I worked at AutoMed from March of 1996 to November of 1997. AutoMed manufactured equipment used to sort, centrifuge and aliquot blood samples in clinical laboratories.
At the time I started, the company had two divisions, one in Richmond, BC and the other in Arden Hills, MN. At that time they were well along in the process of consolidating all operations in Richmond. Everyone in the Arden Hills facility had decided not to transfer to Richmond so my job was to learn all that could be learned about the electronic systems of the three products that they built there before that knowledge slipped away for good.
All machines that AutoMed developed resembled Rube Goldberg devices both mechanically and electronically. For something that mostly moved specimens along a conveyor track, the machines were remarkably complex. As a result, they were also not particularly reliable. I spent much of my time at AutoMed doing trouble shooting, either on the production floor, over the phone or at customer sites. I even did a six-week installation trip to Perth, Australia.
After about a year I had an opportunity to help right the wrongs of the past when I joined AutoMed's Skunk Works, a team assigned to develop AutoMed's next generation of products.
Eight Lane Test Tube Counter
My best assignment as part of the new team was to develop a device that could count the test tubes leaving a sorter. The device would count eight lanes simultaneously, using optical detectors. The previous AutoMed sorters only knew when a sorting lane was full. This newer sorter would be able to log when each barcoded sample was removed from a sorter lane, even if only one sample were removed from the lane at a time.
To implement this project, I used a Microchip PIC microcontroller. It had room for 512 instructions in its program memory, 128 bytes of RAM and no interrupts. Sensor sampling and a packetized serial communication scheme to the host computer were implemented in firmware. Each program loop had to take exactly one half bit period of a 9600 baud serial line in order to maintain reliable communications and counts. To do this I had to keep track of how many instruction cycles each section of code took and either optimize the algorithm or add small delays to adjust the execution time.
The completed counter turned out to be extremely reliable. The acid test came when the Vice President slid a line of several tubes back and forth through a sort lane as fast as he could. The count remained accurate, even while the main computer was querying the microcontroller at the same time.
Sadly not much remains of AutoMed or its products. In my last year there, it was bought out by its main investor, MDS. I was caught in the first of three waves of layoffs and within a year AutoMed was completely gone. The next-generation products that we were developing in the Skunk Works never made it to market. MDS took over the responsibility of servicing the existing installations but I have no idea if any of the AutoMed machines are still in operation.