (typically 25 000 e- are collected for normal incidence in a 300 pm depletion depth). This requires a relatively simple amplifier to obtain a reasonable signal. - A much smaller surface than, for example, silicon microstrips. Thus their active volume and capacitance are lower and their intrinsic noise is reduced. A typical noise figure is less than 100 e-, compared to typically 1500 e- in silicon microstrips with a similar (50 ns) shaping time. As a consequence of the large signal over background ratio, a thinner detector can be used, even at the cost of signal reduction. - Intrinsically, pixel detectors will be more radiation hard. This comes naturally from the good signal to noise ratio. Radiation effects increase the leakage current, which in turn increases the detector noise by an amount proportional to the detector volume. Simultaneously, the depletion voltage also increases and could even reach the breakdown voltage. Working with thinner or under-depleted detectors helps at the expense of an extra loss of the signal. For pixel detectors the signal to noise ratio remains excellent.
[email protected]
0168~9002/95/$09.50 @ 1995 Elsevier Science B.V. All rights reserved SSDIO168-9002(95)00938-8
- Finally, because of their small size (50 ,um x 300 pm pixels are being considered for ATLAS), they have a low occupancy ( <10m4 per bunch crossing at maximum LHC luminosity of 10% cm-*s-l) and provide space-point information which is very useful for pattern recognition in the high multiplicity environment. Pixel detectors are typically built by bonding electronic readout chips (M 1 cm*) onto a detector chip (typically 8 x 2 cm*). The mechanical and electrical connections are made by a “flip chip” or “bump” bonding technique. At present, contacts between the connection pads of the detector and readout chips are made with solder or indium. 2. Pixel detectors in heavy ion physics Pixel detectors are currently being used in the WA97 [ 1] experiment at the CERN Omega spectrometer. The collision of Pb (33 TeV) on a Pb target produces a large number of tracks. The telescope, following the target, contains 4 pixel planes. At present, a 120 cm* system is operational, consisting of 3 x lo5 pixels of dimension 75 x 500,pm*. The readout scheme is based on a delay line, allowing time for the trigger decision. Binary information (the pixel is hit or not) is read out after a trigger. The decision is based on the comparison between the integrated charge and a threshold of around 5000 f 750 e-. The performance of the detector is excellent, with less than 3% dead pixels. The noise level is estimated with a Cd source by counting the number of hits above a varying threshold. By differentiating the curve, one clearly sees the two lines from Cd (at 22 keV and 25 keV) , which allows an estimate of the noise around 170 eAfter reconstruction, the probability of a noisy hit for a given pixel is less than low6 of the rate of hit pixels associated with identified tracks. This allows an easy reconstruction of the event, even at high multiplicity (around 35 tracks).
V. DETECTOR
AND TRIGGER
PROJECXS
214
i? Mouthuy/Nucl.
Instr. and Meth. in Phys. Res. A 368 (1995) 213-216
3. Pixels for LEP detectors The DELPHI [ 21 microvertex detector will be upgraded by the end of 1995. It will include two pairs of pixel cones to cover the forward region (angular ranges 15.6’-25.6” and 12.10-21”). The cones are tiled with (2 x 7 cm’) detector substrates onto which are bonded 16 readout chips: 10 chips reading 24 x 24 pixels, and 6 smaller chips reading 16 x 24 pixels. The pixels (330 x 330 pm*) are read out by a selective “sparse scan” scheme [4]. A hit pixel records its hit in an individual one-bit buffer. The readout scheme searches lines with hit pixels, and then hit pixels in the given line. A clock at 10 MHz is used to transfer out pixel information, giving a maximum readout time of 200 ns per hit pixel. A prototype of such a detector was beam-tested at CERN in June 1995. Pions of 100 GeV were sent through a microstrip telescope and the pixel detector, which had an angle of 40” with respect to the beam. This represents the worst case for LEP operation, in which the charge will often be split between adjacent pixels. Fig. 1 shows the measured efficiency as a function of the applied threshold for the 16 electronic chips of the detector. Good agreement is seen with a Monte Carlo simulation including only geometrical charge sharing. It should be pointed out that the chips bonded to the prototype detector were not pre-sorted before bonding. As the working points of the chips are controlled by currents (thresholds etc.) sent along a common substrate bus to the input connections which exhibit chip-to-chip impedance variations, the performance of the chips are not identical. Hence some chips behave less efficiently. Also one of the 16 chips was found to have a non-working column (due to a
short-circuit on a transistor), which gave a 4% inefficiency. For the final production rnn, all chips will be pre-tested and all chips having more than 1% noisy pixels at the working threshold will be rejected before bonding. Also, the detector substrate will be populated with chips that have been grouped into batches with similar control-line impedances. This should improve the behaviour of the detectors to be installed in DELPHI. 4. Pixels for LHC physics A large amount of activity is taking place at various laboratories relating to ATLAS [ 31 and CMS [ 51 microvertex detector R&D. This paper will concentrate on the development work for the ATLAS pixel detectors. Current CMS R&D is described in Ref. [ 61. Many LHC physics topics require a good inner detector to complement the calorimeter. For example, the measurement of B-decay vertices sets more stringent constraints and requires precision points to be measured very close to the beam axis. The barrel part of the ATLAS detector consists of two silicon pixel cylinders located at radii of 11.5 and 16.5 cm from the beam axis. The forward part is made of four silicon pixel disks [ 71. At high luminosity, a measurement of the top quark mass can be made quite precisely at high &“. In order to reduce the background, B-tagging can be used. An algorithm based on silicon pixel layers as a starting point for track finding shows an efficiency above 97% for tracks with pr above 1 GeV/c. The number of fake tracks crucially depends on the pixel layers. In the ATLAS barrel detector, with crossed strips only, the algorithm finds 160% fake tracks; 35% are found including strips at small stereo angle. The fake track fraction reduces to 4% with the help of the two pixel layers. The impact parameter resolution is usually expressed in the form B VIP = A c3 ~ p&m
Fig. I. Efficiency curves as a function of threshold for the 16 electronic chips of a DELPHI pixel detector. The curve represents the Monte Carlo prediction at 40° incidence angle.
.
(1)
The asymptotic term, A, depends mainly on the spatial resolution of the innermost layer. The second term, B, is the multiple scattering contribution which depends explicitly on the track momentum and on the crossing angle with respect to detector plane, 8. Within jets, a resolution ~QPof 28 @ 272/(pr&?) pm is found with the first pixel layer at 11 cm. With an extra “low luminosity” layer at 4 cm, the resolution becomes 17 @ 113/ (pi&&?) ,um. Fig. 2 shows the efficiency to tag a bjet versus the rejection rate against u-jets. The algorithm tags a jet as a b-jet if this jet contains more than n tracks with an impact parameter above 3c+. For a rejection of 50, the tagging efficiency is around 45%, well above specification [ 31. At the LHC, at a mean luminosity of 10” cm-*se’, the expected dose is 22.5 kGy/year [ 81. The total 1 MeV neutron-equivalent displacement per high luminosity year is
T Mouthuy/Nucl.
Instr. and Meth. in Phys. Res. A 368 (1995) 1
v
t 0
02
“4
Real
(MC)
track
“b
parameters
Eff~ckmcy
b-,&s
1
Fig. 2. B-jet tagging efficiency wcsus u-jet rejection.
estimated to be 6.4 x 1OL3cmm2. The ionising dose will deposit charge in the oxide layers and thus, for example, disturb the behaviour of MOS components. Incident particles also provoke atomic displacements in the bulk silicon crystal. These will degrade bipolar transistors by changing the effective doping concentration of the silicon. For silicon detectors, the consequence of a high radiation environment is a leakage current increase, depletion voltage increase and signal deficit [9,10]. Typically, the depletion voltage could reach 800 V at O’C after 10 years for a 250 pm thick silicon diode at a radius of 11 cm. Charge collection degrades, but remains above 50% at O’C with a working voltage of 200 V, independent of the required depletion voltage. Silicon pixel detectors are less sensitive to radiation effects than silicon microstrips. Due to their small size (50 pm x 300 pm), they have a lower intrinsic noise than strip detectors. The signal to noise ratio (S/N) is quite favourable, about 100. Pixel detectors could then work on a thinner detector substrate for example, 150 pm. This would reduce the required depletion voltage, as it is proportional to the square of the silicon thickness. The signal reduces linearly with the change in thickness. Furthermore, if the working voltage is lower than the depletion voltage, the detector is not fully depleted and the charge collection efficiency decreases. An extra loss of about 50% of the signal could be tolerated as the S/N ratio remains excellent. Different readout schemes are under study [ 111 for the high luminosity environment at LHC. The LBL [ 121 group propose a column-based architecture with peripheral storage FIFOs. A column fast OR is used to store the beam crossing number into peripheral memory. The pointer to the stored BCO is sent back to the hit pixels to be stored in electronic registers. Crossing time and time over threshold are measured. This serves to compensate the jitter caused by the
215
213-216
different amplitudes collected. It also provides an analogue readout of pixel charges. A structure made of one column of 64 pixels should be tested in a beam in 1996. The CERN-Genoa group in the RD19 collaboration proposes a digital readout scheme based on the Omega-2 chip. A fixed delay and a programmable delay line is used to compensate the trigger latency (around 2 ps). The “poor” resolution in the delay line (around 100 ns) is compensated by a fast OR in the column, used to time-stamp the hit with good time resolution (25 ns) . Furthermore, the chip has test input, pixel mask and status check, allowing easy debugging of the chip. A test cell with 16 columns of 128 pixels should be ready by the end of the year and beam tests are foreseen in the near future. The CPPM group is studying a digital readout scheme. This is also a column-based architecture but with a column shift register. The storage and coincidence are peripheral. The address of a hit pixel shifts down the column at the LHC clock frequency. When reaching the peripheral logic, a counter is started, which increments at the same frequency. The trigger latency (a fixed number) is compared with the sum of the counter and the pixel address. A positive comparison indicates a hit pixel in the given beam crossing. There are currently two designs under way in radiation-hard technologies: HSOI3-HD and DMILL. Cells with 8 columns of 15 pixels were tested in September 1995. New chips with 12 columns of 63 pixels are expected by the end of the year. Finally the PSI group is studying an alternative readout scheme for square pixels ( 125 x 125 pm2) for CMS. A column fast OR is used to store the beam crossing number into peripheral memory. The readout of hit pixels immediately starts and the charge is stored in analogue registers. On receipt of a Level 1 trigger, the registers are read out for digitisation, or are reset if no trigger arrives. Chips made in standard and radiation-hard technology (DMILL) are expected by the end of the year. Tests will then be made at CERN and PSI. 5. Conclusions The excellence of pixel detectors in pattern recognition has already been demonstrated in heavy ion experiments and is expected for LEP-200 and LHC applications. Pixels are required for survivability at small radii at LHC. They are essential for good impact parameter resolution, which allow efficient b-tagging. The technology is now well advanced. Pixels are operational at Omega. DELPHI pixels are currently under test and should be integrated into the detector by the end of 1995. Some prototype LHC readout schemes will be tested in a beam before the end of 1995. A lot of work is still required. In particular, the final readout scheme for ATLAS is yet to be decided. It is not yet known whether analogue readout is required. This will determine the layout of pixel diodes on the detector substrate, in order to improve the position resolution using charge shar-
v.
DETECTOR
AND TRlGGER
PROJECXS
216
7: Mouthuy/Nucl.
Instr. and Meth. in Phys. Rrs. A 368 (19951 213-216
ing between pixels. Mechanical structure studies, including cooling, must be pursued to achieve as thin a detector as possible with acceptable temperature stability. The optimum working temperature has yet to be decided. Possible alternative pixel scenarios could be studied. For example the use of CVD diamond as a detector is attractive since it appears to be stable under irradiation and could provide a reasonable signal for pixel electronics [ lo].
I41 P. Delpierre and J.J. Jaeger, Nucl. Instr. and Meth. A 305 ( 1991) 627; J.J. Jaeger et al., IEEE Trans. Nucl. Sci. NS-40 (1993) 400. [ 5 I CMS Collaboration, The Cornpa-t Muon Solenoid Technical Proposal, CERN/LHCC/94-38.
[ 61 171 I81 [9]
I IO] References
I II ]
1I]
F. Anghinolfi et al., IEEE Trans. Nucl. Sci. NS-39 (1992) 650. 121 DELPHI Collaboration, Proposal for the Upgrade of DELPHI in tbe Forward Direction, CERN/LEPC/92-13. DELPHI 92-142 GEN 135. 131 ATLAS Collaboration, ATLAS Technical Proposal for a GeneralPurpose pp Experiment at the Large Hadron Collider at CERN, CERNILHCC/94-43.
[ 121
D. Kotlinski and C. Racca, these Proceedings (3rd Int. Workshop on B-Physics at Hadron Machines, Oxford, UK, 1995) Nucl. Ins& and Meth. A 368 (1995) 115. A. Nisati, ibid., p. 109. Th. Moudmy, Radiation doses expected in LHC inner detectors: an update, ATLAS Internal Note INDET-NO-009 (1992). G. Hall, Ref. [6], p. 199. P. Weilhammer. presented at this Workshop (3rd Int. Workshop on B-Physics at Hadron Machines, Oxford, UK, 1995). P. Fischer et al., Pixel detector back-up document to support the ATLAS Technical Proposal. ATLAS Internal Note, INDET-NO-086 (1994). M. Wright, J. Millaud and D. Nygren, A pixel unit cell targeting 16 ns time resolution and radiation hardness in a column readout particle vertex detector, LBL 32912 and Proc. Int. Conf. on High Energy Physics and Technology, Como. Italy, June 1992).