High-Density Wireless Soil Moisture Sensor
Measuring soil moisture and electrical conductivity (EC) is critical for precision farming. Most grow rooms will only have one or two soil sensors to measure these vital parameters for over 100+ plants in a room. This can often lull growers into a false sense of security when, in reality, several things can go wrong in a drip irrigation network similar to the one shown above. Water leaks are the number one cause of mold and mildew (early stage of mold) formation in indoor grows. Under the right conditions, mold and mildew can spread within 24 hours, and their spores will be carried around the room by the HVAC systems. Water leaks can also lead to the underwatering of plants, which can exacerbate an already bad problem.
Deploying a large number of soil moisture sensors can solve this issue. However, the economic reality of farming precludes growers from adopting this solution. Commercially available soil sensors, particularly the ones marketed as Time Domain Reflectometry (TDR) soil sensors, can run into several hundreds of dollars. However, a family of soil moisture sensors relies on measuring soil capacitance charging/discharging rate as a proxy to measure water content in the soil, which are quite affordable. Traditionally, these sensors suffered from low accuracy in soil with varying EC. The EC change arises from the periodic nutrient injection that is essential for the plants' optimal growth. Edyza has developed an innovative soil sensor that can handle EC variations and is finally changing this cost-benefit equation that has prevented soil sensors' high dense deployment. Below is an example of a small test room with 16 pots containing Edyza’s high-density wireless soil sensor. Now we can provide real-time alerts for leaking water pipes, over or under watering of plants, and so on.
How does Soil Moisture Sensors Work?
The majority of the soil sensors that are commercially available work by inferring the dielectric permittivity of soil by indirect measurement of travel time, impedance, capacitance, resonant frequency, frequency shift, etc. [Blonquist, J. M et al.]. Dielectric permittivity is a physical property of any substance, and it measures its ability to retain an electric charge when presented with an external electrical field. Water molecules have an unusually high dielectric permittivity that allows them to retain charge better than some common soil constituents. The high dielectric permittivity of water arises from two unique properties. First, they are polar in nature, which means a water molecule has a slight negative charge near its oxygen atom compared to the two hydrogen atoms. Second, the tetrahedral formation in which multiple water molecules align themselves helps amplify this polarity [ Kirkwood, J. G. J. Chem. Phys. 1939, 7, 911].
When you compare the dielectric permittivity of major constituents of potting soil as listed in the table to the left, it is clear that water dominates the chart. A simplified linear dielectric mixing model is where the bulk dielectric permittivity of the soil is expressed as a weighted sum of its constituents' dielectric permittivity, as given below.
Where κₐ, κₘ, κₒ, κᵢ and κw are their relative dielectric permittivity, θₐ, θₘ, θₒ, θw and θᵢ are the volumetric concentration(e.g., for air θₐ = Vₐ/(Vₐ+Vₘ+Vₒ+Vᵢ+Vw) of air, minerals, organic materials, water, and ice respectively in soil. Relative permittivity is the ratio of dielectric permittivity of the medium to that of air. From the above equation, if the soil constituents' volumetric concentration is the same, water dominates the soil dielectric calculation. However, if the soil has significant organic growth or experiences mineral concentration changes due to periodic fertigation, it affects permittivity by constituents other than water matter. Now we will present a method that will track changes in volumetric water content to an accuracy better than 1% for indoor drip irrigation systems.
Accounting for mineral accumulation and organic growth
A change in bulk relative permittivity over time is given by
The above equation can be simplified because the volume of dissolved air in the soil hardly changes, i.e., dθₐ/dt = 0, change in mineral concentration given by dθₘ/dt happens during nutrient injection into the soil which for indoor grows is through a process called fertigation (fertilization+ irrigation). Fertigation involves mixing nutrients directly into water and then irrigating the plants with this mixture. Therefore, the change in nutrient volume will be a fraction of the change in water volume, i.e., dθₘ/dt = α dθw/dt where α represents the ratio of nutrient mixing in water, dθₒ/dt represents the organic growth such as plant roots, decaying leaves, microbes, etc. which is practically zero between measuring cycles and finally, ice formation is rare in an indoor grow resulting in dθᵢ/dt = 0. Therefore, the change in bulk dielectric permittivity is given by
Here κₗ=ακₘ+κw is the measured bulk dielectric permittivity of the nutrient water mixture that is used for fertigation. From the above equation, change in κₛₒᵢₗ is proportional to the change in volumetric water concentration in the soil. If κₛₒᵢₗ(0) represent the measured bulk dielectric permittivity of soil at a known time t= 0 and volumetric water content in soil measured at that time represented as θ(0), then the volumetric water content at any time t > 0 is given by
Depending on how often the nutrient regimen changes, the rate of plant growth, accumulation of minerals in the soil, and so on, there is an upper limit to the time duration for which the above equation can be used before re-measuring κₛₒᵢₗ(0) and θ(0). Every other day, this value is to keep the volumetric water content accuracy within 1% from our tests. The best strategy to do this in situ is to have an absolute VWC sensor (e.g., TDR soil sensor) next to our soil moisture sensor for one in every 50 deployments of our sensor. This sensor pair serves as the gold reference. All our other soil sensors will re-calibrate periodically to remove the accumulation of integration error in estimating volumetric water content. The schematic of a drip irrigation network that employs this strategy is shown below.
The above schematic also includes an Edyza sensor monitoring the dielectric permittivity of the nutrient water mixture fed through the drip lines. Now we will explain the inner working of our soil sensor.
Inner workings of Edyza Soil Sensor
Edyza soil moisture sensor uses soil capacitance charge/discharge time to measure water content. Soil capacitance is related to the bulk dielectric permittivity of the soil. Our sensor consists of two cylindrical steel tubes and an insulated pocket, as shown in the figure below. The steel tubes serve as a sensor probe that is in direct contact with the soil medium. The insulated pocket contains all the electronics for moisture sensing. This includes circuitry to generate the excitation wave, a 12-bit Analog to Digital Converter (ADC) to digitize the response signal from the soil, filters to remove noise from sensor probes, etc. and finally, an 8-pin IP-68 rated circular connector that is used to interface with an Edyza’s wireless node that relays the data to our cloud through an Edyza edge gateway.
To measure soil moisture, a high frequency (1 MHz) excitation wave changes its electric field polarity once every 0.5μs is injected into the soil through the steel probes. Due to their polar nature, the water molecules try to align themselves to this rapidly changing electric field, as shown below.
However, the excitation frequency was chosen such that the water molecules will not be able to keep up with the fast-changing electric field. Consequently, this lag in water molecule response introduces a delayed/damped response picked up by our sensor probes. The plot below shows the excitation square wave as dotted lines, while the solid line is the soil response due to water molecules lagging behind the excitation wave.
As the concentration of water molecule increases in the soil, the excitation signals are damped further. Therefore, a straightforward way to measure the extent of damping and consequently estimate the water molecule concentration is to measure the peak response voltage for a single cycle of excitation wave called the peak cycle voltage (Vc). Peak cycle voltage can be extracted from the saw-tooth soil response signal using a peak detector circuit formed from diode D₁, resistor R₄ and capacitor C₁ in the simplified circuit of the Edyza moisture sensor shown below. In this circuit, Rₛ and Cₛ represent the soil's electrical model and are the bulk resistance and capacitance of the soil.
From the above circuit, the peak cycle voltage can be derived as
Where R=R₁||Rₛ is the net resistance due to the soil resistance Rₛ and a resistor R₁ that we introduced to limit the current flow between the steel probes through the soil, T is the 1μs period of the excitation square wave, Cₛ is the soil capacitance, and Vᵣ is the voltage between the steel probe when a known direct current is injected into the soil through the steel probes.
How to correct for soil EC in Capacitance Probes
Typical dielectric based soil moisture sensors experience a higher error in moisture estimation when the soil has salt. This increased error is attributed to the higher electrical mobility of salt ions in the soil. Salt ions act as a short circuit between the sensor probes and manifest as a reduction in bulk resistance Rₛ of the soil. The faster flow of electrical charge between the probes short circuits the electric field charging/discharging the water molecules. This results in water molecules experiencing reduced electrical field excitation in soil with higher salt concentration. Consequently, if not compensated for the EC, such a soil will measure lower water (VWC) than reality.
The straightforward solution to this problem is to estimate the bulk resistance Rₛ of the soil and then use that to derive the soil capacitance Cₛ from the peak voltage equation listed above. However, when Rₛ drops, the soil signal response measured at the sensor probes is dominated by the signal from ion mobility. The signal from the water molecule dielectric charging/discharging is several orders lower than ion mobility. This presents problems to our 12-bit ADC that is tasked with translating the analog peak cycle voltage to a fixed size digital data. E.g., when EC=0, the change in Vc between a dry soil to a fully saturated soil is around 30mV. The smallest signal that our 12-bit ADC can measure is 0.3mV. This gives us 100 steps between the two extremes of soil saturation. However, when EC=1.0 dS/m, the change in Vc between fully saturated and dry soil is only 1.8mV, which results in only six moisture measurement steps between the moisture extremes. With increased EC, our ability to accurately measure soil moisture is severely limited.
Edyza solves this resolution problem using oversampling and averaging. After estimating the soil resistance, we derive the number of over-sampled data needed to result in the same granularity in capacitance calculation when the EC=0. We also employ of sinc digital filter to shape the signal noise to be away from where the signal of interest lies. Luckily, the soil moisture is a very slow-moving value, and hence the complexity of the digital sensor to achieve the desired VWC resolution is minimal.
Update: We have a newer version of soil sensor EZ-SSP101, which has a larger measuring volume compared to the one presented here. Besides, the newer version works with all types of mediums, including Rockwool and Coco Fiber. The details on this newer sensor are available here [EZ-SSP101]
