Our innovative SoC design for precision agriculture revolutionizes field management by deploying a robust mesh network of sensor-based devices, capable of detailed monitoring and swift response to variations in soil health, erosion, drought, and pest activities. This system not only ensures reliability through its mesh architecture—eliminating single points of failure—but also incorporates diverse sensors for comprehensive data acquisition. It's engineered for energy efficiency to sustain operation throughout an entire crop season, significantly optimizing resource use and reducing waste. By providing continuous and granular data directly from the field, our SoC solution surpasses traditional remote sensing methods like drone mapping, offering enhanced decision-making capabilities and operational cost savings for sustainable agriculture.

GitHub Project: https://github.com/users/lewisMW/projects/1/views/1

GitHub Repository: https://github.com/lewisMW/agriculture-SoC

Project Management

We are a 11-person team. Our main project management tool is GitHub Projects, which is public and viewable here: https://github.com/users/lewisMW/projects/1/views/1.

We have distributed the work into analog and digital teams:

Digital Team Update

Currently the digital team is learning about the ARM M0 processor and the AHB and APB buses and how to create a custom wrapper to interface between the bus and our IP - namely the ADC that the analog team are designing. We have also got Verilator working as a simulator for the DesignStart processor and would be interested to get it working with NanoSoC, (if that's possible). We have 4 members are each currently exploring different ARM IP from the CMSDK.

• Sandin: SRAM IP
• Ben: Slave Mux IP
• Patrick: GPIO Peripheral IP
• Lewis: Timer IP

Aside from the above, Neil is currently looking into FIFO IP to buffer sensor data after being sampled by the ADC and Ted (Kezhi) is investigating a custom wrapper file for the ADC, drawing from https://soclabs.org/project/adc-integration-nanosoc.

Analog Team Update

The analog team have created circuits for different sensor conditioning modules and are currently using QSpice to simulate them with an ADC, which is a free mixed-signal simulator. The 5 members on this team have split up to focus on 2 different sensor types:

• Jaime & Adrian: pH Sensor
• Shin, Erick & Hee: Soil Moisture and Temperature

Currently, both teams have started creating circuits in QSpice to simulate.

SoC Architecture

The current system architecture is as follows:

Figure 1: The current system architecture (excluding integration with the AFE).

We have started with a simple system with limited peripherals, restricting our design to the ARM CMSDK IP kit. The motivation behind this is as a learning exercise, such that each team member becomes familiar with the ARM AHB-Lite bus, before moving onto more sophisticated architectures. We plan to migrate to NanoSoC once we get a basic system working (in functional simulation) in Verilator.

Peripherals

• ADC (Analog-to-Digital Converter):Necessary for converting the analog signals from various sensors (such as dielectric soil moisture sensors and electrochemical sensors) into digital data that the processor can handle.
• Real-Time Clock (RTC): Helps in time-stamping the sensor data. depends on which chip we used. (internal or external), Cortex-M0 might not have one inside.
• Timers: We might need Timers as well, it is crucial for managing sensor data collection intervals and controlling the duty cycling of the system, which helps in reducing power consumption.
• Power Management ICs (PMICs): To efficiently manage power distribution and battery charging, especially vital in deploying power-saving techniques like duty cycling.
• GPS Module: For precise geotagging of data, real time positioning for precise watering and management.
• I2C: To consider the cost and low power, we might choose I2C for short connect (interface with RTC, a set of sensors connect over short distances); But we might also need UART for long distances communication.
• Memory Modules : SPI for an SD card - non-volatile memory for data logging.
• Communication (Network) Modules: Depending on the chosen communication protocol (e.g., Zigbee, LoRa, BLE, ZWave, ANT+), for transmitting data across the mesh network.

Memory Mapping

Proposed Memory Map

We connect the peripherals of the SoC to the bus via a bus interface (AHP and APB). Hence, we define the following address map to interface with each peripheral. The ADC of the sensors are fed into a FIFO. Therefore, this FIFO region is responsible for the entire analog frontend of the system.

start-address  end-address  notes

0x40000000   0x40000FFF   Timer

0x40001000   0x40001FFF Real-Time Clock

0x40002000   0x40002FFF PMIC

0x40003000   0x40003FFF GPS

0x40004000   0x40004FFF I2C

0x40005000   0x40005FFF SPI

0x40006000   0x40006FFF SD Card

0x40007000   0x40007FFF LoRa

0x40008000   0x40008FFF   FIFO Region

Current Memory Map

Currently, we have the following memory map implemented:

You can see that we have started with a very minimal memory map. This was inspired by the HellSoC entry (GitHub) from 2023, as we felt starting from the bottom up would be highly beneficial for our education.

Functional Simulation

We have set up Verilator on our current codebase in the GitHub.

• See https://github.com/lewisMW/agriculture-SoC/tree/main.

This requires Verilator 5.026, make and gcc and the ARM DesignStart Eval kit. To simulate the system, run make in the agriculture_soc_main directory. To build software to flash the SRAM with, run make in an appropriate software project directory in the software directory.

Figure 2: Waveform produced by Verilator for the CPU at startup, viewed in the Surfer waveform viewer.

Analog Front End and Sensors

The architecture of the analog front end is summarised in the following diagram:

Figure 3: The Analog Front End connected to the AHB-Lite Bus (via APB).

In the above, we have categorised 4 sensing types and are implementing separate signal conditioning / sensor driving modules:

• Moisture Sensor (TDR-based)
• Temperature Sensor
• Soil PH
• Soil Composition

Moisture Sensor (TDR-based)

The standard used in the industry is based on time domain reflectometry (TDR), thus the technology we're implementing is based on TDR. The soil dielectric constant is strongly dependent on the volumetric water content and is almost independent of soil density, texture, salt content, and even temperature [1]. This approach utilized a metal probe of length L in which an electromagnetic wave propagates with a speed that is dependant to the soil moisture content. In [2] an equation for the propagation speed of the wave is presented:

Where:

ε is the complex dielectric permittivity

σ_b is the bulk electrical conductivity

μ is the magnetic permeability of the soil

⍵ is the angular frequency

ε_o is the dielectric permittivity of the free space

ε_b is the bulk dielectric permittivity

If we consider a relatively low electrical conductivity which is the case for homogeneous soil and a high angular frequency, the equation can be reduced to:

With this expression, we're able to calculate the bulk dielectric permittivity from the time it takes to the electromagnetic wave travel the probe and return, hence 2L.

The proposed diagram for measuring the time is the following

Figure 4: Modules required to implement the AFE for the moisture sensor.

Bibliography

[1] H. Yin, Y. Cao, B. Marelli, X. Zeng, A. J. Mason, C. Cao, Soil Sensors and Plant Wearables for Smart and Precision Agriculture. Adv. Mater. 2021, 33, 2007764. https://doi.org/10.1002/adma.202007764

[2] W. Skierucha, A. Wilczek, A. Szyplowska, C. Swawinski  & K. Lamorski. A TDR-Based Soil Moisture Monitoring System with Simultaneous Measurement of Soil Temperature and Electrical Conductivity. Sensors. Poland: Institute of Agrophysics. 2012. https://doi.org/10.3390/s121013545

Temperature Sensor

Temperature sensors can be broadly divided into three categories based on their operating principles:

1. Temperature Dependent Resistor Sensor
   -    Thermistor: A sensor whose resistance changes with temperature, typically available in NTC (Negative Temperature Coefficient) or PTC (Positive Temperature Coefficient) forms.
   -    RTD (Resistance Temperature Detector): Made using pure metals (usually platinum), these sensors exhibit an increase in resistance as the temperature rises.
2. PN-Junction Temperature Sensor
   Utilizes the PN junction of semiconductors, detecting temperature changes through voltage variations at the junction. Silicon diode temperature sensors are commonly included in this category.
3. Thermoelectric Sensor
   Thermocouple: Measures temperature by detecting the voltage difference at a junction connecting two different metals. If the junction temperatures vary, different voltages are generated, which are used for measurement.

The LM61 is a temperature sensor that utilizes a PN-Junction to detect temperature changes based on variations in semiconductor current with temperature. It can measure temperatures from -55°C to 150°C, and uses a low supply voltage of 3V to measure temperatures from -30°C to 100°C. At temperatures of 0°C and 100°C, the output voltages are 600 mV and 1.6 V, respectively, increasing by 10 mV for every 1°C rise in temperature. The temperature accuracy is within ±2°C.

Figure 5: Functional block diagram of the LM61.

Figure 5 presents the block diagram of the LM61, which is composed of seven transistors and several resistors, divided into three main sections: temperature measurement, signal modulation and buffering, and output. Transistors Q1 and Q2 form a pair that acts as a differential amplifier. Q2 is biased by a voltage divider, resulting in an amplification 10 times that of Q1. This difference generates a voltage called V_PTAT (Voltage Proportional to Absolute Temperature), which is proportional to the absolute temperature.

Figure 6:  Block diagram of conditioning circuit.

Figure 6 is a conditioning circuit created using an example provided by Texas Instruments. The output from the temperature sensor requires minimal modification and consists of a combination of amplification and filtering.

Soil pH Sensor

Three main categories of sensors have been considered for in situ measurements of soil pH. For the purposes of this competition, which focuses on SoC design, the potentiometric sensor type has been chosen as the focus for our design. This is because it is the most widely used in industry for in situ measurements, and therefore the conditioning circuit we design for this type of sensor will be able to interface with a wide range of products which already exist on the market. It should be noted that the conductometric sensor has been ruled out due to concerns about ion selectivity, but the ion-selective FET (ISFET) sensors may be explored further in later design stages. 
A potentiometric pH sensor uses an ion-selective electrode (ISE) pair and measures the potential difference between the reference and the sensing electrodes. The ISE is placed in the soil, and an ion-selective membrane allows H+ ions to pass into and out of the sensing electrolyte. This allows the potential at the sensing electrode to change compared to the reference, depending on the pH of the surrounding soil. The pH of the soil can be inferred from the potential between the electrodes, since voltage can be related to pH via the Nernst equation.  A Texas Instruments product has been used as inspiration for the conditioning circuit of the ion-selective electrode, because it addresses key challenges with ISE measurements. These include the dependence of pH/mV sensitivity on temperature, the fact that the ISE produces a bipolar voltage, and the high resistance of the ion-selective membrane. As shown in the diagram, these issues can be respectively combatted by the inclusion of automatic temperature compensation, high input impedance amplification (e.g. voltage follower op-amp configuration) and an offset voltage applied to the reference. A temperature compensation system which incorporates the temperature sensor on our SoC will be designed. 
This basic conditioning system has been modelled in Qspice, along with a voltage source to represent the sensor. The model will be refined to include temperature variations, input impedance, temperature compensation, other forms of noise and interfaces with digital components (e.g. for the reference voltage and ADC).  A deeper understanding of how each part of the system can be changed to affect key performance metrics (including sensitivity, resolution and accuracy) is needed before the design can be the AFE design can be refined and changed compared to the TI example. Table 1 contains a list of specs taken from existing systems to be used as benchmarks for the performance of our system.

Figure 7: example ion-selective electrode and conditioning circuit, taken from Texas Instruments LMP91200.

Figure 8: Preliminary Qspice model based on Texas Instruments Application Report available here.

Table 1: list of typical performance specifications from a range of pH sensor types (including conductimetric, potentiometric and ISFET).

Bibliography

1. Yin, H., et al. "Soil Sensors and Plant Wearables for Smart and Precision Agriculture," in Advanced Materials, vol. 33, no. 20, pp. 2007764, 2021.
2. J Bausells, undefined., et al. "Ion-sensitive field-effect transistors fabricated in a commercial CMOS technology," in Sensors and Actuators B: Chemical, vol. 57, no. 1, pp. 56-62, 1999.
3. Matthew McCole, undefined., et al. "A low-cost portable system for on-site detection of soil pH and potassium levels using 3D printed sensors," in Results in Engineering, vol. 20, pp. 101564, 2023.