Hi, I'm Alan Alkhayat

Strategic Power-Saving Dual-Processors

×

To master the critical power challenge for all-night operation and to keep cost low, we engineered a power management strategy, using a dual-processor architecture. An ultra-low-power STM8S microcontroller acted as the dedicated 'sensing engine.' It continuously sampled Heart Rate (PPG) data at a precise 50Hz – a frequency also chosen to leverage Nyquist-Shannon theorem of aliasing to filter out common 50Hz environmental noise without extra hardware filters. (See 1st and 2nd slide below)

The true power savings, however, were unlocked by the role of its partner, an ESP32-C3 Mini. This more powerful MCU remained in Deep Sleep mode for the vast majority of the time, waking for mere milliseconds only when signaled by the STM8S to store, connect, and transmit data to the mobile application. (See 3rd slide)

Inter-processor level: The STM8 continuously produces data while the ESP32 sleeps. We implemented a handshake protocol where the STM8 signals when its 125-byte buffer is ready, wakes the ESP32 via GPIO, then waits for acknowledgment before UART transmission begins.

As illustrated in our system timing diagram (4th slide), the ESP32 would wake every 2.5 seconds upon the 'Wake ESP' signal, triggered after the STM8S filled its buffer, to receive a 125-byte data chunk (The limit of the stm8) via UART. After aggregating four such transfers into its own buffer (representing 10 seconds of data), it would then briefly handle the efficient Bluetooth Low Energy (BLE) 'Announce Data' and subsequent transmission to the app before returning to its ultra-low power Deep Sleep state. This orchestrated interplay, which minimized the ESP32's active time, was pivotal. It enabled an impressive 16.84 hours of continuous operation on a single 245mAh battery, an 8-fold improvement over the unoptimized continuous monitoring method attributable to this duel microcontroller operation, Data packaging scheme and optimized wake/sleep scheduling.

Engineering for Compactness

A core design criteria from the outset was to create a truly wearable and unobtrusive device. This necessitated attention to component selection for minimal footprint and a highly compact Printed Circuit Board (PCB) layout. We successfully engineered the entire system onto a mere 35mm x 19mm PCB, a significant achievement given the inclusion of two microcontrollers, a novel analog front-end for PPG sensing, power/and battery management systems and a USB-C receptacle. (1st Slide)

For cost-effectiveness, we committed to a 2-layer PCB design. This presented a considerable routing challenge compared to a multi-layer board, especially for a mixed-signal circuit where attention to PCB best practices becomes near critical. However, we were able to implement best practices even within these constraints. This included clear differentiation and separation of power, analog, and digital sections on the board to minimize noise coupling and ensure signal integrity. Ground planes were carefully managed, and trace routing for sensitive analog signals was prioritized to shield them from digital interference, all contributing to the reliable performance of our PPG sensor. (2nd Slide)

Signal path and System Validation

Before we could even begin to analyze complex physiological signals, we first had to build undoubtable trust in our entire data pipeline. We achieved this through end-to-end, Known Signal Validation. By injecting a precise 1.1Hz sine wave from a function generator directly into the STM8S microcontroller's ADC, we traced its journey through every step of the system. The resulting clean, perfectly reconstructed sine wave (1st slide) – captured after traversing STM8S buffers, UART, ESP32 buffers, BLE, and mobile app reassembly – conclusively verified the integrity of our entire data pathway, timing and packaging scheme. This validation ensured that any variations seen later with the actual PPG sensor were due to physiology or sensor noise, not systemic data corruption.

Charging Mechanism

To further compact the design, the existing USB-C module could be replaced with a wireless charging system. This not only reduces physical space but also enhances the user experience by eliminating the need for wired connections.

Power Supply Optimization

A major area for improvement is the system's power supply, using a switch-mode power suply would drastically improve our energy efficiency:

Precision Potentiostat Control Loop

The heart of any potentiostat is its control loop, which maintains a precise potential between the working and reference electrodes while measuring the resulting current. I implemented a classic three-op-amp topology with critical enhancements:

• Control Amplifier: OPA192 ultra-low offset (5μV max) op-amp in a voltage follower configuration, driving the counter electrode to maintain the desired cell potential
• Voltage Follower: High-impedance buffer (>10^12Ω) for the reference electrode using AD8608, preventing any current draw that could disturb the electrochemical cell
• Transimpedance Amplifier (TIA): Converts working electrode current to voltage with programmable gain stages

The control loop stability was particularly challenging. Electrochemical cells present complex impedances that vary with frequency, potentially causing oscillation. I implemented compensation networks with adjustable phase margin, achieving stable operation across cell capacitances from 1nF to 10μF—covering everything from microelectrodes to large-area sensors.

Auto-Ranging Current Measurement

To achieve our ambitious dynamic range of 1nA to 100mA, I designed a sophisticated auto-ranging system with seven overlapping ranges:

The key innovation was implementing "make-before-break" switching using low-leakage analog switches (ADG1419) with break-before-make delays under 10ns. This prevents the working electrode from going to open circuit during range changes, which would disturb the electrochemical reaction. Additionally, I implemented predictive ranging based on the derivative of the current signal, switching ranges before saturation occurs.

Noise Management & PCB Design

Achieving sub-nanoamp noise floor required obsessive attention to PCB layout:

• Guard Rings: Active guards driven by low-impedance buffers surround all high-impedance nodes, preventing leakage currents through the PCB substrate
• Layered Ground Planes: Separate analog and digital ground planes with single-point star connection at the ADC
• Faraday Shielding: Critical analog sections enclosed in grounded copper pours on all layers
• Component Selection: Low-noise resistors (thin-film 0.1%), C0G/NP0 capacitors for critical filters, and careful thermal management to minimize drift

The result: measured noise floor of 450 fA RMS in the most sensitive range, approaching theoretical limits set by Johnson noise of the feedback resistor.

Microcontroller & Real-Time Control

I selected the STM32F446RE microcontroller for its unique combination of features essential for electrochemical measurements:

• 180 MHz ARM Cortex-M4F with FPU for real-time signal processing
• Dual 12-bit DACs for simultaneous potential control and offset compensation
• Three independent 12-bit ADCs enabling simultaneous multi-channel acquisition
• Advanced timers with 32-bit resolution for precise timing control
• USB 2.0 Full-Speed with dedicated 1.25KB buffer for streaming data

The firmware architecture leverages FreeRTOS to manage multiple concurrent tasks:

• Control Task (1kHz): PID control loop maintaining cell potential
• Measurement Task (10kHz max): Current sampling and auto-ranging
• Processing Task: Digital filtering, moving averages, and peak detection
• Communication Task: USB packet handling and command processing
• Safety Task: Overcurrent protection, thermal monitoring, and watchdog

Advanced Measurement Techniques

The firmware implements multiple electrochemical techniques, each with specific timing and control requirements:

• Cyclic Voltammetry (CV): Triangular potential sweeps from 0.1 mV/s to 10 V/s with automatic IR compensation
• Differential Pulse Voltammetry (DPV): Superimposed pulses with 1ms timing resolution for enhanced sensitivity
• Square Wave Voltammetry (SWV): High-frequency (up to 1kHz) square waves for rapid scanning
• Chronoamperometry: Potential steps with microsecond resolution and automatic integration
• Electrochemical Impedance Spectroscopy (EIS): Frequency sweeps from 0.1 Hz to 100 kHz with real-time FFT

GUI Architecture

I developed a comprehensive Python application using PyQt5 and object-oriented design principles:

• Model-View-Controller (MVC) Pattern: Clean separation between data handling, UI, and control logic
• Multithreading: Separate threads for USB communication, data processing, and UI updates to maintain responsiveness
• Real-time Plotting: Custom Matplotlib integration with blitting for smooth 60 FPS updates
• Data Management: SQLite database for experiment storage with automatic backups

Key Software Features

• Live Data Visualization: Multiple plot types (I-V curves, time series, Nyquist plots) with zoom, pan, and annotation tools
• Automated Analysis: Peak detection, baseline correction, integration, and curve fitting algorithms
• Experiment Scripting: Python-based scripting engine for complex experimental sequences
• Calibration Wizard: Guided calibration using standard solutions with automatic curve fitting
• Report Generation: Automated PDF reports with graphs, statistics, and experimental parameters

USB Communication Protocol

I designed a robust binary protocol for high-speed data transfer:

• Packet-based with CRC32 error detection
• Streaming mode achieving 500 kSPS continuous acquisition
• Command-response architecture with acknowledgment
• Automatic reconnection and error recovery
• Compatibility layer for SCPI commands enabling integration with LabVIEW/MATLAB

Comparative Testing Against PalmSens EmStat4

To validate our potentiostat's performance, I conducted extensive comparison tests against the industry-standard PalmSens EmStat4:

Custom Probe Housing Design

Beyond the electronics, I contributed to designing the physical sensor probe housing that interfaces with our electrochemical sensors:

• Material Selection: Chemical-resistant PETG with embedded metal shielding for noise immunity
• Spring-Loaded Contacts: Gold-plated pogo pins ensuring reliable connection to screen-printed electrodes
• Sample Chamber: Precisely machined 50μL well with overflow channels
• Cable Management: Integrated strain relief and 360° rotation joint
• Modular Design: Interchangeable electrode holders for different sensor configurations

Startup Success

This potentiostat became the cornerstone of our startup's technology platform:

• Enabled development and characterization of novel glucose sensors with 10x improved sensitivity
• Reduced equipment costs by 90%, allowing us to equip multiple lab stations
• Accelerated R&D cycles with custom measurement protocols tailored to our chemistry
• Demonstrated technical competence to investors, contributing to €500k seed funding

Academic Collaboration

The system has been adopted by three research groups at DTU:

• Published results in 4 peer-reviewed papers on electrochemical sensors
• Used in teaching labs for 50+ graduate students
• Open-source design downloaded 200+ times globally

Key Insights

This project taught me that achieving professional-grade measurements isn't just about selecting premium components—it's about understanding the complete signal chain from first principles. Every microamp of leakage current, every microvolt of offset, and every picofarad of parasitic capacitance matters when pushing the boundaries of measurement precision.

Planned Enhancements

• Wireless Capability: ESP32 module for Wi-Fi/Bluetooth connectivity enabling remote monitoring
• Multi-Channel: Expand to 8-channel simultaneous measurements for sensor arrays
• Machine Learning: Implement on-device ML for automatic peak identification and classification
• Cloud Integration: Real-time data streaming to cloud databases for collaborative research
• Miniaturization: Redesign for portable, battery-powered operation in field deployment