Machine Vision Design for Performance and Cost

We all want to pay as little as possible for the things we buy.

  • Sometimes this means we only purchase for the performance we need. We don’t purchase top grade gasoline just because the oil company says it will enhance our engine’s performance.
  • Sometimes this means we purchase what is already available in the mass market. We don’t usually custom tailor our clothes, because mass-manufactured clothing is often good enough.
  • Sometimes this means we forego service and support. We don’t fly first class because we can live without the pampering.
  • And sometimes this means that we purchase from a cheaper source. We buy produce shipped halfway around the world because it costs less than our local produce.

Designing an imaging system is similar. The challenge is to design an inexpensive system and still meet the performance, quality, and support expectations of the market while delivering capability that satisfies the broadest possible range of applications. When designing such a system, there are some things to consider.

Many engineers like to work with leading-edge technology. To some, leading-edge means the best performance possible. To me, this means the best value possible. It is often more technically challenging to design for good enough performance with the lowest possible cost. The technical literature is replete with papers that discuss ways to improve performance, but almost none that discuss ways to improve yield and lower cost. It requires innovative thinking and personal discipline to design for good enough performance, and not more.

Accountants often equate low cost with low labour and material cost, but this is often not the case. Low cost requires good yield. To achieve good yield, one needs the best and brightest engineers, and they are not inexpensive to hire from anywhere in the world as a result of global mobility and regional shortages of highly skilled expertise. In addition, inexpensive materials do not help to achieve lower costs if they result in low yield. Unless their quality is high, cheap materials are typically anything but in the long run.

My advice?

If keeping costs low is important, leverage as much as possible from higher volume, lower cost standard products and customize only when necessary. There are teams of engineers working continuously to improve or sustain the yield of high volume products, while there are inevitably less resources dedicated to low volume products. Invariably, custom products cost more because of lower yield or higher per unit support cost.

We sometimes receive requests for higher grade versions of higher volume imagers. The cost difference will be small, some have reasoned, because we can still sell the lower-grade versions to others. This is only true if someone else is willing to purchase large quantities of lower-grade sensors for a lower price. Often, this is not the case.

We sometimes see requests for custom operating conditions. In one situation, an engineer wants to reuse existing electronics that force a set of unique biases to the imager. By reusing electronics, the engineer reckons that they will save on design costs. The unique biases appear to work for the samples tested. In the end, do we ultimately save cost? No, because this forces us to test the product with the unique biases to ensure that it still fully functions under these custom conditions. In case the product does not function under these conditions, a separate team of engineers will need to address this unique yield issue. Ultimately, custom operating conditions require additional service and support.

A few years ago, many equated CMOS imagers with inexpensive imagers. CMOS wafer costs are lower, some have reasoned, so CMOS imagers should be cheaper. In reality, wafer cost comprises only a fraction of the cost of an imager. Yield ultimately affects the imager cost more often than wafer cost.

In conclusion, I can offer the following advice:

Be clear on the objectives of your imaging system. Establish the performance you truly need and the cost that you have to meet. Expect trade-offs. Then choose a partner who will work with you to determine the right balance between performance and cost, who has designed and manufactured both higher-performance and volume products to be able to provide a wide range of options, and who has learned about the expense of building and leveraging design to mitigate costs from years of experience.





About Nixon

Nixon is often seen with a talented group of researchers, driving them to develop better technologies and products. When not doing this, he explores the Canadian Shield, Canada’s lakes, and Canada’s multicultural cities with his family.
Posted on by Nixon. This entry was posted in Machine Vision. Bookmark the permalink.

Comments are closed.