[argyllcms] Re: DIY reflective spectro
- From: Stephan Bourgeois <strangelv@xxxxxxxxxxx>
- To: argyllcms@xxxxxxxxxxxxx
- Date: Thu, 09 Mar 2006 16:09:25 +0000
Roberto Michelena wrote:
However a DIY reflective spectro with XY table would be most welcome :)
I'm sure the parts list would add up to less than $500, for a device
that usually sells over $2500.
I actually started some research on a DIY spectrometer. As a
proof-of-concept prototype I did the following:
The optical section was a replica of cardboard spectroscopes available
for schools.
* card 0.5 mm slit
* holographic diffraction grating 1000 lines/mm
* acrylic collimating lens +9.4 dioptr. 106mm diam.
* baffles made of foamboard coated with black card
lens and diffraction grating are available from www.astrodienst.de
Sensor
* Philips Toucam Pro II webcam with standard lens. This webcam uses a
Sony HAD 640x480 colour CCD
* linux drivers and control (setpwc, dov4l)
Camera settings: 5fps open shutter, no noise reduction, no compression,
sharpness 0, gamma 2.2, gain 12000 (0..65535)
Light source
* tungsten halogen + Lee Filter Full CTB (tungsten to daylight)
The filter is *not* an attempt to make the light source spectrum similar
to daylight, (I hear laughter at the back), but to avoid using excessive
blue gain (and therefore noise) on the camera.
Calibration
* wavelength calibration done with a compact fluorescent lamp. These
lamps show very strong spectral lines common to fluorescent tubes
* intensity calibration with a white card (Now you can laugh!)
At this stage, the limitations of using a colour CCD sensor become quite
obvious. The response curve shows significant dips at the wavelengths
where RGB filters overlap. I didn't know how to calculate how these dips
affect the accuracy of the instrument.
Processing
* capture with xawtv, processing with ImageMagick, reduce png to 32
values in a text file
* final calculations are done in a spreadsheet (OpenOffice Calc) to plot
spectral response and calculate XYZ values
* XYZ values were converted to sRGB using Bruce Lindbloom's formula and
observed on a computer monitor. Which means very little because the
monitor is not profiled and the sample is not observed under a proper
light box! However, this was useful enough to point me to obvious errors
like gamma correction.
Conclusion
* cost: lens 1.50 Euro, grating 4.5 Euro, webcam £ 60, foamboard+card £10
* the setup is definitely interesting as an educational project, however
highly unpractical to measure a large number of samples
* After this experiement I bought a LED colorimeter (Colormouse)
* It would be interesting to evaluate the accuracy of the instrument and
compare it with the Colormouse.
If anybody is interested, please let me know. I dropped the project
because I felt very much on my own. In any case I hope you enjoyed
reading about this!
Stephan.
Other related posts: