James Berg
CEE 309
26 January 2005
$30 Turbidity Sensor
The function of a turbidity sensor is to measure the amount of particulate matter in a water sample. Once the particulate matter has been determined, that number can be used to estimate the amount of carbon based material in the water. We want to know how much carbon matter is in the water because that can be used to estimate how much bacteria and viruses are in the water. This estimate is necessary for determining what treatments are necessary and how effective they are. The problem is that presently the cheapest turbidity meters are above $600, well out of range for poor villages in developing countries.
It seemed to Professor Weber-Shirk and I that the basic design of a turbidity meter was fairly simple and could be constructed for a small fraction of the present cost. The hard part was that in order to be useful the turbidity sensor would have to be sensitive enough to distinguish a difference of 1 NTU. If it would register an 18 NTU sample at the same voltage as it would register a 23 NTU sample then it would be useless for determining the appropriate amount of chemicals to add.
A turbidity sensor operates by shinning a laser through a sample of water towards two light sensors. One sensor, positioned 180° away from the laser, measures how much light the water transmits. A second sensor at 90° measures how much light is reflected off the particulate matter. Then the measured intensity of light is turned into a voltage that varies with turbidity, as shown below.
As you can see the 180° sensor decays very quickly and the 90° sensor can have two possible turbidity values for one voltage so neither sensor can be used on its own to read very dirty water. Due to the way turbidity is defined when you divide the 90° sensor by the 180° sensor you get a straight line that is perfect for determining turbidity from voltage.
Past Work
I started doing research on the sensor late last semester. We decided to initially forgo a built in readout and power source and just hook the sensor up to a computer data acquisition system. We also decided that, as the intended application of this turbidity sensor would only need to work in the lower turbidity range, up to around 100 NTU, it was unnecessary to have both sensors and that a 90° sensor would work just fine on its own. The goal set was to construct a turbidity sensor that could distinguish a difference of 10 NTU for under $30, at least 20 times cheaper then the cheapest turbidity meters on the market today. These goals were not set as a final design, but simply as an intermediate step to determine the feasibility of making an inexpensive turbidity meter.
It turned out to be relatively easy prospect. All that was needed was a light-tight box, a laser, and a light sensor. The basic design was a box large enough to hold a 2 liter bottle (the sample container), a pen laser and a 90° sensor (placed in the lid of the box). I made the light box out of black foam core which was effective but turned out to have a lot of light leakage around the joints, the sensor could register someone walking by it, this decreased the sensor’s ability to detect light scattered by particles in the water. I was able to increase the sensitivity of the system by adding a light well on the opposite side of the bottle from the laser, where the 180° sensor would have been, that decreased laser light scattered by the opposite wall. The light sensor presented a bit of a problem, though it turned out to be an easy fix. Initially I thought that photodiodes varied resistance with light, this lead me to construct a fairly simple little circuit to turn varying resistance into varying voltage and hook it up to a power source. The results with this system were sporadic, so I went back to the drawing board where I discovered that a photodiode is in fact an extremely sensitive photo voltaic cell that generates its own voltage. I simply removed the circuitry and power supply and connected the photodiode directly to the data acquisition system. I was then able to collect enough data to determine that the sensor was accurate enough to detect as little as a 3 NTU difference in samples. I did not have a lot of time to run tests but I believe that the sensor could actually be more accurate, something which I will have to test this semester.
The accuracy of the turbidity meter was in part due to the long path length of the laser. Using a 2L bottle with a diameter of roughly 10cm, instead of a test vial with a diameter of roughly 1.5cm, meant that the laser came into contact with more particles and thus more light was refracted into the 90° sensor. The laser and the photodiode were choosen based on their wavelengths. I must admit that I don’t know why, but turbidity meters in labs use lasers that are in or very close to the infra red spectrum. None of the inexpensive lasers produce wavelengths in the IR spectrum, so I choose one that was as close as possible, 680nm. The Photodiodes were chosen because they were receptive to that light range.
As the table to the right shows the whole experimental system was put together for less then $30, though this number lacks an output device to tell the user exactly how dirty the water is, most likely a voltmeter, an on/off switch and a power supply (batteries). I expect that a voltmeter, either digital or analogue will cost around $10, a switch should be about $2 and batteries should be around $3, putting the cost at about $35.
Improvements for This Semester
The most important thing to do this semester is to, preferably find, and if necessary make a voltmeter that is sensitive enough to register a difference in the voltage put out by the sensor from two samples with a difference of 1 NTU. I fear that this may prove to be the most difficult part of the whole project. Voltmeters are fairly common and as far as anyone doing electronics is concerned, fairly cheep, however getting the necessary accuracy from a voltmeter that costs $10 could be difficult. Another problem is that the voltage outputted by the sensor is not going to be exactly constant, it will fluctuate some and with a voltmeter sensitive enough for our purposes that fluctuation is going to show. On a digital voltmeter this will cause the numbers on the right side of the readout to change, possibly too quickly for a person to read, making it difficult, but not impossible, for the user to obtain a good averaged, value. An analogue voltmeter would actually work much better for our purposes. It is true that the needle on an analogue voltmeter would also fluctuate rapidly but given their layout it would be much easier for the user to visually determine an average value. Unfortunately analogue voltmeters are becoming increasingly hard to find and so we may end up using a digital one. There are averaging microchips that are available but they are fairly expensive.
Also the possibility of adding a 180° sensor should be explored. As was stated above with just a 90° sensor it is possible for two different turbidities to have the same voltage. If one of these turbidities is out of the range likely to be encountered then it is not worth the expense, nor the effort. Even if a 180° sensor would be beneficial the circuitry necessary to divide one signal by the other may be quite expensive, even if the additional sensor is not.
The next improvement that needs to be done is to improve the box. Both in preventing ambient light from getting in and making it out of something more durable then foam core. I believe that the best solution is to make it out of 1/8” plywood, paint it black and place some foam insulation strips around the joints to keep light out. Similar to improving the box is adding a centralized power supply (batteries) and an on/off switch. Ideally one switch would turn on everything, this may mean taking apart the laser to certain extent, to wire it to the power supply rather then having batteries in it.
An improvement could also be made in the sample container, a 2 Liter bottle was chosen because it increased the distance the laser would travel, increasing the number of particles that the light hit, increasing the amount of light scattered, increasing the amount of light that hit the sensor and improving its sensitivity. Unfortunately a 2L bottle takes a long time to fill through such a small hole and is both large and heavy to carry around. A smaller, lighter turbidity meter would be easier to carry around and less likely to get broken. A 2L bottler may in fact be the best available solution but, it is worth looking into other options. Attention also has to be paid to the ease of use of the device. It would be much less useful if users needed significant training before they could use it.
I will most likely use the same or at least similar laser and photodiodes. The only reasons for changing them would be if I were to find a cheep IR laser (unlikely) or if they do not work with a centralized power supply.
There will be three other written reports this semester. Progress reports will be due on March 4 and April 11, and a final report will be due during finals week.
James,
Check out http://rocky.digikey.com/WebLib/Microchip/Web%20Data/TC500,%20A,%20510,%20514.pdf as an example of an incredibly cheap ($5) analog to digital converter with built in integrator. But it doesn’t look simple to implement since it requires a micro controller and the output device (the display). I am uncertain if a reasonable design for a turbidity meter is going to get this complicated. I forget, what did the engineers at Honeywell advise you? I recommend that before you begin building you spend more time exploring the options and develop a well justified proposal for the system you want to assemble. The progress report of March 4 could be this more specific proposal that includes the specifications of the components.