Andrew Wilson, Editor, [email protected]
The time taken for vendors to solve your specific problem often costs more than the product they are trying to sell you.
Being an engineering manager or systems integrator in the machine-vision industry is a difficult task. You may be asked to solve a number of complex, untried, untested problems and must understand every intricacy of how to integrate networking, machine vision, lighting, optics, and software into automated manufacturing systems.
At first, one might imagine that such tasks, while daunting, are not insurmountable. If you as an engineer lack a specific talent, such as lighting savvy or algorithm development, some vendor will surely help out in an effort to promote his/her solution as part of your system.
Unfortunately, OEM vendors of cameras, lighting equipment, frame grabbers, and software are often reticent to dedicate engineering time because the time taken for them to solve specific problems often costs more than the product they are trying to sell you. With intelligent frame grabbers now costing approximately $2000, and nonrecurring engineering (NRE) charges running at a minimum of $100/hour, a single engineering week will run at twice the cost of the company's product.
In an effort to promote their products, some less-reputable OEM vendors may make promises they find difficult to keep. In the fervor of many trade shows, I have encountered vendors who seem to promise the impossible to attain a sale. Later, after the product is delivered with a modicum of engineering support, it emerges that the technology is not capable of performing the task. The customer is then abandoned in the hope that he/she can somehow solve the problem without any OEM support.
Some rather unscrupulous sales people have even sold products before they exist. As technology changes so rapidly, such sales types can convince some OEMs to wait for products, as "they are only a few months away." In one case, a certain company managed to convince its customer just that. But things became worse when, at the last moment, certain semiconductors on the board could not be delivered. Even so, the customer was shipped a sparsely populated board and told he would have to wait for the delivery of those parts. Needless to say, the OEM switched vendors.
There are other reasons for this lack of interest by OEM vendors in their customer's products. If a lighting manufacturer, for example, knows that you are likely to design just one system using its parts, you may not even get to see a salesperson, let alone be provided any engineering support! Of course, this isn't the manufacturer's fault. In an effort to maximize sales, many OEMs will only provide engineering support to customers willing to pay for NRE costs. So while Phillip Morris, Kulike and Soffa, IBM, and the US Department of Defense can pay for engineering support, medium and small outfits are left to struggle as best they can.
Unfortunately, unlike the computer-graphics industry, there are no "killer apps" such as solids modeling, animation, and desktop publishing to drive the volume of OEM imaging components. Rather, there are a number of applications that span the gamut of packaging and production, food inspection, automotive analysis, vision-guided robotics, and medical and scientific image analysis. Even in these applications, no two are alike. Inspecting a packet of cigarettes traveling at 200 ft per minute, for example, may require a completely different approach than testing whether a bottle of soda is correctly filled.
Faced with such rapidly developing applications, many engineering managers turn to systems integrators—companies with expertise in developing solutions for specific applications. While such companies may have the engineering talent to develop such products, most are scattered across the USA, and few have immediately available resources to place at your disposal. As with most engineering problems, there is no easy way out. Somewhere you are going to pay multiples of the cost of the hardware required to build that automated inspection system.
