Cost modeling or price comparing? Complex tools aren't needed if what you want is to compare prices.

GOING BACK A few years, I recall taking some sketches of an idea, a "black box," to the "model shop" and having a 3D version of my idea formed out of a variety of materials, from plywood, plastic, plaster of Paris or clay. Keep in mind this was only a mechanical part. Today we can build such models with the computer, and if we have to have something to touch we can use 3D lithography. Adding the element of cost to that model was possible, in fact, easy. I would figure out how to manufacture it, choose the materials and send my sketches to the appropriate fabricators for pricing. However, every time I made a small change to the physical model I would have to reanalyze the costs through this same, manual process. Each model that I built would have its particular associated cost. If I wanted to create a cost model for this particular type of black box, I could build a number of physical models (this could be a large number), systematically varying the cost-drivers such as size, thickness, materials of construction, machined features and tolerances, and the manufacturing processes such as casting, hog-out (machining from a solid piece of material), molding, making subassemblies to be bolted together and on and on. With these data I could create a spreadsheet to aid in predicting the cost of various forms of my black box. The beginning of a basic cost model.

A cost model for embedded passives would be infinitely more complex, perhaps even impractical, and most likely inaccurate because the variables are ever in flux.

Consider the OEM, the project engineer or project manager doing the modeling. Their desired modeling process, I believe, would involve inputting a few variables and outputting some end-product cost for the purpose of making comparisons. (The end-product in this case would be an assembled PCB.) The inputs would he specific information extracted from the schematic and BoM. The tool would associate the costs of a standard SMT assembly of passive parts vs. embedding some of these parts directly in the PCB substrate. Based on determining which passives could be embedded, the applicable embedding materials technology, the potential for size reduction of the original PCB which could result in a more efficient PCB process, and the absence of the piece part and assembly costs for the embedded parts, the system cost could be determined from a modeling tool.

I am skeptical of the feasibility and meaningfulness of this computerized process because the tool would have the user analyzing IMS (interconnect manufacturing services providers, aka PCB shops) and EMS materials and manufacturing costs. As such, "PCB cost" is perceived to be the cost of the PCB, when in reality what you pay for a PCB is the price. Today, board prices and board costs are virtually unrelated. There is no reason to believe that this will not hold true for embedded passives.

The key word here is cost. Modeling PCB costs assumes that throughout the IMS industry there are standard costs, implying relatively standard pricing, for PCBs. Anyone fostering this assumption is hopelessly naive. I recognize that my forthcoming comments are generalized and that there are exceptions. (If you are in that category, I applaud you.) Pricing for a given product and quantity in North America may vary over a range of 10X or greater. If brand A prices a board for $1 and brand B prices it for $10, that is a 1,000% difference. This, and everything in between, happens everyday!

There are five drivers to an IMS' cost performance:

* Management commitment to quality.

* Equipment platform.

* Robustness of systems.

* Maturity of staff.

* Depth of technical experience.

All these affect yield, which determines product quality throughput and, ultimately, cost. Weakness in any of these drivers creates a huge variable in cost from IMS to IMS. Relatively few merchant IMS companies have cost systems or, in particular, activity-based costing (ABC). The prices paid for materials, both direct and indirect, vary significantly, as do labor and benefits costs, thus adding more variation.

Within this broad spectrum of price and given the cost performance drivers and material and labor variations, some manufacturers are able to produce profitably at the lower price while others are marginally profitable or even lose money at the higher price. Some, driven by market-share mentality or commission-seeking sales reps, have no idea whether the job will be profitable, regardless of their bid. It's sad, but it's also reality.

Therefore, cost, as contemplated in the embedded passives cost-modeling concept, is really the price of the PCB and/or discrete parts and assembly. These are so variable and so unrelated to anything tangible that I believe it is virtually impossible to create a meaningful cost-modeling tool. My comments on the cost vs. price in the IMS industry are heartfelt and based on years of understanding how we do business. This is my world. I am not as familiar with the EMS side of electronics, but understand that there may be similarities.

If what you want is to compare prices, the desired results may be relatively easy to come by without the aid of a complex tool. Next month, I'll describe how to do this.

RICHARD SNOGREN is a member of the technical staff at Coretec Inc. (coretec-inc.com). He can be reached at [e-mail]. He will speak on embedded passives this month at PCB West 2004 (pcbwest.com)