We have made a CCD detector based spectrometer but are facing a problem while taking the absorption spectra.
We are using tungsten lamp as source and stored its spectrum as Yref. Next we placed a transparent acrylic sheet in front of the source and stored the spectrum as Ysamp. We tried to plot both y1=Yref-Ysamp and y2=Ysamp/Yref.
We should obtain a flat horizontal line for absorption spectra of transparent material (since the absorption is almost equal for all wavelengths). But instead, in both y1 and y2 plots we see the trend of the source's spectrum i.e. a curve similar to the source's spectrum curve.
I have attached the excel files for your reference.
The hollow cathode Tungsten reference source, driven by a biased dc current should be focused by a system of lenses and diaphragms to a beam of narrow diameter for the OAS.
You can use a shutter for cutting the source signal in order to collect a signal from the plasma only, which can allow both emission and absorption measurements to be performed simultaneously for the same conditions.
From the detector side, you may add or use one or two additional diaphragms to reduce the beam diameter as much as possible.
better use quartz window port if possible.
Good luck
With the sheet in then the transmitted intensity or power is lower. Thus the background subtraction is not comparing like with like. You need to subtract the data in line with the reduction in intensity or power.
Compensation of the slope of the illumination is not unusual. If you use a "normal" spectrometer during the variation of the wavelength the hard and software also perform corrections so that the illumination seem to be "horizontal". You might try to mathematically solve your problem (I am not able to evaluate the data) as performed in our study.
Article About the possibility of calibrating optical detectors by so...
Hi Shubham, Can you clarify the data in your excel files? Is the unit of x-axes in wavelength (nm)? Is the y-axes the absorbance (or transmittance? or some unit of intensity)? I ask because, assuming it is absorbance as a function of wavelength, the "source" data (Yref) does not appear to correspond to typical tungsten output profiles (or maybe I am missing something?).
Again, assuming these are wavelength vs absorbance data, the Ysamp data indicates less absorbance than Yref at certain wavelengths, which doesn't make sense. Clearly I am misinterpreting your data, so can you explain?
OK, interpreting the X-axis for "pixel no." - not knowing about the y-axis units:
You have little attenuation with a lot of noise (negative differences should not occur at all but are prevalent). Though the curve is looking somehow "dramatic", regarding the absolute values of the differences the attenuation is close to zero and close to the noise floor. Thus the dramatic look.
What can you do?
Regards
In your case:
Absorbance = log (Yref/Ysample),
according to the definition A= log ( Io/It.)
If the sample is a transparent sample (e.g., an acrylic cuvette) you obtain the baseline.
See attached file. What you see as Io or It is a combination of spectral energy and detector sensitivity.
Try with a coloured solution. Use its absorbance to calibrate the wavelength axis.
I am interested to know more details about your project.
Hello Everyone
Thanks for your replies.
I should have mentioned clearly that x-axis is the sensor's pixel number and y-axis is the output voltage (which corresponds to intensity of light).
One point that I feel might be a mistake is that I have not subtracted the dark noise from the spectrum (you can see that graph begins from 1 not 0) and thus the result of the division operator ("transmission") is not correct.
To know more about the project please see the attached links.
https://www.researchgate.net/post/Would_you_buy_a_cheap_350_good_resolution_Vis-NIR_spectrometer
Dr. Michael, I am searching for amplitude calibration for a long time and have found these (please refer to the attached links).
Will be trying them shortly. Any suggestions would be appreciated.
http://eprints.lancs.ac.uk/6736/1/inproc_326.pdf
https://jethomson.wordpress.com/spectrometer-articles/system-function/
You have an interesting project. A dynamic linear regression of the data points will also reduce the error (origin statistic noise).
To be honest for my applications both approaches contains too many assumptions and are not “experimental” or compressible enough. Additionally our approaches calibrate the fiber optical receiver (the spectrometer and the periphery) so that the number of photons and their wavelength is known at the sample position. This is only possible using a broadband well-known and monitored reference light source (like the sun).Then you have to use another light source if you exchange the fiber or the collimator lens.
Thank you Dr. Schmitt. I am actually going with dynamic linear regression of the data. Will keep you updated on this post.
- Badges
- Science method
- Similar topics
- Methodology
- Absorption Spectroscopy