Public Lab Research note


Repeatability

by viechdokter | April 11, 2016 22:26 | 739 views | 7 comments | #12959 | 739 views | 7 comments | #12959 11 Apr 22:26

Read more: stable.publiclab.org/n/12959


viechdokter was awarded the Basic Barnstar by dhaffnersr for their work in this research note.


viechdokter was awarded the Empiricism Barnstar by warren for their work in this research note.


I am not sure if this word even exists in the English language but well... I browsed this forum and found some notes by "stoft" about rigidness of the PLab spectrometer and proposals how to make the setup more stable. Because I had some stability problems, too, when I put together my PLab Spec. 3.0 I found the proposals and the comments very interesting. One point in this discussion was that results should be repeatable (even for different users with different spectrometers) so that for instance an oil pollution spectra database could make any sense at all.

As I said I am not a scientist, but this discussion urged me to find out how repeatable the results of my own spectrometer were. So I took my table lamp (IKEA style) and aimed it at the slit of my spectrometer and collected 4 spectra within 13 minutes. There was no additional light from any other source and I did not change anything in the path of the light. (I also calibrated them all with the same calibration spectrum.) Here are the results:

Repeatability_-_intensity_curve_comparison.jpg

In the first picture I stitched the four spectra together in Photoshop, lined them up and marked the peak wavelengths and the peak intensities of the R-,G-, and B-channels. The "waterfall" image shows that the light intensity of the bulb or of the camera sensor perception was not constant. (When I find out how to switch off automatic gain control of the camera I might repeat this experiment.)

Then I took the CSV files and looked for exact wavelengths of the peak intensities and marked them with their respective colours (R, G and B).

Repeatability_reds.jpg

Repeatability_greens.jpg

Repeatability_blues.jpg

For anyone interested in the whole data, here is the CSV for all channels:

repeatability_all.csv

So... what do we make of it? Well, the blue peaks happen at 461 and 470 nm. Thats 9 nm difference. Greens 512 to 518 nm, 6 nm difference. Reds about 579 to 589 nm, 10 nm difference. Average peaks 512 to 518 just like the greens.

As I said: I did not change the path of light. The only ways such differences could occur, could be a) a different temperature of the lamp over time (perhaps because of small changes in electrical current) b) an automatic gain control issue concerning the webcam c) both effects overlaying

Comments or further experiments welcome.


7 Comments

I think you may be right about the camera's auto gain control, your using a cellphone right?

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

Reply to this comment...


No, the cellphone I only use for the nice flower pix... ;-) For the spectra I use the Desktop Spectrometry Kit v3.0 with the JDEPC-OV04 webcam.

Reply to this comment...


Ok, then its the same camera that is in mine, very cool! Yes, stoft is correct on the stability issue, I followed his lead on that and took a lot of time to stabilize mine, it makes a world of difference on the repeatability issue.

Reply to this comment...


Hey, I noticed when I downloaded your csv file, the wavelength numbers are here but the RGB channel numbers are all zero's?

did you do a calibration for your initial spectrum?

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

Reply to this comment...


Look further down at about 420 nm. The fun starts there.

Reply to this comment...


This is great. I have been hoping someone would do a study of stability -- I'm curious, is there a discernable trend over time, as I've seen changes especially in the near infrared spectrum of a warm fluorescent lamp (with a filter-stripped webcam). I'm also curious about the stability over time of a halogen lamp and/or a Solux lamp as @stoft has used in his work. I'm thinking a timeframe of 2 hours in 5min intervals would be a really interesting dataset to see, and very important regarding our assumptions in gain correction.

Reply to this comment...


Sounds like I will have an "illuminating evening" tonight (as long as no model girls come along and tear my clothes off...). I don't have a Solux lamp (I'm not sure this brand is sold here in Germany) but I might try some different lamp types over the proposed 2 hours in 5 min steps during the next days. Will get back with the results as they come in.

Reply to this comment...


Login to comment.