Public Lab Research note

Does leaving the dye on a DVD-R grating make any difference?

by MrBumper | February 04, 2015 14:00 04 Feb 14:00 | #11552 | #11552

In a previous research note - Preparing a DVD-R to act as a diffraction grating I described how to remove both the aluminium reflective silvering and the blue/violet photosensitive dye from a DVD-R to produce a section that could be used as a 'clear' diffraction grating.


A section of DVD-R with the dye removed next to a section with the dye still present

Having just completed construction of the Public Lab Spectrometer 3.0 kit I decided to do a quick test to see if I could produce results that would show that the dye has a significant effect on the response of the spectrometer. Using the same light source and the 0.2mm photo printed slit, I captured two plots for comparison. One plot using a grating that still had dye on it and then a second plot using a grating that had the dye removed. As far as possible all the other conditions were unchanged.


A section of DVD-R grating held with a small butterfly clip and one of the chrome wire levers shown removed. For convenience, the black card shown happens to be a slit frame, standing in for the grating support.

Rather than use the double sided tape to stick my gratings to the grating support in the Spectrometer 3.0, I used a small office butterfly clip to clamp the grating to the support. This makes it much easier to adjust the grating for angle, or to swap a grating section with the dye, for a section with the dye removed. The butterfly clip has a folded spring steel body with two chromed wire levers to open the jaws of the body. Once the clip is put in place, the two wire levers can each be squeezed width-wise and removed, leaving just the low profile body behind.

One difficulty with trying to capture two plots from a spectrometer for comparison, is that ideally a wide band light source with a fairly smooth response might be used, so that differences in amplitude across the entire range of wavelengths would clearly show up. I don't have access to such a light source, so I stuck with the old standby, the fluorescent lamp (a cool white) as used for spectrometer calibration. It is at least a familiar source and does emit light across the visible spectrum, albeit in a rather spiky fashion.


Guvcview settings used for the comparison. First factory default was selected, then WB auto off and Exposure manual mode were selected. Sharpness and Backlight comp were set to zero, WB to 5,000. Gain and exposure were set to produce an image just at the point of clipping.

Fortunately the new camera supplied with the Public Lab Spectrometer 3.0 supports the control of Gain and Exposure. I used the Guvcview software controls to adjust the exposure for my two test captures so that the dual green peaks were just clipping in Spectral Workbench. The data for both captures were exported from Spectral Workbench as a .CSV files and imported into LibreOffice Calc to be plotted as an XY Scatter graph. A scaling factor was applied to the data for the 'no dye' capture and this was adjusted so that the green peaks for the two data sets were the same amplitude.

The test result and conclusion

If the dye makes a significant difference what you would expect to see is that the purple/blue dye would reduce the levels of the light from the green and red phosphors in the fluorescent lamp, relative to the blue and violet. This should be demonstrable by making two plots of the light from a fluorescent lamp, one using a grating with the dye and one without, with the exposure adjusted so the longer wavelengths are at the point of clipping. When the plots are compared, and the longer wavelengths equalised by scaling, it should be apparent that the violet and blue peaks for the plot made using a grating with the dye removed are reduced in amplitude, compared to the plot made using a grating with dye remaining.

The test produced two plots that are obviously not the same. The plot made using the grating with the dye has a clear bias towards higher amplitude peaks for the shorter wavelengths and a smaller amplitude peak for the red phosphor at the long wavelength end of the graph. The clearest difference is for the violet peak at around 400nm, which is twice the height of the violet peak for the plot made using the grating with the dye removed.


The plotted test results

My conclusion from this test is that leaving the dye on the grating made from a DVD-R results in a significant distortion of the spectrometers amplitude response. As I eventually found, removing the dye by soaking fragments cut from the bottom layer of a split DVD-R in warm, soapy water, is a simple and safe procedure and I believe is well worth doing.


The two curves also show a small difference in wavelength calibration. At its worst this difference is about 3 to 4nm. This is probably due to the accuracy with which it is possible to select the middle blue and the green peak during the calibration procedure. The wavelength calibration procedure for Spectral Workbench was probably arrived at using much earlier versions of the spectrometer with possibly lower resolution. In my plots I see that the green peak from a fluorescent lamp appears to be two peaks rather than a single peak as shown in the calibration procedure (or this may be a peculiarity of the phosphors in the lamp I used). Also, with the filter bias of the DVD-R dye removed, the most violet of the three blue peaks is much lower in amplitude and can be hard to see, which can make identifying the middle of the three peaks more difficult.

Adjusting the exposure so that the green peak just clips can also cause the green peak / peaks to be distorted, again making it difficult to pinpoint the exact peak for calibration.

A Re-test

Following discussion in the comments section after I first published this research note I decided to repeat the test. This time around I was extremely careful to adjust the exposure so that the captures were right on the point of clipping. This meant I could capture sharper peaks for the calibration points. Tthe double green peak was much clearer and I could select the sample points for the calibration procedure with greater accuracy. I chose to use the longer wavelength of the two green peaks for the calibration for both captures.

The resulting graph I think still supports my conclusion that leaving the purple/blue dye on the DVD-R diffraction grating will significantly distort the spectrometer amplitude response by attenuating the longer wavelengths. It shows just how careful you have to be in adjusting the exposure precisely to the point of clipping because, in this retest, the red end of the graph shows much higher amplitudes for the 'no dye' plot than in my first test. I did allow rather more clipping for the original test and it seems this limited the response more than I realised.


Note - I used a Crompton 22W Cool White circular lamp for all my testing. It uses Crompton's SPECTRA-PLUS Tri-phosphor, is rated at 4,000 K CT and has a colour rendering index of 81. To my eyes it is not a particularly cold/blue light. It is not a new, straight out of the box, lamp. I use it as my everyday workbench lamp.


This is a really great experiment.

It looks like the distortion is a squeezed distortion, where the x-axis is shrunk around a point. That seems more like the result of a physical difference in how the diffraction grating was squeezed. You might want to repeat the experiment two more times and plot all six spectra together (three lines with one color, three lines with another color). That'll give you some visual idea of how much physically manipulating the grating distorts the reading.

The amplitudes would be the most likely way in which the dye would distort the results. You can see that the amplitudes in the shorter wavelengths are more diminished with no dye in the low wavelengths. I have no idea why the dye would cause an increase in low-wavelength diffraction. If anything, I would expect the purple dye to absorb, and thus diminish the amplitude of, the very low and very high wavelengths (red and blue).

Reply to this comment...

The "squeezing" referred to in the previous comment is, I believe, actually a result of the way that Spectral Workbench uses the two reference points and then scales the X axis accordingly. I believe also that the variation in the X axis calibration is due to the variability in accuracy of the Spectral Workbench calibration procedure. The grating in my rig is not bent in any way and in fact I do not have any mechanism in place to intentionally bend the grating. Repeating the capture several times and trying to plot the results together could well only increase confusion because of the SW variation in calibrating the X axis.

The way I interpret the results is that the dye does adsorb the longer wavelengths. Rather than scale the curves so that the green peak for both was the same I could have scaled the 'with dye' curve so that the amplitude peaks for the shorter wavelengths were the same. The longer wavelength peaks for the 'with dye' curve would then be lower in amplitude than the long wavelength peaks for the 'no dye' curve. As I say the variation between the curves in the X or wavelength axis is, I think, just due to variations in the precision of the calibration.

Reply to this comment...

Good point on the potential for software to warp the x-axis.

"Repeating the capture several times and trying to plot the results together could well only increase confusion because of the SW variation in calibrating the X axis." You're right. It could absolutely increase the confusion. That's called discovering how much error is caused by unmeasured factors, such as physical distortion or Mapknitter x-axis mapping. If the results are too confusing to comprehend, then we cannot scientifically say there is any measurable difference between the two causes (dye and no dye). That is scientifically relevant, and an important component of experimental design and analysis.

Reply to this comment...

OK but my intention was not to measure inaccuracy / repeatability in the wavelength measurement, or highlight any errors in the Spectral Workbench calibration procedure. It was rather a fairly quick test to see if I could measure a difference in amplitude between 'dye' and 'no dye'. Since there is a variation between the two test curves in both amplitude and wavelength it makes interpreting the result more difficult than if the variation were in amplitude alone. Unless you just ignore it. In any case, over the visible range, the wavelength error is really quite small. It is under 2%. It just looks obvious when plotted. In retrospect I could have repeated the test several times. As it happens I didn't, although now you have suggested it, I might. I suppose I could just average 3 plots for 'with dye' and 3 plots 'no dye' and still end up with two curves to compare amplitude. I think I have documented all the conditions well enough that perhaps someone else can repeat the experiment to see if they can duplicate the results. I'm more than happy for someone to do that.

Reply to this comment...

I agree that looking over repeats will be difficult. I think it is worth it to draw any conclusions.

As much as I like your experiment, the lack of repeats means your single experiment and single result could be completely anecdotal. The perceived change could be due to unaccounted measurement error rather than the seeming controlled factor of dye or not. That's all I'm getting at.

Reply to this comment...

this is awesome!

there should be two green peaks. we need to re-work the calibration procedure now that our equipment is sensitive enough to see them.

Reply to this comment...

This is a great discussion, and of a type that is rare on this site. Usually the question being asked, the methods used to answer it, and the results of the observations are not articulated well enough for anyone to know whether the conclusions are valid or not. In this case, Terry described a question that arose from an observation he made, then formulated a relevant hypothesis, then devised a method to test the hypothesis, then performed the test and found that the result was consistent with his hypothesis. All this is clearly presented in this note. Bryan is correct that we don't know how much variability there is in the system used to make the test, so we don't know whether the same result would arise if the test were repeated. Bryan thinks that is Terry's problem, but I think the problem belongs to Bryan (and to the rest of us). Unlike most Public Lab authors, Terry has presented all the information needed to reproduce his test. Until someone does that, I consider Terry's result to be the scientific truth. If we have a hypothesis about some alternate explanation for Terry's result, it is up to us to devise a test and perform it. Isn't that the way it works?

Fleischmann and Pons published one paper on cold fusion and sparked hundreds of attempts to reproduce the result. Another paper by Fleischmann and Pons was the last thing we needed to find the truth.

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

Reply to this comment...

Thank you very much. I'm sorry if my initial response seemed overly protective of my results. I hope I don't wind up like Fleischmann and Pons ! I have now repeated the experiment and this time I was super careful about adjusting my captures to be just on the point of clipping, which gave me a clearer view of the green double peak and allowed me to more precisely choose the longer wavelength of the two. This has refined my results. Although this second test still supports my hypothesis about the effect of the dye it shows an even greater increase in the red wavelengths relative to the shorter wavelengths when the 'no dye' grating its used. With the dye and switched to the component colour view I could see the blue and green tending to clip, then with the 'no dye' I could see green and red tending to clip.

Reply to this comment...

replication! nice! I want to try this out too.

I'm going to throw a binder clip in the kit- great idea!- much better than taping it down for re-adjustment and playing around like this.

Reply to this comment...

Following my repeat experiment I have been thinking about the IR response of the PL spectrometer. I haven't been that concerned about it because it was not of immediate interest to me. I haven't tried removing the IR filters from any of the 3 spectrometers I have now built. However given the even more obvious attenuation of the low frequencies in my most recent plots I do wonder if the DVD-R dye would have quite a drastic effect on IR. Of course it might not, the dye might act as a notch filter and let the lower IR through, rather than as a high pass filter. I suspect though that it acts as the latter rather the former. Matthew - does the camera shipping with the Spec 3.0 have the IR filter removed at the factory? Or is it still in there.

Butterfly clips rule!! (or rather they clip)

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

Reply to this comment...

No, it doesn't have the filter removed. We've done it in the past, but when it gets down to it we haven't found the IR response to be good enough to justify the step of removing the filter, really.

that said, its easy to pop out-- just unscrew the lens, its on the bottom.

Reply to this comment...

Thanks Matthew, that's good to know. I suppose the IR response might be better if both the lens filter and the dye on the grating weren't present. My second test plot shows more than double the amplitude for the 570 to 630nm yttrium oxide(?) red phosphor peak from the fluorescent lamp. Which might mean that there is a healthy response even lower down.

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

Reply to this comment...

Login to comment.