Metrika Inc. (Sunnyvale, CA), a manufacturer of single-use, quantitative medical devices for people with diabetes, produces diabetes test meters that measure levels of hemo-globin A1c in blood. It performs an 8-minute analysis on a blood sample, typically obtained from a fingertip. The reliability of test results depends on the blood sample. Once it passes through the meter’s aperture, the sample makes direct contact with a 1-millimeter-wide stripe of dry reagent chemistry on an assay strip inside the unit.

The chemical assay strips are sealed in a fixed position during the meter assembly process. The only variable that could affect blood-to-strip contact is the width and position of the chemistry stripping. This is applied to a foot-long, inch-wide laminate card that is diced into 100 individual chemistry strips. "Basically, we are drawing a line with liquid running the length of the laminate using a precision pump dispenser," explains Joel Blatt, director of research and development. "The laminate is hand-placed onto a table, which moves under the dispenser’s pen tip sitting on a micro-meter stage. If the micrometer is slightly off or the thickness of the laminate varies, the striping may not fall within our set tolerances."

To ensure the width and position of the striping, the company uses an In-Sight 2000 vision sensor from Cognex (Natick, MA). The inspections were previously done manually. An operator would place the laminate under a reticle and perform the measurements. While the required measurement accuracy could be achieved with this approach, the process was slow and became a production bottleneck. "With the sensor, we realized we could get the same accuracy in less time than it was taking under the eyepiece," says Blatt.

After purchasing the sensor, Blatt mounted the In-Sight camera above a backlit inspection station. He then began setting up the application using the sensor’s vision spreadsheet interface. The process involved selecting vision tools and parameters from drop-down menus using a handheld control pad. The vision spreadsheet then automatically generated tool results into worksheet cells, which were then linked together to set up the width and position inspections.

As part of the setup, Blatt trained the sensor to use the centers of sprocket holes, which are punched into the edge of the laminate, as the reference points from which to measure. "The holes are created using a laser process and remain in a fixed position. For that reason, it was an ideal reference point to use," says Blatt.

In run-time mode, the camera, mounted above the inspection point, captures an image of a laminate sitting on a backlit plate immediately after it is striped. The vision processor then analyzes the image, using Segment software tools that calculate the width of the stripe and its position relative to the sprockets. To see a live image of the laminate being inspected and results data, Blatt linked a monitor to the processor’s SVGA output and developed a custom operator interface that displays the values for each measurement. A green or red dot signals a passed or failed image. A personal computer records measurement data for each laminate, eliminating manual recording.

While backlighting is used to help establish better contrast between a freshly dispensed stripe and its laminate background, a number of image variations can occur that are problematic for machine vision. For example, the dispensed fluid normally turns the membrane translucent, resulting in a bright stripe against an opaque background. However, depending on the exact chemistry being striped, the stripe may appear darker than the background, resulting in a confusing scene for the vision system. According to Blatt, these conditions are not problematic for the sensor. "There can be a lot of variation in how the striping appears, but the vision tools are robust enough not to be fooled."

If the vision sensor determines that a stripe is too thick, too thin or outside positional tolerances, the laminate is discarded. If all dimensions are within specifications, the laminate is removed from the inspection station and moved to an oven. The stripe then dries until the fluid has completely evaporated. The laminate is cut into 100 individual pieces, which are later assembled into test meters.

Blatt feels that while the vision sensor helps minimize inspection cycle time, efficiency improvements take a back seat to the ability to get precise, accurate measurements. "Quality is so critical to the performance of these devices, and the vision sensor gives us the accuracy we need. This, in turn, ensures those who use the test meters get the accuracy they need," he says.

For more information on vision sensors, call Cognex Corp. at 508-650-3000 or visit www.cognex.com.