A low-cost bench-top research device for turbidity measurement by radially distributed illumination intensity sensing at multiple wavelengths

Presented here is a new bench-top research device for the measurement of the optical tur- bidity of natural sediment-laden water samples. This prototype device employs 18 unique angular measurement positions and a variety of user-selectable LED light sources. The motivation for this project was the need to generate more parameter-rich data sets per-taining to the light-scattering properties of natural sediment suspensions, and to address the issues raised by Kitchener et al. (2017) concerning the inconsistent calibration method-ologies currently employed to quantify suspended sediment concentration (SSC) by optical turbidity measurement. The mechanical design comprises re-purposed waste plastic materials and 3D-printed parts. The active light-source control and monitoring hardware and ﬁrmware executes on the open-source Arduino embedded microcontroller platform. The modular light sensors plug into any of the angular measurement positions, providing a 0–5 V nominal output signal, which is readable by the user’s choice of data-acquisition system. The device will facilitate the highly detailed characterization of suspended sediment samples, providing 18 voltage output channels for analysis by the user. The precise calibration of the light sensors is by the use of neutral density (ND) ﬁlters in conjunction with light-source electrical current measurements, providing light-source intensity values as required. The empirical data provided by existing turbidity meters are acquired using incommensurate methodologies, and therefore they are not cross-comparable. A new methodology, described here, facilitates the cross-comparability of turbidity measurements.


Hardware in context
Commercial instruments available for the purpose of optical turbidity measurement vary widely in design, operating principle, and cost. Turbidity instruments are now ubiquitous in the water supply sector, and operate widely throughout the physical and engineering sciences (oceanography, fluvial and glacial sciences, civil engineering, chemical engineering etc.). These instruments are designed to operate either in situ (in a river, ocean, harbor etc.), or in a laboratory setting (in-situ in a flume, of off-line on a bench-top). They tend to be expensive, and they are inconsistent with regard to the measurement methodology (e.g. instrument geometry, wavelength of light source). Hence it is possible to obtain completely different and non-comparable measurements of the same water sample measured by two or more different instruments. The ''turbidity units" reported by these devices are inconsistent and do not conform to the SI system of measurement, for example FNU (Formazin Nephelometric Units), NTU (Nephelometric Turbidity Units), FAU (Formazin Attenuation Units) [1].
Open-source and other public-domain instruments for turbidity measurement are available [2][3][4][5][6], but these devices only provide one or two measurement angles, and are limited to a single light source. In order to fully explore the limitations and capabilities of turbidity measurement it is necessary to investigate the physics of light scattering and absorption by suspended sediment, at a comprehensive range of scattering angles and wavelengths of light [1].
Designed for one specific application, these existent public-domain turbidity meters are not versatile research instruments. The list of devices presented in Table 1 is not exhaustive, but it is representative of the open-source state of the art.
Commercial instruments exist that not only measure turbidity, but also provide an estimate of the particle size distribution (PSD) of a suspension in water. In the case of suspended sediment, the determination of the PSD is by small-angle forward scattering measurements by LASER diffraction. These measurements require expensive (1000-20,000 GBP) LASER diffraction devices, which can work in situ or on a bench top in a laboratory.
There is a need for a research device capable of measuring the optical properties of a sediment-laden water sample and presenting the data in an unbiased format, using physically acceptable SI units (notionally mW/sr). Kitchener et al. [1] have shown that the pre-existing units of turbidity measurement (FNU, NTU, FAU etc.) are not physically valid since they are based on the mass-concentrations of arbitrary polymer suspensions, and not on the intrinsic optical properties of the particles in suspension. It was suggested [1] that research devices capable of producing multi-parameter, multi-wavelength measurements of the bulk optical properties of suspended sediment be developed, which will begin to address the issues inherent in the reporting of turbidity data and its relationship to suspended sediment concentration (SSC). TARDIIS is a first attempt at producing such an instrument. Designed for use with off-line water samples, TARDIIS operates on a laboratory bench-top. It will allow the user to observe the sediment-settling process from multiple scattering angles simultaneously, using any wavelength LED available in a 5 mm package. The user is free to interpret the measurement data in their own way, and to gain their own insights into parameters such as the PSD. The user is also encouraged to customize and refine the design to accommodate their own particular measurement requirements, and to challenge the status quo in terms of turbidity measurement methodology. Table 1 Public domain turbidity instrument designs.
With a knowledge of the LED light source characteristics and a direct measurement of the LED current, TARDIIS makes it possible to report the measured light intensity in terms of appropriate SI units (mW/sr). This approach to the reporting of turbidity measurements will facilitate a more meaningful way to cross-compare measurements made on different samples by other similar devices.

Hardware description -TARDIIS
TARDIIS (Turbidity Assessment by Radially Distributed Illumination Intensity Sensing) is a prototype bench-top research device specifically for the examination of the light scattered by suspended sediment at multiple angles during the process of settling in a column of water, over a variable time. It is not, however, restricted to this single use-case.
TARDIIS offers: Turbidity measurement at 18 distinct scattering angles (with 1 LED light-source location and 35 potential sensor locations). Measurement of absorption and scattering in any water sample. User selectable light wavelengths. ND filter calibration. Data reporting in SI units (mW/sr), rather than using incommensurate turbidity units (NTU etc.). [1] Generation of a detailed ''optical sedigraph" at multiple wavelengths.

Design files
The The 3D printed parts were originally created using OpenSCAD, which can be used to open and modify the designs. The STL files were generated by OpenSCAD, and can be directly uploaded to a 3D printer for printing. These designs have also been re-drawn in SolidWorks in order to produce 2D drawings (as PDF files (3.5)) of the parts, as this function is not available in OpenSCAD. The PDF design files can be handed directly to a machinist for production of the parts. Alternatively, the SolidWorks files will allow the user to make modifications to the designs using the SolidWorks CAD package.

PCB fabrication
The PCBs used in this project were fabricated by Ragworm [7]. The price per PCB depends on its size, and the number ordered. We purchased two of the LED Control & Monitoring Shield PCBs for 31.83 GBP (one was a spare). We also purchased 50 of the Photodiode Amplifier PCBs for 39.06 GBP.

Sample cell, sensor collar and mounting system
The sample cell (#2) consists of an extruded acrylic tube (#2a) 491 mm in length, with an internal diameter of 93 mm, and an external diameter of 100 mm. An acrylic disc (#2b) is glued (#2c) to one end of the tube to close it and form the watertight bottom of the sample cell. The sample cell then sits in the recessed base (#1) of the mounting system, surrounded by the three support columns (#7). The height of the nuts (#5 -see Fig. 2 for positions) on the support columns can be adjusted to alter the height of the sensor collar (#3) containing the sensor ring (#4) above the base plate, which slots onto the support columns and encapsulates the circumference of the sample cell (Figs. 1, 2). This arrangement allows turbidity measurements to be performed at any height in the water column containing the sediment sample. The amplifier housing (#8) may contain either a photodiode amplifier PCB, or an LED.

Sample cell
The sample cell is made from an extruded acrylic tube (#2a), and has an acrylic disc glued (#2b) to one end to form the bottom of the cell. We found that the acrylic disc had a slightly larger diameter than the acrylic tube. After gluing the disc onto the tube, we had to perform some manual filing of the disc circumference to bring it flush with the tube diameter. This was necessary to allow the whole tube to fit easily through the sensor collar, thus allowing clearance for its lifting in and out of the instrument assembly without snagging on the sensor ring.

Sensor ring
The sensor ring (Fig. 3) fits over the outer diameter of the sample cell to facilitate the connection of LED light modules and sensor modules at any of the 36 receptacle positions around the circumference of the ring, at 10°intervals. The receptacle positions each consist of an 8.7 mm diameter hole in the outer surface of the ring, bored to a depth of 10 mm. This hole allows a 3D-printed sensor housing (which could contain either a photodiode amplifier or an LED) to be slotted in and held in place by an interference fit. A second concentric hole of 3 mm diameter then penetrates through the remaining thickness of the sensor ring wall and breaks through to the inner surface of the sensor ring, to form a collimated light path. This collimator hole helps to eliminate stray light when a sensor module is connected, and provides a consistent angle of maximum beam divergence for any given type of LED utilized as a light source. The sensor ring was machined from waste nylon plastic (Fig. 4), and the sensor collar into which it fits was made from waste PVC, although both parts could be made from the same stock. Alternative materials could be used to create these parts, e.g. wood or MDF, or any suitably rigid and machineable material.

Sensor collar
The sensor collar (Fig. 5) has lugs with M8 clearance holes that allow it to slot onto the support columns of the main assembly, thus allowing the sensor ring to encircle the sample cell when contained by the collar (Fig. 6).

Base
The base (Fig. 7) has lugs with M8 threaded holes that allow the support columns to be screwed in and then locked in place with M8 nuts and washers. A 5 mm deep recess allows for sample cell lateral retention at the bottom end.

Sensor and LED modules
The small form-factor of the sensor/LED housing measuring only 27.5 Â 10.5 Â 13.5 mm, as shown in Fig. 8. The sensor module consists of the photodiode amplifier board and a plastic housing for it, which was designed using OpenSCAD [8] and printed on a Makerbot 3D printer from PLA plastic (Fig. 9). As the photodiode amplifier utilises an SFH213 photodiode in a standard 5 mm package, the same plastic module can be used to house an LED with the same standard 5 mm package.

Photodiode calibrator and ND filters
The principle of operation of the optical calibrator is the measurement of the attenuation of a light source by the placing of an optical neutral density filter (ND filter) between the light source (LED module) and the detector (sensor module), as in Fig. 10. It is worth noting that the filter number ND is synonymous with absorbance, A (see Kitchener et al., p.628 for an explanation of absorbance [1]).   The percentage of light transmitted by an ND filter is T: For example, a 3.0 ND filter will transmit 0.1% of the incident light, whereas a 0.3 ND filter will transmit 50.12% of the incident light. The transmission formula in terms of the ND filter number is: The photograph in Fig. 11 demonstrates the logarithmic nature of the relationship between percent transmission and ND filter number.

Photodiode calibrator design
The photodiode calibrator consists of two different parts, designed to be printed separately on a filament type 3D printer. Part 1 (Fig. 12) is a receptacle for a photodiode/LED module. Two of these parts form the ends of the calibrator. Sandwiched between them is part 2 (Fig. 13). This section forms the housing for an ND filter.   The optical calibrator model was produced using OpenSCAD [8], and printed by a Makerbot 3D printer from PLA plastic filament (Fig. 14). The orientation of part 1 (2 off) and part 2 (1 off) as shown in Fig. 14(a) is flat. This is to optimize the dimensional precision in the 3D printing of the holes and circular sections.

Active LED control and current monitoring system
In order to perform precise measurements of optical attenuation using light detectors to measure the transmitted light intensity through a sample, it is first necessary to know precisely the intensity of the LED light source, since the optical transmittance T through a sample is equal to the ratio of the measured light intensity I to the incident light intensity I 0 . An instantaneous value for I 0 can be derived from a measurement of the current passing through the LED. The light intensity can then be determined from the luminous intensity vs. forward current graph on the appropriate LED data sheet. This instantaneous current value is required as the current passing through the LED will drift over time due to thermal effects in the control  circuit. A simple circuit based around a ubiquitous LM317 voltage regulator with a digitally controllable feedback section has been designed (Fig. 6) to allow the user to select a desirable LED current, which the circuit will then attempt to establish and maintain over time. Digital control of the active elements in the feedback section is provided by an Arduino UNO microcontroller board running a simple active control algorithm. These active elements are two digital potentiometers (DIGPOTs) and an analogue to digital converter (ADC). The user can communicate with the Arduino UNO via USB to send commands, e.g. the desired current (set-point). The instantaneous LED current measurement can also be polled repeatedly by the user interface, for use in light transmittance calculations.

Conceptual circuit diagram
The diagram in Fig. 15 illustrates the architecture of the active LED control and current monitoring system, which comprises a control algorithm (Fig. 16) and an electronic circuit (Fig. 17). This circuit is implemented as a ''shield" PCB for the Arduino UNO (Fig. 18), where ''shield" refers to a secondary ''daughter" PCB that can be stacked on top of the ''parent" Arduino board and electrically connected by means of SIL headers.
Control of the feedback to the voltage regulator is achieved by the adjustment of two feedback resistors connected in parallel. These feedback resistors are digital potentiometers designated W0 and W1. Each potentiometer wiper can be digitally set to one of 256 discreet values, where the value 0 notionally represents zero resistance, and the value 255 indicates maximum resistance (5 KX in this case). This scale provides a resistance resolution of 5000/256 = 19.53 X per step. By fixing the value of one potentiometer (W0, or ''fine adjust") and then searching for the optimum value of the second potentiometer (W1, or ''coarse adjust"), it is possible to achieve the desired set-point current. When the LED current is close to the set point, the control algorithm (Fig. 16) maintains the current level and compensates for drift continuously (for example see Fig. 20). This level of control can be difficult at low currents, i.e. below 1 mA, due to potentiometer resolution limits, or the presence of protective Zener diodes in the LED package. In addition, the forward-voltage to current characteristics of the LED can have a steep gradient in this region (especially so where infrared LEDs are concerned). The minimum load threshold of the voltage regulator is 3.5 mA, and so operating at load currents below this level is outside the specification of the device, adding to the unpredictability of current control in this low current range (Fig. 20), an effect that is further confounded by the increased dominance of voltage measurement noise in this region. However, for most turbidity measurements this effect will not be problematic, as a higher LED intensity would be more desirable for suspended sediment experiments. It will present some small challenges with respect to sensor calibration, however, as low light intensities are useful for the calibration of the highgain sensors utilised in this application. An obvious solution to the problem of generating a calibration curve at low light levels is to ramp the LED current from a high intensity level to a low intensity level with a high-value ND filter between the LED and the photodiode. We have found that a 3.0 ND filter is ideal for determining the full measurement range of the photodiode amplifier (4.0 ND for k = 940 nm) whilst keeping the LED at a well-controlled intensity.

Circuit schematic
The components used for the LED control and monitoring shield were chosen primarily for ease of assembly, i.e. no surface mount devices (SMDs), only dual-in-line (DIP) package integrated circuits (ICs) and through-hole passive components  were chosen. The voltage regulator is an adjustable LM317P [9] in a TO-220FP package, which can be operated from a DC differential input voltage of up to 40 V. The suggested input voltage range for this application is +9 V to +18 V with respect to a common ground, as these voltages can notionally be provided by one or two 9 V batteries connected in series (if portability is required). However, for development and testing purposes a bench power supply was used to provide a nominal 12 V input to the voltage regulator via a Molex connector (J1).
The circuit board has a second Molex connector for plugging a remote LED into the output of the voltage regulator (Fig. 18,  J2). This connector puts the LED in series with stabilisation resistor R2, and shunt resistor R3. The shunt resistor provides a ground-referenced measurement to the ADC [10]. This voltage measurement represents the feedback signal to the control algorithm on the Arduino UNO (sent via SPI bus), which can then adjust the values of the two DIGPOTs on board the MCP4261 chip [11] -also via the SPI bus -thus forming direct feedback to the voltage regulator. The MAX6225ACPA+ [12] is a precision 2.5 V voltage reference and is required by the ADC for correct operation. Fig. 17 shows the net names of all the connections, including to the headers on the Arduino UNO (ARD1). The schematic was created using DipTrace [13], for which a 300-pin limited freeware version is available for non-commercial use.

Arduino UNO shield PCB
The circuit described by Fig. 17 is realized as a PCB ''shield" for the Arduino Uno microcontroller board (Fig. 18). This shield PCB is designed to have the same footprint as the Arduino Uno PCB, and stacks on top of the Arduino by means of SIL headers (2.54 mm pitch), as in Fig. 19.

Photodiode amplifier design
The photodiode amplifier is a simple transimpedance amplifier that converts the photodiode current into a voltage output. It was designed to be as small as possible to fit in the constrained space available for each of the 18 sensor positions on the outside circumference of the sensor ring. The basic design consists of the SFH213 photodiode [14], an MCP6491 operational amplifier [15], a gain resistor (R1 = 100 MX), and two optional capacitors C1 and C2 (Fig. 20). Given the size constraints, a surface-mount design was implemented (Fig. 21), although a through-hole design would be preferable for ease of construction. It is however possible to solder manually the SMD components onto the PCB with a steady hand and a good pair of tweezers.
Solder one end of a short, screened 3-core cable

Sensor instrumentation and data logging requirements
The photodiode amplifier is particularly sensitive to ambient electric fields due to its high gain configuration. In this design, the amplifier is encased in plastic and potted with epoxy resin. Although this design approach is low-cost, it does not instil any degree of RFI or EMC protection into the circuit. During development, it was observed that the amplifier is particularly sensitive to the presence of ambient 50 Hz interference when a human operator places their hand in close proximity (within a few cm) to the device. The human body is a very good radiator of this background artefact of the domestic power distribution system, hence leaving the device undisturbed during operation is recommended. A countermeasure to this undesirable effect is to sample the photodiode amplifier signal at a rate of 10 k samples per second (SPS), allowing for the application of a digital low-pass filter to remove 50 Hz power line interference and other transients (3rd-order Butterworth infinite impulse response (IIR) filter with 1 Hz cut-off frequency). We applied this approach to all 18 data channels, but only recorded the filtered data to disk at a rate of one sample per second. Also recorded at a rate of 1 SPS was the instantaneous LED current as measured by the active LED control and current monitoring system (Fig. 23). This measurement allows each of the 18 measured data points to be compensated for LED intensity, as inferred from the recorded LED current value.
The resolution at which the sensor voltage is measured is recommended to be less than 1 mV if possible. In some situations, the side-scattered signal from suspended sediment will lead to some very small responses in the detectors situated around the 90°position, so to see any structure in the recorded time-domain signal the best voltage resolution possible is desirable. We used a DAQ system with a voltage input range of ±10 V, and an ADC resolution of 16 bits. This system gave us a voltage measurement resolution of 0.305 mV. However, if the voltage input range of the DAQ were to be limited to 0-5 V, then the exact same voltage resolution could be achieved with only 14 bits of ADC resolution.

User DAQ
This article does not describe a data acquisition and logging system. The operators must therefore select their own system. The data presented in this article were obtained using National Instruments (NI) DAQ equipment, and was programmed using NI LabVIEW graphical programming environment [16]. The production of a low-cost, open-source DAQ system is the subject of future work.

Example LabVIEW GUI
The DAQ system that we implemented using NI equipment has a LabVIEW GUI (Fig. 24) that controls the operation of the LED Control & Monitoring system and reads the instantaneous LED current. It also logs 18 channels of data from the connected NI DAQ hardware, which was in this case 16 channels from an NI USB-6211 [17] plus 2 channels from an NI myDAQ [18].

Hazards associated with high-brightness LEDs
Warning: All LEDs that emit focused IR or UV light can be dangerous to the eyes, and UV LEDs can potentially damage the skin. Even the visible light LEDs when operated at nominal currents (e.g. 30 mA) can cause damage to the eye due to high output intensities (sometimes tens of thousands of millicandela). It is therefore recommended that anyone operating high- intensity visible, IR or UV LEDs should never look directly into the lens of the LED. This is of paramount importance where IR LEDs are concerned, as they will appear to be un-powered even when operating at nominal intensity, since IR light is not visible to the human eye. As such the human eye will have no pupil reflex with which to attenuate the incoming light, and permanent damage can be caused to the retina -including blindness. It is good practice to treat high-brightness LEDs with the same caution as one would a LASER light source.

Sensor calibration
The following list is a complete calibration example. The operations are to be performed in this order: 1. Connect DAQ system together, wire up all sensors etc., apply power and test all DAQ channels and functionality. 2. Choose a light source. In this example an IR LED (k = 940 nm) has been selected [19]. Plug it into J2 on the Control & Monitoring shield. 3. Determine the relationship between LED current and output light intensity in mW/sr by extracting information from the LED data sheet. In Fig. 25 the radiant intensity is stated in milli Watts per steradian (mW/sr). If the radiant intensity is stated in photometric units of millicandela (mcd), then the mcd value must be multiplied by the conversion factor V (k) to obtain the radiometric value in mW/sr (Eq. (3)). Wavelength k is stated in nm [20,21]: For example, a blue LED of wavelength 470 nm, V(k) = 0.106264. This conversion is necessary since the candela is weighted according to the wavelength-dependent sensitivity of the human eye to light. Since photodiodes do not have this non-linear human-eye response, photometric units such as candelas (or mcd) must not be used to report the measured light intensities.
Using data extracted from the graph in Fig. 25, the relationship between the applied current and the 940 nm LED light output can be stated: I OUT (mW/sr) = 1.978 Â I IN (mA), as shown in Fig. 26. The constant multiplier 1.978 shall be referred to as m LED . So the expression for the input current to output light intensity is: 5. Determine the response of each individual sensor module to the light source using the optical calibrator, and plot the relationship between sensor voltage and incident light level (mW/sr). The Arduino software has a ''calibration sweep" function that performs a stepwise drop in LED current from the present set point. The current dwells at each step for one second for stabilization, thus allowing a calibration graph to be plotted. The gradient of the linear regression fit of sensor voltage S V to the measured light intensity I h (h is the scattering angle) yields the sensor calibration coefficient S a , where a is the sensor number. The measurements shown in Fig. 27 were made with a 4.0 ND (3.42 ND 940nm ) filter in place to attenuate the beam (otherwise the sensor output would saturate). Extracted from Vishay TSAL6100 data sheet [19]. Fig. 26. LED current input vs. light output intensity as derived from Vishay TSAL6100 data sheet [19]. The data points were extracted from the data sheet ''by eye". Even so, there is a good fit to a linear regression (R 2 = 0.9947, p = 2.2 Â 10 À16 ) making the data suitable for calibration purposes.
We now have a general calibration equation that can be applied to experimental data to give I h at a given scattering angle h in mW/sr: The calibration sweep was performed by starting at an initial LED current of 30 mA and ramping the current down from there. Applying general calibration Eq. (7) we now obtain a specific calibration equation for sensor #9 at the 90°position on the sensor ring: The I 90 equation (above) will be used in the experimental measurement section.
6. With all the sensor modules in position around the sensor ring, add clean water to the sample cell (in this case tap water). Remove all ambient light (i.e. total darkness) and measure the response of all the sensor modules to the LED light source located at the 180°position. Any response measured at sensor locations from 10°to 170°are due to internal reflections within the instrument (Fig. 28).
The response of the sensor at the 0°position is due to the direct beam, and is likely to cause the sensor to saturate. This effect could also be apparent at high beam intensities at the 10°and 20°positions (and other forward-angle positions), due to the divergence of the incident beam. The LED intensity can be stepped from high to low in order to generate a device geometry baseline (DGB k ) dataset for each sensor. The ''calibration sweep" function in the Arduino firmware can be used again to achieve this functionality. Fig. 29 shows data for high (29.02 mA) and low (5.38 mA) LED intensities. Scattering angles h 0 and h 10 have saturated detector responses at both intensities. The effect of the beam divergence on the forward-scattering sensors can be seen to drop off as h approaches 90°. The effects of internal reflections can be seen to increase in the backscattering sensors as h approaches 170°. The sensors are numbered from 0 to 17, and are placed at corresponding h Â 10°a ngular positions around the sensor ring. For example, sensor #9 at 90° (Fig. 27) is located at the h 90 position.

Performing a sediment-settling experiment
1. Choose a light source. In this example a 940 nm LED (infrared) was selected, at a nominal operating current of 30 mA. The diagrams (Figs. 30 and 31) illustrate the experiment using blue light for clarity, since infrared light is not visible to the human eye. 2. Perform an experiment on a suspended sediment sample. The experiment design is at the discretion of the user, for example the choice of initial agitation method (e.g. stirring, dropping sediment into the water column, shaking the sample, measurement frequency, duration of experiment). The sample was stirred with a magnetic stirrer until a consistent degree of suspension was achieved. The settling experiment begins when the stirrer is switched off.

Preparing the data
It is important to know what events have been recorded during the experiment, and to know when the settling process begins. In our experiments, the entire process including the agitation phase is recorded. Thus, it is necessary to remove manually the beginning section of each data set in order to remove the effects of the stirring process. Also during this start-up phase, the LED is ramping up to its set-point current, and so this initial data is not meaningful. Only when the LED is at a stable current and the stirring has stopped can the data considered to be at the time = 0 position, t 0 (Fig. 32).

Validation, characterization and modelling
TARDIIS allows the researcher to obtain a rich and diverse set of data about sediment as it settles in a column of water. These data do not however show directly the concentration of sediment at a particular height in the water column. The only way to measure exactly the characteristics of the sediment at different positions in the settling column and at different times during the settling process, is by physically sampling the column. Then, by the use of traditional sediment analysis methods, it is possible to determine the concentration, the PSD, the grain shapes, the degree of flocculation (DOF) and so on. The researcher must accept that there is not a well-defined link between sediment properties and turbidity in general. With this  acceptance, the researcher may then challenge existing approaches to turbidity measurements, and use TARDIIS to establish new models that are based on the physics of light scattering and absorption.

Assessment of the active LED control and current monitoring system
The instantaneous current was measured throughout the settling experiment, thus allowing for the correction of the calculated I 0 value. The current is plotted against time (Fig. 33) and shows that the drift in I 0 is very small -i.e. not greater than 72 mA -over the 10.67 h duration of the example experiment. There is a small amount of noise present in the signal; however, the overall degree of drift shows that the Arduino control system is performing very well, as the desired current is 30 mA in this case. There is an overall offset error of approximately 40 mA.

Linearity of device geometry baseline measurements
It has been shown that the response of a specific sensor (#9) to incident light is linear (Fig. 27) when calibrated using an ND filter. We must now check that the response function is linear when TARDIIS is operational. Using data that was collected to produce Fig. 29, we can now show that the same linearity is present under device geometry baseline (DGB) conditions, with only water present in the sample cell. The responses of sensors #6, #9 and #12 are shown as I 60 , I 90 and I 120 respectively in Fig. 34, which confirms that the linearity is present. This DGB linearity must be tested for all sensors at all wavelengths of light that are to be used during settling experiments.

Example of sediment-settling data
The data in Fig. 35 show the measured light intensity at three of the 18 available scattering angles during a sediment settling experiment. The initial peak that occurs at around 30 s into the experiment is due to turbulence in the water column after the magnetic stirrer is switched off. By analysing data gathered in this way at multiple scattering angles and wavelengths of light, the researcher may then: Develop models of light scattering by particles in suspension, and gain new insights into the effects of grain shape and PSD. Explore the impact that water colour has on turbidity measurement. Investigate the conditions under which the particle settling velocities deviate from Stokes Law. Enquire as to the suitability of the recognized turbidity standards. Apply new knowledge to future field measurements.
The ''optical sedigraphs" shown in Fig. 35 look as if they may be giving an indication of Stokes-like settling behaviour. In order to explore this possibility, the researcher must design their experiments carefully, and then develop suitable numerical models to make predictions based on empirical data obtained using TARDIIS. This detailed analysis and model development is however beyond the scope of this article.

Declaration of interest
None declared.