Grantee Research Project Results

Final Report: Smart Turbidimeters for Remote Monitoring of Water Quality

EPA Grant Number: SU835732
Title: Smart Turbidimeters for Remote Monitoring of Water Quality
Investigators: Ball, William P. , Kelley, Chris , Backer, Andrew , Brunner, Logan , Burkland, Alison , Farhat, Michelle , He, Ziwei , Majithia, Vishwesh , Murphy, Sushant , Kahn, Daniel
Institution: The Johns Hopkins University
EPA Project Officer: Page, Angela
Phase: II
Project Period: October 1, 2014 through August 31, 2017
Project Amount: $89,995
RFA: P3 Awards: A National Student Design Competition for Sustainability Focusing on People, Prosperity and the Planet - Phase 2 (2014) Recipients Lists
Research Category: P3 Challenge Area - Safe and Sustainable Water Resources , P3 Challenge Area - Sustainable and Healthy Communities , Sustainable and Healthy Communities

Objective:

During this Phase II project, the Johns Hopkins “Smart” Turbidimeters team worked to improve Phase I outcomes by standardizing our prototype designs, lowering costs, reducing complexity, improving wireless communications capabilities, and making device designs that are easier to manufacture. The team also developed and expanded an international collaboration with our main field implementation partner, the Honduran NGO Agua Para El Pueblo, to distribute devices to field sites for evaluation. Important objectives were to to improve the designs based not only on laboratory testing, but also based on experience with transport to the field and routine field application in the real world environment of poorly resourced developing communities.  In addition, the team worked expand the scope of the hardware design by integrating chlorine residual detection into the handheld turbidimeters and developing an in-line unit that was later (outside the scope of this project) purchased and tested by an NGO working in Kenya.  Finally, the lead graduate student also did work in partnership with the AguaClara drinking water infrastructure design program at Cornell University to develop a mobile app (targeting affordable, low-end Android smartphones) to interface with the prototype handheld and inline turbidimeters via Bluetooth.  This smartphone app has been successfully applied by the Agua Clara team.

Summary/Accomplishments (Outputs/Outcomes):

In 2015 (year 1 of the phase II project), the team produced several iterations of our handheld turbidimeter. Comparing the Phase I and current Phase II versions, the team was able to reduce size by 80%, weight by 75%, and assembly time by 66%, while holding material costs at $25 per device. They upgraded the device's display from a 4-digit “alarm clock” display to a 1.3" OLED screen -- this has greatly expanded the utility of the device's user interface, while reducing the complexity of the circuit board design and lowering passive power consumption by 20%. Disposable alkaline batteries were swapped for rechargeable lithium batteries, improving the environmental impact of the device while lowering the long-term operating costs. Battery life has been significantly improved (by, e.g., switching to a low-dropout linear voltage regulator), allowing 4,000-8,000 measurements per charge cycle. For an additional $5 in materials costs, a real-time clock with battery backup and 256Kb of onboard RAM can now be added to the turbidimeter, allowing automatic timestamping and backup of the most recent 1,000 readings taken. The next version of our handheld turbidimeter will make use of a "2.5-dimension" design plan, in which the case components are sandwiched together and secured with bolts or clips. This will allow case components to be easily cast from polyurethane rather than 3D-printed, which means that cases can be manufactured more quickly, will be stronger, and will be significantly easier to waterproof.

The team also increased the accuracy of the handheld turbidimeter, particularly for lowturbidity measurements, beyond what was documented in our 2014 Sensors paper. Five key design choices helped this effort:

1)      The use of multiple infrared LEDs, both to increase total light available for scattering by suspended sediments of aqueous samples, and to allow for the generation of multiple overlapping calibration curves (corresponding to different combinations of lights).

2)      Intentional misalignment of the LED paths: using three LEDs, one at 270 degrees to the sensor face and the other two offset slightly (at +/- 16 degrees) from 90 degrees, helps to correct for optical distortion of the cuvette glass due to anisotropy.

3)      Incorporating charge regulators for each LED, in addition to a voltage regulator for the circuit board, to minimize voltage and ambient temperature effects on LED brightness.

4)      Adding firmware code that automatically doubles the sampling period for very low turbidity readings, and halves the resulting sensor readings, to compensate for the “stair step” pattern of data produced by discrete sensors at very low signal levels.

5)      Smoothing low-turbidity readings with oversampling: For a given sample measurement, multiple readings are taken – each with a different fractional multiplier applied to a standard one-second reading interval (this multiplier then divided from the numerical reading). This “time stretching” of the sampling window is comparable to adding noise to the signal received by the sensor, and yields a significantly better averaged result at low turbidities than would be obtained from simple replicate sampling. 

Chlorine monitoring was added to the handheld turbidimeter with the addition of a

TCS3200RD color-to-frequency sensor placed directly below the turbidity sensor, and a green

(515 nm wavelength) LED mounted opposite in the sample chamber. This colorimeter setup can be used to detect the presence of the DPD colorimetric indicator, which turns pink in the presence of chlorine. Results show that this colorimeter setup can detect chlorine concentrations in the range of 0.8-10 mg/L.

The inline turbidimeter was implemented and steadily improved over the cou in this ragnge, rse of the first year of Phase II. The updated design included two 30mA infrared LEDs placed on either side of the digital light sensor (which is covered by a visible-light filter). The device is enclosed in a commercial waterproof case (IP67-grade, manufactured by SERPAC) to which a Lexan viewport is epoxied. The focus on year one was primarily on turbidities above 5 NTU and in this range, some important outputs of the research were calibration curves that showed regression lines with R-squared values greater than 0.99995. In addition opportunities for further improvements were identified for the next year, including an increase in the LED output (by increasing the charge regulators to 40mA, and/or adding a third light) and implementation of improved sampling windows for the online unit, as based on successes achieved with the handheld device. Also planned for year two was the implementation of temperature compensation using an externally mounted analog sensor or thermistor.

In 2015, the team also made great strides in improving their knowledge of and familiarity with low-cost GSM modems for SMS communication. This work has reduced the cost of drop-in GSM communications from $90 (with an Arduino Antenova unit) to $20 (with a customdesigned SIM800 board) per device. They successfully tested an inline turbidimeter, immersed in three feet of water, relaying data with the OpenSourceWater SMS protocol via an internal SIM800 GSM modem, and adjusting its sampling parameters in response to remotely issued commands. We have also tested remote communication via Bluetooth paired to an Android phone, and over wifi (using a $3 ESP8266 wifi module). In collaboration with a Cornell computer science master’s student (Ethan Keller), they began the design of a smartphone app for transmitting and receiving device data, communicating the OpenSourceWater server, and providing real-time visualization of water quality data.

One unexpected setback in 2015 was that our proposed implementation partner in India, AguaClara LLC, found themselves unexpectedly overloaded with other work and were unable to help us evaluate our technology. Thankfully, our Honduran partner Agua Para El Pueblo was able to increase (to ten) the number of treatment plants at which our team could evaluate equipment. Handheld models of the turbidimeters were tested  only in Honduras.  In addition the team obtained preliminary feedback on the inline prototype from the Maine office of The Nature Conservancy, who tested the device at a project site in Kenya.  One espeically important output was the software and hardware schematics for all of the Phase II deliverables, all o fwhich were made publicly available in a GitHub repository, under the acronym AWQUA (for Affordable Water Quality Analysis; see <https://github.com/awqua>).

In year two, the team continued development of the low-cost, open-source inline and handheld turbidimeters developed the previous year . The durability and accuracy of the submersible inline turbidimeter was significantly improved through hardware and software modifications, while the per-unit cost reduced (to $40/unit). A photodiode-based prototype of an open-source low-cost ($40) handheld turbidimeter was developed; while somewhat more complicated in design than the monolithic sensor-based turbidimeter developed last year (based on the TSL230RD light-to-frequency sensor), the photodiode-based prototype offers multiple potential advantages in low-light detection and statistical evaluation of turbidity readings. The team developed a prototype open-source low-cost ($150) battery-powered jar tester, and thus completed at least one prototype of each device specified in the project goals. And, the partnership with the AguaClara drinking water infrastructure design program at Cornell University to develop a mobile app (targeting affordable, low-end Android smartphones) to interface with the prototype handheld and inline turbidimeters via Bluetooth. 

In the second year of the Phase II project, the handheld and inline turbidimeter design have iteratively improved without increasing costs, and we have completed a prototype of an opensource jar tester. 

Casing: We reached the limits of what we could reasonably accomplish with 3D-printed (filament deposition) cases – 3D printing is a time-consuming process, and as we have improved the optical instrumentation of our handheld turbidimeters, the slight (thermal and shear) deformability of our 3D-printed cases became increasingly problematic. We have always considered it a key piece of this project is to build water quality monitoring tools that are not only affordable and open-source, but to use low-cost, commonly available manufacturing tools and methods. Our previous cases were built using an open-source RepRap 3D printer that can be built for roughly $700. We have begun building cases with a desktop 3-axis CNC mill, which can be purchased for $1100. To test the limits of low-cost manufacturing, we have also undertaken to machine cases and internal optical instrumentation holders for our handheld turbidimeters using just a drill press ($80) and a band saw ($100). This process should help us to better understand how affordably small-scale manufacturing centers could be set up in developing countries to produce the devices developed in this project.

Inline turbidimeter: The resolution of the inline turbidimeter has been greatly improved by increasing the number of IR LEDs from two to four (arranged in a rectangle around the central sensor) and implementing the “stretchy” sampling window times and turbidity-dependent sample batch sizes that we earlier successfully implemented in the prototype handheld turbidimeter. To summarize the “stretchy” sampling windows:  For a given sample measurement, multiple readings are taken – each with a different fractional multiplier applied to a standard one-second reading interval (this multiplier then divided from the numerical reading). This “time stretching” of the sampling window is comparable to adding noise to the signal received by the sensor, and yields a significantly better averaged result at low turbidities than would be obtained from simple replicate sampling. (We have tried using equal-interval sampling window fractions (e.g. 9/8, 10/8, 11/8… seconds) and coprime sampling window fractions (e.g. 8/7, 6/5, 4/3… seconds); so far, these appear to perform equally well.) With these improvements we have been able to lower the linear turbidity detection threshold to below 2 NTU. 

New optics for handheld turbidimeters: While the light-to-frequency sensor (TSL230RD) we have incorporated in previous handheld and inline turbidimeters provides a simple, hassle-free optical setup, we decided to develop an alternative optical setup using silicon-based photodiodes and an analog signal processing front-end to take advantage of the speed and low-light sensitivity of amplified silicon photodiodes (which are mainstays of high-speed fiber-optics communications systems). In addition to reduced sampling times, this new sensing paradigm makes a ratio turbidimetric sensing setup (utilizing a second photodiode measuring attenuated light through a sample to improve resolution and linearity) more feasible, and offers the ability to increase the number of replicates in a single read by an order of magnitude, making practical the calculation of confidence intervals for turbidity measurements and the statistical detection of potential hardware errors (such as optical drift, or intermittent light leakage in the cuvette chamber). In the coming year we intend to adapt this setup for the inline turbidimeter – with the addition of a chopper driver (to transform the constant brightness of the LED driver in the

current optical setup into a modulated square wave) and synchronization of the chopper driver and the photodetector, the inline backscatter turbidimeter can completely compensate for the presence of significant background light. This is particularly useful as it is infeasible to completely shield the inline turbidimeter sampling window from stray light, and the current solution for ambient light filtering (visible light-blocking film) has a limited life span and does not block IR light from the sun. 

Smartphone-based user interface: The first version of the smartphone app, developed in collaboration with AguaClara, is now available in the Google Play Store

(https://play.google.com/store/apps/details?id=org.aguaclara.post.visualizations). A Bluetooth connectivity module for this app, to allow it to connect to the prototype turbidimeters developed by the “Smart” Turbidimeters team, is under active collaborative development; the code repository is located at https://github.com/AguaClara/post.bluetooth. 

All other software and hardware files for all of the “Smart” Turbidimeters team’s Phase II deliverables continue to be made publicly available in a GitHub repository, under the acronym AWQUA (for Affordable Water Quality Analysis) at https://github.com/awqua

 

Conclusions:

The Johns Hopkins “Smart” Turbidimeter team has built on the work produced during the first year of Phase II of this P3 grant. We have significantly improved upon our previous device designs, and have now completed prototypes of all Phase II deliverables. This work made important strides towards achieving the ready availability of affordable, responsive, remotely accessible, real-time water quality monitoring with wide potential ramifications for the sustainability and public health of water treatment projects in developing regions around the world. The preliminary work done under this P3 funding was extremely important toward launching ideas that we believe should be carried forward – see below.

Journal Articles on this Report : 1 Displayed | Download in RIS Format

Publications Views

Other project views:	All 2 publications	1 publications in selected types	All 1 journal articles

Publications

Type	Citation	Project	Document Sources
Journal Article	Kelley CD, Krolick A, Brunner L, Burklund A, Kahn D, Ball WP, Weber-Shirk M. An affordable open-source turbidimeter. Sensors 2014;14(4):7142-7155.	SU835732 (Final) SU835517 (Final)	Full-text: MDPI-Full Text HTML Exit Abstract: MDPI-Abstract Exit Other: MDPI-Full Text PDF Exit

Supplemental Keywords:

Turbidity, chlorine residual, water quality monitoring, open-source, low-cost technology, sustainable development

Relevant Websites:

Affordable Water Quality Analysis (AWQUA) Initiative - GitHub Exit

Progress and Final Reports:

Original Abstract

2015 Progress Report

2016 Progress Report

P3 Phase I:

"Smart" Turbidimeters for Remote Monitoring of Water Quality  | Final Report