The DAΦNE accelerator is presently under construction at Frascati for the study of physics at the phi resonance. The FINUDA experiment (FIsica NUcleare a DAΦne - FINUDA Technical Report, LNF-95/024 (IR) 18th May 1995) will study the physics of hypernuclei by stopping the kaons produced by the decay of the phi in various nuclear targets and observing the decay products, particularly the prompt pions which emerge at relatively low momenta. The cylindrical geometry tracking system is composed of two planes of double-sided silicon detectors, two planes of low mass drift chambers, and six planes of straw tubes. To minimise the multiple scattering, which would otherwise significantly limit the momentum resolution of the experiment, the inactive volume within the tracker is filled with a helium atmosphere. In addition to the tracking system there is a time-of-flight detector system comprising two layers of scintillators, one mounted close to the beam pipe and the other mounted beyond the straw tubes. The entire detector apparatus is mounted inside a superconducting solenoid magnet which generates a field of 1.1 Tesla.
The silicon vertex detector is made up of two cylindrical detector layers. The inner layer is referred to as ISIM and the outer layer as OSIM. A single module design is used for all the modules of both ISIM and OSIM, ISIM consisting of eight modules and OSIM consisting of ten modules. The nuclear targets, thin slabs of various elements, are mounted between the two silicon layers and thus ISIM and OSIM play distinctly different roles. ISIM tracks the particles produced from the interaction point to determine the point of impact in the nuclear target. Since the phi does not decay exclusively to kaon pairs, ISIM additionally provides particle identification using dE/dx since the energy loss of the kaons within the silicon is more than a factor of ten greater than minimum-ionising. Any particles emerging in the backward direction from the nuclear interaction will also be seen by ISIM. OSIM tracks the charged particles emerging from the target, mostly pions and protons. Again dE/dx information can be used to discriminate between particles. The position resolution necessary to achieve the required physics performance is around 50 microns, a figure well within the capability of the silicon detector design used.
Each module consists of a ladder of three identical silicon detectors glued end-to-end which constitute the active area of the module. At both ends of the ladder are glued beryllia ceramics separated from the silicon by quartz spacers. The ceramics provide the support for the readout electronics hybrids and also assist in the removal of heat from the electronics because of their high thermal conductivity. The quartz spacers are necessary to thermally isolate the silicon detectors from the readout electronics, since otherwise the heat would lead to increases in the reverse currents of the detectors and thus the noise in the readout electronics. The gluing of the detector modules is made using custom support jigs mounted on a high precision coordinate measuring machine to ensure that the deviation from linearity of the strips is no more than a few microns from one end of the ladder to the other.
On the "phi-side" - named as such because it measures the [phi]-coordinate in the cylindrical r-[phi]-z coordinate system - the strips run along the length of the ladder and are read out by a single readout hybrid at one end (the strips of each detector are wire-bonded together in a daisy-chain style). On the "z-side" the strips are orthogonal to the length of the ladder and so additional readout routing is required in order to connect the strips to the electronics. This is done using additional fanout pieces glued to the silicon detectors which allow the strips to be bonded to the fanouts and then the fanouts glued to the readout electronics. Two z-side readout hybrids are used, one at each end of the ladder.
The silicon detectors were produced by C.S.E.M. (Centre Suisse d'Electronique et de Microtechnique SA, Neuchatel, Switzerland) using the design already used for the ALEPH microvertex upgrade (D. Creanza et al., "Construction and Performance of the new ALEPH Vertex Detector", presented at the 5th International Conference on Advanced Technology and Particle Physics", Como, Italy 7-11, 1996). They are double-sided devices with orthogonal strips. The phi-side, which is the junction side of the detector, has 1021 readout strips which are 6.4cm long and have a pitch of 50 microns. Between every readout strip there is an additional floating strip read out via capacitive charge-division leading to a measurement precision capability of better than 10 microns (for vertical, straight particle trajectories). The phi-side strips are biased via the punch-through effect which gives an extremely high dynamic bias resistance and thus no significant contribution to the readout noise. The z-side, which is the ohmic side of the detector, has 1280 strips which are 5.1cm long and have a pitch of 50 microns. The fanouts connect to every second strip leading to a readout pitch of 100 microns with a floating strip in-between but this still allows a measurement precision capability of 10-15 microns. The z-side uses the p-stop isolation technique, with accumulation layer biasing which typically results in bias resistance values in the range of 10-30M[Omega]. All detectors are tested using a semi-automatic probe station to determine the leakage current and, on the z-side, the isolation resistance, bias resistance, and isolation voltage of every strip. In this way they can be selected and matched before gluing in order to achieve the best module performance.
The readout electronics are based on the VA1 chip, a commercial variant of the Viking chip (Toker et al., Nucl. Instr. & Meth. A340 (1994) 572), which are provided already mounted and tested on hybrids by I.D.E. A.S. (Integrated Detector and Electronics AS, Høvik, Norway). Each amplifier channel is made up of a low noise pre-amplifier followed by a shaping filter with a tunable shaping time of 1-2 microseconds and then a sample/hold circuit which is activated for all channels simultaneously by an external trigger. The voltages stored by the sample/hold circuit can then be read out sequentially by a serial analogue multiplexer circuit. Each VA1 chip has 128 amplifier channels on a pitch of 48 microns read out by a single multiplexer output channel.
The hybrids have eight VA1 chips, eight coupling capacitor chips, and a small number of discrete surface-mount components. The coupling capacitor chips and VA1's are glued directly to an aluminium nitride ceramic support, to which is also glued the printed circuit board which holds the discrete components and provides the power and control connections to the VA1's via wire bonds. A connector at the back of the pcb connects the hybrid via a flat cable to repeater electronics outside the tracking volume. The hybrid connection scheme allows all eight chips to be read out sequentially into a single flash ADC channel at a readout rate of 1 channel/microsecond.
Two fanout options will be used for the modules. In both cases the track layout is identical, only the fabrication techniques and practical implications differ.
- Copper on Upilex
Upilex fanouts were already developed at C.E.R.N. (Product Engineering and Support Group, CERN, Geneva for use with other experiments, such as the ALEPH microvertex upgrade already mentioned, and are made with copper tracks on a Upilex substrate which has a thickness of typically 50 microns. Although the production of the fanouts is relatively straightforward, the gluing of the fanouts to the silicon detectors is a critical step since all bond pads must have glue directly beneath them for wire bonding to be possible. At the same time, however, the limits on the maximum bonding distance mean that even small errors in the application of the glue will lead to the glue spreading beyond the fanout and also covering the bonding pads of the silicon detector. A further disadvantage of the Upilex approach arises from the fact that the Upilex is flexible and thus provides no mechanical stiffness to the modules. Since the material budget constraints preclude the use of additional stiffening, modules made with Upilex fanouts and mounted in a horizontal or near-horizontal orientation would suffer from a significant gravitational sag at the centre, thus complicating the data analysis somewhat.
- Gold on glass
Because of the negative aspects of the Upilex fanouts, an effort has been made, in conjunction with the MIPOT company (MIPOT S.p.A., Cormons), to develop a reliable process for producing fanouts with gold tracks on glass substrates of typically 100 micron thickness. The advantages of a rigid substrate would seem to imply an easier and more reliable process in comparison with that of the Upilex fanouts, and also make the gluing less critical. In reality, however, certain technological aspects - notably the cutting of the substrates - have required a significant amount of work to achieve an acceptable yield. Aluminium tracks were initially tried but problems with substantial series resistance lead to the adoption of gold as the conductor, with an under-layer of chromium being used to ensure good adherence of the gold tracks to the substrate. Small scale production has now been achieved with a satisfactory yield and so it is foreseen to make ten modules with glass fanouts for the positions where the modules are horizontal or near-horizontal and the remaining eight with Upilex fanouts for the more vertical orientations.
Ultrasonic wire-bonding is a crucial aspect of the construction of the modules. Each module involves approximately 10000 wire bonds. Good yield is vital to achieve good detector efficiency, and therefore make a good physics measurement. Cleanliness is fundamental, since a single dust particle on a bond pad will - at best - prevent the bonding of that pad and interrupt the automatic bonding sequence. All of the construction is made in clean room conditions. In addition, substantial effort has been made in the development of precision bonding supports. These have to protect the modules during handling and transport, but also are critical to the bonding process since, if the surface to be bonded is not held absolutely rigid then the ultrasonic energy is not transmitted to the bond wire but instead dispersed leading to weak or failed bonds. Since it is usually not possible to make more than two attempts to bond a given pad, errors for whatever reason must be avoided as much as possible.
When received, the hybrids are inspected and tested (including a 12 hour test under power to check for stability) before being sent to MIPOT for bonding of the connections between the VA1 inputs and the coupling capacitor chips. After bonding the hybrids are tested again, and then glued to the modules using a special epoxy loaded with aluminium nitride which ensures the maximum heat transfer between the electronics and the support ceramic. The modules are then sent to MIPOT for the complete bonding between the strips, the fanouts, and the electronics. Finally the modules are tested both by making a standard readout to check the noise performance and test pulse response of every channel, and by scanning with an infra-red light spot to determine the detector performance.
For every hybrid a flat cable connects the control and power lines to the data acquisition via a repeater card which buffers the analogue output signals and also handles the incoming digital hybrid control signals. The repeater card is, in turn, connected via an optical decoupler card to the ADC and to the readout control sequencer.
Space restrictions dictate that the repeater electronics cards are mounted at a distance of approximately one metre from the hybrids. This therefore requires that the hybrids alone must be able to drive the capacitive load of the cable without degradation of the noise performance. Tests have shown that this is possible for cables up to at least 1.5 metres long, although it was discovered that a selective shielding of the cables was necessary in order to prevent oscillation. At present it is believed that the shielding solves the problem because it introduces a high frequency damping effect for certain control lines to the hybrid. In the beam test of prototype modules, to be described later on, cables of 60cm length were used for all three hybrids without problems of instability.
All repeater cards are connected to the data acquisition via decoupler cards which optically decouple both the digital hybrid control signals and the multiplexed analogue outputs. This allows the hybrids to be maintained at the voltage of the detector strips, which otherwise would have to be held across the decoupling capacitors, and also provides electrical isolation from supply-borne electrical interference which can otherwise lead to severe readout noise problems.
The flash ADC's used are 10-bit 2-channel VME units from C.A.E.N. (V550 CRAM from CAEN, Costruzioni Apparecchiature Elettroniche Nucleari, S.p.A., Viareggio, Italy). These have on-board pedestal subtraction and thresholding capability, thus substantially reducing the data traffic required to read out a system with so many channels. It should, however, also be noted that this system of hardware sparsification is not capable of common-mode subtraction and therefore that this places a significant extra burden on the quality of the power supplies and shielding to ensure that detection efficiency does not suffer (adequate common mode correction can still be made by maintaining a limited number of "spy" channels with zero threshold, but any channel which remains below threshold due to negative-going common mode will inevitably be lost).
In February of this year, two prototype modules - one with Upilex fanout, one with glass fanout - were tested in a particle beam at TRIUMF, Vancouver. The aim was firstly to verify the operation of the modules and then secondly to study the dE/dx response of the modules to both pions and protons of various momenta. In particular, it was important to demonstrate that the dynamic range of the VA1's would be adequate for the heavily-ionising kaons which will be present in FINUDA. The figure at the left shows the experimental set-up in the beam. Three scintillators were used to generate a trigger using time-of-flight information for particle selection. A silicon telescope of three planes, each plane a single double-sided silicon detector, was used for measurement of the particle tracks. As can be seen from the figure, to maintain reasonable position measurement precision it was necessary to mount the telescope and prototype modules as close together as possible in order to minimise the effects of multiple scattering due to the low momentum beam.
Analysis of the data is underway, but histograms made using on-line monitoring can already be shown. The two figures below show histograms taken from the phi-side of one of the modules for two different beam momenta. For the low energy protons the separation is very clear, and the symmetrical form of the histogram shows that no saturation or non-linearity in the readout is occurring even with such large signals. For the higher energy protons, which are approximately three times minimum-ionising, one still sees a good separation from the pions.
A simple simulation of the beam test configuration for comparison with the results obtained was made using GEANT (GEANT, A Detector Description and Simulation Tool, part of the CERN Program Library), and the results are shown in the two pictures below.
The distributions are reproduced well. In the higher momentum data one sees a noticeable number of events between the region between the main peaks, this can at least in part be attributed to the simplistic common mode correction algorithm used in the on-line monitoring which did not correct completely for the substantial signals caused by the protons. The number of dead channels was also non-negligible for these modules due to the time pressure of being ready for the test, and so some events do not include all the energy deposited; since that time an important part of the bonding machine has been upgraded and the necessary improvements in bonding quality have been achieved. For the lower momentum results the absolute value of the energy deposited is very sensitive to the precise description of all material in the beam and so the agreement is considered to be acceptable. Finally, the pions in GEANT give a value which is too high for the most probable energy deposition of a minimum ionising particle, ~120keV instead of the correct value of around 85keV. Since 120keV is essentially correct as the average energy loss including delta rays this discrepancy probably arises from the treatment of delta rays in the simulation code.
Figure 9 (not available) shows first results from the off-line analysis, a scatter plot of the correlation between phi-side signals and z-side signals. The separate pion and proton peaks are clearly seen. It should be noted that the z-side signals in this plot are taken from one hybrid only, and therefore, with the beam positioned in the centre of the module, approximately one half of the events from the phi-side do not have a corresponding z-side signal.
All the important technological aspects of constructing and operating the modules for the Finuda silicon vertex detector have been verified successfully. Production is now underway of the modules for the experiment.