Some thoughts on SCTDAQ as a Multi-module Production Test System

version 2, 21 July 2000 GFM version 1, 11 June 2000 GFM
Contents
Introduction

SCTDAQ is a project by Peter Phillips, Gareth Moorhead and John Hill to profit from the development by Martin Morrissey and Maurice Goodrick of the two VME modules MuSTARD and SLOG for SCT readout and control. Mustard and Slog were developed to supply multi-channel readout capability for applications such as beamtest, systemtest, and hybrid and module production testing in the period prior to the availability of ROD prototypes. This has evolved as a project to provide ancillary hardware and a useful software environment to complement these readout modules, providing a reasonably complete system for testing both individual modules and hybrids and small systems of up to 6 or 12 hybrids and modules. This work commenced in late 1998 and has been continually evolving since, with implementations tailored for the systemtests, for testbeams at KEK and CERN, for single-event upset irradiation studies, and for detailed characterisation of individual modules. Our approach has been to provide general purpose components, both hardware and software, which can be built into systems meeting specific requirements. The software is built on the ROOT object-oriented physics analysis package developed at CERN which among other things provides a C++ interpreter as its native scripting language, a feature we use to provide macros for specific DAQ and analysis tasks.

This note discusses possible implementations with Mustard, Slog, SCTLV and SCTDAQ of multi-module production test systems for hybrids and modules, one of the main applications originally envisaged for Mustard and Slog. It does not disuss a priori the requirements or specifications for such tests. In particular, it assumes the modularity of 6 or 12 modules implicit in the design of Mustard and Slog which each have 12 channels. It should be noted that larger systems can easily be constructed (finances permitting) by using more Mustards and Slogs since from the outset we followed coding conventions which allow for multiple VME modules forming an arbitrarily large system (within the VME addressing constraints). This has been demonstrated with a system for 12 modules in the June 2000 testbeam at CERN.

The multi-module production-test application is envisaged to be routine, largely unattended, and technician or student operated, for HV soak-testing, thermal cycling, or burn-in under operating conditions to accelerate infant mortality of components. The test software needs to be complete in the sense of exercising the relevant circuits and functions for the duration of the test, robust against failures, and provide diagnostics and standard test results. Run macros for these functions need to be developed. As an example, a small production centre might manufacture five modules a week, which could then undergo a four- or five-day burn-in test cycle. This requirement can be comfortably met with a six-module test station.

For detailed testing and characterisation of individual modules or hybrids after major production steps (assembly, bonding, burn-in, transport etc.) it is envisaged that labs would have a separate Mustard/Slog/SCTDAQ system so as not to interrupt the unattended burn-in/soak system. This would also be used for debugging in the re-work cycle. Since full characterisation of a hybrid or module normally takes several hours, such a system should also cope with a similar production rate.

Since the systems for multi-module soak/burn-in testing and detailed single module characteristion are essentially identical, differing only in ancillary components, a lab with both also has a degree of redundancy and commonality.


Discussion of Existing SCTDAQ Features for Multi-module Readout

A basic SCTDAQ system consists of a Mustard and a Slog in a VME crate connected via a National Instruments NI-VXI interface to a PC running ROOT. Present implementations run under Windows NT or 95, but ROOT and SCTDAQ are largely platform-independent and have also run under various flavours of UNIX (linux, HP-UX etc.) for analysis. It should not be difficult to build an online version for linux once drivers for the NI-VXI interface are available. Ancillary hardware usually includes a CLOAC prototype SCT timing VME module for global clock, trigger and reset generation (essential in applications with external triggers requiring clock synchronisation such as beamtests, source or cosmic tests and perhaps laser scanning), and sufficient SCTLV prototype SCT low-voltage modules for the number of hybrids or modules in the system. Each SCTLV supplies low-voltage power, control signals and current monitoring for two hybrids or modules, so that s system of 6 modules would require 3 SCTLV units. Six fully-populated, double-sided modules, with two sets of clock and control inputs and two data outputs each, are supported by a one each of the 12-channel Mustard and Slog in their simplest arrangement. Multiples of 6 modules can be accommodated in the same architecture by multiple sets of Mustard, Slog and three SCTLVs. Only one CLOAC is required in any system for global timing of fast commands (triggers and resets). Modules must also interface to the VME equipment, either via optical readout components as at the systemtests, or by electrical repeater cards such as the SC99 support card.

Although in principle a module uses only one set of clock and control inputs at any time (controlled by the SELECT low-voltage control signal), complete testing requires connections to both sets of input. In the systemtest and other applications with optical readout, switching between the two input sets is done at the optical transmitter, eg, the BiLED optical driver unit, so one SLOG can drive more than six modules. In electrical readout systems using the existing Melbourne patch panel (PPR) and support cards (SC99) , two sets of clock and control are connected to each module, providing full test functionality but limiting each Slog in the system to six modules.

A six-module electrical test stand for barrel modules in QMW-style test boxes or hybrids connected to SC99 support cards can be itemised as follows, and has already basic software support for testing multiple modules:

Implementation of systems for multiples of 6 modules using such equipment sets is straight forward, and already demonstrated in the testbeam. However, it may not be practical to acquire duplicates of the more expensive VME modules. Options for more cost-effective solutions for more than six modules are discussed below.

Caveats

Cheaper Multi-module Readout

Several options exist to reduce the cost of systems of more than 6 modules, such as might be required at larger production centres that do not want to purchase more than two Mustard/Slog systems (one each for their multi-module and detailed single-module test stands). However, there are some observations which can be made:

Firstly, there is not much to be gained at the low-voltage side, since SCTLV provides all the necessary specialised supply voltages, current monitoring and control voltages in a cost-effective way.

Secondly, support or repeater cards are inexpensive items, around CHF120 per module at present. While simplifications might be achieved through placing the circuitry for multiple modules on one PCB, most of the cost is in the components. The natural geometry for a stack of modules is at right-angles to the orientation of their connectors, so intermediate cables would in any case be required. It may prove that such a solution is favoured for mechanical or thermal control reasons, but it is unlikely to be cheaper.

Some saving can be made with SLOG, which has 12 clock and control channels. A minor variation in the repeater card design, with no cost penalty, would allow use of the SELECT control line to switch a single set of clock and control supplied to either of the two sets of module clock and control inputs. This would allow full redundancy testing of 12 modules from one SLOG at no extra cost. New PPR and SCs would need to be designed and purchased.

A significant cost in the SCTDAQ hardware collection is the MuSTARD. For detailed module characterisation, and most other applications of SCTDAQ (beamtest, systemtest etc.) simultaneous monitoring of both data links of all modules is essential. However, for soak/burn-in type testing, it may be that occasional or cyclic monitoring is sufficient. If cyclic monitoring is deemed acceptable, then one can envisage multiplexing the LVDS signals from many modules to Mustard. A particularly simple case, requiring only one control bit, would be to switch between two sets of 12 data links, allowing 12 modules to be cyclically monitored. This should cater to the requirements of larger production centres. This could be achieved by daisy-chaining pairs of support cards, or "twinning" modules onto one support card per pair, or by having a switching board at the VME side, replacing the passive patch panel. If the switching occurred at the support card, it could be controlled (in the case of barrel modules) by an unused SCTLV low-voltage line intended for the opto-harness, or it could be derived from the SELECT state changes controlling a flip-flop (which would be reset by RESET to ensure it was tracked by the controller). If more extensive corner testing requires that a more complex support card be designed, then such concerns can be addressed in a common design.

Forward Module Optical and Electrical Readout

Existing solutions for forward modules without opto chips bonded (DORIC and VDC) include:

An important question for forward multi-module testing is whether or not it is necessary to include the DORIC and VDC in the test chain during the long-term multi-module testing, or whether electrical tests of the rest of the system are sufficient, with tests of the opto chips before and after burn-in. It is always possible to supply clock and control electrically to a forward hybrid using its LVDS redundancy inputs; it should be possible to read the data out electrically without disturbing VDC by monitoring the SPY connector with some kind of high-impedance receiver. This possibility is currently under active investigation but is not yet demonstrated.

If it is determined to be a requirement for soak/burn-in that DORIC and VDC must be tested during the long-term cycle, then several solutions may be available: