8 replies to this topic

[Ayaan Hashim]

Y'all know that a high end spectrometer costs 5000pounds+.

We're making a rubbish one with a CD and a webcam which also is adequate for a barbeque of one's fingers.

Now we're having trouble with Theremino and we're tryin to use a software developed on instructables, files on github.

I don't know how to read this graph-

https://www.instruct...t-Spectrometer/

Any help would be appreciated ASAP. We have to complete in 2 days, but best to show the exhibit tomorrow.

Thanks in advance

Ayaan and a buncha geeks

 

 

[BGRE]

Without calibration of the relative spectral response of the red green and blue pixels the spectrum will be distorted. A source with a known wideband spectrum such as a quartz iodine lamp or perhaps sunlight may suffice for calibration.

[Ayaan Hashim]

Thanks from 4 geeks.

Anyone knows how to calibrate it?

[BGRE]

To get the vertical scale correct use something with a known fairly flat spectrum such as a tungsten halogen lamp.

To calibrate the wavelength scale use a source with known wavelength such as a green DPSS laser pointer.

[Ayaan Hashim]

Now the webcam is showing a greensih tint in the feed. Has the bayer matrix burnt out or what?

Thanks

[Mike I. Jones]

Image mercury and low-pressure sodium lights.  Nice bright, narrow spectral lines with very precisely known wavelengths.

[hamishbarker]

if that's a fluorescent lamp, the blue line is probably 4358 Angstrom and the red line is probably about 6113 Angstrom. Scale and translate the horizontal axis to get this. then try a solar spectrum. What are you using for slit? the spectrum seems rather out of focus. it's probably too close to the webcam - make the enclosure between slit and grating longer, and adjust the webcam focus if possible.

 

there are two calibrations to make : wavelength (i.e. what horizontal pixel value corresponds to what wavelength) and photometric response. 

 

the first calibration is pretty easy to make a rough calibration. Once you have done that, try taking the spectrum of an incandescent, tungsten bulb. If you can find a new one, look on the box for what colour temperature it is rated at (usually typical values are 2700-2900K). Don't use an LED or compact fluorescent bulb.  The spectrum of a tungsten filament bulb will be fairly close to a black body spectrum of the rated colour temperature. you can get the curve here:

https://lampx.tugraz...body.php?T=2800

 

note that webpage shows wavelength in um (microns) which is Angstroms / 10000.

 

note that at 2800K, the output is getting very low at wavelengths shorter than 0.5micron (5000 Angstrom).

 

If you have wavelength calibrated your tungsten lamp spectrum, then you divide the values by the theoretical blackbody curve.  that gives you the response function for the spectrograph - how sensitive it is to each wavelength.

 

Then if you take spectra of other objects, you wavelength calibrate them, then multiply by the response function to get the photometrically calibrated spectrum of the new object.

 

taking a spectrum of a star means getting enough light from the star through a  telescope and focusing it onto the slit of your spectrograph and keeping it there. if you don't have a slit on the spectrograph, then wavelength calibration is more complicated, and also more background light contaminates your target spectrum.

[gr5org]

The guy in "instructables" used a CFL light that you can see there.  The left of the double green line is right at 546nm.  So if you have a similar spectrum from a CFL then you can pretty safely use that.  But it has to be mostly identical and include the same strong red line (plus weaker lines) and the same strong two blue lines.  Warning - every flourescent light uses different chemicals (I think Yttrium is popular now?  Used to be mercury which was easier) and different phosphors but many are similar so if you post a spectrum I can probably tell you exactly what the wavelength is of most of the spikes.

 

You can use the sun (just use the sky or a cloud or something white on the ground - don't actually point it at the sun) - it has dark lines - much harder to see but they should show up in the computer plot.  The most prominent ones are:

393, 397, 434, 486, 517, 527, 589, 590, 656nm

 

If you post a plot I can tell you which ones are which.

 

You can use a laser.  Most lasers say what their wavelength is.  Typical red laser pointers - those are all "650nm" but in reality closer to 655.  They can be anywhere 650 to 660nm.

 

Actually using a red laser is a great idea for a first estimate.  You really need at least two points (2 colors) to calibrate properly.

[gr5org]

I'm not sure what you were asking though.  So in the graph it is just showing the result of the camera RGB colors and then the yellow line on top is the total brightness (the sum of the 3 colors) and that's really the only line you should look at.  The scale along the bottom is arbitrary and is just distance in pixels which need to be converted to wavelength in nm by knowing what wavelength some or all of the peaks are at.

 

If you post a wavelength of a source (such as a CFL or a laser or sunlight) then I can probably tell you the wavelength of some or all of the peaks and you can use that information to calibrate your spectrometer.

 

I have a decent spectrometer (only accurate to about 1 or 2 nm).

 

You can also look up the values for each chemical.  For example that "sun" line I mentioned above at 656nm is also known as the "hydrogen alpha" spectral line and is actually at 656.281nm.  The one from the sun at 527 is easier for me to see and is from Iron.