Looks like a weak cavity resonance.  Possibly an air gap of around 0.005 mm between two glass surfaces.

Could you have dust or debris preventing two PC fibre connectors from making physical contact, or an angled APC connector mated to a PC connector?  http://www.thefoa.org/tech/ref/basic/term.html

SMA connectors have an air gap by design, so it is more difficult to avoid reflections http://www.thefoa.org/tech/ref/termination/names.html

Try cleaning and re-making any connectors.  If you have mechanical splices or fibres aligned by micro-positioner, consider using index matching gel (or even a drop of liquid paraffin) to suppress reflections.  https://www.thorlabs.com/newgrouppage9.cfm?objectgroup_id=354

Good luck!

Dear Zare,

Most probably, the interference pattern is due to some fingerprints remained on some of the optical pieces that you are using in the setup. Clean up the surfaces, the unwanted interference will disappear. Good Luck!

It is hard to identify root cause without careful analysis of the whole measurements system. One of the potential causes is diffraction on the slit. To check it try with increased slit. In that case you will lose some resolution, but you will know the cause.

Another and more probable cause is that you have strong source so the higher orders of diffraction on your diffraction grating has an influence on signal. Try to use cut-on filter that transmits only radiation with wavelength higher than 400-450 nm, that will eliminate influence of the low wavelength and UV contents in quartz halogen lamp radiation have. This will limit your spectral measurement bandwidth, but you will see the influence and rearrange measurement system accordingly, measurements in the short wavelength region without filter and measurement in longer wavelength region with filter.

Hope that this will help. Also, try the previous recommendations. It is important to troubleshoot this issue carefully and eliminate this anomaly from future measurements.

Good luck

Branko,

Thanks of all.

I cleaned all of elements but the problem remains.

I think it should be because of diffraction of light from slit. So, at first  I thought I could get rid of it by applying a smoothing formula (Savitzky Golay). This solution was helpful for HW spectrum but there was another problem. Other discrete spectra like Hg lamp  spectrum became terrible.

Second, I thought the interference comes from the  second order of the spectrum, as Livanda mentioned.  I used an order sorting filter on the surface of CCD. But unfortunately, the problem also existed.

Third, I removed the slit and repeated the experience with a 400 micron core fiber optics. The problem was solved and I got a smooth curve. So, I guess by widening the slit the problem should be resolved. I decide to prepare some slits  with 50 or 100 micron width and replace them with 25 um one.

As forth experience, I changed the CCD with a 2048 pixels one. Again, I got a smooth worthy curve.

In conclusion, I think replacing CCD with a less pixels one and widening the slit (if it give a reasonable response) should solve the problem due to decreasing the resolution. 

Does anyone agree or have another idea?

A smaller slit may develop more diffraction, but the problem is the interference! Unless a complete setup is given, one can not trace the problem. 

Unless a complete setup is given, one can not trace the problem. 

One rough guess, maybe some back reflection from somewhere in the setup. 

Dear Dr. Seraji,

Thank you for your response. Because changing other elements is difficult for me, I will use wider slit to decrease the resolution. 

I think that your problem comes from the interference pattern that can occur due to the thin protective glass which is usually placed over the CCD pixels in order to protect them (in optics, you generate fringes of equal inclination inside the glass thickness). When you changed your CCD , for another one, it seemed that the problem has disappeared.  If you want to avoid these fringes, one usually removes this protective glass, which is rather delicate to do, but it can be done; or try another CCD from the spectrometer maker (with the same resolution); It may have some differences in its glass placement.

Dear Zare,

Is it possible in your setup to see the fringes (interference patterns) on a screen before detection by CCD? 

No, because the fringes are generated inside the protective glass and are seen by the CCD pixels which in close contact to the glass.

You must try to remove this protective glass or try another CCD which could have a different construction. For me it's the only solution.

The same fringe frequency pattern would be generated by air gap of 0.0042 mm between two glass surfaces. Protective glass with refraction index of 1.51 has to be 0.0028 mm thick to create such pattern, that is impossible.

However the amplitude of modulation cannot be explained just by interference between two glass surfaces. I can imagine possible interference in air gap between fiber and order selection filter on spectrometer entrance.

Can add just if it need to get exact emission source spectra (tungsten lamp for example) entire optical setup should be calibrated against light source with known spectra, like black body emission reference source.

Dear Zare,

Without changing the setup condition, you may perform all the suggestions given by the colleagues, one by one. Then in each step, check the output. 

Thank you Dr. Seraji.

I changed the slit yesterday but unfortunately the curve was unchanged.

Dear Zare,

Try to remove any "Back-Reflection". Could you give a schematic of the setup?

The standard spectrum of H-W light source is shown here for better consideration.

Dear Dr. Seraji,

The setup is in the black box and I think there is no back reflection. I attached the schematic of the setup configuration here.

Dear researchers,

Could it be due to etaloning effect?