Issue link: https://iconnect007.uberflip.com/i/1531663
36 PCB007 MAGAZINE I JANUARY 2025 Many of the best shops I have visited have some type of mechanism for passing on skills and knowledge from experienced people. Usu- ally, it's as an apprenticeship, although it's oen not called that. Employees do not move to a new job, promotion, or retirement until they have trained their replacement, so it takes some foresight and planning from the manage- ment to implement this. If this doesn't happen, it puts more strain on the company. I have found things are smoother when the operators have some understanding of the pro- cesses. I've seen too many cases where oper- ators kept feeding panels into the etcher and producing scrap simply because they didn't realize there was a problem. Most of the pro- cesses in producing a circuit board are not that complicated and don't require a college degree to understand them. Most of the successful shops I have seen combine training offered by their suppliers with internal training and promotion. ey then trust those workers to recognize potential problems and bring them to the attention of their supervisors before things get out of hand. Many of the not-so-good shops try to cut expenses by reducing the number of skilled (and higher-paid) workers to a bare mini- mum, depending on them to direct lower-paid unskilled workers. is seldom works because spreading skilled supervisors too thin over- whelms and burns them out. Minimum-wage workers don't care because every problem is above their wage level (and I find it hard to fault them for this). I don't believe you can produce high-tech products with minimum-wage labor. Finding and training skilled workers requires companies to train them to become more invested in the industry. When they are invested, workers will want to advance their skills, seek promotions, and help train oth- ers. It's a symbiotic process, and it works when everyone is aligned with the same goals. PCB007 Don Ball is a process engineer at Chemcut. To read past columns or contact Ball, click here. Improving Brain–Machine Interfaces With Machine Learning Brain–machine interfaces (BMIs) have enabled a handful of test participants who are unable to move or speak to communicate simply by thinking. BMIs currently in an experimental phase may also con- sist of robotic limbs that can execute manual tasks as instructed by a disabled person's thoughts alone. At Caltech, researchers use implants that con- sist of arrays of 100 microelectrodes mounted on a 4x4 mm chip. Unfortunately, the performance of these microelectrode arrays is not consistent and degrades over time. To overcome this challenge, Caltech's Azita Emami and her colleagues have used machine learning to effectively interpret the neuronal signals picked up by older implants. Now Emami and her colleagues have found that by applying machine learning, BMIs can be trained to interpret data from neural activity even after the signal from an implant has become less clear. The team's algorithm is called FENet, for Fea- ture Extraction Network. Remarkably, it can be trained on data from one patient and then used successfully in another. "This means that there is some fundamental type of information in the neu- ral data that we are picking up," Emami says. Not only that, FENet can generalize across different brain regions and types of electrodes and be eas- ily incorporated into existing BMIs. This research was published in Nature Biomedi- cal Engineering under the title "Enhanced control of a brain-machine interface by tetraplegic participants via neural-network-mediated feature extraction." (Source: Caltech)