Public Lab Research note


This is an upgrade for spectrometry. Try building it and report back how it goes.

Ebert

by cfastie | February 20, 2013 01:09 20 Feb 01:09 | #6061 | #6061

The video below suggests that I had a plan for this new spectrometer, but that plan changed every day as the device came together in the last couple of weeks. Stuff I found in the garage and at the hardware store altered the design, and my observations on the SpectroBench 2000 completely changed the final dimensions and angles. The angles above could still change, but I'm not planning on changing the slit-to-grating distance.

Below is the first calibration spectrum of a compact fluorescent lamp. I think I can get a little bit more resolution out of it, mostly by narrowing the entrance slit. We will see.

EbertCFLTrace1280

Ebert's first calibration spectrum of a compact fluorescent lamp. The data were downloaded from SpectralWorkbench.org as csv and graphed in Excel then overlain on the spectral image.

Here are some links to parts:

1000 lines/mm diffraction grating film

3/4" PVC conduit body

1" PVC 45 degree connector

Vacuum cleaner wand I am not sure the wand above is the same as the one I used.

Brooxes Better Camera Keeper


I made this Help out by offering feedback!


People who made this (0)

None yet. Be the first to post one!


15 Comments

It looks awesome. I look forward to trying lots of these tweaks!

Reply to this comment...


rather nice. I need to lift my game ! :)

Reply to this comment...


If i watch the video, will I know why the post is named "Ebert"?

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

Reply to this comment...


You tell me.

Reply to this comment...


Wow, i did not regret that.

Reply to this comment...


Good job. I am inspired to work on a similar project. I love the way you recycle what most see as junk! A man after my own heart.

I made a spectrometer for a digital camera a few years ago, but it was much bulkier than yours, and included a collimating lens. One thing I noticed with it was that the relative line intensities would change depending on the angle between the slit and the object I was trying to measure. To overcome that I added an imaging lens in front of the slit to make an image of the object on the slit, and that seemed to overcome the problem. Without the imaging lens, some of the CFL lines at the end of the spectrum could completely drop off. I do not understand why. Anyway, the thing was too bulky and I did not do much with it. Now I am motivated to incorporate some of your fabrication methods to try again.

Reply to this comment...


Chris, How did you overlay the spectral image on the Excel graph?

There is another way to extract data from the camera image: Use a software that I think is variously called "NIH image" and "ImageJ". With imageJ one can draw a line or rectangle across the spectrum and then ask the software to plot a profile that can then be copied and pasted into Excel. Using a known spectrum like a CFL you can then do a pixel vs. wavelength calibration. So, if you don't have access to spectralworkbench you can use this technique. ImageJ is freeware.

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

Reply to this comment...


Tom, How did the CFL lines drop off in your earlier spectra? I have noticed that my spectral images taken with P&S and DSLR cameras do not record a strong signal for the mercury line at 405 nm compared to images made by the Public Lab webcam spectrometers. I hypothesized that either 1) my Nikon and Canon lens coatings were blocking UV too vigorously, or 2) the CCD sensors were not very sensitive to that end of the spectrum compared to the CMOS sensors in webcams or 3) both.

Good to know that ImageJ can do this easily. We use that for registering normal and NIR images from the Public Lab infrared camera and making new multispectral images.

Overlay graph: 1) Make a simple Excel graph with bold lines, no gridlines, no y axis, etc. Make the line color what you want and the background some other color. 2) capture the graph image with the Windows Snipping tool (in Accessories), save as png. 3) Open in Photoshop, use the background eraser tool to remove the background color. This tool has settings that make this easy (e.g., "discontiguous"). Save as png with transparency (“Save for Web and Devices”). 4) Open spectral image in Photoshop and “Place” graph png file over it. Resize to match. 5) I added a drop shadow to the graph because some of it was not so visible.

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

Reply to this comment...


Chris, I also noticed the 405nm line was all but absent in your spectrum, and that is one reason I brought up my previous experience. I came across that data recently, but now I cannot find the file where I recorded it. What dropped out for me were some of the red lines, but in some recent configuration studies I saw the 405 line go away. In that case I was using the reflective part of the dvd (when you split the dvd one piece is reflective and the other has no reflector on it but they both have grooves). I am really not sure what is going on here. In the experiments with my DSLR in 2008 my best recollection is that it had to do with whether the slit was illuminated by the light bulb straight on or somewhat at an angle. You might want to play with the illumination angle and see what happens.

Thanks for the info on the overlay.

Reply to this comment...


How do you calibrate anything? I haven't figured this out yet. I don't have 3 blue wavelengths to choose from...I don't know which other green wavelength to pick. I tried though. Other spectra that I do that have blue wavelengths that show up in the ultraviolet range even though I tried to calibrate the bright Hg at 435. I don't get it. Maybe the rewritable DVD gratings that I used are off? I'm using the foldable and an iphone so I don't expect perfection but I can't even get started.

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

Reply to this comment...


I should have also said that I don't understand this website. It's all over the place. Is there one of those standardized forums you could use instead?

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

Reply to this comment...


Hi Spectralod, Your long tube CFL spectrum is calibrated correctly. I am not sure why you are not getting more distinct peaks with regions between them which are dark. You should be able to get better resolution by making your entrance slit a little narrower. It might be that your camera is not focusing very well. But I found that an iPhone 4 was able to focus really well all by itself, although I was using 1000 lines/mm film instead of a DVD. It could also be that the image is over exposed. I got a free app that allowed the iPhone exposure to be set for a small area of the image where you touch. That helped some. If you embed that spectrum in a post at the plots-spectrometry Google group, maybe someone will recognize the problem. Good luck

Reply to this comment...


Do you have a data source for the intensity, pixl, wave length for the CFL in this image. Pictures are beautiful. I have a similar setup using some 1000 line/mm film a Canon A620, plumbing tube and fittings and razor blade for slit, and shopping bag for diffuser, some black construction paper, Ubuntu, gphoto2 (some scripts) and ImageJ. I've found a way to identify the peaks in the data in Excel / Open Office but I don't have any calibrated CFL data points to compare to. I've been using the same camera and Ubuntu to take astronomical photos (mainly for surveying).

Reply to this comment...


Sunwukong, You can download the data from any spectrum at SpectralWorkbench.org. So at this page for one of the nicer CFL spectra from Ebert, you can get JSON, CSV, or XML data for the intensity per wavelength. You can get the expected wavelengths of the peaks in a CFL at this Wikipedia page. I tried to correct this spectrum to better match the expected wavelengths with a third order polynomial which you could apply to the data.

There is a new macro at SpectralWorkbench which identifies peaks or valleys in a spectrum. It can be run on any spectrum at SpectralWorkbench. The variables in the macro might need to be manually tweaked to get the best results.

Reply to this comment...


Sunwukong - I would love to see a pic of your setup; is it connected to a telescope?

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

Reply to this comment...


Login to comment.