This thesis explores the modification of an EPSON EcoTank ET-8500 desktop inkjet printer to facilitate printing on CMOS chips for the cost-effective production of capacitive sensors. By reengineering both the hardware and software of the ET-8500, we aim to repurpose this standard desktop printer into a tool capable of advanced, precision printing for electronic applications.

The report details the development process, beginning with an overview of printer drivers and the Epson Standard Code for Printers (ESC/P) language. Utilising this foundation, custom software was developed in Python, featuring a user-friendly interface for seamless operation. The customised system was rigorously tested to assess the modified printer's capabilities, with results thoroughly analysed to determine feasibility and performance. The thesis concludes with a discussion of findings, along with recommendations for future enhancements and applications.

The Epson ET-8500 inkjet printer was modified to deposit precise quantities of ink on CMOS chips, enabling the fabrication of capacitive sensors. This project explores the potential of using affordable, consumer-grade printing technology for advanced sensor development, typically requiring expensive equipment. The research focused on understanding the interaction between printed ink and environmental variables such as humidity, light, and temperature. Three measurement setups were developed to detect capacitance changes under varying conditions. The setups for light and humidity successfully generated usable data, while the temperature setup needs redesigning for consistency. Key findings include the exponential relationship between humidity and capacitance in water-based inks. Light exposure generally decreased capacitance, except for white, zinc oxide and tin oxide under UV light. A machine learning model, using a neural network with a cross-entropy loss function, effectively identified inked electrodes but requires more variance in the datasets.

The manufacturing of sensors often requires the precise deposition of small volumes of liquid. In order to do this, expensive industrial equipment is often used, giving this form of manufacturing a high financial barrier to entry. As standard desktop inkjet printers operate by precisely depositing small volumes of paper onto a substrate, typically paper, the goal of the project was to modify a desktop inkjet printer, the Epson EcoTank ET-8500, to allow it to be used to deposit small droplets of fluid onto a CMOS sensor chip.

This report covers the design process of the hardware modifications that were necessary in order to allow the Epson EcoTank ET-8500 to print onto CMOS sensor chips packaged in a QFP- 44 package. These modifications include a custom chip tray to safely house the chips as they go into the printer, as well as an external fluid system to allow the user to easily use ink other than the stock ink provided by Epson or to put a cleaning fluid into the system in order to clear out any residue deposited by the inks used.

The results obtained indicated high levels of precision, and the system’s accuracy was sufficient for the sensor chips used. The external fluid system was also tested and successfully allowed the printer to work with fluids not placed into the built-in tanks.

Artificial Intelligence has become a dominant part of our lives, however, complex artificial intelligence models tend to use a lot of energy, computationally complex operations, and a lot of memory resources. Therefore, it excluded a whole class of hardware in its applicability. Namely, relatively resource-constrained low-cost hardware. This paper investigates learning methods that are potentially better suited for these types of devices: the forward-forward algorithm and Hebbian learning rules. The results are compared to backpropagation with equivalent network configurations, training hyperparameters and internal data types on different types of low-cost hardware. Backpropagation has consistently outperformed other algorithms in various tests. It exhibits higher accuracy, faster training, and faster inference compared to forward-forward models. However, forward-forward models can come close to matching backpropagation's performance, but they suffer from longer training times and decreased performance with multi-layer networks. Additionally, a poorly trained forward-forward model is sensitive to quantization, resulting in a significant drop in accuracy. On the other hand, forward-forward models offer the benefit of independently training each layer, allowing for more flexibility in optimizing the training process. Hebbian models were not found to be competitive, displaying performance below the required threshold. Smaller models for MCU and FPGA would likely perform even worse.

The creation of effective computational models that function within the power limitations of edge de- vices is an important research problem in the field of Artificial Intelligence (AI). While cutting-edge deep learning algorithms show promising results, they frequently need computing resources that are many orders of magnitude more than the available power and memory budgets for these devices. During the thesis, two unique learning algorithms (backpropagation and forward-forward) were developed and compared using the Teensy 4.1, a low-cost microcontroller board. This work seeks to bridge the gap between the necessary computing efficiency and the hardware’s restricted resources.

By creating and analyzing these algorithms, with the Fashion MNIST dataset as a validation set, this thesis creates a baseline for AI efficiency on microcontrollers, with performance targets set at a mini- mum of 80% test accuracy. The microcontroller software, implemented in C++, is limited to using less than 512 kB RAM for all online training methods. In addition, the potential of transfer learning was also explored.

Key performance parameters, including memory utilization, training and inference times, and accu- racy, were analyzed in a comparative study of the backpropagation and forward-forward algorithms. For each learning algorithm, several configurations were explored (such as topologies, and optimizers) to determine the most effective and efficient way for AI implementation on low-cost hardware. The key conclusions of this study reveal that backpropagation demonstrates superior performance in terms of both accuracy and computational efficiency. However, it requires more memory for storing variables, which may be a constraint in on-edge environments. Conversely, the forward-forward algorithm, while achieving lower accuracy, is more memory-efficient, making it a potential choice for less complex tasks or systems with severe RAM limitations.

The application of transfer learning showed potential to accelerate the learning process and to improve the final accuracy, hinting at an effective strategy for deploying advanced AI models on resource-limited edge devices.