SimWitch: A Software Tool for Assessing Cost Distribution of Materials Streams
Application to the FLIPAC Esprit Project
 SimWitch®, acronym for Simulate Where Is The Cost Hidden, is a software tool aiming at simulating a material process, in terms of time and cost. Not pretending to compete with dedicated commercial and proprietary tools, we designed it as a first approach tool for engineers to assess the cost distribution of a process step and perform cost sensitivity analysis. It was used within the “cost analysis” work package of the EC-funded Esprit project FLIPAC, to assess the cost breakdown of MCM technology demonstrators.
SimWitch lets you draw your piece of system with interconnected pre-characterised modules, which operating parameters (delay, batch size, yield, throughput, uptime...) can be tuned to match your actual system (Figure 1).
Early awareness of cost issues
We believe that taking into account and anticipating costing aspects, during the initial phase of a product or a process, are not systematically given the precocious consideration that they deserve, by the teams in charge of designing and implementing the products or processes. This may be caused primarily by the difficulty to handle the cost attributes of objects when compared to purely technical aspects, which probably prove easier to be described and quantified.
Now, the cost of operating, owning an object, a product, a tool, is oftentimes highly dependant on the basic and early choices and options that are selected as soon as is organised the overall chain of design-production-delivery. Some of these options may also be dictated from mere habit and reluctance to changes, without being objectively confronted to possible alternatives, without seeking potential productivity gains or costreductions.
A complementary approach to spreadsheets
We see that the process of elaborating a material object can often be formally described, but it is not so often the case for describing the corresponding cost distribution. Depending on personal and corporate habits and practices, the cost estimation is oftentimes realised by mixing three approaches:
- empirical computation, rule-of-thumb and fundamental estimations,
- utilisation of spreadsheet programs,
- utilisation of proprietary software, linked to accounting and production data bases (ERPs).
Empiricism
Empiricism and estimations are irreplaceable, as they are part of the individual and collective experience of a company related to its specific products and market; they sustain feeling and intuition, but they may not allow an objective and documented comparison, and may not always be suitable for introducing new methods or new technologies.
Spreadsheets
Spreadsheets are of a very widespread usage and bring daily help at every level of an organisation, thanks to their ability to handle large quantities of numbers. Their purchasing price is small and they suit many applications. They are made to handle linked series of numbers, but they can hardly integrate the non-synchronism and random time effects that are impacting all industrial activities: bottlenecks, delays, overload, supply break....
Proprietary software tools
Proprietary software tools are designed to fulfil specific needs of a company or of a department, but their utilisation may be restricted to a certain category of users. If correctly designed, these tools may be of great help and achieve powerful control. On the other hand, they may not be as user-friendly as the PC-user can expect, and, more annoying, their maintenance and evolution can be costly or even impossible.
Here comes SimWitch...
Noting frequent and repetitive facts, we devised a simple and generic method of computing streams of items and related costs, implemented in a software tool. Its development was led within the frame of the Thematic Network "COMPETE" (CO-ordination of Microelectronics Packaging & interconnection projects: Environment and Trends for the development of European solutions), funded by the European Commission and co-ordinated by MTA (www.compete.tm.fr). This method, based on a time-simulator, does not pretend to substitute or replace the other ones, here above mentioned. We rather see it as a useful complement for engineers, linked with personal estimations and knowledge on one hand, and with spreadsheets on the other hand.
Why a time simulation?
When a phenomenon involving non linear behaviour has to be studied, two basic approaches can be considered:
- If you are skilled in mathematics, you can attempt to modelise your system by devising clever transfer functions, that you assemble so that they can represent the behaviour of the system. This is easily achievable if you concentrate on a single system that you are not likely to change every morning.
- If your mathematical background has fade away, or if it proves practically not possible to obtain these mathematical relations, particularly if you change the structure of your system every other day, then you simulate your system, step by step, replacing complex behavioural modelling by simpler computing of the evolution of the state of every node of your system.
This is what SimWitch actually does: as we are unable to predict the non-correlated interactions caused to streams by local delays, different throughputs, bottlenecks and cyclic uptime periods for a variable diagram, SimWitch implements a few basic behaviours that your computer will be able to repeat hundred of times, to simulate various systems.
More than a spreadsheet
Being a true time-simulator, SimWitch implements functionality such as delays and throughput-bottlenecks, which cannot be taken into account by spreadsheets applications. It is a general purpose tool, using a handful of generic modules, not restricted for a specific application or industrial domain.
Modules receive input streams of items and deliver an output stream of other items, through the connection between them. This time simulation of the internal streams of items modelises the interdependency of all modules, and permits a realistic computation of the costs incurred by the system.
The "value" of streams is increasing from upstream to downstream as modules are adding their own costs to the streams they handle or convert.
Being the end of the line, most downstreams modules are having unconnected output ports. The user then imposes onto these floating ports a volume requirement in terms of output quantity, a constraint that SimWitch transfers backwards to the upstream blocks.
The cost distribution is then analysed for every module, which weights the streams of items crossing it with costs. One can compare the unit cost, the total cost of the input items to the unit cost and total cost of output items:
Assessing the impact of a parameter
The possibility of « sweeping » (sequentially assigning a set of values) any parameter of a module is intended to perform a rough cost-sensitivity analysis, and is very useful for comparing the impact of any parameter in the item's unit cost at every level of the ine.
Targeted users
SimWitch is particularly helpful during the early phases of a project, when:
- choices and options (organisation, materials, tasks sequencing...) are still possible,
- the local and global impacts of these choices must be compared on a quantitative basis.
The versatility of the SimWitch method meets various levels of abstraction and corresponding modelisation: designers can modelise the basic steps of a design job, production, planning and engineering may optimise their streams, while marketing teams may describe timing of product introduction and associated costs.
This is why SimWitch was successfully used within the Cost Analysis of the FLIPAC project, here after described.
Application within the FLIPAC project
FLIPAC, Fine Line Interconnection Packaging, is a 30 month ESPRIT project, started for 30 months in January, 1998, totalling a bud-get of 2,8M euros and 38 men-years. It brings together 9 partners:
Bull SA (France - coordinator), Bull HN (Italy), Cimulec (France), Dicryl (Spain), ACREO (Sweden), IMEC (Belgium), Italtel (Italy), AMP-AKZO (The Netherlands) and Technische Universität Berlin (Germany).
The main objectives of the programme are:
- to develop and compare build-up printed circuit boards using enhanced processes and materials,
- design, build and test of functional MCM-L/D demonstrators for Data Processing and Telecommunication industry.
(More information is available on the FLIPAC web site at http://mst.tu-berlin.de/flipac/sites/index.htm)
Demonstrators are made of active and passive chips, mounted on a laminate substrate created with additive technologies, the resulting carrier looking like a BGA.
The cost breakdown parameters of the telecom demonstrator module to be delivered by FLIPAC were then injected into SimWitch.
Several modules are laid out and tied together to simulate the various operations from dicing to final test, through substrate manufacturing and chip bumping and bonding.
Global expenses per module demonstrate that a single operation (FPGA rerouting) is the major cost contributor, followed by laminate manufacturing, then bonding and bumping of various chips.
About the authors
Olivier de SAINT LEGER received his MS in electronics in 1979 and held various engineering and marketing positions in hybrid and semiconductor companies, focusing on analog-digital custom Ics, GaAs devices and thin-film modules. He joined MTA, Marketing & Technologies Avancées, in 1997 as a consultant, where he conducts specific marketing studies for customers of the electronics and microelectronics industry. He co-ordinates the Thematic Network COMPETE, for which he designed the SimWitch software.
Daniel LAMBERT received a MS in Physical Sciences in 1975 and PhD degree in Material Sciences in 1980 from university P. & M. Curie (Paris VI).
In 1980 he joined CII Honeywell Bull where he first worked as a photolithographic process development engineer for the custom Integrated Circuit proptotype line (HMOS and CMOS technologies).
From 1986 to 1992, he worked on the development of a Polyimide/Copper thin film process on cofired ceramic substrates (MCM-C/D). He was responsible for: photolithography and polyimide processes; design of test vehicules and he had in charge the capacitance test procedures applied to bare MCM-D. He contributed to the design and the manufacture of MCM-D built on cofired ceramic substrates.
Since 1993, he has been in charge the development of High Density Interconnection packaging technologies: built up on PWB with applications to PBGA and MCM-L and he is project leader of MCM projects for harsh environment applications.
He is involved in European R/D programs: FLIPAC, DONDODEM, AUDICO and EPMCM.
Pascal GUILBAUT received the electronics engineering degree from ENSERB (Bordeaux) in 1997. He is currently doing a PhD in industrial environment at Bull SA (France), in relation with the IXL laboratory (Bordeaux), where he works on high density interconnect substrates.
| Table of Contents | Previous Page | Next Page |
|
