Public Lab Research note


Reproducibility test of data using an IR lamp

by viechdokter | | 1,044 views | 4 comments |

Read more: unstable.publiclab.org/n/12962


This was my first in-depth data reproducibility test with the PLab spectrometer 3.0. I used a Philips 250 W infrared lamp, let it shine (let it shine, let it shine...) for about 2 hours and took the "spectrum" roughly every 5 minutes. You can see the setup on the photo. I had an absolutely dark room without any stray light. Oh, you might stumble upon the time gap in my data... well, let me tell you: it is no good doing an experiment while watching StarWars...

However, you can take the CSVs from the spectra named "Infrared lamp 1 ...24" if you want to have a closer look at the basic data. I only used the peak values for average, red and green, stitched together a an Excel worksheet and saved it as pdf and jpg. Here are the results:

reproducibility_test_IR_3.jpg

reproducibility_test_IR_3.pdf

The peak wavelength is stable within a 7 nm range.

The average light intensity seems to change a bit while the lamp gets warmer but then stays pretty much the same.

The reds and the greens (yes there were greens!) peaked at the same wavelengths.

The red intensity goes down while the lamp heats up but after a while it stabilizes.

The values in the CSV file are not in %!

So far it looks good for reproducibility of data, although one should repeat the test on another day with absolutely identical setup to be sure. Even better if different guys in different places with different spectrometers got the same results while sharing only the same ONE lamp for their experiments.

The next reproducibility test I will perform with a CFL lamp because there are thinner lines to detect.


4 Comments

Wow, this is so great to see very close to @stoft's post: https://publiclab.org/notes/stoft/04-14-2016/spectrometer-stability

Were you able to assess intensity stability over time? See @stoft's post for his approach to measuring intensity stability.

Great post, thanks!

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

Reply to this comment...


Yeah I had a look at stoft's post and was very impressed about his orderly and organised way of doing things. I built my spectrometer 3 weeks ago and was basically playing around with it at first before I wondered how good a cardboard apparatus really can be and how reproducible it's data really are. I must admit, so far I am pretty impressed by this low-cost science. The velcro design did not look too promising at first, because the tension on the webcam wire lifted the webcam up a bit thus changing the angle between dvd grating and camera. The double sided sticky tape also did not hold the camera in place, so I used single sided sticky tape and wound it around the webcam base. Then it bworked and I got my first spectra. And then I wanted to find out more. I read some of stofts posts about stability issues and started my own tests. How does the RGB thing work, what kind of sensors are we using, where is always the yellow gone...? One problem is I don't have many lamps to choose from at the moment, esp. not a super-stable Solux. But although I had a 7 nm peak wavelength "wobble" here in IR spectra over time, I was very impressed by the CFL lamp test with almost matching curves:

cfl_repro_results.jpg

But I must admit that I was more looking for the peak wavelengths than for the intensities because the places of the peaks usually are the fingerprints (for material spectra) not the heights.

I will have another look at stofts statistics to see if I can evaluate my data more in-depth. And now that the time-lapse capture macro is working I might do some more tests.

What I would like to find out one of these days is, how often we have to re-calibrate the spectrometer - every week? Once a month? Only once? The longterm drift in wavelengths over weeks or months would be interesting. But that would mean to set up the thing and leave it untouched for the whole test period. Absolutely stable environment conditions (no displacement of the apparatus, regular light source pre-heating, light ray angle stability, stray light prevention, vibration prevention, temperature stability and so on...) over weeks to capture maybe one spectrum a week or so. Then calibrate all the spectra with only one calibration spectrum and look at the differences. I guess it could be interesting for the oil testers here, too. And of course, it would be great if anyone who can get hold of a Solux could re-do stoft's tests and statistics to see if the results are the same in Australia or the Phillipines. Because thats exactly what would make a oil pollution database workable: different spectrometers producing the same "fingerprint evidence". Perhaps you could even exchange or pass around some oil samples to as many spectrographers here in the forum as possible and compare the results.

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

Reply to this comment...


Great questions. I do think @stoft is mostly interested in intensity stability for the moment but both are good questions.

I'm interested in whether the stability of the solux is unique or if all incandescent bulbs are that stable. Or if the 12v power source @stoft is using is relevant.

Reply to this comment...


The last point would be easiest to test: test a no-brand lamp with the same setup and the same power source and do the same stats on the results.

Reply to this comment...


Login to comment.