An emission spectrum in IR wavelength region consists of interference fringes (after spectral correction). How do I smooth such spectra? When collected through fiber, absorption around 950 nm makes the situation worst. Is there any way to smooth or fit it?
May be one of the best ways is to use sliding window smoothing average FIR filter.
You can use the DSP tool kit in matlab to select a suitable filter parameters.
You can use straight forward averaging process as:
y[n] = x[n] + x[n-1] + x[n-2] + ... up to N terms which is the window length. You experimentally find the proper value of N.
You can also make spectrum analysis of the signal to get the lowest frequency to be removed. Then the length of the window must at least equal one cycle time of the lowest frequency to be removed.
Best wishes
A very similar question was asked just a couple weeks ago.
https://www.researchgate.net/post/Could_anyone_suggest_how_to_remove_interference_patterns_from_reflectance_and_PL_spectra
This seems applicable.
Mark Schnittker Sir, I just went through the answers. I have tried doing FFT but this does not seem to be working around 950 nm since frequency is very different around this wavelength.
Abdelhalim Zekry Sir, can you explain the last paragraph of your answer a bit more?
Dear Raju,
When i inspect your signal i fined that there is a relatively high frequency oscillations are superimposed on the original signal. You want to remove these high frequency oscillations. You can use a low pass filter as i explained in my first post.
You can also make FFT analysis of your signal to locate the original signal at the lower frequency region and the high frequency contaminating signal.
Then you can determine the cut off frequency of your low pass filter by inspecting the FFT spectrum. I can help you to locate the cut off frequency when you make FFT and display it.
Best wishes
what is the sampling frequency you used and what is the value of the samples N you used to calculate your FFT. Can you enlarge the x axis near the maximum of the M plot? what is the repetition rate of the high frequency oscillations? If you enlarge the spectrum figure you can determine it. I counted about 15 peaks in 100A , then the frequency will be 15/100 A. The low freq period is about 150 A. Then a low pass filter having cut off frequency about 3/150 =2/100 will remove the high frequency oscillations and pass the required low frequency signal.
Best wishes
Abdelhalim Zekry Sir, I have used 775 samples to perform FFT, while the total input data points are 1548. The sampling interval in frequency is 0.00192. Since I am doing with origin, I am unable to extract those parameters, I may need some time. Why did you use '3'? Thank you for your explanation.
This is just an estimation of the cut off frequency you can change it around the value which i gave as possible value you can make it 2/150 A or 4/150 and see its effect on the high frequency oscillations.
Best wishes
Spectra can be smoothed in ORIGIN package.
Analysis: Smoothing
What information do you want to extract from the spectrum?
Why do you need to smooth it?
One can easily see the maximum and the bandwith
- Badges
- Science topic
- Similar topics
- Physical Sciences
- Physical Phenomena
- Luminescence
- Fluorescence