Back to Project Kymarion
🌊 Project Kymarion

Sensor Design V1

Design doc May 2026 8 min read

Sensor Architecture and Analytical Payload

To overcome the limitations of open-water optical sensing—where ambient sunlight, wave action, and depth variations severely compromise data integrity—the Autonomous Surface Vehicle (ASV) is equipped with a highly controlled internal analytical payload. Rather than exposing sensors directly to the open water, the ASV pumps water through a custom, light-tight optical flow cell to measure particle fluorescence and turbidity under identical, repeatable conditions.

Plumbing and Fluid Sequence

The fluid path is designed to ensure valid, measured sample volumes while protecting the sensors from macroscopic debris. The sequence operates as follows:

Optical Flow-Cell Geometry

The flow cell is the core scientific instrument of the ASV. It is a custom 3D-printed chamber (made of matte-black PETG or ABS) designed to eliminate ambient light and standardize the physical geometry of every reading.

Planned Sensors

Category Component Role Est. Cost Interface
Microcontroller & Data ESP32 Microcontroller 32-bit system brain; handles multitasking, interrupt pins for flow counting, I2C sensor polling, and data routing. ~$5–$10 Main Controller
Microcontroller & Data U-blox NEO-M8N GPS Provides fast, high-accuracy satellite locks to map precise geospatial coordinates of sensor readings. ~$15–$20 UART / I2C
Microcontroller & Data ADS1115 16-bit ADC Bypasses the ESP32's noisy internal ADC to provide ultra-high-resolution, linear analog readings from the turbidity sensor. ~$5 I2C
Microcontroller & Data MicroSD Card Module + 16GB Card Provides a redundant, localized storage backup for all CSV log files to prevent data loss if a Wi-Fi connection drops. ~$10 SPI
Optical Detection TSL2591 HDR Light Sensor Primary fluorescence reader; I2C digital interface, highly sensitive with separate IR/full-spectrum channels. ~$7 I2C
Optical Detection TCS34725 RGB Color Sensor Secondary fluorescence reader; tracks RGB channel ratios to confirm the target spectral band. ~$8 I2C
Optical Control 470nm Blue LED + Driver High-stability excitation source; constant-current regulation ensures reliable calibration. ~$15 GPIO / MOSFET
Optical Control Emission Filter (Longpass Gel) Blocks direct 470nm excitation light so sensors only read the plastic's emitted fluorescence. ~$10 Physical Optical
Baseline / Context DFRobot Gravity SEN0189 Measures total light scattering (non-plastic baseline control) through the ADS1115 ADC. ~$10 Analog (to ADC)
Baseline / Context DS18B20 Temp Probe Kit Logs ambient water temperature via a 1-Wire bus. ~$7.50 1-Wire
Fluidics YF-S201 Flow Sensor Uses a Hall-effect turbine to output electrical pulses, translating to precise sample volume. ~$10 Pulse / Interrupt
Fluidics 12V Diaphragm Pump Draws a consistent water sample into the flow cell. ~$20 Power / Relay
Mechanical & Actuation Rows left blank in the source document (to be specified)
The source document includes two images here — illustrations of the sensor payload and flow-cell layout — not mirrored on this page.

1. Internal Flow-Cell Fluidics and Hardware Architecture

To overcome the limitations of open-water optical sensing—where ambient sunlight, wave action, and depth variations severely compromise data integrity—the Autonomous Surface Vehicle (ASV) utilizes a highly controlled internal analytical payload.

[Intake & Pre-Screen] ➔ [12V Pump] ➔ [Flow Sensor] ➔ [Optical Flow Cell] ➔ [50μm Filter]

Water is drawn through a 1mm–2mm coarse intake screen by a 12V pump and measured volumetrically by an inline YF-S201 Hall-effect flow sensor to establish precise quantitative metrics (particles per liter). The fluid then enters a custom, light-tight, matte-black 3D-printed optical flow cell before passing through a terminal 50-micron mesh filter cartridge, which captures a physical ground-truth sample for laboratory microscope validation.

The flow cell utilizes a classic 90-degree fluorometer geometry to isolate target signals:

2. Nile Red Fluorometric Chemistry and Seawater Compatibility

The analytical sequence relies on the lipophilic, hydrophobic fluorescent dye Nile Red, which undergoes solvatochromism—a property where a molecule's photoluminescence alters based on the polarity of its immediate environment.

3. Signal Conditioning and False-Positive Mitigation Matrix

Because Nile Red bonds to any hydrophobic substance, natural marine lipids (e.g., algae, zooplankton, fish eggs) will absorb the dye and fluoresce, posing a distinct risk for false-positive readings. The ASV circumvents this limitation through integrated hardware signal conditioning and multi-sensor data cross-referencing.

Hardware Signal Conditioning

The raw analog signal from a downstream DFRobot Gravity SEN0189 turbidity sensor is routed through an external ADS1115 16-bit analog-to-digital converter (ADC) connected via I2C. This bypasses the noisy, non-linear native ADC of the ESP32 microcontroller, upgrading the system's baseline particle resolution to a highly precise, linear depth capable of detecting minute changes in suspended solids.

Mathematical & Data Filtering Matrix

Mitigation Layer Technical Mechanism Analytical Target
Turbidity-to-Fluorescence Ratio Compares total light scattering (mass) against emission intensity. High turbidity paired with moderate fluorescence indicates organic mud or sediment. Low turbidity paired with a sharp fluorescence spike flags a sparse, highly reflective microplastic fragment. Filters out general organic/inorganic suspended sediment.
Spectral Fingerprinting Tracks the Red-to-Green (R/G) channel ratio via the TCS34725 sensor. Nile Red bound to natural biological lipids shifts toward a yellow-gold wavelength, while bonds with synthetic polymers emit a distinct red-orange wavelength. Differentiates biological lipids (algae/plankton) from synthetic plastics.
Physical Ground-Truth Backup Captures the exact water sample scanned by the sensors on a terminal 50-micron physical mesh. Allows for manual laboratory microscope verification to validate digital sensor calculations.

Systemic Constraints and Analytical Limitations

The primary challenges encountering the Autonomous Surface Vehicle (ASV) center on the sharp engineering trade-off between fluidic sample volume and data integrity. Because microplastics are mathematically sparse in natural aquatic environments, dropping the sample volume to a highly manageable micro-scale (such as 250 mL) creates an extreme risk of false-negative readings, as the statistical probability of capturing an individual floating particle within such a small volume is low. Furthermore, at standard pumping rates, a micro-sample reduces the data-acquisition window to less than ten seconds per waypoint, leaving an exceptionally narrow frame of time for the digital sensors to register transient fluorescence spikes before the volume is exhausted.

Compounding this volume constraint is the persistent issue of biological matrix interference within natural water bodies. Because Nile Red is an indiscriminately lipophilic dye, it aggressively binds to natural marine lipids found in algae, zooplankton, and organic detritus. In highly productive environments, these biological materials fluoresce under a 470nm light source right alongside target synthetic polymers, generating severe optical background noise that can trigger false-positive spikes. The system must therefore rely entirely on downstream multi-sensor data processing—cross-referencing total light scattering against specific spectral color ratios—to mathematically differentiate natural organic matter from actual synthetic microplastics.

Beyond hardware and biological limitations, the vehicle's operational architecture is strictly bounded by environmental law. The direct discharge of Nile Red dye and its volatile carrier solvents (such as acetone or ethanol) into aquatic ecosystems is illegal under environmental protection regulations like the Clean Water Act. Unregulated chemical discharge introduces toxic hazards to local habitats and violates stringent field-testing safety protocols. Consequently, the ASV cannot utilize a simple open-loop, in-situ staining mechanism; the system is forced to rely on an ex-situ laboratory staining protocol, where the ASV physically harvests the samples in the field while all chemical handling is strictly isolated to a controlled lab environment.