Public Lab Research note

  • 1

Filtered tap H2O vs unfiltered tap H2O spectral graph comparison

by Suiris |

I wanted to put my water filter up to a test. Two identical transparent bottles were filled with tap water, one bottle with filtered tap H2O and the other unfiltered tap H2O. Two tests were taken to make sure the results were conclusive and in both tests all parameters were exactly the same except for camera shutter speed.

Location: Sweden

Camera: Canon IXUS 120 IR filter removed, ISO 80

Slit width: 0.08mm

Distance from slit to grating: 30 cm

Distance from grating to camera lens: 3mm

Illumination: Warm white IKEA LED

Filter: Espring (carbon + UV)

Test # 1

Camera shutter speed 8 sec


Filtered tap H2O spectral image

Unfiltered tap H2O spectral image


Test # 2

Camera shutter speed 16 sec


Filtered tap H20 spectral image

Unfiltered tap H2O spectral image


Result: In each test the graph for filtered H2O peaked higher than the graph for unfiltered H2O

Conclusion The filtered H20 is more transparent due to less contamination, therefore peaking higher then the more contaminated unfiltered H2O. The filter is working!

spectrometer spectroscopy water-quality water-monitoring sweden transmittance


Excellent test. What do you think is in your water? Particulate matter? Or something dissolved, like a salt? Either way, if you boiled some down to half its volume the transmittance should decrease. If the contaminant is volatile, boiling would probably not decrease transmittance.

Is this a question? Click here to post it to the Questions page.

Hi cfastie. Thanks! The water I tested came from the faucet in Stockholm, Sweden. The water treating plants here have a hard time filtrating medicinal, contraceptive and chemical residue. Besides that after the water is treated it travels through mile long old pipes that often had been repaired with led and along the way pick up particulate matter. Chlorine is also added to the water. The filter I use will filter everything except for chlorine, instead it will add an extra electron (Cl--->Cl-) and turn it into Chloride. The filtered water in this test is supposed not to contain any particulates and chemical residue.

Boiling the samples down to half their volume is a great idea! Will try that next time and see if there are any volatile contaminants.

I wonder about the peaks appearing to shift between different exposures. Looking at the left most peak the filtered 8 second exposure is to the right of the unfiltered exposure while the 16 second exposure is the opposite case. What is the cause? I also wonder what the spectrum would look like if you subtracted the "Warm white IKEA LED" spectrum.

Is this a question? Click here to post it to the Questions page.

I noticed the shift too danbeavers. My thoughts are that when I cut out the spectrum from my image and pasted it as a new image I wasn´t carefull with cuting at the exact same place each time. The cuts were of different hight and length. Could that be it? How can I subtract the LED spectrum?

Is this a question? Click here to post it to the Questions page.

OK then I would like to know about your work flow. Are you using a camera attached to your computer or are you taking a picture and then processing the image?

I think "use as baseline" is the key to the subtraction. The instructional video is the clue. Also you could use the unfiltered water as a baseline.

Is this a question? Click here to post it to the Questions page.

I am using a Canon IXUS 120 IS for taking pictures, then I cut out the spectrum from the black background, paste it as a new image and upload it to spectralworkbench. Can you post a link to the instructional video?

Is this a question? Click here to post it to the Questions page.

I hope your system is mechanically stable so as to get reproducible spectra. With regard to the illumination spectrum (the LED), it is not truly white i.e. it is not uniform across the range of wavelengths that your camera is sensitive to. The sensitivity of the camera is probably not all that uniform either. To normalize your sample spectrum, you will want you to divide (not subtract) the sample spectrum by the illumination (LED) or reference spectrum on a pixel by pixel basis. Use an sample empty bottle for the illumination spectrum. Then add your water sample to the bottle for your sample spectrum. This will largely cancel the effects of the LED spectrum and the color sensitivity characteristics of the camera, as well as the effects of the bottle. Use the same shutter speed for both spectra. Dividing the sample spectrum by the illumination spectrum gives what is known as the transmission spectrum, which varies from 0 to 1 for each pixel. It ideally is a characteristic of the sample alone, the "instrument function" having been eliminated by the above normalization. As long as you are doing math on the spectra, you might as well go for the whole enchilada. Take the negative logarithm of the transmission spectrum, again on a pixel-by-pixel basis. These values will vary linearly with concentration of the light absorbing species in your sample. Look up Beer's law on the web or in any introductory chemistry book.

Now for the bad news. What ever may be in your water samples may not absorb in the visible range. Further, unless your water is very bad and your spectrometer very sensitive, you may not have enough sensitivity to see anything in the spectra. As a test, try adding some food coloring to your water sample. Try several concentrations to see how the spectral amplitudes vary. If you want to get even fancier, you can take a "dark spectrum" i.e. no illumination and subtract it from both sample and reference spectra before normalizing to correct for any ambient light leakage or baseline errors in the camera.

Some typo's in my previous post:
To normalize your sample spectrum, you will want (delete "you") to divide... Use an (delete "sample") empty bottle...

One other test to characterize the stability of your spectrometer is to just repeat identical measurements after a certain time span, say minutes, or even hours. Have you tried to repeat the measurement just of one sample with a couple of minutes or hours in between, in order to test the reproducibility of your spectrometer? do you get the same spectrum? If for example the LED fixture warms up over the time of minutes, the absolute brightness may vary, and peaks might shift...

Is this a question? Click here to post it to the Questions page.

Hi Ernie, Thanks for the great advice and information. To be honest I am completely new to te spectrometer hobby and most of what you wrote baffles my mind. I am glad you wrote all that stuff though, some day when I have the time and interest I will go through it all carefully and try to follow your advice step by step. Will be a fun project. I thought about it and to make it more simple I could have just taken a red laser and see the difference in water clarity that way, it would be just one curve instead of a hole spectrum of wavelengths. I never thought as far as to see what exactly is in my water, the experiment was done just to see if there will be any difference at all in the spectrum; and there was! :) Greetings from Sweden.

Hi drabanus, The time span between each spectra was less than 30 seconds and the LED light had been on for quiet some time before I began the tests. I never tried to repeat the experiment with hours in between. The spectrometer I am using is the same as in my other post, "New spectrometer build and CFL lines", it is a box with black filt inside, the camera was never moved through out the testing, all I did was replaced the bottles. I had an idea to use a laser instead of a LED, then I would just get one curve and could measure the difference in clarity between the different water samples that way. Thanks.

I afraid that using a laser will not tell you much about the chemical composition of any contamints in your water. The laser spectrum is much too narrow. You might be able to measure turbidity (suspended particulates, etc.) with the laser. You will need to change the optical geometry. Place the optical detector (camera) so that its axis is perpendicular to the laser beam passing through your sample. This way, the detector will only receive light that is scattered by the sample, rather than light that is transmitted as in the spectrometer set up. You will still need a reference sample to compensate for the istrument function i.e. the part of the measurement that is not due to your sample of interest. In this case, a sample of known clean water would be the reference. This time you will want to subtract the readings from the reference measurement from the readings from the sample measurement. As a test sample, try a little whole milk diluted with water. The tiny globules of fat in the milk will act as suspended particles to scatter the laser light. i would try one part milk to nine parts water as a starting point. DO NOT LOOK DIRECTLY INTO THE LASER BEAM. You will probably have to adjust the dilution factor to get the measurement readings in the operating range of your instrument.

The test was just to measure the difference in clarity between tap water and filtered water, using a single wavelength instead of a white warm LED would give me just one curve and the the graph would look simpler and easier to read. Diluting water with milkis a great idea.

Hi Suiris, it good to see how filtering the tap water works. How do you take transmittance measurements using spectralworkbench? I am looking into taking transmittance measurements of a glass coated with a thin transparent layer. I would appreciate if anyone could give me a link to an instructional video. Thanks.

Is this a question? Click here to post it to the Questions page.

You must be logged in to comment.