Physics, Praha; Bergische Universitat, Wuppertal. OCR OutputlN2P3/C.P.P.Marseille; INFN/Univ. di Milano; lN2P3/CdF/LPC, Paris; FZU/Inst. of

2lN2P3/ISN, Grenoble; Institut fiir Exp. Kernphysik, Karlsruhe;1sauvage@cppm.in2p3.fr

Ein Gedi, Dead Sea, ISRAEL. June 1995

VERTEX 95

Talk presented for the Fourth International Workshop on Vertex Detectors

modules, will be reported, and future prospects will be addressed.and packaging will be presented. Also, early beam tests with fully equippedreadout chip will be described. The very peculiar aspect of interconnectionelectronics, will cover the forward region, down to 12°. The detector and theLEP200. A total area of 1300 cm2, read out by true 2D sparsing on-chipDELPHI MVD, to be installed at the beginning of 1996, for the starting of

Silicon pixel diodes have been proposed as an upgrade detector for the

Abstract

This work is a joint effort ofthe Delphi VFT Pixel Group

F-13288 Marseille Cedex 9, France

Université d’/liz-Marseille II, Luminy, Case 907,

Centre de Physique des Particules de Marseille - INQPB - CNRS

D. SAUVAGE 1, J.C. CLEMENS, M. COHENSOLAL, M. RAYMOND, P.A. DELPIERRE

for the ’95 Upgrade of the DELPHI MicroVertex Detector

A Pixel Detector

important reasons like tests, detector price or chip interconnections layout, OCR Outputobviously want to built detector modules of rather large sizes, and for othertronics sits on a different chip, on top of the detector (fig. Because weimplanted on the same piece of silicon), and the hybrid one (where the elecapproaches: the monolithic one (where the detector and its electronics arePixel detectors are being studied and developped over the world using two

2 A detector module: the hybride technique

of 24><24 cells, and the other of l6><24 cells (Fig. 3).only constraint being to have electronic chips of two sizes (in our case), oneallows to design petal-shaped modules in an easy and convenient way; therectangular modules, as it is mostly the case with strips, the pixel option(and costly as far as material thickness is concerned) to tile a cone withthe detector itself (see chapter 2). Also, while it is geometrically difficultby the detector is made of sensitive material, with the electronics sitting onthe space requirement in the sense that the full geometrical area covered

provide two—coordinate projective readout Pixel detectors suit very well

other two layers are made out of ministrip detectors, glued back to back topixel layers are located where the space constraint is the most severe. Thebeen added (Fig. 2), of which two layers out of 4, are made of pixels. The

To extend the angular coverage of the existing detector, endcaps havemeasurement in front of the forward RICH detectors (Fig. 1).for 4-jet event tagging, and help for particle identification, with momentumchallenging detector will improve the pattern recognition capabilities, e.g.the performance of the DELPHI MVD in the very forward region This

a full detector has been proposed, and is under construction, to improveseveral planes of such modules are already taking data within WA97 ,lllis presently the best answer to solve most of these limiting factors. Whilesome limits, even when double-sided detectors are used. The pixel detectortrack reconstruction, for example, is also growing and silicon strips do haveluminosity of accelerators, the demand for accurate, fast and unambiguousof high energy physics instrumentation. However, with the ever increasingSilicon strip detectors have shown their huge capabilities for several aspects

1 Introduction

The front-end (analog) part of the circuit has been developped previOCR Outputin 2 parts : the analog chain, and the digital processing and readout stage.minimized for the other 3 sides. The discussion on electronics naturally splitsbetween adjacent chips, as well as the dead space on the detector, can behave to be layed out only on one side ofthe chip, so that the approach distanceto tile as many chips as possible on a single detector module, all the I/O padsof identical cells are duplicated to build up the 2D array. Finally, to be ableincluded on the same area. Crosstalk is a major concern, as several hundredsthan the cell size required by the physics. Also the contact pad has to beother hand, the available area to implement the electronics cannot be largeron the power consumption and on the signal-to—n0ise performance. On thebut also some contraints. Pixel capacitance is smaller, which has incidenceAs far as electronics is concerned, pixel detectors introduce some advantages,

3 The readout electronics

bus, and a reduction in the amount of signals by a factor greater than 6.I/O signals were fed through tri-state registers, allowing multiplexing on thethe 8 chips on each side of the module. When necessary, the electronic chipinterstrip capacitance, and line resistance) and used to route a bus connectingprocess, as proposed by our detector founder, was improved (in terms of yield,quite new in the hybrid microelectronics field. To do this, a double—metalwhere the substrate is active, and acts as the mechanical structure, which islines that connect all the chips. We have made a MCM (multichip module),not used a ceramic circuit underneath the detector to route the bus and powermaterial thickness is of prime concern in this region of the detector, we havealso the fact that no extra support has been added to the module. Because

A quite novel aspect of packaging, for the construction of the modules, isthe choice of the best chip sorting and matching to build each module.in the next chapter, the hybrid choice has the another advantage to allowcost of production for extra wafers, for example. Furthermore, as developpedis process-related. A compromise has to be found, taking into account thesize, which is a true geometrical constraint, and by the yield per wafer, whichcost and yield. On the electronic side, the chip size is limited by the reticlelarge as what we are used to for a standard strip detector, at a reasonnablewe chose the hybrid technique. With this option, we can have a detector as

is transmitted to the next hit line. OCR Output

skipped, and when the line is read, it is automatically reset, and the priorityallowing the column number to be decoded as well. The non-hit columns aresame signal propagates to the corresponding column, through the hit pixel,line. While this line number is decoded and buffered as a 5-bit word, thean asynchronous signal runs through the lines, until it finds the first hitwhile a second switch shorts an x-y bus inside the pixel matrix. At readout,that signals to the end of the corresponding line that this line contains a hit,then by column within a line. When a pixel is hit, a first switch is closed,(Fig. 5). The matrix is scanned by a priority encoding circuit, first by line,suppression-like technique. We call this technique Sparse Data Scan (SDS)[6]are hit per b.c.o., it is worthwhile to skip all the other pixels, using a zeroselective readout scheme was implemented. As only few hundreds pixels

Because of the large amount of pixels (over 1.2 million) to be read, aBeam test results).success of this selection procedure can be seen in Fig. 9; see in chapter 5From that, we finally extract groups of 16 well-matched circuits (the goodin a scatter plot, the axes being the slope-of—the—threshold vs. the threshold.selected sensitive parameters is recorded. The chips can then be representedvarious polarisations. Then, during a second pass, detailed dependance ofsic satisfactory functionning (in digital mode), and average values for thescanned twice : once to look for general features of the chip behaviour, bawe have developped an automatic test procedure. A full undiced wafer isout bus, special care must be taken to match these characteristics. Hence,threshold. Because 8 chips are connected together on the integrated readespecially in the preamplifier shaping and gain, and for the discriminatorues, different chips troughout a same wafer will exhibit different behaviours,4 current polarisations. However, with these currents set to nominal val

Several internal chip parameters can be tuned by adjusting externallybe opened only at beam cross-over (b.c.o.) time

match the larger bonding pad capacitance, and a gate was added, which willbit (D-latch) memory. The amplifier input transistor has been modified topreamplifier, a shaping discriminator, followed by a comparator and a oneously, mostly within the framework of RD19, at CERN It consists of a

the procedure for mass production. OCR Outputsuccessfully built by this manufacturer and steps are being taken to set upcheaper than previously used processes. Several modules have been alreadyinternal production, but recently released to the public market, and muchprocess (Fig. 7). This is a 20—year old IBM patent, mostly dedicated toand proven solution was chosen, the C4 (controlled collapse chip connection)

these solutions in the construction of the final detector. A more industrial

chips to a single bus, including all the I/O pads, has not allowed for usingproblems are still under study. Unfortunately, the yield when connecting 16success has been achieved with the conductive glue process, while some basicdreds of contacts) have been achieved with this second technique. Some

Good contact resistance and repeatibility (~ 200mQ, over several hundifficulty was to make this process work for a rigid—to—rigid connection.

While this is made easier when one of the two ”circuits” is soft, the wholetrical contact the two opposite pads, with the help of the metallic grains.thermo-compression process is used to melt the polymer, and put into elecbeen shaped to size and sandwiched between the detector and the chip, apolymer, including small grains of conducting material. When the tape hasthe connecting media is provided as a stabilized film/tape of some insulatingtors One advantage of this method, besides its low price, is the fact that

(Fig. 6), and has been tested for passive connections in microstrips detections between LCD displays, flex kapton cables and/or motherboard PCBstive adhesives. It is used in portable appliances, for connectorless connec

We, at CPPM, have developped a technique based on anisotropic conducdeposition by the screen printing technique.

Our colleagues from INFN Milano have concentrated in conductive glue

connection to the detector is essential to validate the readout response.

low-cost prototyping at the laboratory level, and chip evaluation, where theder to produce modules at an attracting cost, but also to provide fast andtechniques have been investigated by members of the collaboration, in orexpensive, a special effort was devoted to this point and several ”cheap”within RDIQ being that a conventional ilip—chip technique could be ratherment to the electronic chain can also be rather large (150,um). Our experienceBecause the size of the sensitive cell is relatively large, the connecting ele

4 Packaging overview

tern recognition. However, we already foresee to install several pixel layers OCR OutputThis detector will demonstrate the excellent capabilities of pixels for pat

next LEP shutdown.

tion of the VFT assumes that we will install this detector as planned at thepackaging, and detector integration. The proposed planning for the construcadvances have been made in electronics designs, interconnection and modulePHI experiment, demonstrates the maturity of this new technique. Major

The decision to install pixel detectors in the LEP200 version of the DEL

6 Conclusions and future prospects

enhanced charge sharing, due to this tilt.

within respect to the beam axis, this resolution is much improved by thedirection, because we use data from the run where our module was tilteddivided by \/12. These is clearly the case in the y direction. For the 2:tracks well inside the pixel, the expected resolution is equal to the pitch,discriminator threshold. ln the case of digital readout, with perpendicular

The detector resolution is shown in Fig. l0, plotted as a function of thea source scan of a ”matched” module).bus; the situation has improved since, as shown in Fig. 9, which representswhich causes some differences of response between several chips on the samebe noted that at that time, the chip matching procedure was not optimized,simulation including the charge sharing between adjacent pixels (it shouldof the applied threshold. A good agreement is seen with a Monte Carlo

Measured efficiency for the 16 chips, is presented in Fig. 8, as a function

detector is installed in DELPHI.

ular to the beam, and at 40°, which represents the worst situation when thecrostrip telescope. The full module was scanned in two positions, perpendicpion beam, last June, and tracks were reconstructed using a 4-plane miules in beam, at CERN. A full module detector was installed in a l00 GeVreconstruction capability, we have tested several single- and multi-chip mod

However, in order to measure the detector efficiency, as well as the track

nals and a radioactive 9°Sr source.Extensive tests have been performed in the laboratory, using calibration sig

5 Beam test results

ings. OCR Output

[3] M.Krammer: Ministrips in the DELPHI VFT Detector, these proceed

and ATLAS at LHC. Nucl. Instr. and Methods A342 (1994) 233.P.Delpierre et al.: Large Scale Pixel Detectors for DELPHI at LEP20UForward Direction, CERN/LEPC/92-13, DELPHI 92-142 GEN 135,DELPHI Collaboration: Proposal for the Upgrade of DELPHI in thet2l

(1994) 138.con Micropattern Pixel Detector Array, Nucl. Instr. and Methods A349E.H.M.Heijne et al.: First Operation of a 72k Element Hybrid Silitl}

References

during the developpment of this project.DELPHI collaboration, and in particular .].E. Augustin, for their supportwell as for the scheduling ofthe production. Finally, we want to thank thepersons: D. Parison and J.J. Ichai) during the whole prototyping period, aseffort made by IBM—France, Corbeil, and Montpellier Technologies (contactP. Weiss), where all detector wafers are produced. We want to stress the greatin collaboration with CSEM, Neufchatel (contact persons: A. Perret and

The two—metal process used for the detector modules was developpedinspection machine, and for the production tests.months (including some nights!) for the installation of the automatic waferIt is a pleasure to thank Petr Sicho, from FZU, Prague, who spent several

Acknowledgements

CPPM, and tested recently in beam, bonded to a detector.eral RadHard readout matrices implanted in DMILL technology, designed athave already been provided with viable solutions, as demonstrated by the sevintegration, within the thinnest possible structure. Some of these problemsradiation survival ofthe electronics and the detector, cooling and mechanicalto solve several remaining problems, like analog readout and time stamping,in the inner tracking system of ATLAS, and a huge R&D effort is underway

IEEE Trans. Nucl. Sci. NS—41 (1994) 1192. OCR Outputicon Micro—Vertex Detector Unit using the Flip-Chip Bonding Method,Y.Saitoh et al.; Development of Novel Fabrication Techniques for a Sil[7]

IEEE Trans. Nucl. Sci. NS-40 (1993) 400.J.J.Jaeger et al.: A Sparse Scan Circuit for Pixel Detector Readout,Chips, Nucl. Instr. and Methods A305 (1991} 627,P.Delpierre et .l.J.Jaeger: A Sparse Scan Circuit for Pixel Detector[6]

tions, to be published in Nucl. Instr. and Methods.

conductor X and 7 Ray Detectors Associated Electronics and Applicasented at the 9th International Workshop on Room Temperature SemiM.Cohen-Solal et J.C.Clemens: Electronics for Pixel Detectors, talk pre[5]

strobed binary output, IEEE Trans. Nucl. Sci. NS-39 (1992) 650.F.AnghinolH et al.: A 1006 element hybrid silicon pixel detector withMicropattern Detectors, CERN/DRDC/P22/90-81,RD19 Collaboration: Development of Hybrid and Monolithic Silicon[4]

p > 4 GeV(b). OCR Outputand without (dashed crosses) the VFT, for jets with p < 4 GeV (a.) andFigure 1: Track reconstruction efriciency vs. angie, with (solid crosses)

‘O 'E. ZC. 2; BC 'U `5 20 Z5 500;.;..;,4 `_)` 1 . A A 0 1 M; .,...4 . .-4;.. . . . .4..

0.1

22 }—0.2 f· -%·

* ¤

C 1 L. :

0.4 04% I

¥· t·Z

C.5 = L. ? ·· 2 · :E i

f ·a- I · :

J

¤_7¤.i`

{

ce

-1- IJf asc.s Y- G) 0.9 ¥— b> +

Figure 4: Principle of the hybrid pixel technique. OCR Output

AI

t_·_ S; t_;`

P +··"“';F

bvmnAlsolder

particle

and the buses.

Figure 3: Layout of a detector module, showing the 2 diiierent matrix sizes,

OCR Output•I'L·Hy¤-I

Figure 5: Sparse Data Scan Readout Diagram OCR Output

+1 +1 V +1 COLUMN SCAN

0.2 ps x (nb of hit + nb of lines)READOUT TIME =

also usable as a trigger signalNO => SKIP

= >

MII||I|I llllvlllllllmllfbllllllflllllltlltlllltllllltlllllflllllmllImllllllllllhnnnllllhlllunllvulnrmm+1

II•HrI¤II•||€IlW°I\I&I1I!IIIII$|@IIi I•IIC\“II|•$+1+1

or after readout

reset before BC0

e ............;?.

Ona

Hnriaqprdnnunn+I"I ggql $ Il Illlllxlt W mv -. *

+1."`"*""`

Q."T tl

LINE SCAN

Figure 10: Resolution in the x and y direction, vs. adjusted threshold, in k.e

Thrsskad

2 s s 75 10 ·2.s *5 ~7.5 20 22.5 25

OA A OQQJ 01% U ID

20

Over 5,***bcw ’0r 2 pi¤€\s c usters

syrwbo s for ‘ pixe cfusters

OCR OutputOCR OutputBO

7 4** I I I "A W r .

Trvasholc

7.5 5 22 wc *2.5 ·5 is 20 22.5 25

ww ~ * @¤z@ lm A A

, W;vn?

AAI I ·¢‘ '

Oper 5),/···bo\5 ‘0r 2 ;>}><e1s c ustsrs

Va.! symbcs for ‘ paxe c\ust€*s