This looks interesting!

From a simple spectrometer like the PublicLab spectrometer I would expect the intensity results to be lower than from a professional spectrometer; but the shapes of the graphs are what I think are interesting. I do not know how to smooth the graphs from your PublicLab spectrometer. Perhaps Jeff at PublicLab can help with the smotthing.

Guillaume8507

Hi, @coight! This looks like a spectacular (for real!) project and I'd love to learn more.

There are two ways of smoothing that I know of. One is "time-averaging" which can reduce noise and preserve finer features, and that's available via the "More tools" menu in the current version, as in the screenshot below:

There's also "rolling average" smoothing, which instead of averaging repeated spectra over time, just averages neighboring pixels along a single line of data. This is fast, and useful for some things, but does destroy data and smooths indiscriminately over both noise and finer non-noise features. This is going to be featured in the new v2.0 which we're hoping to launch a preview of this week.

Also, are you open sourcing your rover design? Do you have more pics of how you've attached the spectrometer and details on how youre operating it?

Doing division between sample and light source sounds promissing. How exactly do you shield the sample from outer light? Do you snapshot both in one picture, or do you take pictures of white reflected light source and sample reflection one after one?

the preview of Spectral Workbench v2 with subtraction is now live -- i'll be posting about it shortly. @ThomasS - how would subtraction and division of two spectra differ, as methods?

I do think simultaneous measurement is probably best -- so you know you have as close to identical light, and at least identical camera settings.

Hi all,

thanks for comments. I have added a word doc outlining my methodology and initial results. I am unsure how to divide spectra of a sample by spectra of a white card reading from a light source using Spectral Workbench. I have had to use another program for this. Also how can you export a spectra as a .csv or excel plot from Spectral Workbench? This is critical for my work. Additionally I cannot take simultaneous measurements - this is not an option in the field. Is there another way to do this in Spectral Workbench?

Spectrometer laboratory setup Fig. 1A illustrates the design of our spectrometer. Our spectrometer was based on the design used for the Public Labs webcam instrument and consisted of a 20 X 20 X 150 mm balsawood box containing a ~1 mm slit. We used a section of DVD for a diffraction grating. We had previously trialled acetate diffraction gratings though found that these warped and lost integrity from thermal expansion and contraction in outdoor environments. This issue adversely affected the ability to calibrate the instrument. We found that DVDs were much more resistant to temperature changes, facilitating greater stability in spectral measurements. The diffraction grating was mounted at a 45o angle to the incident light in order to align the diffracted spectra with our imaging device. In liu of a webcam as suggested by Public Labs we used a Raspberry Pi NOIR camera with a pixel resolution of 2592 x 1944. This camera uses a Samsung Omnia CCD chip which has the IR cut filter removed. This allows the camera to capture visible and NIR wavelengths in its red, green and blue channels. We chose this camera because of its higher imaging resolution as compared to a webcam, greater ability to control exposures, as well as its ability to be controlled by the Raspberry Pi. The latter point was important for us as we intended this spectrometer to be used on small ground robots, such as our Little Blue that was trialled in Arkaroola in 2014. Using the Raspberry Pi enabled a light weight system that could be controlled wirelessly. The interior of our spectrometer was painted matte black in order to minimize stray light reflecting inside the instrument. Test setup Our instrument was intended to collect reflected spectra from materials of interest in order to assist in identifying mineralogical or biological composition. We thus designed two test scenerios, one indoors using a controlled, tungsten light source (Fig. 1B) and outdoors, using unfiltered sunlight (Fig. 1C). Fig. 2A-B illustrates our two spectrometer test setups. For the indoor setting we used a Miniblitz studio flash for our controlled light source. We used the tunsten-based photomodelling light on this flash unit as our light source. We desired a continuous spectrum light source capable of transmitting visible and NIR wavelengths suitable for analysis with our spectrometer. Our light source was placed at a distance of 25 cm from the sample and set at a 90o angle. The spectrometer was placed ~ 5 cm from the sample to prevent spectral mixing from the surroundings while precluding the spectrometer’s shadow falling on the sample. In the case of the outdoor setting illumination was achieved using direct sunlight on a clear day with no cloud cover Following a 2 minute interval to allow both the photo lamp and camera ccd to warm up and stabilise, capturing of spectra was achieved by commanding the NOIR camera to take a photograph of the sample, and also an additional photograph of a specular reflector. Additional images were taken with the spectrometer slit covered in order to capture noise generated by the camera electronics, or dark frame current. This process enabled calibration of the spectral amplitude, obtaining the reflectance spectra of the sample, Rs, via the following formula: Rs = (Im – Id)/(If – Id) (1) Where Im is the raw captured spectra of the sample, Id is the dark frame current and If is the spectra obtained from sampling the specular reflector lit by the photo lamp. In the case of the outdoor environment If is the spectra obtained from the specular reflector lit by sunlight. Calibration We used a compact fluorescent lamp (CFL) in order to undertake wavelength calibration of our spectrometer We noted the spectral range of our instrument precluded identification of all peaks emitted by the CFL, thus after waiting 2 minutes for the CFL to warm up and stabilize, we captured its spectra and used the peaks listed in Table 1 for calibration.

Colour Origin Peak wavelength (nm) Dim blue Mercury 436 Turquoise Terbium 485-490 Green doublet Terbium and Mercury 544, 546 Red Europium 611 Table 1: CFL spectral lines used for wavelength calibration The spectral sensitivity characteristics of the NOIR camera is not available for public release by the manufacturers. We thus used the If derived from sunlight to infer spectral sensitivity of each of the camera bands. The spectra of unclouded sunlight is well known and enabled determination of our camera sensitivity. 2. Results

For some reason I am unable to upload any file here - I have tried eight times without success. I have pasted the text for the project above - no images sorry.

Hi coight, sorry, I think it doesn't allow word docs for upload. Do you have a PDF or can I help you convert it? jeff@publiclab.org