If I touch the fiber probe which transfers the light to the spectrometer, these oscillations change dramatically. I can't filter them out as in the FFT of the spectrum, there is no clear peak for these oscillations and they only disappear if I apply a low pass filter with a cutoff frequency very close to central peak (zero frequency).

These oscillations are not random as they don't go away by averaging the scans and in fact over time they are quite stable.

But the problem is if I put the spectrometer's fiber probe (with a collimating lens attached) right in front of the collimated light before adding any other optical elements, the shape of the spectrum shows very strong high frequency oscillations. I have a high resolution spectrometer (50pm from Ocean Optics) to do it. My aim is to measure the change in the spectrum profile as the light travels through some optical elements.

In my setup (see attached file), I have a fiber coupled broadband light source (SLED) which I have collimated to use in free space.

Siemens star testchart